28.11.2014 Views

Removal of DBP Precursors by GAC Adsorption - Water Research ...

Removal of DBP Precursors by GAC Adsorption - Water Research ...

Removal of DBP Precursors by GAC Adsorption - Water Research ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

American <strong>Water</strong> Works Association<br />

RESEARCH FOUNDATION<br />

<strong>Removal</strong> <strong>of</strong><br />

<strong>DBP</strong> <strong>Precursors</strong> <strong>by</strong><br />

GAG <strong>Adsorption</strong><br />

Subject Area:<br />

<strong>Water</strong> Treatment


<strong>Removal</strong> <strong>of</strong><br />

<strong>DBP</strong> <strong>Precursors</strong> <strong>by</strong><br />

GAG <strong>Adsorption</strong>


The mission <strong>of</strong> the A WWA <strong>Research</strong> Foundation is to advance the science <strong>of</strong> water to<br />

improve the quality <strong>of</strong> life. Funded primarily through annual subscription payments from<br />

over 900 utilities, consulting firms, and manufacturers in North America and abroad,<br />

AWWARF sponsors research on all aspects <strong>of</strong> drinking water, including supply and<br />

resources, treatment, monitoring and analysis, distribution, management, and health<br />

effects.<br />

From its headquarters in Denver, Colorado, the A WWARF staff directs and supports<br />

the efforts <strong>of</strong> over 500 volunteers, who are the heart <strong>of</strong> the research program. These<br />

volunteers, serving on various boards and committees, use their expertise to select and<br />

monitor research studies to benefit the entire drinking water community.<br />

<strong>Research</strong> findings are disseminated through a number <strong>of</strong> technology transfer activi<br />

ties, including research reports, conferences, videotape summaries, and periodicals.


<strong>Removal</strong> <strong>of</strong><br />

DBF <strong>Precursors</strong> <strong>by</strong><br />

<strong>GAC</strong> <strong>Adsorption</strong><br />

Prepared <strong>by</strong><br />

Douglas M. Owen and Zaid K. Chowdhury<br />

Malcolm Pirnie, Inc.<br />

703 Palomar Airport Road, Suite 150<br />

Carlsbad, CA 92009<br />

R. Scott Summers, Stuart M. Hooper, and<br />

Gabriele Solarik<br />

Department <strong>of</strong> Civil and Environmental Engineering<br />

University <strong>of</strong> Cincinnati, Cincinnati, OH 45221<br />

Kimberly Gray<br />

Department <strong>of</strong> Civil Engineering<br />

Northwestern University, Evanston, IL 60208<br />

Sponsored <strong>by</strong>:<br />

AWWA <strong>Research</strong> Foundation<br />

6666 West Quincy Avenue<br />

Denver, CO 80235-3098<br />

Published <strong>by</strong> the<br />

AWWA <strong>Research</strong> Foundation and<br />

American <strong>Water</strong> Works Association


Disclaimer<br />

This study was funded <strong>by</strong> the AWWA <strong>Research</strong> Foundation (AWWARF).<br />

AWWARF assumes no responsibility for the content <strong>of</strong> the research study<br />

reported in this publication or for the opinions or statements <strong>of</strong> fact expressed in the<br />

report. The mention <strong>of</strong> trade names for commercial products does not represent or imply<br />

the approval or endorsement <strong>of</strong> AWWARF. This report is presented solely for informational purposes.<br />

Library <strong>of</strong> Congress Cataloging-in-Publication Data<br />

<strong>Removal</strong> <strong>of</strong> DBF precursors <strong>by</strong> gac adsorption / prepared <strong>by</strong> R. Scott<br />

Summers ... [et al.]; sponsored <strong>by</strong> AWWA <strong>Research</strong> Foundation,<br />

xxxiii, 284 p. 21.5x28cm.<br />

Includes bibliographical references.<br />

ISBN 0-89867-941-9<br />

1. <strong>Water</strong>-Purification-Organic compounds removal. 2. <strong>Water</strong>-<br />

Purification <strong>Adsorption</strong>. 3. Carbon, Activated. 4. <strong>Water</strong>-<br />

Purification-Disinfection-By-products. I. Summers, R. Scott.<br />

II. AWWA <strong>Research</strong> Foundation.<br />

TD427.07R46 1998<br />

628.1 '64 dc21 97-31908<br />

CIP<br />

Copy right 1998<br />

<strong>by</strong><br />

AWWA <strong>Research</strong> Foundation<br />

and<br />

American <strong>Water</strong> Works Association<br />

Printed in the U.S.A.<br />

ISBN 0-89867-941 -9 Printed on recycled paper.


Contents<br />

List <strong>of</strong> Tables ............................................... xi<br />

List <strong>of</strong> Figures ............................................... xv<br />

Foreword ................................................. xxv<br />

Acknowledgments ........................................... xxvii<br />

Executive Summary .......................................... xxix<br />

Chapter 1 Introduction ...................................... 1<br />

Regulatory Agenda, 2<br />

<strong>GAC</strong> Treatment Issues, 2<br />

Adsorbability <strong>of</strong> Natural Organic Matter, 2<br />

Blending <strong>of</strong> Contactor Effluents, 3<br />

DBF Speciation and Formation Conditions, 3<br />

Evaluation <strong>of</strong> Process Variables, 3<br />

Study Objectives, 4<br />

Chapter 2 Materials and Methods<br />

Experimental Approach, 7<br />

Testing Program, 7<br />

Sample Collection and Handling, 8<br />

Experimental Procedures, 8<br />

Analytical Methods, 12<br />

Standard Methods, 12<br />

Nonstandard Methods, 15<br />

Quality Assurance-Quality Control, 18<br />

Chapter 3 Using the RSSCT for Prediction <strong>of</strong> Field-Scale<br />

NOM <strong>Removal</strong> and DBF Control <strong>by</strong> <strong>GAC</strong><br />

<strong>Adsorption</strong> ....................................... 21<br />

Objectives and Approach, 22<br />

Results and Discussion, 23<br />

Ohio River <strong>Water</strong>, 23<br />

Lake Gaillard <strong>Water</strong>, 23


vi <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

Mississippi River <strong>Water</strong>, 24<br />

Passaic River <strong>Water</strong>, 24<br />

Summary, 24<br />

Chapter 4 Comparison <strong>of</strong> Breakthrough Patterns for<br />

Simulated Distribution System and Uniform<br />

Formation Condition DBFs ......................... 37<br />

Experimental Conditions, 37<br />

Results and Discussion, 38<br />

Ohio River <strong>Water</strong>, 38<br />

Lake Gaillard <strong>Water</strong>, 39<br />

Mississippi River <strong>Water</strong>, 39<br />

Summary, 40<br />

Chapter 5 NOM Characterization ............................ 47<br />

Objectives, 47<br />

Raw <strong>Water</strong> NOM Characterization, 48<br />

Approach, 48<br />

Results and Discussion, 48<br />

Field-Scale NOM Characterization, 51<br />

Approach, 51<br />

Results and Discussion, 51<br />

Bench-Scale NOM Characterization, 54<br />

Approach, 54<br />

Results and Discussion, 54<br />

Summary, 60<br />

Chapter 6 Design and Operational Considerations .............. 81<br />

Impact <strong>of</strong> EBCT on <strong>Adsorption</strong> Behavior, 82<br />

Objectives and Approach, 82<br />

Results and Discussion, 83<br />

Impact <strong>of</strong> Blending, 85<br />

Objectives and Approach, 87<br />

Results and Discussion, 87<br />

NOM Desorption and the Impact <strong>of</strong> Backwashing, 88<br />

Objectives and Approach, 89<br />

Results and Discussion, 90<br />

Summary, 90<br />

Impact <strong>of</strong> EBCT on <strong>Adsorption</strong> Behavior, 90<br />

Impact <strong>of</strong> Blending, 91<br />

NOM Desorption and the Impact <strong>of</strong> Backwashing, 91


Contents vii<br />

Chapter 7 The Impact <strong>of</strong> Optimized Coagulation on NOM<br />

<strong>Removal</strong> and DBF Control <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong> ....... 109<br />

Objectives and Approach, 109<br />

Experimental Conditions, 110<br />

Results and Discussion, 110<br />

Effect <strong>of</strong> Coagulation Pretreatment on<br />

Specific DBF Yield <strong>of</strong> <strong>GAC</strong> Effluent, 114<br />

Impact <strong>of</strong> Pretreatment on NOM Fraction Characteristics<br />

After <strong>GAC</strong>, 117<br />

Humic-Nonhumic Fractionation, 117<br />

Molecular Size Fractionation, 118<br />

PY-GC-MS Fractionation, 119<br />

Summary, 120<br />

Chapter 8 The Impact <strong>of</strong> Ozonation and Biological Filtration on<br />

NOM <strong>Removal</strong> and DBF Control <strong>by</strong> <strong>GAC</strong> ............ 135<br />

Objectives and Approach, 135<br />

Experimental Conditions, 136<br />

Impact <strong>of</strong> Pretreatment on <strong>GAC</strong> Breakthrough Behavior:<br />

Results and Discussion, 136<br />

Effect <strong>of</strong> Ozonation and Biological Filtration on Specific<br />

<strong>DBP</strong> Yield <strong>of</strong> <strong>GAC</strong>, 140<br />

Impact <strong>of</strong> Ozonation and Biological Filtration on NOM<br />

Fraction Characteristics After <strong>GAC</strong>, 142<br />

Humic-Nonhumic Fractionation, 142<br />

Molecular Size Fractionation, 143<br />

PY-GC-MS Fractionation, 144<br />

Summary, 146<br />

Chapter 9 Impact <strong>of</strong> Treatment on <strong>DBP</strong> Speciation .............. 167<br />

Results and Discussion, 167<br />

<strong>GAC</strong> Effluent <strong>DBP</strong> Speciation, 167<br />

Bromine Incorporation Factor, 168<br />

Optimized Coagulation, 169<br />

Ozonation and Biological Filtration, 170<br />

Summary, 172<br />

Chapter 10 The Impact <strong>of</strong> NOM Preloading on SOC <strong>Adsorption</strong><br />

Behavior ........................................ 181<br />

Objectives and Approach, 181<br />

Results and Discussion, 182<br />

Summary, 184


viii <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

Chapter 11 Relationship Among NOM Characteristics, <strong>Removal</strong><br />

<strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> Treatment, and<br />

DBF Formation .................................. 189<br />

Overview <strong>of</strong> Characterization Techniques, 189<br />

Molecular Size Fractionation, 190<br />

Humic-Nonhumic Fractionation, 191<br />

The Role <strong>of</strong> TOC and UV2J4 , 192<br />

Pyrolysis GC-MS, 192<br />

Raw <strong>Water</strong> Characteristics, 193<br />

Effect <strong>of</strong> Coagulation, 193<br />

Effect <strong>of</strong> Ozone and Bi<strong>of</strong>iltration, 194<br />

Chapter 12 Evaluation <strong>of</strong> Design Criteria and Costs .............. 199<br />

<strong>GAC</strong> Design Parameters, 199<br />

Empty Bed Contact Time, 200<br />

Carbon Usage Rate, 200<br />

Effect <strong>of</strong> Pretreatment, 201<br />

Determination <strong>of</strong> Design Criteria, 201<br />

<strong>Water</strong> Quality Goals, 201<br />

Single-Contactor Breakthrough Curves, 202<br />

Effect <strong>of</strong> Blending From Multiple Contactors, 204<br />

Cost Curve Development, 204<br />

Process Components for <strong>GAC</strong>, 204<br />

Process Components for Pretreatment, 205<br />

Design Parameters, 205<br />

Cost Factors, 206<br />

Cost Curves, 208<br />

Verification <strong>of</strong> Costs, 208<br />

<strong>GAC</strong> Treatment Costs for Participating Utilities, 210<br />

Cincinnati <strong>Water</strong>works, 210<br />

City <strong>of</strong> Phoenix, 212<br />

Passaic Valley <strong>Water</strong> Commission, 212<br />

Florida Cities <strong>Water</strong> Company, 213<br />

Implications for Process Optimization, 213<br />

Empty Bed Contact Time, 213<br />

Pretreatment Optimization, 214<br />

Chapter 13 Conclusions and Recommendations .................. 223<br />

Verification <strong>of</strong> RSSCT Results, 223<br />

Impact <strong>of</strong> <strong>GAC</strong> Treatment on NOM Characteristics and<br />

<strong>DBP</strong> Yield, 224


Contents ix<br />

Impact <strong>of</strong> Operational Characteristics and Pretreatment<br />

on <strong>GAC</strong> Performance, 225<br />

Estimation <strong>of</strong> Design Criteria and Conceptual Costs<br />

for <strong>GAC</strong> Treatment, 226<br />

Appendix A: RSSCT Design .................................. 229<br />

Appendix B: A Statistical Method for the Comparison <strong>of</strong><br />

RSSCT and Field-Scale <strong>GAC</strong> Breakthrough<br />

Curves ........................................ 231<br />

Appendix C: Development <strong>of</strong> a New Test for Disinfection<br />

By-Product Formation Assessment: Uniform<br />

Formation Conditions ............................ 247<br />

References ................................................. 277<br />

List <strong>of</strong> Abbreviations ......................................... 283


Tables<br />

ES.l Participating utilities and source water characteristics xxxi<br />

ES.2 Summary <strong>of</strong> experimental program xxxi<br />

ES.3 Parameters for the UFC test xxxiii<br />

2.1 Summary <strong>of</strong> water quality parameters for water sources 6<br />

2.2 Field-scale <strong>GAC</strong> systems 7<br />

2.3 Summary <strong>of</strong> experimental program 8<br />

2.4 Laboratories performing sample analyses for experimental<br />

program 9<br />

2.5 Field-scale and RSSCT verification operation parameters 10<br />

2.6 Summary <strong>of</strong> analytical methods 13<br />

2.7 UFC and SDS chlorination conditions 15<br />

2.8 Specifications for PY-GC-MS analysis 18<br />

2.9 Summary <strong>of</strong> detection limits and precisions for analytical<br />

methods 19<br />

3.1 RSSCT and field-scale <strong>GAC</strong> contactor design parameters 22<br />

4.1 UFC and SDS chlorination conditions 38<br />

4.2 DBF formation in <strong>GAC</strong> influent under UFC and SDS<br />

conditions 39<br />

5.1 Summary <strong>of</strong> raw water fractionation and BDOC results 49<br />

5.2 Specific TTHM yields <strong>of</strong> raw waters 49<br />

5.3 PY-GC-MS classification <strong>of</strong> raw waters 49<br />

5.4 Summary <strong>of</strong> seasonal water quality monitoring BDOC results 50<br />

5.5 PY-GC-MS classification <strong>of</strong> seasonal samples 51<br />

5.6 Summary <strong>of</strong> field-scale <strong>GAC</strong> influent humic-nonhumic<br />

fractionation results for ORW, LOW, MRW, and PRW 52<br />

5.7 Summary <strong>of</strong> ORW and PRW field-scale BDOC results 53<br />

5.8 Summary <strong>of</strong> bench-scale <strong>GAC</strong> influent fractionation results 55<br />

5.9 Specific TTHM yields <strong>of</strong> bench-scale <strong>GAC</strong> influents 56<br />

5.10 Summary <strong>of</strong> BDOC results for PRW at bench scale 5 8<br />

5.11 PY-GC-MS classification <strong>of</strong> Ohio River water 5 8<br />

5.12 PY-GC-MS classification <strong>of</strong> Lake Gaillard water 58<br />

5.13 PY-GC-MS classification <strong>of</strong> Mississippi River water 59<br />

5.14 PY-GC-MS classification <strong>of</strong> Passaic River water 59


xii <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> GA C <strong>Adsorption</strong><br />

5.15 PY-GC-MS classification <strong>of</strong> Salt River Project water 61<br />

5.16 PY-GC-MS classification <strong>of</strong> Florida groundwater 61<br />

5.17 Average DOC distribution <strong>of</strong> river waters with<br />

breakthrough: Nonhumic and humic fractions 62<br />

5.18 Average DOC distribution <strong>of</strong> river waters with<br />

breakthrough: Molecular size fractions 62<br />

6.1 Summary <strong>of</strong> bed lives for different effluent criteria 84<br />

6.2 Impact <strong>of</strong> blending on bed life 89<br />

7.1 Raw water characteristics for Chapter 7 experiments 111<br />

7.2 Treated water characteristics for Chapter 7 experiments 111<br />

7.3 Summary <strong>of</strong> <strong>DBP</strong> surrogate and DBF concentrations prior to<br />

<strong>GAC</strong> treatment 112<br />

7.4 Effect <strong>of</strong> coagulation pretreatment on <strong>GAC</strong> breakthrough<br />

characteristics 113<br />

7.5 Effect <strong>of</strong> coagulation pretreatment on relationship between<br />

<strong>GAC</strong> effluent TOC and <strong>DBP</strong> formation 116<br />

7.6 Impact <strong>of</strong> coagulation treatment on humic-nonhumic and<br />

molecular size fractionation 117<br />

7.7 PY-GC-MS classification <strong>of</strong> ORW: Enhanced coagulation 120<br />

7.8 PY-GC-MS classification <strong>of</strong> SRPW: Enhanced coagulation 121<br />

8.1 Summary <strong>of</strong> experimental conditions for ozonation and<br />

biological filtration 137<br />

8.2 Summary <strong>of</strong> <strong>GAC</strong> influent characteristics after ozonation<br />

and biological filtration 137<br />

8.3 Summary <strong>of</strong> impact <strong>of</strong> ozonation and biological filtration on<br />

<strong>GAC</strong> performance 138<br />

8.4 Summary <strong>of</strong> the effect <strong>of</strong> ozonation and biotreatment on the<br />

relationship between TOC and <strong>DBP</strong> formation 141<br />

8.5 Impact <strong>of</strong> ozonation and biotreatment on humic-nonhumic<br />

and MS fractionation results 143<br />

8.6 PY-GC-MS classification <strong>of</strong> ORW: Ozonation and<br />

biotreatment 145<br />

8.7 PY-GC-MS classification <strong>of</strong> PRW: Ozonation and<br />

biotreatment 145<br />

8.8 PY-GC-MS classification <strong>of</strong> FGW: Ozonation and<br />

biotreatment 145<br />

9.1 Effect <strong>of</strong> optimized coagulation on <strong>DBP</strong> speciation 169<br />

9.2 Effect <strong>of</strong> ozonation and biological filtration on <strong>DBP</strong><br />

speciation 171<br />

10.1 Summary <strong>of</strong> operating conditions for preloaded <strong>GAC</strong> 182


Tables<br />

xiii<br />

10.2 Comparison <strong>of</strong> <strong>GAC</strong> residual capacity for DCE 183<br />

12.1 Estimates <strong>of</strong> CURs for the participating utilities 203<br />

12.2 Treatment plant flows considered for cost curves 206<br />

12.3 Indices used in developing cost curves 207<br />

12.4 Factors applied to construction cost estimates 207<br />

12.5 Unit costs for building and operating materials 207<br />

12.6 Comparison <strong>of</strong> costs for the <strong>GAC</strong> processes 209<br />

12.7 Comparison <strong>of</strong> costs for optimized coagulation 209<br />

12.8 Comparison <strong>of</strong> ozone facilities costs 210<br />

12.9 Estimates <strong>of</strong> <strong>GAC</strong> treatment costs for the participating<br />

utilities 211


Figures<br />

3.1 Full-scale and RSSCT TOC breakthrough for Ohio River<br />

water 26<br />

3.2 Full-scale and RSSCT UV254 breakthrough for Ohio River<br />

water 26<br />

3.3 Full-scale and RSSCT SDS-TOX breakthrough for Ohio<br />

River water 27<br />

3.4 Full-scale and RSSCT SDS-TTHM breakthrough for Ohio<br />

River water 27<br />

3.5 Full-scale and RSSCT SDS-HAA6 breakthrough for Ohio<br />

River water 28<br />

3.6 Full-scale and RSSCT SDS-CH breakthrough for Ohio River<br />

water 28<br />

3.7 Pilot-scale and RSSCT TOC breakthrough for Lake Gaillard<br />

water 29<br />

3.8 Pilot-scale and RSSCT SDS-TOX breakthrough for Lake<br />

Gaillard water 29<br />

3.9 Pilot-scale and RSSCT SDS-TTHM breakthrough for Lake<br />

Gaillard water 30<br />

3.10 Pilot-scale and RSSCT SDS-HAA5 breakthrough for Lake<br />

Gaillard water 30<br />

3.11 Pilot-scale and RSSCT TOC breakthrough for Mississippi<br />

River water 31<br />

3.12 Pilot-scale and RSSCT UV254 breakthrough for Mississippi<br />

River water 31<br />

3.13 Pilot-scale and RSSCT SDS-TOX breakthrough for<br />

Mississippi River water 32<br />

3.14 Pilot-scale and RSSCT SDS-TTHM breakthrough for<br />

Mississippi River water 32<br />

3.15 Pilot-scale and RSSCT SDS-HAA6 breakthrough for<br />

Mississippi River water 33<br />

3.16 Pilot-scale and RSSCT SDS-CH breakthrough for<br />

Mississippi River water 33<br />

3.17 Pilot-scale and RSSCT TOC breakthrough for Passaic<br />

River water 34<br />

3.18 Pilot-scale and RSSCT UV 254 breakthrough for Passaic<br />

River water 34<br />

XV


xvi <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

3.19 Pilot-scale and RSSCT SDS-TOX breakthrough for Passaic<br />

River water 35<br />

3.20 Pilot-scale and RSSCT SDS-TTHM breakthrough for Passaic<br />

River water 35<br />

3.21 Pilot-scale and RSSCT SDS-HAA6 breakthrough for Passaic<br />

River water 36<br />

3.22 Pilot-scale and RSSCT SDS-CH breakthrough for Passaic River<br />

water 36<br />

4.1 Effect <strong>of</strong> chlorination conditions on RSSCT TTHM<br />

breakthrough for Ohio River water 41<br />

4.2 Effect <strong>of</strong> chlorination conditions on RSSCT HAA6<br />

breakthrough for Ohio River water 41<br />

4.3 Effect <strong>of</strong> chlorination conditions on RSSCT TOX<br />

breakthrough for Ohio River water 42<br />

4.4 Effect <strong>of</strong> chlorination conditions on RSSCT TTHM<br />

breakthrough for Lake Gaillard water 42<br />

4.5 Effect <strong>of</strong> chlorination conditions on RSSCT HAA5<br />

breakthrough for Lake Gaillard water 43<br />

4.6 Effect <strong>of</strong> chlorination conditions on RSSCT TOX<br />

breakthrough for Lake Gaillard water 43<br />

4.7 Effect <strong>of</strong> chlorination conditions on RSSCT TTHM<br />

breakthrough for Mississippi River water 44<br />

4.8 Effect <strong>of</strong> chlorination conditions on RSSCT HAA6<br />

breakthrough for Mississippi River water 44<br />

4.9 Effect <strong>of</strong> chlorination conditions on RSSCT TOX<br />

breakthrough for Mississippi River water 45<br />

4.10 Effect <strong>of</strong> chlorination conditions on RSSCT CH<br />

breakthrough for Mississippi River water 45<br />

5.1 Seasonal variation <strong>of</strong> nonhumic and humic fractions for<br />

Ohio River raw water 63<br />

5.2 Specific TTHM yields <strong>of</strong> nonhumic and humic fractions for<br />

Ohio River seasonal raw water samples 63<br />

5.3 Seasonal variation <strong>of</strong> nonhumic and humic fractions for<br />

Passaic River raw water 64<br />

5.4 Specific TTHM yields <strong>of</strong> nonhumic and humic fractions for<br />

Passaic River seasonal raw water samples 64<br />

5.5 Full-scale nonhumic and humic fraction breakthrough for<br />

Ohio River water 65<br />

5.6 Full-scale nonhumic and humic fraction breakthrough for<br />

Lake Gaillard water 65<br />

5.7 Full-scale nonhumic and humic fraction breakthrough for<br />

Mississippi River water 66


Figures xvii<br />

5.8 Pilot-scale nonhumic and humic fraction breakthrough for<br />

Passaic River water 66<br />

5.9 Pilot-scale BDOC breakthrough for Mississippi River water 67<br />

5.10 Example curve showing sampling points for NOM<br />

characterization at bench scale 67<br />

5.11 Bench-scale nonhumic and humic fraction<br />

breakthrough for Ohio River water 68<br />

5.12 Bench-scale specific TTHM yields <strong>of</strong> humic<br />

and nonhumic fractions for Ohio River water 68<br />

5.13 Bench-scale nonhumic and humic fraction<br />

breakthrough for Lake Gaillard water 69<br />

5.14 Bench-scale specific TTHM yields <strong>of</strong> humic<br />

and nonhumic fractions for Lake Gaillard water 69<br />

5.15 Bench-scale nonhumic and humic fraction<br />

breakthrough for Mississippi River water 70<br />

5.16 Bench-scale specific TTHM yields <strong>of</strong> humic and<br />

nonhumic fractions for Mississippi River water 70<br />

5.17 Bench-scale nonhumic and humic fraction<br />

breakthrough for Passaic River water 71<br />

5.18 Bench-scale specific TTHM yields <strong>of</strong> nonhumic<br />

and humic fractions for Passaic River water 71<br />

5.19 Bench-scale nonhumic and humic fraction<br />

breakthrough for Salt River Project water 72<br />

5.20 Bench-scale specific TTHM yields <strong>of</strong> nonhumic<br />

and humic fractions for Salt River Project water 72<br />

5.21 Bench-scale nonhumic and humic fraction<br />

breakthrough for Florida groundwater 73<br />

5.22 Bench-scale specific TTHM yields <strong>of</strong> nonhumic<br />

and humic fractions for Florida groundwater 73<br />

5.23 Molecular size fraction breakthrough for Ohio River water 74<br />

5.24 Specific TTHM yields <strong>of</strong> MS fractions for Ohio River water 74<br />

5.25 Molecular size fraction breakthrough for Lake Gaillard water 75<br />

5.26 Specific TTHM yields <strong>of</strong> MS fractions for Lake Gaillard water 75<br />

5.27 Molecular size fraction breakthrough for Mississippi River<br />

water 76<br />

5.28 Specific TTHM yields <strong>of</strong> MS fractions for Mississippi River<br />

water 76<br />

5.29 Molecular size fraction breakthrough for Passaic River water 77<br />

5.30 Specific TTHM yields <strong>of</strong> MS fractions for Passaic River water 77<br />

5.31 Molecular size fraction breakthrough for Salt River Project<br />

water 78<br />

5.32 Specific TTHM yields <strong>of</strong> MS fractions for Salt River Project<br />

water 78<br />

5.33 Molecular size fraction breakthrough for Florida groundwater 79<br />

5.34 Specific TTHM yields <strong>of</strong> MS fractions for Florida groundwater 79


xviii <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

5.35 Effect <strong>of</strong> BR~:DOC on DOC fractions for <strong>GAC</strong> influent for<br />

six waters 80<br />

6.1 Model simulation <strong>of</strong> the effect <strong>of</strong> EBCT on TOC breakthrough<br />

for the Ohio River water 92<br />

6.2 Model simulation <strong>of</strong> the effect <strong>of</strong> EBCT on normalized TOC<br />

breakthrough for Ohio River water 92<br />

6.3 Effect <strong>of</strong> EBCT on TOC breakthrough for Ohio River water 93<br />

6.4 Effect <strong>of</strong> EBCT on normalized TOC breakthrough for Ohio<br />

River water 93<br />

6.5 Effect <strong>of</strong> EBCT on normalized UV254 breakthrough for Ohio<br />

River water 94<br />

6.6 Effect <strong>of</strong> EBCT on normalized UFC-TOX breakthrough for<br />

Ohio River water 94<br />

6.7 Effect <strong>of</strong> EBCT on normalized UFC-TTHM breakthrough for<br />

Ohio River water 95<br />

6.8 Effect <strong>of</strong> EBCT on normalized UFC-HAA6 breakthrough for<br />

Ohio River water 95<br />

6.9 Effect <strong>of</strong> EBCT on TOC breakthrough for Passaic River water 96<br />

6.10 Effect <strong>of</strong> EBCT on normalized TOC breakthrough for Passaic<br />

River water 96<br />

6.11 Effect <strong>of</strong> EBCT on normalized UFC-TTHM breakthrough for<br />

Passaic River water 97<br />

6.12 Effect <strong>of</strong> EBCT on TOC breakthrough for Salt River Project<br />

water 97<br />

6.13 Effect <strong>of</strong> EBCT on normalized TOC breakthrough for Salt<br />

River Project water 98<br />

6.14 Effect <strong>of</strong> EBCT on normalized UFC-TOX breakthrough for<br />

Salt River Project water 98<br />

6.15 Effect <strong>of</strong> EBCT on TOC breakthrough for Florida groundwater 99<br />

6.16 Effect <strong>of</strong> EBCT on normalized TOC breakthrough for Florida<br />

groundwater 99<br />

6.17 Effect <strong>of</strong> EBCT on normalized UFC-TTHM breakthrough for<br />

Florida groundwater 100<br />

6.18 Operation <strong>of</strong> multiple <strong>GAC</strong> contactors in parallel 100<br />

6.19 Individual breakthrough curves <strong>of</strong> multiple contactors operated<br />

in parallel 101<br />

6.20 Blended <strong>GAC</strong> effluent from multiple contactors operated in<br />

parallel 101<br />

6.21 Effect <strong>of</strong> blending on TOC breakthrough for Passaic River<br />

water 102<br />

6.22 Effect <strong>of</strong> blending on UFC-TTHM breakthrough for Passaic<br />

River water 102<br />

6.23 Effect <strong>of</strong> blending on TOC breakthrough for Salt River<br />

Project water 103


Figures xix<br />

6.24 Effect <strong>of</strong> blending on UFC-TTHM breakthrough for Salt River<br />

Project water 103<br />

6.25 Effect <strong>of</strong> blending on TOC breakthrough for Florida<br />

groundwater 104<br />

6.26 Effect <strong>of</strong> blending on UFC-TTHM breakthrough for Florida<br />

groundwater 104<br />

6.27 NOM desorption for Passaic River water: TOC levels 105<br />

6.28 NOM desorption for Passaic River water: UV254 levels 105<br />

6.29 Impact <strong>of</strong> backwashing on TOC breakthrough for Passaic<br />

River water 106<br />

6.30 Impact <strong>of</strong> backwashing on UV254 breakthrough for Passaic<br />

River water 106<br />

6.31 Impact <strong>of</strong> backwashing on UFC-TTHM breakthrough for<br />

Passaic River water 107<br />

7.1 Effect <strong>of</strong> coagulation pretreatment on TOC breakthrough for<br />

Ohio River water 122<br />

7.2 Effect <strong>of</strong> coagulation pretreatment on UV254 breakthrough for<br />

Ohio River water 122<br />

7.3 Effect <strong>of</strong> coagulation pretreatment on TTHM breakthrough for<br />

Ohio River water 123<br />

7.4 Effect <strong>of</strong> coagulation pretreatment on HAA6 breakthrough for<br />

Ohio River water 123<br />

7.5 Effect <strong>of</strong> coagulation pretreatment on TOX breakthrough for<br />

Ohio River water 124<br />

7.6 Effect <strong>of</strong> coagulation pretreatment on CH breakthrough for<br />

Ohio River water 124<br />

7.7 Effect <strong>of</strong> coagulation pretreatment on TOC breakthrough for<br />

Salt River Project water 125<br />

7.8 Effect <strong>of</strong> coagulation pretreatment on UV254 breakthrough for<br />

Salt River Project water 125<br />

7.9 Effect <strong>of</strong> coagulation pretreatment on TTHM breakthrough for<br />

Salt River Project water 126<br />

7.10 Effect <strong>of</strong> coagulation pretreatment on HAA6 breakthrough for<br />

Salt River Project water 126<br />

7.11 Effect <strong>of</strong> coagulation pretreatment on TOX breakthrough for<br />

Salt River Project water 127<br />

7.12 Effect <strong>of</strong> coagulation pretreatment on CH breakthrough for<br />

Salt River Project water 127<br />

7.13 Effect <strong>of</strong> coagulation pretreatment on the relationship between<br />

TOC and TTHM formation during <strong>GAC</strong> breakthrough for<br />

Ohio River water 128<br />

7.14 Four possible outcomes based on regression analysis with<br />

categorical variables 128


xx <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

7.15 Effect <strong>of</strong> coagulation pretreatment on the relationship between<br />

TOC and HAA6 formation during <strong>GAC</strong> breakthrough for<br />

Ohio River water 129<br />

7.16 Effect <strong>of</strong> coagulation pretreatment on the relationship between<br />

TOC and TOX formation during <strong>GAC</strong> breakthrough for Ohio<br />

River water 129<br />

7.17 Effect <strong>of</strong> coagulation pretreatment on the relationship beween<br />

TOC and TTHM formation during <strong>GAC</strong> breakthrough for<br />

Salt River Project water 130<br />

7.18 Effect <strong>of</strong> coagulation pretreatment on the relationship between<br />

TOC and HAA6 formation during <strong>GAC</strong> breakthrough for<br />

Salt River Project water 130<br />

7.19 Effect <strong>of</strong> coagulation pretreatment on the relationship between<br />

TOC and TOX formation during <strong>GAC</strong> breakthrough for<br />

Salt River Project water 131<br />

7.20 Effect <strong>of</strong> coagulation pretreatment on humic and nonhumic<br />

fraction breakthrough for Ohio River water 131<br />

7.21 Effect <strong>of</strong> coagulation pretreatment on the specific UFC-TTHM<br />

yields <strong>of</strong> humic-nonhumic fractions for Ohio River water 132<br />

7.22 Effect <strong>of</strong> coagulation pretreatment on humic and nonhumic<br />

fraction breakthrough for Salt River Project water 132<br />

7.23 Effect <strong>of</strong> coagulation pretreatment on molecular size fraction<br />

breakthrough for Ohio River water 133<br />

7.24 Effect <strong>of</strong> coagulation pretreatment on the specific UFC-TTHM<br />

yields <strong>of</strong> molecular size fractions for Ohio River water 133<br />

7.25 Effect <strong>of</strong> coagulation pretreatment on molecular size fraction<br />

breakthrough for Salt River Project water 134<br />

7.26 Effect <strong>of</strong> coagulation pretreatment on the specific UFC-TTHM<br />

yields <strong>of</strong> molecular size fractions for Salt River Project water 134<br />

8.1 Effect <strong>of</strong> ozonation and biotreatment on normalized TOC<br />

breakthrough for Ohio River water 148<br />

8.2 Effect <strong>of</strong> ozonation and biotreatment on TOC breakthrough for<br />

Ohio River water 148<br />

8.3 Effect <strong>of</strong> ozonation and biotreatment on UV254 breakthrough for<br />

Ohio River water 149<br />

8.4 Effect <strong>of</strong> ozonation and biotreatment on UFC-TOX breakthrough<br />

for Ohio River water 149<br />

8.5 Effect <strong>of</strong> ozonation and biotreatment on UFC-TTHM<br />

breakthrough for Ohio River water 150<br />

8.6 Effect <strong>of</strong> ozonation and biotreatment on UFC-HAA6<br />

breakthrough for Ohio River water 150<br />

8.7 Effect <strong>of</strong> ozonation and biotreatment on normalized TOC<br />

breakthrough for Passaic River water 151


Figures xxi<br />

8.8 Effect <strong>of</strong> ozonation and biotreatment on TOC breakthrough for<br />

Passaic River water 151<br />

8.9 Effect <strong>of</strong> ozonation and biotreatment on UV254 breakthrough for<br />

Passaic River water 152<br />

8.10 Effect <strong>of</strong> ozonation and biotreatment on UFC-TOX<br />

breakthrough for Passaic River water 152<br />

8.11 Effect <strong>of</strong> ozonation and biotreatment on UFC-TTHM<br />

breakthrough for Passaic River water 153<br />

8.12 Effect <strong>of</strong> ozonation and biotreatment on UFC-HAA6<br />

breakthrough for Passaic River water 153<br />

8.13 Effect <strong>of</strong> ozonation and biotreatment on normalized TOC<br />

breakthrough for Florida groundwater 154<br />

8.14 Effect <strong>of</strong> ozonation and biotreatment on TOC breakthrough for<br />

Florida groundwater 154<br />

8.15 Effect <strong>of</strong> ozonation and biotreatment on UV254 breakthrough for<br />

Florida groundwater 155<br />

8.16 Effect <strong>of</strong> ozonation and biotreatment on UFC-TOX breakthrough<br />

for Florida groundwater 155<br />

8.17 Effect <strong>of</strong> ozonation and biotreatment on UFC-TTHM<br />

breakthrough for Florida groundwater 156<br />

8.18 Effect <strong>of</strong> ozonation and biotreatment on UFC-HAA6<br />

breakthrough for Florida groundwater 156<br />

8.19 Effect <strong>of</strong> ozonation and biotreatment on the relationship<br />

between TOC and TOX formation during <strong>GAC</strong><br />

breakthrough for Ohio River water 157<br />

8.20 Effect <strong>of</strong> ozonation and biotreatment on the relationship<br />

between TOC and TTHM formation during <strong>GAC</strong><br />

breakthrough for Ohio River water 157<br />

8.21 Effect <strong>of</strong> ozonation and biotreatment on the relationship<br />

between TOC and HAA6 formation during <strong>GAC</strong><br />

breakthrough for Ohio River water 158<br />

8.22 Effect <strong>of</strong> ozonation and biotreatment on the relationship<br />

between TOC and TOX formation during <strong>GAC</strong> breakthrough<br />

for Passaic River water 158<br />

8.23 Effect <strong>of</strong> ozonation and biotreatment on the relationship<br />

between TOC and TTHM formation during <strong>GAC</strong><br />

breakthrough for Passaic River water 159<br />

8.24 Effect <strong>of</strong> ozonation and biotreatment on the relationship<br />

between TOC and HAA6 formation during <strong>GAC</strong><br />

breakthrough for Passaic River water 159<br />

8.25 Effect <strong>of</strong> ozonation and biotreatment on the relationship<br />

between TOC and TOX formation during <strong>GAC</strong> breakthrough<br />

for Florida groundwater 160


xxii <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> GA C <strong>Adsorption</strong><br />

8.26 Effect <strong>of</strong> ozonation and biotreatment on the relationship<br />

between TOC and TTHM formation during <strong>GAC</strong><br />

breakthrough for Florida groundwater 160<br />

8.27 Effect <strong>of</strong> ozonation and biotreatment on the relationship<br />

between TOC and HAA6 formation during <strong>GAC</strong><br />

breakthrough for Florida groundwater 161<br />

8.28 Effect <strong>of</strong> ozonation and biotreatment on humic-nonhumic<br />

fraction breakthrough for Ohio River water 161<br />

8.29 Effect <strong>of</strong> ozonation and biotreatment on humic-nonhumic<br />

fraction specific yields for Ohio River water 162<br />

8.30 Effect <strong>of</strong> ozonation and biotreatment on humic-nonhumic<br />

fraction breakthrough for Passaic River water 162<br />

8.31 Effect <strong>of</strong> ozonation and biotreatment on humic-nonhumic<br />

fractionation for Florida groundwater 163<br />

8.32 Effect <strong>of</strong> ozonation and biotreatment on humic-nonhumic<br />

fraction specific yields for Passaic River water 163<br />

8.33 Effect <strong>of</strong> ozonation and biotreatment on humic-nonhumic<br />

fraction specific yields for Florida groundwater 164<br />

8.34 Effect <strong>of</strong> ozonation and biotreatment on MS fraction<br />

breakthrough for Ohio River water 164<br />

8.35 Effect <strong>of</strong> ozonation and biotreatment on specific yields for MS<br />

fractions for Ohio River water 165<br />

8.36 Effect <strong>of</strong> ozonation and biotreatment on MS fraction<br />

breakthrough for Passaic River water 165<br />

8.37 Effect <strong>of</strong> ozonation and biotreatment on MS fraction<br />

breakthrough for Florida groundwater 166<br />

8.38 Effect <strong>of</strong> ozonation and biotreatment on specific yields for<br />

MS fractions for Passaic River water 166<br />

9.1 <strong>GAC</strong> breakthrough <strong>of</strong> THM species in Ohio River water after<br />

conventional treatment 173<br />

9.2 <strong>GAC</strong> breakthrough <strong>of</strong> THM species in Ohio River water after<br />

optimized coagulation 173<br />

9.3 <strong>GAC</strong> breakthrough <strong>of</strong> THM species in Salt River Project water<br />

after conventional treatment 174<br />

9.4 <strong>GAC</strong> breakthrough <strong>of</strong> HAA species in Salt River Project water<br />

after optimized coagulation 174<br />

9.5 <strong>GAC</strong> breakthrough <strong>of</strong> THM species in Passaic River water after<br />

conventional treatment 175<br />

9.6 <strong>GAC</strong> breakthrough <strong>of</strong> THM species in Passaic River water after<br />

ozonation and biotreatment 175<br />

9.7 <strong>GAC</strong> breakthrough <strong>of</strong> HAA species in Florida groundwater after<br />

conventional treatment 176<br />

9.8 <strong>GAC</strong> breakthrough <strong>of</strong> HAA species in Florida groundwater after<br />

ozonation and biotreatment 176


Figures xxiii<br />

9.9 Effect <strong>of</strong> treatment on THM bromine incorporation factor (n)<br />

forORW 177<br />

9.10 Effect <strong>of</strong> treatment on THM bromine incorporation factor (n) for<br />

SRPW 177<br />

9.11 Impact <strong>of</strong> pretreatment on THM bromine incorporation factor (n)<br />

for Passaic River water 178<br />

9.12 Impact <strong>of</strong> pretreatment on THM bromine incorporation factor (n)<br />

for Florida groundwater 178<br />

9.13 Effect <strong>of</strong> treatment on HAA bromine incorporation factor (n') for<br />

ORW 179<br />

9.14 Effect <strong>of</strong> treatment on HAA bromine incorporation factor (n') for<br />

Salt River Project water 179<br />

9.15 Effect <strong>of</strong> pretreatment on HAA bromine incorporation factor (n')<br />

for Passaic River water 180<br />

9.16 Effect <strong>of</strong> pretreatment on HAA bromine incorporation factor (n')<br />

for Florida groundwater 180<br />

10.1 Comparison <strong>of</strong> <strong>GAC</strong> adsorption capacity for DCE under different<br />

matrix conditions <strong>of</strong> Passaic River water 185<br />

10.2 Impact <strong>of</strong> preloading <strong>of</strong> Ohio River water at full and small scale<br />

on DCE adsorption capacity 185<br />

10.3 Impact <strong>of</strong> preloading <strong>of</strong> Lake Gaillard water at pilot and small<br />

scale on DCE adsorption capacity 186<br />

10.4 Impact <strong>of</strong> preloading <strong>of</strong> Mississippi River water at small scale<br />

on DCE adsorption capacity 186<br />

10.5 Impact <strong>of</strong> preloading <strong>of</strong> Salt River Project water at small scale<br />

on DCE adsorption capacity 187<br />

10.6 Impact <strong>of</strong> preloading <strong>of</strong> Florida groundwater at small scale on<br />

DCE adsorption capacity 187<br />

10.7 Correlation between solid phase DCE concentration and<br />

throughput in bed volumes for six sites 188<br />

11.1 Comparison <strong>of</strong> relative breakthroughs for NOM and UFC-<strong>DBP</strong>s<br />

for Ohio River water 196<br />

11.2 Comparison <strong>of</strong> relative breakthroughs for NOM and UFC-<strong>DBP</strong>s<br />

for Mississippi River water 196<br />

11.3 Comparison <strong>of</strong> relative breakthroughs for NOM and UFC-<strong>DBP</strong>s<br />

for Passaic River water 197<br />

11.4 Comparison <strong>of</strong> relative breakthroughs for NOM and UFC-<strong>DBP</strong>s<br />

for Salt River Project water 197<br />

11.5 Comparison <strong>of</strong> relative breakthroughs for NOM and UFC-<strong>DBP</strong>s<br />

for Florida groundwater 198<br />

12.1 Schematic <strong>of</strong> process components for <strong>GAC</strong> 216<br />

12.2 Schematic <strong>of</strong> process components for optimized coagulation 216


xxiv <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> C AC <strong>Adsorption</strong><br />

12.3 Process schematic for ozonation and bi<strong>of</strong>iltration treatment 217<br />

12.4 Cost curves for <strong>GAC</strong> facilities 218<br />

12.5 Cost curves for on-site reactivation 219<br />

12.6 Cost curves for <strong>of</strong>f-site reactivation 220<br />

12:7 Cost curves for optimized coagulation 221<br />

12.8 Cost curves for ozonation 222


Foreword<br />

The AWWA <strong>Research</strong> Foundation is a nonpr<strong>of</strong>it corporation that is<br />

dedicated to the implementation <strong>of</strong> a research effort to help utilities respond to<br />

regulatory requirements and traditional high-priority concerns <strong>of</strong> the industry. The<br />

research agenda is developed through aprocess <strong>of</strong> consultation with subscribers and<br />

drinking water pr<strong>of</strong>essionals. Under the umbrella <strong>of</strong> a Strategic <strong>Research</strong> Plan, the<br />

<strong>Research</strong> Advisory Council prioritizes the suggested projects based upon current<br />

and future needs, applicability, and past work; the recommendations are forwarded<br />

to the Board <strong>of</strong> Trustees for final selection. The foundation also sponsors research<br />

projects through the unsolicited proposal process; the Collaborative <strong>Research</strong>,<br />

<strong>Research</strong> Applications, and Tailored Collaboration programs; and various joint<br />

research efforts with organizations such as the U.S. Environmental Protection<br />

Agency, the U.S. Bureau <strong>of</strong> Reclamation, and the Association <strong>of</strong> California <strong>Water</strong><br />

Agencies.<br />

This publication is a result <strong>of</strong> one <strong>of</strong> these sponsored studies, and it is hoped<br />

that its findings will be applied in communities throughout the world. The following<br />

report serves not only as a means <strong>of</strong> communicating the results <strong>of</strong> the water<br />

industry' s centralized research program but also as a tool to enlist the further support<br />

<strong>of</strong> the nonmember utilities and individuals.<br />

Projects are managed closely from their inception to the final report <strong>by</strong> the<br />

foundation's staff and large cadre <strong>of</strong> volunteers who willingly contribute their time<br />

and expertise. The foundation serves a planning and management function and<br />

awards contracts to other institutions such as water utilities, universities, and<br />

engineering firms. The funding for this research effort comes primarily from the<br />

Subscription Program, through which water utilities subscribe to the research<br />

program and make an annual payment proportionate to the volume <strong>of</strong> water they<br />

deliver and consultants and manufacturers subscribe based on their annual billings.<br />

The program <strong>of</strong>fers a cost-effective and fair method for funding research in the<br />

public interest.<br />

A broad spectrum <strong>of</strong> water supply issues is addressed <strong>by</strong> the foundation's<br />

research agenda: resources, treatment and operations, distribution and storage,<br />

water quality and analysis, toxicology, economics, and management. The ultimate<br />

purpose <strong>of</strong> the coordinated effort is to assist water suppliers to provide the highest<br />

possible quality <strong>of</strong> water economically and reliably. The true benefits are realized<br />

when the results are implemented at the utility level. The foundation's trustees are<br />

pleased to <strong>of</strong>fer this publication as a contribution toward that end.<br />

Granular activated carbon (<strong>GAC</strong>) has been specified <strong>by</strong> the United States<br />

Environmental Protection Agency (USEPA) as a best available technology (BAT)<br />

for removing many contaminants in drinking water. It is known that <strong>GAC</strong> has the<br />

potential to remove precursor compounds for disinfection <strong>by</strong>-products (DBFs), an<br />

important classification <strong>of</strong> contaminants that are currently being regulated, and<br />

therefore it is important to improve the industry's current understanding <strong>of</strong> the<br />

XXV


xxvi<br />

<strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> GA C <strong>Adsorption</strong><br />

capability <strong>of</strong> <strong>GAC</strong> to remove precursor compounds. This report examines the<br />

capacity <strong>of</strong> <strong>GAC</strong> to adsorb DBF precursors for several source waters, provides<br />

information relative to the impact <strong>of</strong> pretreatment on <strong>GAC</strong> adsorption, and discusses<br />

the relationship between DBF precursor characteristics and the precursors' ability<br />

to be removed <strong>by</strong> <strong>GAC</strong> treatment.<br />

Julius Ciaccia Jr.<br />

Chair, Board <strong>of</strong> Trustees<br />

AWWA <strong>Research</strong> Foundation<br />

James F. Manwaring, P.E.<br />

Executive Director<br />

AWWA <strong>Research</strong> Foundation


Acknowledgments<br />

The authors <strong>of</strong> this report wish to thank the following participants for their<br />

contributing efforts to this project:<br />

A WWARF Project Managers<br />

Jeffrey L. Oxenford and Joseph Roccaro<br />

Participating Utilities<br />

Cincinnati <strong>Water</strong> Works, Cincinnati, Ohio<br />

City <strong>of</strong> Phoenix <strong>Water</strong> and Wastewater Department, Phoenix, Ariz.<br />

Florida Cities <strong>Water</strong> Company, Sarasota, Fla.<br />

Jefferson Parish Department <strong>of</strong> <strong>Water</strong>, Jefferson Parish, La.<br />

Passaic Valley <strong>Water</strong> Commission, Clifton, N.J.<br />

South Central Connecticut Regional <strong>Water</strong> Authority, New Haven,<br />

Conn.<br />

Other Participants<br />

Kimberley A. Gray, University <strong>of</strong> Notre Dame, Notre Dame, Ind.<br />

David A. Reckhow, University <strong>of</strong> Massachusetts-Amherst, Amherst,<br />

Mass.<br />

John Tobiason, University <strong>of</strong> Massachusetts-Amherst, Amherst, Mass.<br />

Project Advisory Committee<br />

Robert Beaurivage, Manchester, N.H.<br />

Pierre La France, City <strong>of</strong> Laval, Que., Canada<br />

Benjamin Lykins Jr., USEPA, Cincinnati, Ohio.<br />

Technical Review Committee<br />

John C. Crittenden, Michigan Technological University, Houghton,<br />

Mich.<br />

Jack DeMarco, Cincinnati <strong>Water</strong> Works, Cincinnati, Ohio<br />

Philip C. Singer, University <strong>of</strong> North Carolina, Chapel Hill, N.C.<br />

XXV'II


Executive Summary<br />

Background<br />

Granular activated carbon (<strong>GAC</strong>) adsorption is universally recognized as<br />

one <strong>of</strong> the most effective processes for the removal <strong>of</strong> organic matter. <strong>GAC</strong><br />

adsorption has demonstrated multiple uses in water treatment, such as removal <strong>of</strong><br />

(1) taste-and-odor-causing compounds, (2) synthetic organic chemicals (SOCs)<br />

that may be present in industrially impacted source waters, and (3) natural organic<br />

matter (NOM), which serves as a precursor to the formation <strong>of</strong> disinfection <strong>by</strong><br />

products (DBFs). In many instances <strong>GAC</strong> is also considered to be a preferred<br />

medium to support biodegradation <strong>of</strong> some fractions <strong>of</strong> the NOM matrix.<br />

Various aspects <strong>of</strong> the <strong>GAC</strong> adsorption process have been studied in<br />

isolation <strong>by</strong> many researchers, focusing on key aspects <strong>of</strong> this treatment process and<br />

developing great insights into the NOM removal processes. A comprehensive and<br />

systematic study addressing wide variations <strong>of</strong> raw water characteristics and the<br />

effect <strong>of</strong> pretreatment on the removal <strong>of</strong> NOM <strong>by</strong> <strong>GAC</strong>, however, has not been<br />

undertaken thus far. In addition, conclusions regarding design criteria and costs<br />

have <strong>of</strong>ten been based upon bench- or pilot-scale studies composed <strong>of</strong> singlecolumn<br />

experiments evaluating individual compounds. In full-scale applications,<br />

multiple <strong>GAC</strong> contactors are routinely employed to produce the desired water<br />

volume, resulting in significant cost savings from the blending <strong>of</strong> effluents from<br />

<strong>GAC</strong> contactors in different stages <strong>of</strong> maturation with respect to the breakthrough<br />

<strong>of</strong> organic material. The goal <strong>of</strong> this study was to investigate optimization <strong>of</strong> the<br />

<strong>GAC</strong> adsorption process utilizing the most advanced analytical protocols over a<br />

wide range <strong>of</strong> raw water qualities.<br />

Regulatory Implications_________________<br />

The role <strong>of</strong> <strong>GAC</strong> for water treatment has been made prominent <strong>by</strong> recent<br />

regulatory activities that are expected to lower the maximum contaminant level<br />

(MCL) for some DBFs and establish new MCLs for other compounds. These<br />

regulations also recognize the <strong>GAC</strong> adsorption process as a best available technol<br />

ogy (BAT) for DBF precursor removal. As a part <strong>of</strong> the newly promulgated<br />

Information Collection Rule (ICR), the United States Environmental Protection<br />

Agency (USEPA) will require larger utilities treating relatively poorer quality<br />

source water to collect information on <strong>GAC</strong> treatability <strong>by</strong> conducting bench-scale<br />

or pilot-scale <strong>GAC</strong> adsorption studies. Clearly, greater scrutiny is being placed on<br />

<strong>GAC</strong> as a treatment process for controlling DBFs.


xxx <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> GA C <strong>Adsorption</strong><br />

The results presented in this report in terms <strong>of</strong> both adsorption behavior<br />

and cost implications provide a preview <strong>of</strong> the national database that will be<br />

developed as a result <strong>of</strong> these USEPA activities. In addition, the results will provide<br />

a more thorough understanding <strong>of</strong> the adsorption characteristics <strong>of</strong> <strong>GAC</strong> adsorption,<br />

with and without pretreatment optimization.<br />

Approach_________________________<br />

There are several factors that affect the design <strong>of</strong> a <strong>GAC</strong> process, including<br />

(1) the breakthrough characteristics <strong>of</strong> the naturally occurring DBF precursors, (2)<br />

the mode <strong>of</strong> operation <strong>of</strong> multiple <strong>GAC</strong> contactors, (3) the type <strong>of</strong> pretreatment<br />

before the adsorption process, and (4) the presence <strong>of</strong> other, competing adsorbing<br />

material. All <strong>of</strong> these factors are addressed in this study. In addition, a protocol was<br />

developed for evaluating the formation <strong>of</strong> DBFs on a uniform basis when multiple<br />

source waters are being compared. This protocol provides a means for directly<br />

comparing precursor material among different sources while at the same time using<br />

criteria that are typical <strong>of</strong> many treatment and distribution systems. Previous to this<br />

study, DBF formation potential (<strong>DBP</strong>FP) was routinely used for comparing precur<br />

sor materials among multiple sources, whereas simulated distribution system (SDS)<br />

conditions were used to assess the conditions expected for a given system. Tests<br />

using the uniform formation conditions (UFC) that were developed during this study<br />

provide a method for addressing both issues simultaneously.<br />

The adsorption behavior <strong>of</strong> NOM onto <strong>GAC</strong> was evaluated <strong>by</strong> conducting<br />

both bench-scale and pilot-scale experiments for a wide variety <strong>of</strong> source waters<br />

from participating utilities. Some raw water characteristics for the participating<br />

utilities are shown in Table ES.l. The NOM before and after <strong>GAC</strong> treatment was<br />

characterized <strong>by</strong> such traditional parameters as NOM surrogates (e.g., total organic<br />

carbon [TOC] and ultraviolet adsorption at 254 nm [UV254]) and DBF formation,<br />

as well as <strong>by</strong> more advanced techniques, including humic-nonhumic fractionation,<br />

molecular weight fractionation, and pyrolysis gas chromatography-mass spectrometry<br />

(PY-GC-MS).<br />

In addition to an evaluation <strong>of</strong> the removal <strong>of</strong> NOM <strong>by</strong> <strong>GAC</strong> treatment, the<br />

fate <strong>of</strong> organic matter through pretreatment (e.g., optimized coagulation; ozonation<br />

and biological filtration) and pretreatment's impact on <strong>GAC</strong> adsorption were<br />

characterized. The effects <strong>of</strong> various process parameters on <strong>GAC</strong> adsorption, such<br />

as empty bed contact time (EBCT), backwashing (which <strong>of</strong>ten results in<br />

restratification <strong>of</strong> adsorption zones), and the competitive adsorption <strong>of</strong> SOCs and<br />

NOM were also evaluated. Finally, experimental and modeling techniques were<br />

developed to use single-column experiments to determine breakthrough character<br />

istics resulting from the blending <strong>of</strong> multiple contactors operating in parallel in a<br />

full-scale facility.<br />

Table ES.2 summarizes the testing program evaluated during this research.<br />

The costs <strong>of</strong> <strong>GAC</strong> treatment with and without optimized coagulation and biological<br />

pretreatment were estimated to provide guideline to utilities that are considering<br />

<strong>GAC</strong> adsorption as a treatment process for DBF precursor removal.


Executive Summary xxxi<br />

Table ES. 1 Participating utilities and source water characteristics<br />

Raw water quality<br />

Utility<br />

name<br />

Cincinnati <strong>Water</strong> Works<br />

Source<br />

water<br />

Ohio River (ORW)<br />

TOC UFC-THM UFC-HAA6<br />

(mg/L) (ug/L) (ug/L)<br />

2.2 87 41<br />

South Central Connecticut<br />

Regional <strong>Water</strong> Authority<br />

Lake Gaillard<br />

(LGW)<br />

2.9 131 I09*<br />

Jefferson Parish <strong>Water</strong> Department<br />

Mississippi River<br />

(MRW)<br />

5.2 102 133<br />

Passaic Valley <strong>Water</strong> Commission<br />

Passaic River<br />

(PRW)<br />

5.0 135 171<br />

City <strong>of</strong> Phoenix<br />

Salt River Project<br />

(SRPW)<br />

2.5 76 36<br />

Florida Cities <strong>Water</strong> Company<br />

Groundwater<br />

(FGW)<br />

13 135 457<br />

HAAS = sum <strong>of</strong> five haloacetic acids<br />

HAA6 = sum <strong>of</strong> six haloacetic acids<br />

THM = trihalomethane<br />

* UFC-HAA5<br />

Table ES.2 Summary <strong>of</strong> experimental<br />

Program stage<br />

Field-scale <strong>GAC</strong> (full- or pilot-scale)<br />

RSSCT verification<br />

Design and operational considerations<br />

Impact <strong>of</strong> EBCT<br />

Impact <strong>of</strong> blending<br />

Impact <strong>of</strong> backwashing<br />

Impact <strong>of</strong> pretreatment<br />

Optimized coagulation<br />

Ozonation and biotreatment<br />

SOC adsorption<br />

program<br />

ORW LGW<br />

X X<br />

X X<br />

X<br />

X<br />

X<br />

X X<br />

Source<br />

MRW PRW SRPW FGW<br />

X X<br />

X X<br />

XXX<br />

XXX<br />

X<br />

X<br />

X X<br />

X XXX<br />

RSSCT = rapid, small-scale column test


xxxii <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> GA C <strong>Adsorption</strong><br />

Conclusions and Recommendations<br />

The following are the significant conclusions and recommendations devel<br />

oped as a result <strong>of</strong> the experimental program conducted during this research:<br />

Based upon visual and statistical comparisons, bench-scale<br />

experiments using the rapid small-scale column test (RSSCT) were<br />

found to adequately predict NOM breakthrough behavior observed<br />

during pilot- or full-scale applications. Collection <strong>of</strong> a representative<br />

sample was found to be a key to the success <strong>of</strong> RSSCTs. For sources<br />

that show significant seasonal variation in NOM content, multiple<br />

RSSCT experiments are recommended to understand adsorption<br />

behavior subject to seasonal variability. It is also recommended that<br />

RSSCTs be conducted as rapidly as possible for waters with<br />

significant concentrations <strong>of</strong> biodegradable dissolved organic carbon<br />

(BDOC). Any delay due to shipping or other concerns may result in<br />

breakthrough characteristics that will not be similar to those in pilotor<br />

full-scale tests.<br />

The UFC test developed during this research is represented <strong>by</strong> the<br />

parameters in Table ES.3. This protocol can be used for comparing<br />

DBF formation across unique waters and can be used to predict DBF<br />

formation in a distribution system if the system characteristics are<br />

similar to the UFC.<br />

The raw waters evaluated during this study were found, in general,<br />

to be slightly more humic than nonhumic and dominated <strong>by</strong><br />

intermediate molecular size (500 to 3,000 D) fractions. The aromatic<br />

and aliphatic fractions varied widely among sources. Some <strong>of</strong> the<br />

sources that are believed to be industrially impacted also showed<br />

signatures <strong>of</strong> chlorinated fragments, a telltale sign <strong>of</strong> industrial<br />

activities.<br />

Breakthrough <strong>of</strong> dissolved organic carbon (DOC) during <strong>GAC</strong><br />

adsorption was found to be a conservative predictor for the<br />

breakthrough <strong>of</strong> DBF precursors. The early <strong>GAC</strong> effluent samples<br />

containing the nonadsorbable NOM were found to be almost entirely<br />

nonhumic, aliphatic, and <strong>of</strong> small molecular size. Samples later in<br />

the breakthrough curve progressively showed the appearance <strong>of</strong><br />

humic and aromatic fractions and larger molecular size fractions.<br />

The DBF yields <strong>of</strong> the humic fractions were found to be slightly<br />

higher compared to those <strong>of</strong> the nonhumic fractions. No clear<br />

differences in DBF yield were observed with respect to molecular<br />

size.<br />

Pretreatment with coagulation was found to be almost equally<br />

effective on the humic and nonhumic fractions whereas the<br />

intermediate molecular size fractions were preferentially removed.<br />

Ozonation and bi<strong>of</strong>iltration preferentially removed larger and humic<br />

material over smaller and nonhumic material.<br />

Evaluations <strong>of</strong> various EBCTs indicated that the breakthrough curve<br />

could be prolonged with longer EBCTs; however, the production <strong>of</strong><br />

water per unit mass <strong>of</strong> <strong>GAC</strong> (water throughput, or number <strong>of</strong> bed


Executive Summary xxxiii<br />

Table ES.3 Parameters for the UFC test<br />

Parameter Value<br />

Incubation time (hours)<br />

Incubation temperature (°C)<br />

PH<br />

24-hour chlorine residual (mg/L)<br />

24±1<br />

20.0±1.0<br />

8.0±0.2<br />

1.0+0.3<br />

volumes) was not affected <strong>by</strong> EBCTs between 10 and 20 minutes.<br />

Consequently, shorter EBCTs could be preferred in contactor design<br />

because they will result in lower capital costs. Shorter EBCTs will,<br />

however, require more frequent replacement or regeneration <strong>of</strong><br />

carbon.<br />

Based upon experimental and modeling results, blending <strong>of</strong> effluents<br />

from multiple <strong>GAC</strong> contactors was found to significantly prolong<br />

the breakthrough curves and substantially lower the operational<br />

costs. The modeling approach used in this study was found to<br />

adequately predict the blended effluent breakthrough characteristics<br />

based upon experimental results from a single contactor.<br />

Based upon simulated experiments, completely restratifyirrg the<br />

adsorption zone during backwashing may shorten the useful life <strong>of</strong><br />

<strong>GAC</strong> in an adsorption contactor. The capacity <strong>of</strong> <strong>GAC</strong> for adsorbing<br />

SOCs was also found to be reduced as a result <strong>of</strong> NOM adsorption.<br />

The reduction in SOC adsorption was proportional to the amount <strong>of</strong><br />

NOM adsorbed.<br />

Optimizing coagulation pretreatment may result in longer <strong>GAC</strong> life,<br />

whereas the use <strong>of</strong> ozonation and biological filtration prior to <strong>GAC</strong><br />

adsorption may not improve <strong>GAC</strong> adsorption this latter<br />

pretreatment causes a shift to less adsorbable fractions in the <strong>GAC</strong><br />

influent in addition to some removal <strong>of</strong> precursors due to<br />

biodegradation. However, it is recognized that there may be other<br />

advantages to this pretreatment (relative to <strong>GAC</strong> adsorption alone)<br />

that were not evaluated in this work.<br />

The conceptual cost estimates developed during this study suggest<br />

that lower EBCTs may result in lower overall system costs. Based<br />

on these cost estimates, it was also concluded that optimization <strong>of</strong><br />

coagulation pretreatment may result in lower system costs if the<br />

CUR can be reduced <strong>by</strong> 50 percent at a coagulation cost <strong>of</strong> 5 cents<br />

or less per 1,000 gal (3,785 L). Pretreatment with ozonation and<br />

biological filtration also lowered the overall system costs for the<br />

utility with the highest source water NOM content, suggesting that<br />

this treatment strategy may be more effective for <strong>GAC</strong> adsorption as<br />

the concentration <strong>of</strong> organic material in the source water increases.


Chapter 1<br />

Introduction<br />

Granular activated carbon (<strong>GAC</strong>) adsorption is a process that has been used<br />

extensively throughout the world for adsorption <strong>of</strong> organic compounds. Some <strong>of</strong><br />

these compounds are naturally occurring in the environment, such as taste-andodor-causing<br />

compounds, while other compouds known as synthetic organic<br />

chemicals (SOCs) are anthropogenic. Taste-and-odor-causing compounds affect<br />

the aesthetic perception <strong>of</strong> water quality, while certain SOCs in drinking water<br />

are reported to be associated with adverse health effects. Because <strong>of</strong> the wellrecognized<br />

efficiency <strong>of</strong> <strong>GAC</strong> for removing taste-and-odor-causing compounds<br />

and SOCs, <strong>GAC</strong>'s use in the water industry has focused almost exclusively on the<br />

removal <strong>of</strong> these compounds.<br />

More recently, however, utilities have been focusing on the cost-effective<br />

control <strong>of</strong> disinfection <strong>by</strong>-products (DBFs), which also are reported to have<br />

potential adverse health effects. Utilities have a wide variety <strong>of</strong> options available to<br />

reduce the formation <strong>of</strong> DBFs, each with their associated costs and efficiencies. In<br />

general, DBFs can be reduced <strong>by</strong> three methods:<br />

1. Reduce the concentration <strong>of</strong> DBF precursor material prior to adding<br />

a disinfectant.<br />

2. Use alternative disinfectants to free chlorine to reduce the formation<br />

<strong>of</strong> halogenated DBFs.<br />

3. Remove the DBFs after they are formed.<br />

The first two methods have been shown <strong>by</strong> the water industry to be costeffective<br />

in the majority <strong>of</strong> applications. DBF precursor removal strategies include<br />

increasing coagulant dosages in conventional treatment, as well as applying<br />

advanced precursor removal processes such as the use <strong>of</strong> <strong>GAC</strong> or membranes.<br />

Alternative disinfectants to free chlorine include chloramines, typically used as a<br />

residual disinfectant in distribution systems, as well as stronger disinfectants such<br />

as ozone or chlorine dioxide. Although these latter disinfectants substantially<br />

reduce the formation <strong>of</strong> halogenated organic DBFs (such as bromate and chlorate),<br />

there are other DBFs that are formed <strong>by</strong> their use.<br />

Consequently, <strong>GAC</strong> may be cast in a new role in water treatment, and this<br />

report focuses on the optimization <strong>of</strong> <strong>GAC</strong> adsorption for the removal <strong>of</strong> DBF<br />

precursors. Such a study is timely in light <strong>of</strong> the current regulatory agenda.


2 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

Regulatory Agenda<br />

At the time <strong>of</strong> this writing, only one group <strong>of</strong> DBFs is regulated. A<br />

maximum contaminant level (MCL) <strong>of</strong> 100 \ig/L has been established <strong>by</strong> the United<br />

States Environmental Protection Agency (USEPA) fortotal trihalomethane (TTHM),<br />

which is composed <strong>of</strong> four individual chemical compounds. Recently, however, the<br />

USEPA proposed a Disinfectants-Disinfection By-Products (D-<strong>DBP</strong>) Rule de<br />

signed to control the use <strong>of</strong> disinfectants and the formation <strong>of</strong> <strong>DBP</strong>s (USEPA 1994).<br />

This rule has two proposed stages In the first stage, the MCL for TTHM is lowered<br />

to 80 ug/L, and an MCL <strong>of</strong> 60 ug/L has been proposed for the sum <strong>of</strong> five haloacetic<br />

acid (HAAS) species. In the second stage, it is proposed to reduce these two <strong>DBP</strong><br />

groups <strong>by</strong> one-half, to 40 and 30 ug/L for TTHM and HAA5, respectively. Although<br />

this Stage 2 regulation is subject to revision, the intent <strong>of</strong> the regulatory community<br />

is clear: to reduce exposure <strong>of</strong> the public to the potential adverse health effects <strong>of</strong><br />

the <strong>DBP</strong>s.<br />

Furthermore, USEPA is focusing on the removal <strong>of</strong> precursor material as<br />

an initial approach to reducing <strong>DBP</strong> production. In the D-<strong>DBP</strong> Rule, a treatment<br />

technique is proposed for conventional treatment plants to remove <strong>DBP</strong> precursors<br />

<strong>by</strong> improving the efficiency <strong>of</strong> existing coagulation processes. This process, termed<br />

enhanced coagulation, is also a best available technology (BAT) along with <strong>GAC</strong>.<br />

Although many systems may be able to comply with the Stage 1 requirements<br />

without implementing advanced precursor removal strategies or such alternative<br />

disinfectants as ozone or chlorine dioxide, Stage 2 compliance will be more difficult<br />

with conventional water treatment processes. In recognition <strong>of</strong> this difficulty, a<br />

companion regulation, the Information Collection Rule (ICR), has been proposed<br />

to evaluate the removal <strong>of</strong> <strong>DBP</strong> precursors and the formation <strong>of</strong> <strong>DBP</strong>s in large water<br />

treatment systems. A subset <strong>of</strong> these systems will also have to evaluate the removal<br />

<strong>of</strong> <strong>DBP</strong> precursors <strong>by</strong> <strong>GAC</strong> or membranes on a bench or pilot scale to assist in<br />

understanding the efficiency and costs associated with advanced precursor removal<br />

processes. Consequently, there is a great interest in optimizing <strong>GAC</strong> treatment to<br />

improve efficiency and reduce overall implementation costs.<br />

<strong>GAC</strong> Treatment issues__________________<br />

Adsorbability <strong>of</strong> Natural Organic Matter<br />

<strong>DBP</strong> precursor compounds are a subset <strong>of</strong> natural organic matter (NOM),<br />

found in natural waters. NOM is ubiquitous in source waters and is composed <strong>of</strong><br />

organic substances that vary in chemical nature and complexity. This heterogeneity<br />

is also evident from the differences in adsorbability <strong>of</strong> these compounds <strong>by</strong> <strong>GAC</strong><br />

(Sontheimer et al. 1988). Normally, 80 to 90 percent <strong>of</strong> the NOM measured in raw<br />

water sources can be removed <strong>by</strong> <strong>GAC</strong> adsorption (Roberts and Summers 1982).<br />

Recent work has also shown that NOM adsorbed on <strong>GAC</strong> does not desorb as a result<br />

<strong>of</strong> a decrease in the liquid-phase concentration (Summers and Roberts 1988;<br />

Summers et al. 1989, 1992). This may also contribute to the loss <strong>of</strong> <strong>GAC</strong> capacity<br />

for target SOCs that has been found to occur after pre-exposure <strong>of</strong> <strong>GAC</strong> to NOM<br />

(Arora 1989; Sontheimer et al. 1988; Summers et al. 1989, 1992). This <strong>GAC</strong>


Introduction 3<br />

"fouling" has not been evaluated on a wide range <strong>of</strong> NOM sources (1) with the same<br />

SOC or (2) under similar preloading circumstances.<br />

Blending <strong>of</strong> Contactor Effluents<br />

In the full-scale application <strong>of</strong> <strong>GAC</strong> at utilities treating more than 1 mgd<br />

(3.785 ML/d), multiple contactors will be used, with the effluents from the<br />

contactors blended together. The contactors are placed on-line in a staggered<br />

manner (i.e., starting times are different); thus, the effluent from each contactor has<br />

a different concentration, and the blended water concentration is the average <strong>of</strong><br />

individual effluents. For single solute compounds, blending can be effectively<br />

modeled (Roberts and Summers 1982; Sontheimer et al. 1988). However, DBF<br />

formation in blended <strong>GAC</strong> effluents has not been investigated or modeled. As a<br />

result <strong>of</strong> the nonlinear nature <strong>of</strong> the reaction <strong>of</strong> chlorine and NOM, the modeling may<br />

not be as straightforward as with individual compounds.<br />

DBF Speciation and Formation Conditions<br />

Recently, considerable attention has been given to the brominated species<br />

<strong>of</strong> TTHM and the haloacetic acids (HAAs). The speciation is known to be a function<br />

<strong>of</strong> the relative concentrations <strong>of</strong> bromide ion, dissolved organic carbon (DOC), and<br />

applied chlorine, as well as the reaction time. Because the NOM effluent from a<br />

<strong>GAC</strong> adsorber initially is low, the bromide-to-DOC ratio (Br:DOC) is higher than<br />

that in the influent. This leads to the production <strong>of</strong> a higher percentage <strong>of</strong> brominated<br />

DBFs in the <strong>GAC</strong> effluent. Depending on the chlorination approach, a similar<br />

pattern <strong>of</strong> more brominated DBFs may emerge following chlorination. If a constant<br />

chlorine dose is used, the C12:DOC ratio will decrease as the DOC incrases with<br />

time, which may not represent the conditions <strong>of</strong> practice when blended waters are<br />

used or when chlorine is dosed to achieve a constant residual.<br />

The variability in the conditions used to represent DBF formation has been<br />

a major obstacle to the uniform assessment <strong>of</strong> <strong>GAC</strong> and other treatment processes.<br />

Two approaches for DBF formation conditions have been taken in the past. The<br />

formation potential (FP) conditions are somewhat standardized but are thought to<br />

be too extreme to represent DBF formation conditions in practice and may<br />

misrepresent treatment efficiencies. The simulated distribution system (SDS)<br />

conditions are representative <strong>of</strong> practical conditions, but they normally vary from<br />

utility to utility, bcause they are meant to be site-specific. A controlled comparison<br />

<strong>of</strong> several water sources under uniform, realistic conditions is needed to allow a<br />

comprehensive assessment <strong>of</strong> the efficiency <strong>of</strong> <strong>GAC</strong> treatment.<br />

Evaluation <strong>of</strong> Process Variables<br />

To optimize <strong>GAC</strong> use, such process variables as empty bed contact time<br />

(EBCT) and pretreatment (i.e., coagulation; ozonation followed <strong>by</strong> biotreatment)<br />

need to be investigated in combination. For some waters, coagulation can be<br />

effectively optimized for DBF precursor removal, which can impact the perfor<br />

mance <strong>of</strong> the <strong>GAC</strong> system. If the <strong>GAC</strong> system is operated without a disinfectant<br />

residual in the influent, biological activity will occur before and during <strong>GAC</strong><br />

treatment. This bioactivity can contribute to the removal <strong>of</strong> DBF precursors.


4 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> GA C <strong>Adsorption</strong><br />

Evaluating the impact <strong>of</strong> optimized pretreatment and EBCT on <strong>GAC</strong><br />

performance at a pilot scale is time-consuming and costly. The rapid small-scale<br />

column test (RSSCT) for the simulation <strong>of</strong> field-scale NOM removal <strong>by</strong> <strong>GAC</strong> was<br />

shown to be an effective and economical alternative for eight different waters <strong>by</strong><br />

Summers and Crittenden (1989) and Crittenden et al. (1991). More recently, the<br />

success <strong>of</strong> RSSCT has been demonstrated for simulating removal <strong>of</strong> TTHM and<br />

total organic halide (TOX) precursors <strong>of</strong> a groundwater and a surface water<br />

(Summers et al. 1992).<br />

Study Objectives_____________________<br />

The primary objective <strong>of</strong> this study is to evaluate the use <strong>of</strong> <strong>GAC</strong> for the<br />

control <strong>of</strong> DBF formation under uniform, representative conditions over a range <strong>of</strong><br />

water sources. The project was planned to achieve the following specific objectives:<br />

1. Examine a wide range <strong>of</strong> water sources at both field scale and<br />

laboratory scale to assess <strong>GAC</strong> breakthrough behavior.<br />

2. Characterize the NOM in raw and <strong>GAC</strong> treated waters.<br />

3. -Evaluate the formation <strong>of</strong> DBFs under regulatory focus using<br />

representative but uniform conditions (i.e., uniform formation<br />

conditions [UFC]):<br />

4. Optimize <strong>GAC</strong> performance based upon<br />

empty bed contact time<br />

coagulation pretreatment<br />

ozonation-bi<strong>of</strong>iltraion pretreatment<br />

5. Assess the impact <strong>of</strong> the preadsorption <strong>of</strong> NOM on <strong>GAC</strong>'s capacity<br />

for SOCs.<br />

6. Establish relationships between NOM parameters and DBF<br />

formation with respect to <strong>GAC</strong> treatment.<br />

7. Produce optimization guidelines and cost estimates for <strong>GAC</strong><br />

treatment.


Chapter 2<br />

Materials and Methods<br />

Five surface waters and one ground water were sampled for this study. The<br />

waters represent a wide range <strong>of</strong> water quality characteristics, as shown in Table 2.1<br />

for raw and treated water quality. The treatment processes used <strong>by</strong> the participating<br />

utilities that contributed the source waters are summarized in Table 2.2.<br />

Ohio River <strong>Water</strong> (ORW) was collected prior to the full-scale <strong>GAC</strong><br />

contactor at Cincinnati <strong>Water</strong> Works (CWW) (Cincinnati, Ohio). ORW is an<br />

industrially impacted surface water with moderate color, medium alkalinity,<br />

medium hardness, average levels <strong>of</strong> organic matter, and intermittent presence <strong>of</strong><br />

SOCs. The water was pretreated <strong>by</strong> alum coagulation (8 mg/L), sedimentation, and<br />

filtration.<br />

Lake Gaillard <strong>Water</strong> (LOW) was obtained from the South Central Con<br />

necticut Regional <strong>Water</strong> Authority (SCCRWA) (New Haven, Conn.), which<br />

operated a pilot-scale postfilter adsorber. This water represents a high-quality<br />

surface water with low color, low alkalinity, low hardness, medium levels <strong>of</strong> organic<br />

matter, and intermittent presence <strong>of</strong> taste-and-odor-causing compounds. The water<br />

was pretreated <strong>by</strong> direct filtration and ozonation.<br />

Mississippi River <strong>Water</strong> (MRW) was obtained from the Jefferson Parish<br />

<strong>Water</strong> Department (JPWD) (Jefferson Parish, La.), which operated a pilot-scale<br />

filter adsorber. MRW is an industrially impacted surface water with moderate color,<br />

high alkalinity, high hardness, medium to high levels <strong>of</strong> organic matter, and<br />

intermittent presence <strong>of</strong> SOCs. The water was pretreated <strong>by</strong> alum coagulation,<br />

sedimentation, and ozonation.<br />

Passaic River <strong>Water</strong> (PRW) was obtained from the Little Falls <strong>Water</strong><br />

Treatment Plant in Totowa, N.J. The Passaic Valley <strong>Water</strong> Commission (PVWC)<br />

(Clifton, N.J.) operated a pilot-scale <strong>GAC</strong> postfilter adsorber. PRW, also an<br />

industrially impacted surface water, has moderate color, high alkalinity, high<br />

hardness, high levels <strong>of</strong> organic matter, and intermittent presence <strong>of</strong> taste-and-odorcausing<br />

compounds and SOCs. The water was sampled after alum coagulation (30<br />

mg/L), sedimentation, and filtration.<br />

Salt River Project <strong>Water</strong> (SRPW), a blend <strong>of</strong> the Salt and Verde Rivers, was<br />

sampled from the City <strong>of</strong> Phoenix <strong>Water</strong> and Wastewater Department (Phoenix,<br />

Ariz.) after alum coagulation (10.6 mg/L), sedimentation, and filtration. SRPW has<br />

low color, high dissolved solids, moderate hardness, moderate alkalinity, and<br />

medium levels <strong>of</strong> organic matter.<br />

Florida groundwater (FGW) was obtained from the Barefoot Bay <strong>Water</strong><br />

Treatment Plant (Florida Cities <strong>Water</strong> Company, Sarasota, Fla.). FGW has high<br />

color, high hardness, high alkalinity, and high levels <strong>of</strong> organic matter. Raw water<br />

was sampled and batch coagulated (48 mg/L alum) and s<strong>of</strong>tened (280 mg/L lime)


Florida<br />

groundwater<br />

Raw<br />

13<br />

0.48<br />

276<br />

7.7<br />

240<br />

300<br />

135<br />

457<br />

1,070<br />

26<br />

<strong>GAC</strong><br />

influent<br />

10<br />

0.29<br />

276<br />

8.6<br />

130<br />

80<br />

238<br />

142<br />

1,046<br />

7.8<br />

s? s<br />

o<br />

§<br />

to<br />

5 n<br />

1<br />

5'<br />

Table 2.1 Summary <strong>of</strong> water quality parameters for water sources<br />

s<br />

2<br />

Ohio River<br />

Lake Gaillard<br />

Mississippi River<br />

Passaic River<br />

Salt River<br />

Project<br />

3<br />

•S"<br />

Parameter<br />

Raw<br />

<strong>GAC</strong><br />

influent<br />

Raw<br />

<strong>GAC</strong><br />

influent<br />

Raw<br />

<strong>GAC</strong><br />

ifluent<br />

Raw<br />

<strong>GAC</strong><br />

influent<br />

Raw<br />

<strong>GAC</strong><br />

influent<br />

TOC (mg/L)<br />

UV254 (1/cm)<br />

Br (ug/L)<br />

pH<br />

Alkalinity (mg/L as CaCO 3 )<br />

Hardness (mg/L as CaCO 3)<br />

UFC-TTHM (ug/L)<br />

UFC-HAA6 (ug/L)<br />

UFC-TOX (ug CM.)<br />

UFC-CH (pg/L)<br />

2.2<br />

0.053<br />

106<br />

7.9<br />

40<br />

150<br />

87<br />

41<br />

NA<br />

9.3<br />

2.1<br />

0.044<br />

132<br />

7.8<br />

66<br />

163<br />

86<br />

26<br />

170<br />

3.9<br />

2.9<br />

0.093<br />

12<br />

8.1<br />

12<br />

21<br />

131<br />

109*<br />

490<br />

10<br />

1.5<br />

0.012<br />

14<br />

7.9<br />

NA<br />

NA<br />

31<br />

29*<br />

131<br />

2.8<br />

5.2<br />

0.115<br />

51<br />

8.2<br />

96<br />

132<br />

102<br />

133<br />

690<br />

5.9<br />

2.8<br />

0.022<br />

68<br />

7.9<br />

NA<br />

NA<br />

50<br />

22<br />

168<br />

7.9<br />

5.0<br />

0.10<br />

80<br />

7.8<br />

94<br />

143<br />

135<br />

171<br />

480<br />

15<br />

3.2<br />

0.076<br />

80<br />

7.7<br />

80<br />

142<br />

73<br />

70<br />

277<br />

15<br />

2.5<br />

0.055<br />

89<br />

8.2<br />

135<br />

178<br />

76<br />

36<br />

196<br />

4.4<br />

2.2<br />

0.047<br />

89<br />

7.8<br />

NA<br />

NA<br />

72<br />

28<br />

170<br />

4.1<br />

CH = chloral hydrate<br />

HAA6 = sum <strong>of</strong> six haloacetic acids<br />

NA = not analyzed<br />

UV254 = ultraviolet absorbance at 254 nm<br />

*UFC = HAAS


Materials and Methods 7<br />

Table 2.2 Field-scale <strong>GAC</strong> systems<br />

Source<br />

water<br />

Utility<br />

Pretreatment<br />

<strong>GAC</strong><br />

contactor size<br />

<strong>GAC</strong> EBCT<br />

contactor type (minutes)<br />

Ohio River<br />

Cincinnati <strong>Water</strong> Works<br />

Alum coagulation,<br />

sedimentation,<br />

filtration<br />

Full scale<br />

Postfilter 15<br />

adsorber<br />

Lake Gaillard South Central Connecticut Direct filtration,<br />

Regional <strong>Water</strong> Authority ozonation<br />

Pilot scale<br />

Postfilter<br />

adsorber 15<br />

Mississippi River Jefferson Parish <strong>Water</strong><br />

Department<br />

Alum coagulation,<br />

sedimentation,<br />

ozonation<br />

Pilot scale<br />

Filter adsorber 6.25<br />

Passaic River Passaic Valley <strong>Water</strong><br />

Commission<br />

Alum coagulation,<br />

sedimentation,<br />

filtration<br />

Pilot scale Postfilter<br />

adsorber<br />

20<br />

Salt River<br />

Project<br />

Florida<br />

groundwater<br />

City <strong>of</strong> Phoenix<br />

Florida Cities <strong>Water</strong><br />

Company<br />

Alum coagulation,<br />

sedimentation,<br />

filtration<br />

Prechlorination;<br />

lime s<strong>of</strong>tening and<br />

alum coagulation;<br />

sedimentation<br />

No field-scale <strong>GAC</strong> system at this site.<br />

at the University <strong>of</strong> Cincinnati (UC) (Cincinnati, Ohio) using the chemical doses <strong>of</strong><br />

the utility at the day <strong>of</strong> sampling.<br />

Experimental Approach<br />

Testing Program<br />

A summary <strong>of</strong> the experimental program for all waters is presented in Table<br />

2.3. Four pilot- or full-scale <strong>GAC</strong> contactors were operated, and RSSCTs were<br />

conducted to verify the ability <strong>of</strong> the RSSCT for predicting NOM and DBF control<br />

under SDS conditions at the field-scale level. Experiments were conducted to<br />

examine the effects <strong>of</strong> design and operational parameters <strong>of</strong> EBCT, blending, and<br />

backwashing using the RSSCT and UFC. The impact <strong>of</strong> (1) optimized coagulation<br />

and (2) ozonation and biological filtration prior to the <strong>GAC</strong> column were investi<br />

gated using the RSSCT and UFC. To determine the impact <strong>of</strong> prior exposure to<br />

NOM on the residual adsorption <strong>of</strong> cis-l,2-dichloroethene (DCE) on <strong>GAC</strong>, batch


8 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

Table 2.3 Summary <strong>of</strong> experimental program<br />

Source<br />

Program stage______________ORW____LGW MRW PRW____SRPW FGW<br />

Field-scale <strong>GAC</strong> (full or pilot scale) X X X X<br />

RSSCT verification X XXX<br />

Design and operational considerations<br />

Impact <strong>of</strong> EBCT X XXX<br />

Impact <strong>of</strong> blending X X X<br />

Impact <strong>of</strong> backwashing X<br />

Impact <strong>of</strong> pretreatment<br />

Optimized coagulation<br />

Ozonation and biotreatment<br />

SOC adsorption<br />

X X<br />

X XX<br />

X X X X X X<br />

isotherms were run with <strong>GAC</strong> from the field and RSSCT columns after the<br />

completion <strong>of</strong> column operation. A matrix linking the experimental program<br />

analyses with the laboratories at which the analyses were performed is shown in<br />

Table 2.4.<br />

Sample Collection and Handling<br />

Treated ORW, MRW, PRW, and SRPW were sampled in 55-gal (208-L)<br />

drums and shipped to UC on the same day from utilities. These waters were used (1)<br />

as influent to RSSCTs performed at UC for verification studies and (2) to assess the<br />

impact <strong>of</strong> design and operational parameters. For ORW, PRW, and SRPW, a portion<br />

<strong>of</strong> the plant-treated water received was subjected to secondary bench-scale treat<br />

ment at UC, consisting either <strong>of</strong> coagulation with additional alum doses or <strong>of</strong><br />

ozonation and biological filtration, to examine the effects <strong>of</strong> pretreatment.<br />

The verification RSSCT for LGW was performed on-site at SCCRWA. No<br />

other RSSCTs were conducted for this water. For FGW, raw water only was<br />

sampled and shipped to UC in 55 gal (208-L) drums. This water was batch treated<br />

at UC. For additional analyses, raw water from all six utilities was sampled on the<br />

same day as treated water. The raw water was shipped to UC <strong>by</strong> overnight courier.<br />

The delivery time for water in 55 gal (208-L) drums was 4 to 6 days for<br />

MRW, SRPW, and FGW. For PRW the water was delivered in 2 days, while for<br />

ORW the water was delivered on the same day because <strong>of</strong> the proximity <strong>of</strong> CWW<br />

toUC.<br />

Experimental Procedures<br />

Rapid Small-Scale Column Test<br />

The RSSCT was used to predict pilot- or full-scale NOM removal and <strong>DBP</strong><br />

control. The development <strong>of</strong> the RSSCT for NOM and <strong>DBP</strong> control is described <strong>by</strong><br />

Summers et al. (1995). The RSSCT utilizes small <strong>GAC</strong> particle sizes, typically 0.08


Materials and Methods 9<br />

Table 2.4 Laboratories performing sample analyses for experimental program<br />

<strong>Water</strong><br />

source<br />

Experiment<br />

TOC<br />

uv254<br />

SDS<br />

UFC<br />

Analysis<br />

TOX<br />

THM<br />

HAA<br />

CH<br />

MRW<br />

Pilot-scale <strong>GAC</strong><br />

RSSCT verification<br />

JPWD<br />

UC<br />

JPWD<br />

UC<br />

JPWD<br />

UC<br />

NA<br />

UC<br />

JPWD<br />

UC<br />

JPWD<br />

JPWD<br />

JPWD<br />

JPWD<br />

JPWD<br />

JPWD<br />

ORW<br />

Full-scale <strong>GAC</strong><br />

RSSCT verification<br />

Impact <strong>of</strong> EBCT<br />

Optimized coagulation<br />

Ozonation and<br />

biotreatment<br />

CWW<br />

UC<br />

UC<br />

UC<br />

UC<br />

UC<br />

UC<br />

UC<br />

UC<br />

UC<br />

CWW<br />

UC<br />

NA<br />

NA<br />

NA<br />

NA<br />

UC<br />

UC<br />

UC<br />

UC<br />

CWW<br />

UC<br />

UC<br />

UC<br />

UC<br />

CWW<br />

CWW<br />

CWW<br />

CWW<br />

CWW<br />

CWW<br />

CWW<br />

CWW<br />

CWW<br />

CWW<br />

CWW<br />

CWW<br />

CWW<br />

CWW<br />

CWW<br />

LGW<br />

Pilot-scale <strong>GAC</strong><br />

RSSCT verification .<br />

UM<br />

UM<br />

NA<br />

NA<br />

UM<br />

UM<br />

NA<br />

UC<br />

UC<br />

UC<br />

UM<br />

UM<br />

UM<br />

UM<br />

UM<br />

UM<br />

PRW<br />

Pilot-scale <strong>GAC</strong><br />

RSSCT verification<br />

Impact <strong>of</strong> EBCT<br />

Ozonation and<br />

biotreatment<br />

Impact <strong>of</strong> backwashing<br />

Impact <strong>of</strong> blending<br />

UC<br />

UC<br />

UC<br />

UC<br />

UC<br />

UC<br />

UC<br />

UC<br />

UC<br />

UC<br />

UC<br />

NA<br />

NA<br />

NA<br />

NA<br />

NA<br />

NA<br />

NA<br />

PVWC<br />

UC<br />

UC<br />

UC<br />

UC<br />

UC<br />

MWL<br />

MWL<br />

MWL<br />

MWL<br />

NA<br />

NA<br />

MWL<br />

MWL<br />

MWL<br />

MWL<br />

UC<br />

UC<br />

MWL<br />

MWL<br />

MWL<br />

MWL<br />

NA<br />

NA<br />

MWL<br />

MWL<br />

MWL<br />

MWL<br />

NA<br />

NA<br />

SRPW<br />

Impact <strong>of</strong> EBCT<br />

Optimized coagulation<br />

Impact <strong>of</strong> blending<br />

UC<br />

UC<br />

UC<br />

UC<br />

UC<br />

NA<br />

NA<br />

NA<br />

NA<br />

UC<br />

UC<br />

UC<br />

UC<br />

UC<br />

NA<br />

CWW<br />

CWW<br />

UC<br />

CWW<br />

CWW<br />

NA<br />

CWW<br />

CWW<br />

NA<br />

FGW<br />

Impact <strong>of</strong> EBCT<br />

Ozonation and<br />

biotreatment<br />

Impact <strong>of</strong> blending<br />

UC<br />

UC<br />

UC<br />

UC<br />

UC<br />

NA<br />

NA<br />

NA<br />

NA<br />

UC<br />

UC<br />

UC<br />

UC<br />

UC<br />

NA<br />

CWW<br />

CWW<br />

UC<br />

CWW<br />

CWW<br />

NA<br />

CWW<br />

CWW<br />

NA<br />

CH = chloral hydrate<br />

MWL = Montgomery Watson Laboratories, Pasadena, Calif.<br />

NA = not analyzed<br />

TOC = total organic carbon<br />

UM = University <strong>of</strong> Massachusetts, Amherst, Mass.<br />

UV254 = ultraviolet absorbance at 254 nm<br />

to 0.2 mm in diameter, and is designed to decrease the time and effort needed to<br />

assess the concentration pr<strong>of</strong>ile in a <strong>GAC</strong> column (Crittenden et al. 1986, 1987,<br />

1991). A proportional diffusivity design is used (Summers etal. 1989,1994b, 1995),<br />

so the EBCT and run time <strong>of</strong> the RSSCT are inversely proportional to the ratio <strong>of</strong><br />

particle sizes in the field-scale and RSSCT columns, as described in Appendix A.<br />

A statistical procedure for comparing RSSCT to pilot- or full-scale breakthrough<br />

curves is described in Appendix B.<br />

To perform the tests, commercially available 8.0-, 11.0-, or 15.0-mm glass<br />

chromatography columns with polytetrafluoroethylene (PTFE) or stainless steel<br />

fittings were used. After being grinded to appropriate size (see Table 2.5) <strong>by</strong> mortar<br />

and pestle, the <strong>GAC</strong> was washed with laboratory clean water to remove fines. An


10 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

ultrasonic vibrator was used to aid in the removal <strong>of</strong> fines. Finally, a vacuum was<br />

applied to remove air from the pores <strong>of</strong> the <strong>GAC</strong>. The <strong>GAC</strong> was packed into the<br />

column to the appropriate bed length as determined <strong>by</strong> scaling equations (Summers<br />

et al. 1995), which are described in Appendix A. The RSSCT columns were fed from<br />

a 19-L glass carboy containing the prefiltered (1.0-um cartridge filter) influent at<br />

room temperature, which was refilled with the batch influent water as needed. The<br />

influent water was pumped into a stainless steel gas cylinder, which served to<br />

dampen the pulsing effect <strong>of</strong> the diaphragm pump used to push water through the<br />

columns. Fresh <strong>GAC</strong> was obtained from the full- or pilot-scale contactor and ground<br />

to smaller size at UC. Field-scale and RSSCT operational and design parameters are<br />

described in Table 2.5.<br />

Bench-Scale Coagulation and S<strong>of</strong>tening<br />

For optimized coagulation pretreatment, the bench-scale alum coagulation<br />

process was conducted at UC for ORW and SRPW. It was repeated three to five<br />

times on conventionally treated water contained in different 55-gal (208-L) drums.<br />

Table 2.5 Field-scale and RSSCT verification operation parameters___________<br />

Field scale<br />

________________ORW______LGW______MRW_____PRW_____SRPW* FGW*<br />

Carbon type F 400 F 400 <strong>GAC</strong> 40 F 400 F 400 F 400<br />

dLC (mm) 1.05 1.49t 1.05 1.05 1.05 1.05<br />

EBCTLC (minutes) 15 15 6.25 20 15 15<br />

LLC (m) 3.2 — 0.52 2.67 — —<br />

13 5.0 8.0<br />

RSSCT verification<br />

Carbon type<br />

dsc (mm)<br />

SF<br />

EBCTSC (minutes)<br />

Lsc (m)<br />

° e SC,min<br />

vsc (m/h)<br />

Qsc (mL/min)<br />

DSC (mm)<br />

F400<br />

0.08<br />

12.6<br />

1.2<br />

0.193<br />

0.5<br />

9.7<br />

8.1<br />

8.0<br />

F400<br />

0.20<br />

7.4<br />

2.0<br />

0.127<br />

0.5<br />

4.1<br />

6.4<br />

11.0<br />

<strong>GAC</strong> 40<br />

0.11<br />

9.4<br />

0.66<br />

0.16<br />

1.0<br />

15<br />

23<br />

11.0<br />

F400<br />

0.08<br />

12.6<br />

1.6<br />

0.257<br />

0.5<br />

9.7<br />

8.1<br />

8.0<br />

F400<br />

0.11<br />

9.4<br />

1.6<br />

0.193<br />

0.5<br />

7.3<br />

6.1<br />

8.0<br />

F400<br />

0.20<br />

5.3<br />

2.8<br />

0.193<br />

0.5<br />

4.1<br />

12<br />

15.0<br />

d = particle diameter<br />

D = column diameter<br />

EBCT = empty bed contact time<br />

L = <strong>GAC</strong> bed length<br />

LC = large particle column<br />

Q = flow rate<br />

ReSCmin = minimum Reynold's number for small particle column<br />

SC ='small particle column<br />

SF = scaling factor<br />

v = superficial velocity<br />

— indicates not known<br />

* Field-scale column was not operated; RSSCT design was based on the given field-scale parameters,<br />

t Sieved to yield a larger size.


Materials and Methods 11<br />

A variable speed shaft mixer was operated at high speed prior to alum addition and<br />

for 4 minutes after alum addition. The water was then flocculated for 12 minutes at<br />

low speed. After settling, the water was passed through a 1.0-um cartridge filter<br />

prior to <strong>GAC</strong> treatment.<br />

Bench-scale coagulation and s<strong>of</strong>tening <strong>of</strong> FGW were performed in a similar<br />

manner to bench-scale coagulation. Both alum and lime were added during rapid<br />

mix. If the pH did not reach 10.3 immediately after chemical addition, sodium<br />

hydroxide was added to achieve this pH level. This procedure was designed to<br />

simulate the procedure used for FGW at the full-scale treatment facility. The<br />

remaining procedure was the same as that for alum coagulation.<br />

Bench-Scale Ozonation and Biotreatment<br />

The impact <strong>of</strong> ozonation and biological filtration was investigated for<br />

ORW, PRW, and FGW. Conventionally treated water was passed through a benchscale<br />

ozonation system (Allgeier and Summers 1995). Ozonation was performed at<br />

room temperature. A 1 -L contactor with a detention time <strong>of</strong> 7 minutes was used as<br />

a reactor with a porous diffuser at the bottom. Ozonation was carried out in a<br />

countercurrent mode. Ozone was generated using an electric discharge ozonizer<br />

(Model 200, Sander Aquarientechnik, Vetze-Eltze, Germany). The gas flow rate<br />

used was 60 mL/min, and the liquid flow rate used was 143 mL/min. The dose was<br />

applied to produce a residual <strong>of</strong> 0.4 mg O3/L on average. Applied ozone and <strong>of</strong>f-gas<br />

were measured <strong>by</strong> iodometry, Method 422 (APHA et al. 1995) and dissolved ozone<br />

was measured spectrophotometrically using the indigo trisulfonate method (APHA<br />

etal. 1995).<br />

After ozonation the water was passed through a biologically active sand<br />

filter. The bench-scale filter column had a diameter <strong>of</strong> 3 cm and a bed depth <strong>of</strong> 0.30<br />

m. The filter EBCT was 5 minutes and was designed to remove the quickly<br />

biodegradable fraction <strong>of</strong> NOM. The biologically active sand was acclimated<br />

(steady-state BDOC removal) with the water to be treated prior to the run. After<br />

biotreatment, the water was filtered through a 1 -urn cartridge filter and treated <strong>by</strong><br />

<strong>GAC</strong>.<br />

Blending Experiments<br />

In practice, multiple <strong>GAC</strong> contactors are operated in a parallel configura<br />

tion, and the effluents from all contactors are blended prior to disinfection. Parallel<br />

contactors are operated in a staggered mode (i.e., starting times are different)<br />

because each contactor will have been in operation for a different length <strong>of</strong> time,<br />

yielding different levels <strong>of</strong> breakthrough. The RSSCT, however, represents only a<br />

single contactor. To simulate the effect <strong>of</strong> blending on <strong>GAC</strong> performance, an<br />

integral breakthrough experiment was performed. The effluent from the RSSCT<br />

was continuously collected in a large reservoir, from which samples were taken as<br />

the volume and NOM concentration in the reservoir increased. Blending experi<br />

ments were conducted for PRW, SRPW, and FGW. Five to six samples were taken<br />

and analyzed for TOC and THM formation under UFC.<br />

DCE Isotherms<br />

In order to evaluate the residual <strong>GAC</strong> capacity for removing DCE after prior<br />

exposure to NOM, bottle-point batch isotherm studies were performed using<br />

different carbon doses (Randtke and Snoeyink 1983). DCE was chosen because it


12 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> GA C <strong>Adsorption</strong><br />

has intermediate adsorbability <strong>by</strong> <strong>GAC</strong> and is not biodegradable within a short<br />

period <strong>of</strong> time under ambient conditions. <strong>GAC</strong> was sampled both from RSSCTs<br />

performed for all six waters and from two field-scale columns after the completion<br />

<strong>of</strong> the adsorption run. The preloaded <strong>GAC</strong> was removed from the columns and dried<br />

in an oven at 100 C until there was no detectable weight change. The field-scale<br />

<strong>GAC</strong> was first dried and then ground to the same size as the respective RSSCT <strong>GAC</strong>.<br />

The dried carbon was weighed and placed in 250-mL bottles and filled headspacefree<br />

with the prepared DCE solution. To prepare the DCE solution, a highconcentration<br />

aqueous stock solution was added to 200 mL <strong>of</strong> low-organic-content<br />

water and mixed for 24 hours. The prepared bottles were placed on a tumbler at 12.5<br />

rpm for 7 days. Four controls were made without carbon for each set <strong>of</strong> isotherms.<br />

After 7 days, samples were taken from each bottle and filtered before analysis using<br />

a 25-mL glass syringe with 0.45-um-pore-diameter membranes (Millipore HV,<br />

Millipore Corp., Bedford, Mass.)<br />

Backwashing<br />

To investigate the impact <strong>of</strong> backwashing on <strong>GAC</strong> performance, an RSSCT<br />

column was backwashed and run in parallel to a nonbackwashed column treating<br />

PRW. The RSSCT column was backwashed once every 11 full-scale equivalent<br />

days <strong>by</strong> completely removing all carbon and mixing the carbon in a beaker using<br />

effluent water collected just prior to backwashing. This represents a worst-case<br />

scenario because, in practice, a carbon bed will not become completely restratified<br />

after backwashing. The column was then repacked and filled with influent water.<br />

After backwashing, the first 20 mL <strong>of</strong> water in the column and tubing was discarded.<br />

The total down time for each backwashing was 30 minutes. After backwashing, one<br />

bed volume <strong>of</strong> effluent was discarded before sampling. Samples were collected<br />

before and after backwashing and analyzed for total organic carbon (TOC) and<br />

ultraviolet absorbance at 254 nm (UV254). DBF formation was assessed <strong>by</strong><br />

trihalomethane (THM) formation under UFC.<br />

Analytical Methods___________________<br />

Standard Methods<br />

A summary <strong>of</strong> standard analytical methods used in this study is presented<br />

in Table 2.6. The following is a description <strong>of</strong> the analytical procedures followed<br />

at the University <strong>of</strong> Cincinnati.<br />

Total Organic Carbon<br />

Organic carbon is the most commonly used nonspecific parameter for<br />

quantitative measurements <strong>of</strong> NOM. TOC is composed <strong>of</strong> dissolved organic carbon<br />

and a fraction containing particulate and colloidal organic carbon. TOC analysis<br />

performed at UC utilized a Dohrmann Model DC-180 organic carbon analyzer<br />

(Rosemount Analytical Corp., Santa Clara, Calif.) that employed persulfate and<br />

ultraviolet light oxidation. The methodology followed in the operation <strong>of</strong> this<br />

instrument was in accordance with the manufacturer's recommendations and<br />

Method 5310 C (APHA et al. 1995). Samples were acidified to pH < 2 <strong>by</strong>


Materials and Methods 13<br />

Table 2.6 Summary <strong>of</strong> analytical methods<br />

Parameter<br />

Laboratory<br />

Method<br />

Detection limit Precision (%)<br />

TOC<br />

UC<br />

JPWD<br />

CWW<br />

UM<br />

MWL<br />

SM 5310 C (APHA et al. 1995)<br />

—<br />

SM 5310 C (APHA et al. 1995)<br />

SM 5310 C (APHA et al. 1995)<br />

SM 5310 (APHA et al. 1995)<br />

0.1 mg/L


14 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

thoroughly rinsed with laboratory clean water (LCW) and prerinsed once with the<br />

sample to be filtered prior to use.<br />

Ultraviolet Light Absorbance at 254 nm<br />

NOM absorbs light in the range <strong>of</strong> 200 to 700 nm, and absorbance at 254<br />

nm has been the industry standard for representing NOM. UV254 measurements<br />

performed at UC utilized a Hewlett Packard (Palo Alto, Calif.) Model 8452A<br />

diode-array spectrophotometer. A 5-cm-path-length, 15-mL quartz cell was used for<br />

sample analysis. The absorbance at 254 nm was measured, and all reported values<br />

were expressed in units <strong>of</strong> I/cm. Method 5910 B (APHA et al. 1995) was followed<br />

except for ORW, MRW, and SRPW samples, which were acidified to pH < 2 with<br />

o-phosphoric acid and analyzed within 7 days. LCW acidified to pH < 2 with<br />

ophosphoric acid was used as a blank. For PRW and FGW, samples were not<br />

acidified and were analyzed within 24 hours. Nonacidified LCW was used to obtain<br />

a blank reading. In all cases, samples were stored at 4 C prior to analysis and were<br />

warmed to room temperature for analysis.<br />

Trihalomethanes<br />

THM analysis includes chlor<strong>of</strong>orm, bromodichloromethane, dibromochloromethane,<br />

and brom<strong>of</strong>orm. After specified chlorination time, samples were quenched<br />

with 5-mg/L sodium sulfite and held headspace-free until analysis. Samples were<br />

stored at 4 C for up to 1 week. Analysis was performed following USEPA Method<br />

524.2 Revision 3.0 (USEPA 1992) using a gas chromatograph (Model 3400, Varian,<br />

Palo Alto, Calif.) equipped with a purge-and-trap system (LSC 2000, Tekmar,<br />

Cincinnati, Ohio) and an electrolytic conductivity detector (Model 1000, Tracer,<br />

Austin, Texas). A blank sample was run daily for baseline correction, and the<br />

instrument was calibrated daily with a single standard containing known concentra<br />

tions <strong>of</strong> the four THM species. Standard calibration curves were obtained for<br />

concentrations <strong>of</strong> 5, 10, 25, and 50 ug/L <strong>of</strong> each THM species, and r2 values were<br />

typically greater than 0.99.<br />

Haloacetic Acids and Chloral Hydrate<br />

HAA and chloral hydrate (CH) analyses were conducted using the methods<br />

outlined in Table 2.6. These analyses were not performed at UC. The HAA species<br />

analyzed that constitute HAAS are monochloroacetic acid, dichloroacetic acid,<br />

trichloroacetic acid, monobromoacetic acid, and dibromoacetic acid. HAA samples<br />

analyzed at the University <strong>of</strong> Massachusetts (UM), Amherst, Mass., for LOW were<br />

reported as HAAS. However, most analyses included bromochloroacetic acid and<br />

were reported as the sum <strong>of</strong> six haloacetic acids (HAA6).<br />

Total Organic Halide<br />

TOX is a surrogate measure <strong>of</strong> adsorbable halogenated <strong>DBP</strong>s measured as<br />

Cl~. TOX analysis at UC was conducted according to Standard Method 5320<br />

(APHA et al. 1995) using a TOX analyzer (Dohrmann DX-20, Rosemount Analyti<br />

cal Corp., Santa Clara, Calif.). After quenching <strong>of</strong> the sample with 5-mg/L sodium<br />

sulfite and acidifying to pH < 2 with concentrated nitric acid, the sample was stored<br />

headspace-free at 4 C for up to 2 weeks. The instrument was calibrated <strong>by</strong> direct<br />

injection <strong>of</strong> a known concentration <strong>of</strong> 2,4,6-trichloropenol, and standard recoveries<br />

ranged from 90 to 110 percent. Blank samples were also analyzed daily.


Materials and Methods 15<br />

cis-l,2-Dichloroethene<br />

DCE analysis was performed according to USEPA Method 502.2, Revision<br />

2.0, using the same equipment as described earlier for THM analysis. Samples were<br />

held headspace-free in 15-mL vials until analysis, which was performed within 2<br />

weeks. Standard calibration curves at 5, 10, 25, and 50 ug/L were created daily.<br />

Nonstandard Methods<br />

Chlorination<br />

Chlorination was conducted under uniform formation conditions, de<br />

scribed in detail in Appendix C. The UFC test, developed as part <strong>of</strong> this study,<br />

provides constant Chlorination conditions that are representative <strong>of</strong> nationwide<br />

average distribution system conditions. These conditions are as follows: a 24-hour<br />

incubation time at 20.0 1.0 C, buffered at pH 8.0 0.2, and a 24-hour free chlorine<br />

residual <strong>of</strong> 1.0 0.3 mg/L as Clr UFC test conditions allow for DBF formation<br />

comparisons to be made between different waters or between different treatments<br />

performed on the same water. SDS test conditions were utilized in addition to UFC<br />

for the field-scale <strong>GAC</strong> operation and RSSCT verification study for MRW, ORW,<br />

and LGW. The SDS test conditions used for the RSSCT verification study were<br />

based on conditions utilized <strong>by</strong> each utility during the field-scale <strong>GAC</strong> operation<br />

and are summarized in Table 2.7. For the PRW study, UFC and SDS conditions were<br />

the same.<br />

Humic-Nonhumic Fractionation<br />

Humic-nonhumic fractionation was performed using XAD-8 resin (Rohm<br />

& Haas, Philadelphia, Pa.; see Thurman and Malcolm 1981). The resin was packed<br />

in an 8.0- or 11.0-mm-diameter glass chromatography column, and samples were<br />

pumped through the column at a flow rate <strong>of</strong> 4.0 mL/min. A k' (ratio <strong>of</strong> resin void<br />

volume to volume <strong>of</strong> sample passed) <strong>of</strong> 50 was used for all separation experiments.<br />

Prior to fractionation, samples were prefiltered through a 0.45-nm hydrophilic<br />

membrane filter (Millipore HV, Millipore Corp., Bedford, Mass.) and acidified<br />

with concentrated sulfuric acid to pH 2.0. The DOC measured in the effluent was<br />

operationally defined as the nonhumic fraction, while the resin retained the humic<br />

fraction. The DOC <strong>of</strong> the humic fraction was obtained <strong>by</strong> difference between the<br />

unfractionated sample and the nonhumic fraction. The resin was cleaned <strong>by</strong> passing<br />

one bed volume each <strong>of</strong> 0. IN sulfuric acid and 0.1/V sodium hydroxide three times.<br />

Table 2.7 UFC and SDS Chlorination conditions<br />

Parameter<br />

UFC<br />

MRW-SDS<br />

ORW-SDS<br />

LGW-SDS<br />

Incubation time (days)<br />

Incubation temperature (°C)<br />

PH<br />

Chlorine residual (mg CI^L)<br />

1<br />

20<br />

8.0<br />

1.0<br />

5<br />

28,20<br />

8.4<br />

1.0<br />

3<br />

30,23, 10<br />

8.0<br />

1.0<br />

2<br />

10<br />

7.3<br />

Dose: 2.8*<br />

* Initial dose; test was not base on a target residual


16 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

A sample <strong>of</strong> the humic fraction retained <strong>by</strong> the resin was obtained <strong>by</strong> eluting<br />

with 0.1 N sodium hydroxide in an upflow mode. Mass balance calculations<br />

indicated that the measured humic fraction was within 90 to 110 percent <strong>of</strong> the<br />

calculated humic fraction. For chlorination, the unfractionated sample and the<br />

humic fraction were diluted to the DOC <strong>of</strong> the nonhumic fraction using low-organiccontent<br />

water (


Materials and Methods 17<br />

DOC/100 mL <strong>by</strong> a combination <strong>of</strong> rotary vacuum evaporation and room temperature<br />

evaporation under N2 atmosphere. Using a Buchi rotavapor (Model R114withB480<br />

<strong>Water</strong>bath), 1 L <strong>of</strong> sample was reduced to 30 mL at 22 C and under a vacuum<br />

pressure <strong>of</strong> approximately 27.5 in. Hg (70 mm Hg). This volume reduction was<br />

conducted in a successive manner where<strong>by</strong> the concentrate was decanted from a<br />

larger vessel into a smaller flask, allowing the physical separation <strong>of</strong> precipitated<br />

salts. If the salt precipitate appeared colored, it was rinsed with water and this wash<br />

was combined with the concentrate. Upon reaching a volume <strong>of</strong> 30 mL, the<br />

concentrate was placed in small sample vials (50 mL); using ultra-high-purity N2 gas<br />

with a Pierce Reacti-Therm III Heating Stirring Module, the sample was concen<br />

trated to a final volume <strong>of</strong> approximately 100 mL. In this volume, the amount <strong>of</strong><br />

DOC was targeted at 1 mg. If salt precipitation was visible and if the precipitate was<br />

colored, the precipitate was rinsed in an attempt to minimize DOC loss. DOC losses<br />

were typically less than 5 percent.<br />

Pyrolysis samples were prepared in quartz capillary tubes from this final<br />

concentrate <strong>by</strong> coating the tubes with 20 mL <strong>of</strong> the concentrate (200 mg DOC) and<br />

allowing them to dry overnight at room temperature. Care was taken to evenly<br />

distribute the sample along the surface area <strong>of</strong> the tubes and not to overload the tubes<br />

with sample. In order to produce reproducible and uniform pyrolytic conditions,<br />

thermal gradients within the sample tube were avoided (Irwin 1982). Overloaded<br />

tubes could lead to the formation <strong>of</strong> secondary pyrolysis product because <strong>of</strong> a<br />

recombination <strong>of</strong> primary fragments. When there was a high concentration <strong>of</strong> salt<br />

in the sample (i.e., hard water source or high-alkalinity water), the quartz tubes were<br />

placed in an oven at 50 C overnight. Once the tubes were dried, deactivated glass<br />

wool was placed in each end <strong>of</strong> the tube.<br />

Specifications for each component <strong>of</strong> the analysis are shown in Table 2.8.<br />

The quartz tubes <strong>of</strong> samples were placed inside a coiled platinum filament <strong>of</strong> the<br />

Pyroprobe 2000. This unit employed heated filament pyrolysis and had flexible<br />

temperature-ramping capabilities up to a maximum temperature <strong>of</strong> 1,500 C. This<br />

instrument also provided coupled temperature ramping <strong>of</strong> the interface and the<br />

pyrolysis filament. Often the first pyrolysis tube <strong>of</strong> a set <strong>of</strong> replicates was sacrificed<br />

in order to screen any special features <strong>of</strong> a particular sample, particularly salt<br />

influences. In the absence <strong>of</strong> salt effects (i.e., for low carbonate concentration),<br />

sample tubes were loaded at an initial interface temperature <strong>of</strong> 70 C. The interface<br />

was then ramped from 70 to 250 C at 30 C/s simultaneously with the temperature<br />

ramp <strong>of</strong> the pyrolysis filament.<br />

For carbonate salts, an additional effect was observed that involved a<br />

swamping effect <strong>by</strong> CO 2 gas as it sublimed from the pyrolysis sample. A method to<br />

minimize this effect was developed and involved loading the pyrolysis sample into<br />

the interface at an initial temperature <strong>of</strong> 120 C and holding the sample at this<br />

temperature for 90 minutes. This procedure drove <strong>of</strong>f inorganic CO2 and associated<br />

water without loss <strong>of</strong> organic carbon. The interface was then ramped from 120 to<br />

250 C at 33.3 C/s simultaneously with the pyrolysis ramp as described earlier.<br />

Once the pyrolysis run was completed, the interface temperature was held at 250 C<br />

for 2 minutes and then reduced to 50 C. By a lowering <strong>of</strong> the temperature <strong>of</strong> the<br />

interface, CO2 sublimation was minimized. Although these steps help to eliminate<br />

salt effects, it was also necessary to increase the frequency <strong>of</strong> cleaning the mass<br />

spectrometry (MS) source and interface. Because all the first <strong>GAC</strong> breakthrough<br />

samples and Salt River samples had very high salt levels, the source was cleaned


18 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

Table 2.8 Specifications for PY-GC-MS analysis<br />

Analytical step Instrument Conditions<br />

Pyrolysis<br />

Chemical Data Systems<br />

Pyroprobe 2000<br />

Heated filament<br />

Temperature ramp: 100-720°C<br />

at 20°C/millisecond<br />

Hold 720°C for 20 seconds<br />

Internal temperature 625±5°C<br />

Gas chromatography Fison 8030<br />

Supelcowax 10, 60-m column<br />

Splitless injection<br />

45-260°C at 2°C/min;<br />

at 260°C hold 45 minutes<br />

Mass spectrometry Fison MD 800<br />

Operated at 70 eV,<br />

scanning 20-400 amu at 1 scan/s<br />

Data analysis<br />

Digital 433dxLp<br />

NIST Library<br />

NIST = National Institute <strong>of</strong> Standards and Technology<br />

after every two runs, and the interface was cleaned with a cotton swab after each run<br />

in order to avoid salt deposition throughout the instrument.<br />

The pyrolysis temperature was ramped resistively <strong>by</strong> a booster current from<br />

an initial temperature <strong>of</strong> 70 or 120 C to 725 C at 20 C/millisecond. The final<br />

temperature was held for 20 seconds. This internal temperature was critical for<br />

reproducibility <strong>of</strong> organic fragmentation consistent with predicted patterns. A minithermometer<br />

(Cole-Palmer 8508-45 Type K) was used to verify these conditions.<br />

Pyrolysis was conducted in an inert helium atmosphere (ultrahigh purity);<br />

with the temperature pr<strong>of</strong>ile just described, flash pyrolysis <strong>of</strong> the parent organic<br />

compounds occurred. The pyrolysis fragments were quickly swept onto the gas<br />

chromatography (GC) column <strong>of</strong> the Fisons 8030 gas chromatograph. A Supelcowax<br />

10,60-m polar column was used with splitless injection (allowing use <strong>of</strong> very small<br />

sample mass). A 60-m length was chosen because better peak separation was<br />

achieved. The following temperature gradient was used: 45 C held for 15 minutes,<br />

45-260 C ramping at 2 C/min and 260 C held for 45 minutes.<br />

A Fisons MD 800 mass spectrometer served as the detector. It was operated<br />

at 70 eV and typically scans from 20-400 amu at 1 scan/s.<br />

Quality Assurance-Quality Control<br />

The quality assurance-quality control steps stipulated <strong>by</strong> the respective<br />

analytical methods were followed. Table 2.9 summarizes the analytical precision<br />

for the methods followed at UC and the laboratories <strong>of</strong> the participating utilities.


University<br />

<strong>of</strong> Cincinnati<br />

Detection % Samples Detection % Samples Detection % Samples Detection %<br />

limit recovery replicated (%) limit recovery replicated (%) limit recovery replicated (%) limit recovery<br />

Samples<br />

replicated (%)<br />

100<br />

100<br />

—<br />

10<br />

—<br />

—<br />

—<br />

—<br />

I<br />

c<br />

a<br />

ex.<br />

a-<br />

I<br />

Table 2.9 Summary <strong>of</strong> detection limits and precisions for analytical methods<br />

Cincinnati Jefferson Parish University<br />

<strong>Water</strong> Works <strong>Water</strong> Department <strong>of</strong> Massachusetts<br />

TOC or DOC<br />

uv254<br />

CD<br />

TOX<br />

0.2mg/L >90 10 10 pg/L 98<br />

— — — — —<br />

0.01 mg/L 98 — 0.07 mg/L —<br />

5M9/L 93-107 10 0.27 pg Cl~ 99.2<br />

100 50 pg/L ±5<br />

— 0.01/cm —<br />

— — —<br />

100 — —<br />

100<br />

100<br />

—<br />

—<br />

0.1 mg/L ±5<br />

0.01/cm —<br />

0.1 mg/L —<br />

3 M9/L 95-105<br />

THM<br />

0.1 ng/L — 10 0.001-0.04 (jg/L >95<br />

5 0.5 pg/L 90-110<br />

100<br />

1 pg/L >95<br />

HAA<br />

0.3-0.5 \ig/L — 10 — —<br />

— 0.5pg/L 90-110<br />

10<br />

— —<br />

CH<br />

Aldehydes<br />

Ketoacids<br />

0.5 Mg/L — 10 — —<br />

— — — — —<br />

— — — — —<br />

— 0.5pg/L 90-100<br />

— — —<br />

— 1 \ig/L 80-120<br />

5<br />

—<br />

5<br />

— —<br />

— —<br />

— —<br />

10-100<br />

CD = chlorine demand<br />

— indicates not known


Chapter 3___________<br />

Using the RSSCT for Prediction <strong>of</strong><br />

Field-Scale NOM <strong>Removal</strong> and <strong>DBP</strong><br />

Control <strong>by</strong> GAG <strong>Adsorption</strong><br />

Studies to ascertain and optimize <strong>GAC</strong> performance are typically con<br />

ducted at pilot-scale prior to full-scale application. Pilot-scale studies, however, are<br />

costly and time-consuming. Consequently, it is desirable to be able to predict <strong>GAC</strong><br />

performance <strong>by</strong> alternative methods that may reduce both cost and time. Sontheimer<br />

et al. (1988) described both successful and unsuccessful applications <strong>of</strong> mathemati<br />

cal models to predict field-scale (both pilot- and full-scale) breakthrough. The<br />

modeling approach presents various difficulties. There are many limitations in<br />

accurately modeling a mixture such as NOM, and many model parameters are<br />

needed.<br />

An alternative to mathematical modeling is to use the RSSCT, which was<br />

developed to decrease the time and effort needed to assess the concentration<br />

breakthrough in a <strong>GAC</strong> column. To directly predict field-scale behavior, similitude<br />

to field-scale <strong>GAC</strong> systems has been maintained (Sontheimer et al. 1988). The<br />

relationship for the EBCT, hydraulic loading, column length, and operation time <strong>of</strong><br />

the small-scale column and large-scale (pilot- or full-scale) column is determined<br />

through the use <strong>of</strong> dimensional analysis and is a function <strong>of</strong> the ratio <strong>of</strong> <strong>GAC</strong> particle<br />

sizes utilized in the two columns. The RSSCT utilizes smaller <strong>GAC</strong> particle sizes,<br />

which yield shorter EBCTs, column lengths, and operation times compared to the<br />

large-scale columns. Successful application <strong>of</strong> the RSSCT produces breakthrough<br />

curves that are equivalent to those <strong>of</strong> a full- or pilot-scale adsorber. The method was<br />

pioneered <strong>by</strong> Frick (1982) and further developed into the RSSCT and extensively<br />

applied <strong>by</strong> Crittenden and coworkers (Crittenden et al. 1986, 1987, 1989, 1991;<br />

Hineline et al. 1987). A survey <strong>of</strong> past comparisons <strong>of</strong> the RSSCT to field-scale data<br />

for many waters and a variety <strong>of</strong> parameters can be found in Summers et al. (1995).<br />

The RSSCT has a number <strong>of</strong> advantages over other methods used to predict<br />

or simulate full-scale <strong>GAC</strong> performance, such as pilot-scale tests or predictive<br />

models. Depending on the conditions, the RSSCT can be conducted in


22 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

capacity and kinetics are not required, nor is the use <strong>of</strong> numerical or analytical<br />

models.<br />

However, the RSSCT has drawbacks due to the nature <strong>of</strong> the influent water.<br />

Because there is <strong>of</strong>ten variability in the quality <strong>of</strong> the influent water, the use <strong>of</strong> a<br />

batch influent for the RSSCT may lead to results that are not representative <strong>of</strong> the<br />

long-term operation <strong>of</strong> full- or pilot-scale systems, which show the effects <strong>of</strong><br />

seasonal water quality variability. Thus, the selection <strong>of</strong> a representative sample to<br />

serve as the influent to the RSSCT for representative comparisons is critical to the<br />

success <strong>of</strong> RSSCT simulations. Seasonal trends can be addressed <strong>by</strong> conducting a<br />

series <strong>of</strong> RSSCTs using batch influent water sampled at different times <strong>of</strong> the year.<br />

Furthermore, removal <strong>of</strong> DBF precursors in a <strong>GAC</strong> column due to long-term<br />

biodegradation is not simulated <strong>by</strong> the RSSCT, since the microorganisms may not<br />

have sufficient time to acclimate.<br />

Objectives and Approach________________<br />

The objectives <strong>of</strong> the work presented in this chapter were to verify the use<br />

<strong>of</strong> the RSSCT to predict full- or pilot-scale NOM removal and DBF control <strong>by</strong> <strong>GAC</strong><br />

adsorption. For this study, the appropriateness <strong>of</strong> the RSSCT was verified utilizing<br />

four waters from utilities operating either full- or pilot-scale <strong>GAC</strong> facilities: ORW,<br />

LGW, MRW, and PRW. Detailed information about each utility and pretreatment<br />

conditions is presented in Chapter 2. The RSSCT was designed based on either fullor<br />

pilot-scale operational conditions using the proportional diffusivity approach.<br />

This design approach, in which the scaling factor (SF) is the ratio <strong>of</strong> the particle<br />

diameter in the large column to the particle diameter in the small column, has been<br />

found appropriate for the scaling <strong>of</strong> columns used for NOM and DBF precursor<br />

adsorption (Summers et al. 1989, 1994c, 1995). Field-scale and RSSCT design<br />

parameters are presented in Table 3.1; RSSCT design equations are presented in<br />

Appendix A.<br />

The RSSCT and the full- or pilot-scale contactors were monitored for NOM<br />

removal <strong>by</strong> the surrogate parameters TOC and UV254 . DBF formation at both field<br />

and bench scale was assessed <strong>by</strong> the formation <strong>of</strong> TOX, TTHM, HAA6, and CH<br />

under utility-specific SDS conditions. During the RSSCT verification run, influent<br />

and effluent samples were also chlorinated under UFC. These uniform conditions<br />

Table 3.1 RSSCT and field-scale <strong>GAC</strong> contactor design parameters<br />

<strong>Water</strong><br />

<strong>GAC</strong> type<br />

Particle diameter (mm)<br />

RSSCT,<br />

"so<br />

Field scale,<br />

"LC<br />

Scaling<br />

factor<br />

EBCT (minutes)<br />

RSSCT,<br />

EBCTSC<br />

Field scale,<br />

EBCTLC<br />

ORW<br />

MRW<br />

LGW<br />

PRW<br />

F400<br />

<strong>GAC</strong> 40<br />

F400*<br />

F400<br />

0.08<br />

0.11<br />

0.20<br />

0.08<br />

1.05<br />

1.05<br />

1.49<br />

1.05<br />

12.6<br />

9.4<br />

7.4<br />

. 12.6<br />

1.2<br />

0.66<br />

2.0<br />

1.6<br />

15<br />

6.25<br />

15<br />

20<br />

* Sieved to yield a larger size


Using the RSSCTfor Field-Scale Prediction 23<br />

were used for all other RSSCT runs. The comparison between DBF formation under<br />

SDS conditions and UFC is presented in Chapter 4. Full- or pilot-scale chlorination<br />

was performed <strong>by</strong> the utility, while bench-scale chlorination was performed at UC.<br />

Results and Discussion_________________<br />

Ohio River <strong>Water</strong><br />

The TOC results <strong>of</strong> the full-scale contactor and RSSCT are shown in Figure<br />

3.1 for ORW. For this figure and the remaining figures in this chapter, the full- or<br />

pilot-scale effluent values are normalized with respect to the influent concentration<br />

<strong>of</strong> the same sample day. The full- or pilot-scale influent values are normalized with<br />

respect to the average influent concentration. The RSSCT influent concentration is<br />

the average value <strong>of</strong> two or three measurements <strong>of</strong> the batch influent, and the RSSCT<br />

effluent values are normalized to that value. The RSSCT run time is expressed as<br />

that <strong>of</strong> the full-scale column <strong>by</strong> using the SF; i.e., scaled operation time = RSSCT<br />

run time x SF.<br />

The full-scale influent TOC concentration <strong>of</strong> ORW varied <strong>by</strong> nearly a factor<br />

<strong>of</strong> 2 over the course <strong>of</strong> the 210-day run, but the average influent concentration was<br />

the same as that <strong>of</strong> the batch RSSCT influent, which was collected on day 120. In<br />

spite <strong>of</strong> the influent concentration variation, the full-scale breakthrough behavior<br />

was extremely well predicted <strong>by</strong> the RSSCT. Biodegradation, which probably<br />

occurred in the full-scale <strong>GAC</strong> column and presumably not in the RSSCT, did not<br />

have a detectable impact on the comparison. The BDOC levels <strong>of</strong> the full-scale and<br />

RSSCT influents were only 10 percent <strong>of</strong> the TOC.<br />

Similar success was found for UV254 breakthrough behavior, another<br />

indicator <strong>of</strong> NOM breakthrough behavior, as shown in Figure 3.2. Figures 3.3 to 3.6<br />

illustrate breakthrough behavior at both bench and full scale for SDS-TOX, SDS-<br />

TTHM, SDS-HAA6, and SDS-CH, respectively. DBF formation breakthrough was<br />

similar to NOM breakthrough, and the RSSCT successfully predicted full-scale<br />

conditions for these parameters.<br />

Lake Gaillard <strong>Water</strong><br />

The RSSCT and pilot-scale <strong>GAC</strong> performance for LOW is shown in<br />

Figures 3.7 through 3.10. The RSSCT was performed on-site at SCCRWA. Figure<br />

3.7 shows TOC breakthrough. The ozonated pilot-scale influent remained relatively<br />

constant over the course <strong>of</strong> the run, and the average pilot-scale and batch RSSCT<br />

influent TOC concentrations were essentially the same. Although the influent<br />

concentrations were the same, the pilot-scale contactor showed earlier break<br />

through.<br />

Figures 3.8 to 3.10 compare DBF control for SDS-TOX, SDS-TTHM, and<br />

SDS-HAA5, respectively. CH concentrations were at or below the detection limit.<br />

The RSSCT predicted DBF control at the pilot-scale level extremely well. For<br />

LGW, <strong>GAC</strong> very effectively removed DBF precursors: After 4 months the effluent<br />

concentration reached only approximately 30 percent <strong>of</strong> the influent concentration<br />

for the three parameters investigated.


24 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> GA C <strong>Adsorption</strong><br />

Mississippi River <strong>Water</strong><br />

The TOC and UV254 breakthrough behaviors for MRW are compared in<br />

Figures 3.11 and 3.12, respectively. The ozonated pilot plant influent was relatively<br />

constant over the 75-day run, but the average TOC and UV254 values were 30 per<br />

cent higher than those <strong>of</strong> the RSSCT. The batch sample for the RSSCT influent was<br />

taken on day 15, at a time when the influent TOC and UV254 values were 10 percent<br />

lower than for the average pilot plant influent. In addition, during the 6-day shipping<br />

time from JPWD to UC, the influent was biodegraded no BDOC was detected in<br />

the shipped influent, while the pilot plant influent had averaged about 25 percent<br />

BDOC following ozonation. This change in TOC and UV254 level and composition<br />

affected the NOM breakthrough, as shown in Figures 3.11 and 3.12. As expected,<br />

the higher influent values <strong>of</strong> the pilot plant column led to earlier breakthrough than<br />

for the RSSCT. Another factor possibly contributing to the earlier breakthrough in<br />

the pilot plant is that the pilot-scale column was operated as a filter adsorber and was<br />

therefore backwashed regularly, while the RSSCT was not backwashed. However,<br />

Hong and Summers (1994) have shown that backwashing normally does not impact<br />

NOM breakthrough behavior! After about 50 percent breakthrough, the RSSCT and<br />

pilot plant results coincided. Similar behavior was found for SDS-TOX, SDS-<br />

TTHM, SDS-HAA6, and SDS-CH, as shown in Figures 3.13 to 3.16, respectively.<br />

The higher influent concentrations to the pilot-scale contactor resulted in earlier<br />

breakthrough.<br />

Passaic River <strong>Water</strong><br />

The TOC breakthrough behavior for PRW is shown in Figure 3.17. The<br />

sample for the RSSCT batch influent was collected on the first day <strong>of</strong> operation <strong>of</strong><br />

the pilot plant column; the pilot plant influent TOC decreased for the first 100 days,<br />

then increased during the next 100 days. Still, the normalized breakthrough patterns<br />

match up extremely well. Given the 30 percent higher average TOC concentration<br />

in the RSSCT influent, an earlier RSSCT breakthrough was expected. For UV254 ,<br />

shown in Figure 3.18, the pilot-scale showed earlier breakthrough. The results for<br />

DBF precursor breakthrough are shown in Figures 3.19 to 3.22. With the exception<br />

<strong>of</strong> SDS-TOX, the pilot-scale <strong>GAC</strong> contactor showed earlier breakthrough.<br />

Summary_________________________<br />

The RSSCT can be successful in predicting full- or pilot-scale NOM<br />

removal and DBF control. The RSSCT predictions do not seem to be affected <strong>by</strong> the<br />

small amount <strong>of</strong> biodegradation that normally occurs in field-scale operation.<br />

However, the breakthrough behaviors discussed in this chapter illustrate the<br />

importance <strong>of</strong> gathering representative samples for the batch RSSCT influent.<br />

Influent water quality in river water sources can significantly change throughout the<br />

year; thus, several RSSCTs are needed, spaced to capture seasonal trends. For<br />

waters that are ozonated, the RSSCT should be performed on-site as quickly as<br />

possible after ozonation. Delaying the RSSCT because <strong>of</strong> shipping or for other<br />

reasons can lead to biodegradation and changes in the NOM because ozonated<br />

samples typically have higher levels <strong>of</strong> biodegradable fractions. If these limitations


Using the RSSCT for Field-Scale Prediction 25<br />

are recognized, the RSSCT can still be used to investigate operational consider<br />

ations and the impact <strong>of</strong> various pretreatment options under controlled conditions.<br />

In this study, RSSCT predictions were compared to field-scale behavior on<br />

a purely qualitative basis. To determine whether there is a significant difference<br />

between field-scale and RSSCT breakthrough behavior, however, it is desirable to<br />

statistically compare the data. A rigorous statistical analysis to compare RSSCT and<br />

field-scale data has been attempted in the past, but various statistical alternatives<br />

have not led to an accepted approach. As part <strong>of</strong> this project, a statistical method for<br />

the comparison <strong>of</strong> RSSCT and field-scale <strong>GAC</strong> breakthrough curves was developed<br />

and tested using the data presented in this chapter.<br />

The statistical method developed was based on the paired t-test. The<br />

procedure was applied to the four waters and range <strong>of</strong> parameters that were<br />

discussed qualitatively. A detailed description <strong>of</strong> the method and results is presented<br />

in Appendix B.


o\<br />

•<br />

c<br />

i<br />

O<br />

8-<br />

'<br />

-n<br />

(Q*<br />

3<br />

u<br />

NJ<br />

Normalized concentration, C/C0<br />

(Q<br />

§<br />

CO<br />

Normalized concentration, C/CD<br />

CO<br />

O<br />

SL<br />

CD<br />

0)<br />

10<br />

Q.<br />

O<br />

c<br />

o SL<br />

CD<br />

D><br />

Q.<br />

3)<br />

(A<br />

(/)<br />

O<br />

O<br />

O<br />

co<br />

s<br />

3 QJ<br />

S<br />

CD<br />

0)<br />

5<br />

(Q 3-<br />

—«•<br />

O<br />

O<br />

o'<br />

3<br />

CD<br />

(Q 3"<br />

—h<br />

O<br />

O 3; 5'<br />

3<br />

CD<br />

CD"<br />

03<br />

i


o<br />

b<br />

CO<br />

D<br />

I<br />

(D<br />

O<br />

ro<br />

en<br />

-i<br />

'<br />

re' 53<br />

£<br />

t>5<br />

a. o'<br />

s<br />

KJ<br />

( 02<br />

co<br />

Normalized concentration, C/CQ<br />

CO<br />

CO<br />

Normalized concentration, C/C0<br />

CO<br />

oSL<br />

o><br />

0)<br />

Q.<br />

73<br />

V)<br />

(O<br />

O<br />

CO<br />

O<br />

(0<br />

(0<br />

o<br />

SL<br />

(D<br />

B)<br />

a<br />

33<br />

CO<br />

CO<br />

o<br />

CO<br />

D<br />

CO<br />

o<br />

X<br />

CO<br />

8<br />

(D<br />

Q.<br />

O<br />

T3<br />

en<br />

o<br />

•vl<br />

en<br />

p<br />

o<br />

o<br />

en<br />

CO<br />

0)<br />

5 3-<br />

o c<br />

(Q 3"<br />

CO E.CO E.<br />

OT ~8*<br />

Hg<br />

0)<br />

0)<br />

I<br />

o<br />

c<br />

(D<br />

O<br />

a;<br />

3 (D<br />

0)<br />

N><br />

tn<br />

O<br />

T<br />

O<br />

3; o'<br />

en<br />

o<br />

O<br />

5<br />

(D^<br />

3<br />

CD<br />

en<br />

I<br />

&<br />

CD


Oo<br />

§<br />

(Q<br />

3<br />

CO<br />

b> Normalized concentration, C/CD n Normalized concentration, C/C0<br />

-n<br />

(0<br />

o<br />

SL<br />

(V<br />

a><br />

a<br />

33<br />

to<br />

(A<br />

o<br />

D<br />

to<br />

6<br />

X<br />

(D<br />

SB<br />

S 3"<br />

O<br />


Ol n<br />

Ji TO'<br />

£<br />

a.<br />

r>'<br />

o<br />

s<br />

3<br />

CO<br />

CO Normalized concentration, C/CQ<br />

TJ<br />

I<br />

(A<br />

O<br />

Si<br />

<br />

0.<br />

31<br />

CO<br />

CO<br />

O<br />

(O<br />

D<br />

CO<br />

or<br />

3<br />

D)<br />

S<br />

<br />

ff<br />

O<br />

Q)<br />

57<br />

a<br />

I<br />

(D<br />

O<br />

ca<br />

s<br />

&o •o<br />

(D<br />

ro<br />

cn<br />

cn<br />

o<br />

cn<br />

o<br />

ro -<br />

cn<br />

cn<br />

o<br />

Normalized concentration, C/C0<br />

O •<br />

33 T37J "D<br />

£§85<br />

O J>0 J,<br />

l|»f<br />

-vl O>


1<br />

2<br />

I<br />

'<br />

(Q<br />

3<br />

31<br />

I<br />

(A<br />

O<br />

SL<br />

(D<br />

o><br />

a<br />

33<br />

0)<br />

(/)<br />

O -J- -» -* -»•<br />

31<br />

I<br />

V)<br />

o<br />

0)<br />

Q.<br />

o<br />

en<br />

(Q<br />

Normalized concentration, C/C0<br />

I<br />

U<br />

<br />

to<br />

5<br />

o<br />

(O<br />

ff<br />

o2.<br />

5"<br />

(D ex<br />

o<br />

ST.<br />

O<br />

eft<br />

(D<br />

(ff<br />

(0<br />

D<br />

i<br />

0)<br />

I<br />

B><br />

?T<br />

(0<br />

O<br />

Q><br />

5T<br />

a<br />

I<br />

(D<br />

I<br />

O<br />

(D<br />

"S<br />

0)


a<br />

0


3<br />

o<br />

§<br />

to<br />

i<br />

(Q<br />

I<br />

31<br />

Normalized concentration, C/CQ<br />

(Q C<br />

(D<br />

W<br />

Lt<br />

w<br />

TJ<br />

Normalized concentration, C/Cn<br />

o<br />

Q><br />

3 a<br />

S<br />

(0<br />

o 2J.<br />

(D<br />

Q)<br />

3<br />

a<br />

33<br />

to<br />

o<br />

i<br />

CD<br />

0)<br />

(Q 3"<br />

o><br />

3<br />

(Q<br />

>»4i<br />

O<br />

(0<br />

' W<br />

•5' (0<br />

•o<br />

TO<br />

Si<br />

a><br />

W<br />

(0<br />

55'<br />

w •5'<br />

•o<br />

I<br />

Q) i-^<br />

(D


00<br />

'<br />

i<br />

(Q<br />

Iw<br />

O)<br />

Normalized concentration, C/Cn<br />

O CO *. O> 00 O<br />

*> O) 00<br />

(Q C<br />

3<br />

CO<br />

en<br />

Normalized concentration, C/CQ<br />

ooooo-*-*-*<br />

(0<br />

o<br />

SB<br />

3<br />

O.<br />

3J<br />

0)<br />

0)<br />

O<br />

0)<br />

O<br />

(/)<br />

6 I<br />

(D<br />

0)<br />

(Q 3"<br />

(D<br />

Q.<br />

O<br />

T3<br />

(D<br />

I O=1<br />

=±<br />

(D<br />

"3<br />

0)<br />

«<<br />

(A<br />

v><br />

o 9L<br />

(D<br />

o><br />

Q.<br />

0)<br />

D<br />

O)<br />

u<br />

m<br />

o><br />

o<br />

(Q 3T<br />

(0<br />

52. 35'<br />

•5' o><br />

-a m<br />

3J


I<br />

w<br />

00<br />

3;<br />

I<br />

w<br />

o<br />

91.<br />


a'<br />

OQ<br />

n^<br />

s<br />

(Q<br />

C<br />

a>3<br />

(0<br />

o SL<br />

(D<br />

0)<br />

Q.<br />

30<br />

W<br />

O<br />

0)<br />

cr<br />

3 o>5<br />

3-<br />

o c<br />

(Q<br />

CO<br />

s(D<br />

Q.<br />

O •o<br />

(D<br />

O<br />

E±<br />

(D<br />

"S<br />

0)<br />

Q.<br />

(0<br />

(A<br />

o<br />

0)<br />

o<br />

s<br />

oX<br />

tt><br />

0)<br />

I<br />

o<br />

CQ<br />

0)<br />

S<br />

(D<br />

Q.<br />

O<br />

T3<br />

<<br />

(A<br />

0)<br />

V)<br />

a> o> o'<br />

33


u><br />

>3<br />

1<br />

T;<br />

8-<br />

i<br />

Is<br />

TJ<br />

6"<br />

Normalized concentration, C/C0<br />

o<br />

SL<br />

(0<br />

a<br />

Q.<br />

3D<br />

O en<br />

(/><br />

S<br />

(D<br />

o Q.<br />

(/)<br />

•D O<br />

6 X<br />

(D<br />

cr §'<br />

(0<br />

E*<br />

(D<br />

O<br />

(Q 0)<br />

o<br />

•D<br />

(0<br />

o'<br />

5<br />

CD<br />

I<br />

O<br />

(Q<br />

u<br />

Normalized concentration, C/C0<br />

fo<br />

(0<br />

o<br />

D><br />

Q.<br />

3)<br />

V)<br />

0)<br />

O)<br />

a<br />

or<br />

S Q)<br />

|<br />

sill<br />

"* OD CO<br />

3<br />

|<br />

(Q<br />

(0<br />

(A<br />

5'<br />

0)<br />

5<br />

CD<br />

! (D


Chapter 4___________<br />

Comparison <strong>of</strong> Breakthrough Patterns<br />

for Simulated Distribution System and<br />

Uniform Formation Condition DBFs<br />

In Chapter 3, SDS conditions were utilized to verify the RSSCT <strong>by</strong> com<br />

parisons <strong>of</strong> field-scale and RSSCT breakthrough curves for a given water. However,<br />

SDS conditions change from utility to utility, so results from different waters cannot<br />

be directly compared. To facilitate comparisons among different waters, the UFC<br />

chlorination test was developed. The UFC test can also be used to analyze the effects<br />

<strong>of</strong> pretreatment, EBCT, and other operational parameters on <strong>GAC</strong> breakthrough for<br />

a single water.<br />

The concentration <strong>of</strong> DBFs formed under SDS conditions may differ<br />

compared to that under UFC, because <strong>of</strong> variations in incubation time, temperature,<br />

pH, or chlorine dose and residual. After <strong>GAC</strong> treatment, however, the effluent DBF<br />

concentrations are <strong>of</strong>ten normalized <strong>by</strong> being divided <strong>by</strong> influent concentrations,<br />

there<strong>by</strong> increasing the potential for a close match <strong>of</strong> the breakthrough patterns using<br />

UFC and SDS conditions. Of interest is whether normalized DBF breakthrough<br />

curve patterns under UFC are similar to those produced under SDS conditions. If<br />

there are no differences in the normalized breakthrough curves under UFC and SDS<br />

conditions, then studies conducted with UFC can be used <strong>by</strong> utilities to ascertain the<br />

adsorption breakthrough under SDS conditions.<br />

Experimental Conditions_________________<br />

For three <strong>of</strong> the four waters used in the verification study (Chapter 3), <strong>GAC</strong><br />

influent and effluent samples were chlorinated under both UFC and SDS conditions,<br />

allowing for comparisons <strong>of</strong> the normalized DBF breakthrough patterns observed<br />

under both chlorination tests. These waters were ORW, LOW, and MRW. The<br />

treatment conditions for these waters are described in Chapter 2. The three waters<br />

represent a range <strong>of</strong> water quality conditions. TOC concentrations prior to <strong>GAC</strong><br />

treatment range from 1.7 to 2.7 mg/L,whileBr concentrations range from 14to 132<br />

ug/L. Thus, a range <strong>of</strong> DBF formation is expected. Table 4.1 compares UFC to<br />

water-specific SDS chlorination conditions.<br />

SDS incubation times were higher in all three cases, ranging from 2 to 5<br />

days. For both ORW and MRW, SDS incubation temperature was varied during<br />

breakthrough to better simulate field-scale conditions. The pH values used for the<br />

37


38 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

Table 4.1 UFC and SDS chlorination conditions<br />

Parameter<br />

UFC<br />

ORW-SDS<br />

LGW-SDS<br />

MRW-SDS<br />

Incubation time (days)<br />

Incubation temperature (°C)<br />

PH<br />

Chlorine residual (mg CI2/L)<br />

1<br />

20.0<br />

8.0<br />

1.0<br />

3<br />

30, 23, 10<br />

8.0<br />

1.0<br />

2<br />

10<br />

7.3<br />

Dose: 2.8*<br />

5<br />

28,20<br />

8.4<br />

1.0<br />

* Initial dose; test was not based on a residual<br />

SDS tests were similar to UFC, ranging from 7.3 to 8.4. For ORW and MRW, the<br />

chlorine residual was the same as that required for UFC, while LGW used a chlorine<br />

dose-based test resulting in a different residual after the incubation period.<br />

Results and Discussion________________<br />

For the three waters examined, the range <strong>of</strong> SDS chlorination conditions<br />

outlined in Table 4.1 led to differences in DBF formation in the influent to <strong>GAC</strong> as<br />

compared to UFC. Table 4.2 compares the TTHM, HAA6 or HAAS, TOX, and CH<br />

concentrations formed under both UFC and SDS conditions in the influent <strong>GAC</strong><br />

samples for all three waters.<br />

It is evident that the SDS conditions used for ORW and MRW increased the<br />

formation <strong>of</strong> all DBFs analyzed as compared to UFC, because <strong>of</strong> longer incubation<br />

times and higher incubation temperatures. For LGW, SDS conditions resulted in<br />

lower DBF concentrations. Although incubation times were longer and chlorine<br />

residuals higher, the SDS incubation temperature was much lower than for UFC.<br />

The effects <strong>of</strong> incubation time, temperature, pH, and chlorine residual on DBF<br />

formation are discussed for ORW and SRPW in Appendix C.<br />

These differences in DBF yield were expected given the differences in<br />

chlorination between SDS conditions and UFC. After <strong>GAC</strong> treatment, weakly<br />

adsorbed DBF precursors show earlier breakthrough, followed <strong>by</strong> more strongly<br />

adsorbed precursors. It was important to determine whether the change in DBF<br />

precursor levels as a rsult <strong>of</strong> <strong>GAC</strong> treatment was proportionally different for UFC<br />

and SDS conditions. To evaluate this, a comparison <strong>of</strong> the <strong>GAC</strong> breakthrough<br />

patterns on a normalized basis was made for each DBF chlorinated under both types<br />

<strong>of</strong> conditions. If the pattern generated using UFC was representative <strong>of</strong> that using<br />

SDS conditions for each water, then the UFC breakthrough curves could be used to<br />

estimate breakthrough at different chlorination conditions.<br />

Ohio River <strong>Water</strong><br />

Figures 4.1, 4.2, and 4.3 show the effect <strong>of</strong> chlorination conditions on the<br />

<strong>GAC</strong> breakthrough <strong>of</strong> TTHM, HAA6, and TOX, respectively, for ORW. CH<br />

concentrations were difficult to compare because most effluent samples chlorinated<br />

under both types <strong>of</strong> conditions were below the detection limit. The breakthrough<br />

patterns observed for TTHM and TOX were very similar for both UFC and SDS


Breakthrough for SDS and UFC DBFs 39<br />

Table 4.2 <strong>DBP</strong> formation in <strong>GAC</strong> influent under UFC and SDS conditions<br />

ORW LGW MRW<br />

Diff. Diff. Diff<br />

<strong>DBP</strong> UFC SDS UFC SDS UFC SDS<br />

TTHM (ug/L)<br />

HAA6 (ug/L)<br />

TOX (ug CI-/L)<br />

CH (ug/L)<br />

86.1<br />

25.5<br />

172<br />

3.9<br />

131<br />

50<br />

228<br />

11.3<br />

+52<br />

+96<br />

+33<br />

+290<br />

28.9<br />

31.6*<br />

132<br />

NA<br />

22.3<br />

18.1*<br />

109<br />

2.8<br />

-23<br />

-43<br />

-17<br />

NA<br />

50.2<br />

21.2<br />

169<br />

7.9<br />

85.1<br />

43.1<br />

255<br />

10.9<br />

+69<br />

+103<br />

+51<br />

+38<br />

Diff. = difference<br />

NA = not available<br />

*HAA5<br />

conditions. Conversely, a rapid increase in breakthrough was seen for HAA6 under<br />

UFC as compared to SDS conditions. The departure between these two curves may<br />

possibly be attributed to the fact that the HAAs were only partially quantified—<br />

analytical standards are available only for six <strong>of</strong> a total <strong>of</strong> nine HAA species. Those<br />

not analyzed in this evaluation are the mixed chloro-bromo species, which have<br />

been shown to represent a significant percentage <strong>of</strong> total HAA formation in waters<br />

containing high levels <strong>of</strong> bromide. This comparison is further confounded <strong>by</strong> the<br />

facts that (1) brominated DBFs typically form faster than their chlorinated analogs,<br />

which is relevant given that the UFC and SDS conditions represent different holding<br />

periods; (2) early in the breakthrough period, the bromide: DOC ratio (Br:DOC)<br />

and consequently bromide:chlorine (Br~:Cl 2) ratios were high, resulting in higher<br />

bromine substitution in DBFs; and (3) this particular water contained a relatively<br />

high bromide concentration (132 ug/L). These factors combine to hamper the<br />

comparison <strong>of</strong> normalized SDS and UFC HAA6 breakthrough curves.<br />

Lake Gailiard <strong>Water</strong><br />

The influent <strong>DBP</strong> precursor concentrations measured for LGW were<br />

usually much lower than for either ORW or MRW. Still, overall breakthrough<br />

patterns generated <strong>by</strong> the two types <strong>of</strong> chlorination conditions compared extremely<br />

well (except for two points around 30 and 60 days for THM breakthrough), as shown<br />

in Figures 4.4,4.5, and 4.6 for TTHM, HAA5, and TOX, respectively. Is should be<br />

noted here that LGW has much lower bromide levels compared to ORW, and thus<br />

some <strong>of</strong> the confounding factors were not apparent.<br />

Mississippi River <strong>Water</strong><br />

The effect <strong>of</strong> chlorination conditions on the <strong>GAC</strong> breakthrough patterns for<br />

MRW are shown in Figures 4.7,4.8,4.9, and 4.10 for TTHM, HAA6, TOX, and CH,<br />

respectively (except for four points on the CH breakthrough curve). Even though<br />

influent <strong>DBP</strong> formation was consistently much higher under SDS conditions than<br />

under UFC, effluent breakthrough curves, when normalized to the respective


40 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

influent values, showed very similar patterns for both types <strong>of</strong> conditions. Similar<br />

to LOW, MRW also had lower bromide levels compared to ORW.<br />

Summary_____________________________<br />

This study showed that chlorination under UFC resulted in normalized<br />

<strong>GAC</strong> breakthrough patterns that were very similar to those generated under the sitespecific<br />

SDS conditions for three waters encompassing a range <strong>of</strong> water quality<br />

parameters. Although the SDS conditions differed from UFC, as shown <strong>by</strong> the<br />

differences <strong>of</strong> up to a factor <strong>of</strong> 4 in <strong>GAC</strong> influent DBF formation, the normalized<br />

DBF breakthrough curves generated under UFC were very similar to those gener<br />

ated under SDS conditions. In 9 <strong>of</strong> the 10 cases studied, including four DBFs and<br />

three waters, UFC and SDS normalized breakthrough patterns compared extremely<br />

well. This indicates that the DBF precursors remaining after <strong>GAC</strong> treatment reacted<br />

proportionally the same way as before <strong>GAC</strong> treatment under both types <strong>of</strong> chlori<br />

nation conditions.<br />

It should be noted here that, for waters with elevated bromide levels, the<br />

comparisons may not be adequate, especially if all the halogenated species <strong>of</strong> a<br />

particular DBF are not considered. For example, the use <strong>of</strong> six species out <strong>of</strong> a total<br />

<strong>of</strong> nine HAAs resulted in differences between UFC and SDS conditions for ORW,<br />

a source with elevated bromide levels.<br />

The impact <strong>of</strong> bromide concentrations on this finding is important. For<br />

water with relatively high bromide concentrations, the speciation <strong>of</strong> DBFs within<br />

a specific class <strong>of</strong> DBFs may vary depending on chlorination conditions because <strong>of</strong><br />

varying Br~:Cl2 ratios between SDS conditions and UFC. The difference in HAA6<br />

breakthrough patterns between SDS conditios and UFC for ORW may be due in<br />

large part to the relatively high bromide concentrations in this water. For the other<br />

waters, which contained moderate to low bromide concentrations, the normalized<br />

breakthrough curves were more similar.


s<br />

S<br />

o?5-<br />

3-<br />

3<br />

OQ<br />

s-<br />

IOn<br />

a<br />

g.<br />

I<br />

to<br />

Co<br />

3J -n<br />

ff<br />

5*<br />

oS *><br />

CD M<br />

^<br />

m3:<br />

(D<br />

O<br />

o<br />

0)<br />

rt><br />

o'<br />

o<br />

a<br />

r+o'<br />

3<br />

tn<br />

O<br />

V)<br />

8<br />

(D<br />

Q.<br />

O<br />

T3<br />

(D<br />

I<br />

O<br />

:±.<br />

(D<br />

o<br />

IS)<br />

en<br />

s-J<br />

tn<br />

o<br />

o<br />

o<br />

Normalized concentration, C/C0<br />

p p<br />

Ko *•<br />

i__,__i<br />

p p<br />

b> bo<br />

i__,__i<br />

3? 3!


K)<br />

3<br />

o<br />

§ p-«<br />

^Co<br />

O<br />

•$- 3<br />

2<br />

8-<br />

o<br />

•a<br />

5'<br />

O 1<br />

a> (Q<br />

Is<br />

a 4*<br />

S ^<br />

S<br />

5 S<br />

o^5'<br />

a 5<br />

O<br />

O3<br />

Q.<br />

(A<br />

O<br />

3<br />

33<br />

CO<br />

CO<br />

O<br />

(D<br />

I<br />

O c<br />

(Q 3"<br />

—h<br />

O<br />

I-<br />

I<br />

s<br />

W<br />

ISO<br />

O<br />

CD °<br />

Q.<br />

O<br />

TJ<br />

(D<br />

Q> O)<br />

I °<br />

cr.<br />

(D<br />

-a g<br />

Q)<br />

"5<br />

o o<br />

Normalized concentration, C/C0<br />

p p p p p<br />

o KJ *. b> o»<br />

Ilfl<br />

m =6<br />

(D<br />

O<br />

O -4t<br />

O<br />

3;<br />

o<br />

I<br />

o3<br />

O<br />

O<br />

Q.<br />

o'<br />

r*!<br />

0)<br />

O<br />

3<br />

CO<br />

CO<br />

O<br />

X<br />

01<br />

o<br />

e(Q<br />

(0<br />

0)<br />

S 3-<br />

o<br />

(Q<br />

O<br />

•*<br />

O 3; o'<br />

i' 3D<br />

Normalized concentration, C/C0


p<br />

bo<br />

U-n WC<br />

0-n en c<br />

® a »<br />

3. 3. =<br />

-i CO<br />

oo r*<br />

_L CQ I<br />

I<br />

ui<br />

Co<br />

S<br />

a ?7-<br />

S-<br />

|<br />

00<br />

8<br />

to -o<br />

O I! O 3!<br />

<br />

•**+ BW<br />

o m<br />

-• =8<br />

(0<br />

a<br />

o •*<br />

o<br />

p<br />

o<br />

Normalized concentration, C/Cn D) Normalized concentration, C/C0<br />

p<br />

*.<br />

I<br />

p p<br />

b> bo<br />

i i<br />

(D n<br />

o<br />

o<br />

p b<br />

p p '*»• p<br />

ro 0)<br />

i I<br />

3<br />

O<br />

O<br />

Q.<br />

r3t<br />

o'<br />

to<br />

O<br />

CO<br />

CO<br />

O<br />

O<br />

X<br />

(D<br />

0)<br />

5<br />

(Q 3-<br />

3-<br />

CO<br />

&<br />

a<br />

o<br />

(D<br />

s±.<br />

<br />

O<br />

c<br />

(Q<br />

V)<br />

» & o<br />

o •o<br />

I §<br />

o<br />

cr.<br />

(D<br />

D)<br />

09<br />

O<br />

o<br />

ro<br />

o<br />

O<br />

0)<br />

Si<br />

^<br />

o


44 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> GA C <strong>Adsorption</strong><br />

UFC influent 50.2<br />

SDS influent 85.1<br />

UFC effluent<br />

SDS effluent<br />

Scaled operation time (days)<br />

Figure 4.7 Effect <strong>of</strong> chlorination conditions on RSSCT TTHM breakthrough for<br />

Mississippi River water<br />

1.0<br />

0.8 -<br />

HAA6 C0 (ug/L)<br />

UFC influent 21.2<br />

SDS influent 43.1<br />

• UFC effluent<br />

0 SDS effluent<br />

10 20 30<br />

Scaled operation time (days)<br />

Figure 4.8 Effect <strong>of</strong> chlorination conditions on RSSCT HAA6 breakthrough for<br />

Mississippi River water


Breakthrough for SDS and UFC DBFs 45<br />

8<br />

o<br />

1.0<br />

0.8 -<br />

0.6 -<br />

TOX<br />

UFC influent<br />

SDS influent<br />

• UFC effluent<br />

0 SDS effluent<br />

C0 (ugCI7L)<br />

169<br />

255<br />

I 0.4 -<br />

o<br />

T3<br />

0)<br />

N<br />

1 0.2 H<br />

o<br />

z<br />

0.0<br />

10 20 30<br />

40 50<br />

Scaled operation time (days)<br />

Figure 4.9 Effect <strong>of</strong> chlorination conditions on RSSCT TOX breakthrough for<br />

Mississippi River water<br />

CD (ug/L)<br />

UFC influent 7.9<br />

SDS influent 10.9<br />

• UFC effluent<br />

0 SDS effluent<br />

Scaled operation time (days)<br />

Figure 4.10 Effect <strong>of</strong> chlorination conditions on RSSCT CH breakthrough for<br />

Mississippi River water


Chapter 5<br />

NOM Characterization<br />

NOM is composed <strong>of</strong> a broad spectrum <strong>of</strong> organic substances with a wide<br />

range <strong>of</strong> chemical characteristics. Because <strong>of</strong> its heterogeneous nature, NOM<br />

cannot be measured directly; rather, it must be quantified using nonspecific surro<br />

gate parameters. This heterogeneity is also evident in the differing degrees <strong>of</strong><br />

adsorbability <strong>of</strong> these compounds <strong>by</strong> <strong>GAC</strong> (Sontheimer et al. 1988). Owen et al.<br />

(1992) indicated that treatment can be tailored for targeted removal <strong>of</strong> NOM<br />

fractions. In general, the easily adsorbable (and thus easily removable) fraction is<br />

thought to be more nonpolar and more humic in nature than the nonadsorbable<br />

fraction, as well as having intermediate molecular size and specific ultraviolet (UV)<br />

absorbance. The nonadsorbable fraction <strong>of</strong> NOM is thought to be composed <strong>of</strong> the<br />

polar nonhumic NOM fraction that has a smaller molecular size and low specific U V<br />

absorbance (Jackson et al. 1993). Semmens and Staples (1986) showed that 70 to<br />

90 percent <strong>of</strong> organic compounds passing through <strong>GAC</strong> were predominantly<br />

hydrophilic (nonhumic) and had a small MS. Earlier work with isolated humic<br />

substances showed that adsorbability decreased with increasing MS because <strong>of</strong> size<br />

exclusion in the smaller pores (Roberts and Summers, 1982). In that work, the<br />

different MS fractions <strong>of</strong> the humic substances were shown to be chemically<br />

homogenous. For natural waters, NOM is a mixture <strong>of</strong> humic and nonhumic<br />

substances, and the small MS fractions may be nonhumic in nature (Koechling and<br />

Summers 1995).<br />

In addition to a characterization <strong>of</strong> NOM into DOC fractions, the DBF<br />

precursor content in each fraction is also <strong>of</strong> interest. Semmens and Staples (1986)<br />

showed that the THM precursors in raw MRW were almost evenly split between the<br />

hydrophobic (humic) and hydrophilic (nonhumic) fractions. Koechling and Sum<br />

mers (1995) and Shukairy (1994) demonstrated that the small MS fraction contained<br />

similar specific DBF yields as compared to the larger MS fraction when all fractions<br />

were adjusted to the same DOC. Dryfuse et al. (1995), however, showed lower<br />

specific DBF yields for the small MS compounds compared to those for the larger<br />

MS fractions when the DOC was not adjusted.<br />

Objectives________________________<br />

The objectives <strong>of</strong> the work presented in this chapter are (1) to systematically<br />

characterize NOM in both raw and treated waters, as well as its removal <strong>by</strong> <strong>GAC</strong>,<br />

and (2) to relate it to DBF formation for a range <strong>of</strong> waters. The chapter is subdivided<br />

to discuss NOM characterization <strong>of</strong> raw water, field-scale treatment systems, and<br />

bench-scale treatment systems.<br />

47


48 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

Raw <strong>Water</strong> NOM Characterization<br />

Approach<br />

Raw water was sampled at the same time that water for the <strong>GAC</strong> influent<br />

was collected for ORW, LOW, MRW, PRW, SRPW, and FGW. Raw water NOM<br />

was characterized in terms <strong>of</strong> humic and nonhumic fractions, MS fractions, and<br />

biodegradability (as determined <strong>by</strong> BDOC). The unfractionated and fractionated<br />

waters were characterized in terms <strong>of</strong> DOC and specific UFC-TTHM yields accord<br />

ing to the methods outlined in Chapter 2. Seasonal water quality monitoring was<br />

performed for ORW and PRW. Raw water was collected four times, and samples<br />

were characterized in terms <strong>of</strong> humic and nonhumic fraction composition, biode<br />

gradability, and PY-GC-MS fractions.<br />

Results and Discussion<br />

Source <strong>Water</strong> Characteristics<br />

Table 5.1 summarizes the DOC fractionation and BDOC results for the six<br />

raw waters examined. The raw waters were more humic (53 to 75 percent) than<br />

nonhumic (25 to 47 percent) in nature. The relative percent composition <strong>of</strong> each<br />

fraction was dependent on the water source. FGW was the most humic in nature (75<br />

percent <strong>of</strong> the DOC was found in the humic fraction), while SRPW was the least<br />

humic in nature (53 percent). The intermediate MS fraction dominated the MS<br />

distribution, with most waters having an equal distribution between the largest and<br />

smallest MS fractions. The BDOC varied from below detection limit (


NOM Characterization 49<br />

Table 5.1<br />

Summary <strong>of</strong> raw water fractionation and BDOC results<br />

DOC (%)<br />

BDOC<br />

<strong>Water</strong><br />

Sample<br />

date<br />

DOC<br />

(mg/L)<br />

Nonhumic<br />

Humic<br />

3,000<br />

MS<br />

(mg/L)<br />

%<br />

<strong>of</strong> DOC<br />

ORW<br />

1 1/4/93<br />

2.2<br />

45<br />

55<br />

19<br />

60<br />

21<br />

0.5<br />

23<br />

LGW<br />

MRW<br />

5/25/94<br />

10/30/93<br />

2.6<br />

3.7<br />

40<br />

34<br />

60<br />

66<br />

10<br />

14<br />

60<br />

76<br />

30<br />

10<br />

0.4<br />

0.9<br />

15<br />

24<br />

PRW<br />

11/16/94<br />

3.9<br />

44<br />

56<br />

16<br />

63<br />

21<br />

1.0<br />

26<br />

SRPW<br />

FGW<br />

8/10/94<br />

2/28/95<br />

13<br />

2.2<br />

47<br />

25<br />

53<br />

75<br />

20<br />

8<br />

65<br />

90<br />

15<br />

2<br />

BDL<br />

2.2<br />

NA<br />

17<br />

BDL = below detection limit<br />

NA = not applicable<br />

Table 5.2<br />

<strong>Water</strong><br />

Specific TTHM yields <strong>of</strong> raw waters<br />

Br-:DOC<br />

(mg/mg)<br />

Specific Yield (\ig TTHM/mg DOC)<br />

Unfractionated<br />

Nonhumic<br />

Humic<br />

Br-:DOC<br />

(mg/mg)<br />

Specific Yield (ug TTHM/mg DOC)<br />

Unfractionated<br />


50 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> GA C <strong>Adsorption</strong><br />

simpler for FGW, indicating the greater age <strong>of</strong> the FGW humic material compared<br />

tothat<strong>of</strong>LGW.<br />

Seasonal Variations<br />

The impact <strong>of</strong> seasonal variation on the unfractionated DOC and the humic -<br />

nonhumic distribution forORW is illustrated in Figure 5.1. The DOC varied from<br />

1.7 to 3.3 mg/L. As with the data in Table 5.1, the raw water NOM was more humic<br />

than nonhumic in nature for all four samples. The relative distributions <strong>of</strong> DOC did<br />

not change significantly at different times in the year. The humic fraction compo<br />

sition varied from 52 to 63 percent. Figure 5.2 shows the specific TTHM yields for<br />

the fractions and the unfractionated sample. The specific TTHM yields <strong>of</strong> the humic<br />

fraction were slightly higher than those <strong>of</strong> the nonhumic fraction in two <strong>of</strong> the four<br />

seasons.<br />

The humic-nonhumic DOC distribution as a function <strong>of</strong> seasonal variability<br />

is shown in Figure 5.3 for PRW. The raw water DOC varied from 2.8 to 5.4 mg/L.<br />

As with other raw waters, the NOM was more humic in nature, which contributed<br />

to the formation <strong>of</strong> more DBFs in all four samples. The humic fraction composition<br />

varied from 53 to 59 percent <strong>of</strong> the total DOC. The specific TTHM yields for the<br />

fractions are shown in Figure 5.4. The humic fraction showed higher TTHM yields<br />

in three <strong>of</strong> the four seasons. The seasonal BDOC values <strong>of</strong> both ORW and PRW are<br />

presented in Table 5.4 and show that the average BDOC was about 18 percent <strong>of</strong> the<br />

total DOC, with no apparent seasonal pattern.<br />

As shown in Table 5.5, the relative distribution <strong>of</strong> NOM fragments varied<br />

significantly among the sampling dates. More variations were observed for ORW<br />

than for PRW samples. Benzonitrile and benzoic acid, markers <strong>of</strong> industrial or<br />

wastewater influence, were found in all ORW samples. Also, the ORW chromatographs<br />

were relatively simpler, with fewer peaks—another indication <strong>of</strong> a lack <strong>of</strong><br />

natural biological activity. PRW samples, although relatively more stable, showed<br />

a strong presence <strong>of</strong> halogen substituted fragments—another indication <strong>of</strong> indus<br />

trial influence.<br />

Table 5.4 Summary <strong>of</strong> seasonal water quality monitoring BDOC results_________<br />

ORW<br />

Raw Raw Raw Raw<br />

sample 1 sample 2 sample 3 sample 4<br />

Sample date 11/4/93 7/29/94 3/24/95 8/2/95<br />

BDOC (mg/L) 0.5 0.4 0.2 0.5<br />

BDOC (% <strong>of</strong> DOC) 23 23 9 17<br />

PRW<br />

Sample date 11/16/94 2/16/95 5/16/95 8/2/95<br />

BDOC (mg/L) 1.0 0.6 0.4 1.0<br />

BDOC (% Of DOC) 22 17 15 18


NOM Characterization 51<br />

Table 5.5 PY-GC-MS classification <strong>of</strong> seasonal samples<br />

Percent based upon peak height<br />

<strong>Water</strong><br />

Sample<br />

date<br />

DOC<br />

(mg/L)<br />

Aliphatic<br />

Aromatic<br />

Nitrogen<br />

containing<br />

Halogen<br />

substituted<br />

Unknown<br />

ORW<br />

1 1/4/93<br />

7/29/94<br />

3/24/95<br />

8/2/95<br />

2.2<br />

1.7<br />

2.2<br />

2.9<br />

34.5<br />

24.3 .<br />

24.6<br />

5.3<br />

15.8<br />

40.3<br />

51.5<br />

67.6<br />

35.9<br />

19.5<br />

9.0<br />

7.1<br />

13.9<br />

4.0<br />

6.0<br />

17.3<br />

BDL<br />

12.0<br />

8.9<br />

2.6<br />

PRW<br />

11/16/94<br />

2/16/95<br />

5/16/95<br />

8/2/95<br />

4.5<br />

3.5<br />

2.7<br />

5.5<br />

23.0<br />

24.3<br />

39.0<br />

34.3<br />

3.9<br />

4.0<br />

6.4<br />

16.6<br />

1.2<br />

6.6<br />

16.5<br />

12.8<br />

67.0<br />

61.5<br />

33.1<br />

31.2<br />

4.3<br />

3.7<br />

4.9<br />

5.0<br />

BDL = below detection limit<br />

Field-Scale NOM Characterization<br />

Approach<br />

<strong>GAC</strong> influent and effluent were sampled from pilot- or full-scale <strong>GAC</strong><br />

contactors for ORW, LOW, MRW, and PRW. NOM was characterized at each site<br />

for humic and nonhumic fractions and biodegradability.<br />

Results and Discussion<br />

Humic-Nonhumic Fractionation<br />

The <strong>GAC</strong> influent concentrations and the humic-nonhumic fractionation<br />

results for the pilot or full scale are shown in Table 5.6. In general, the <strong>GAC</strong> influent<br />

was more nonhumic than humic in nature. When these values are compared to the<br />

raw water results in Table 5.1 and Figures 5.1 and 5.3, it is evident that the<br />

pretreatment process at all four locations preferentially removed the humic fraction.<br />

This shift can be attributed to removal <strong>by</strong> coagulation for all waters except LGW,<br />

where the decrease in the humic fraction was most likely due to ozonation. The<br />

results presented in Table 5.6 also illustrate the variability found in field-scale <strong>GAC</strong><br />

systems. The <strong>GAC</strong> influent DOC varied from water to water. For ORW and PRW,<br />

DOC varied over the run time <strong>of</strong> the pilot- or full-scale <strong>GAC</strong> contactor. The <strong>GAC</strong><br />

influent DOC for ORW varied from 1.3 to 2.6 mg/L, and the nonhumic fraction<br />

ranged from 56 to 67 percent <strong>of</strong> the total DOC. The <strong>GAC</strong> influent for LGW remained<br />

very constant over the course <strong>of</strong> the run, as did the humic and nonhumic fractions.<br />

For MRW, the <strong>GAC</strong> influent was sampled only once and yielded a 60 percent<br />

nonhumic fraction (Figure 3.11 [p. 31] showed that the influent concentration did<br />

not vary much over the course <strong>of</strong> the run). PRW exhibited significant variability in<br />

DOC concentration in the <strong>GAC</strong> influent, from 1.8 to 4.3 mg/L, although the<br />

percentage <strong>of</strong> DOC that was nonhumic in nature varied from only 51 to 59 percent.


52 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

Table 5.6 Summary <strong>of</strong> field-scale <strong>GAC</strong> influent humic-nonhumic fractionation results<br />

for ORW, LGW, MRW, and PRW<br />

Nonhumic fraction<br />

Humic fraction<br />

<strong>Water</strong><br />

Sample<br />

date<br />

Day <strong>of</strong><br />

run<br />

DOC<br />

(mg/L)<br />

DOC<br />

(mg/L)<br />

% <strong>of</strong> total<br />

DOC<br />

DOC<br />

(mg/L)<br />

% <strong>of</strong> total<br />

DOC<br />

ORW<br />

9/3/93<br />

10/6/93<br />

64<br />

97<br />

2.6<br />

2.3<br />

1.4<br />

1.4<br />

56<br />

64<br />

1.2<br />

0.9<br />

44<br />

36<br />

1(3/29/93<br />

120<br />

2.3<br />

1.6<br />

67<br />

0.7<br />

33<br />

1/27/94<br />

Average<br />

210<br />

1.3<br />

0.9<br />

66<br />

63<br />

0.4<br />

36<br />

37<br />

LGW<br />

4/15/94<br />

4/1 9/94<br />

29<br />

43<br />

1.5<br />

1.5<br />

0.8<br />

0.8<br />

53<br />

53<br />

0.7<br />

0.7<br />

47<br />

47<br />

5/3/94<br />

57<br />

1.5<br />

0.8<br />

53<br />

0.7<br />

47<br />

5/17/94<br />

71<br />

1.5<br />

0.8<br />

53<br />

0.7<br />

47<br />

6/27/94<br />

113<br />

1.4<br />

0.8<br />

57<br />

0.6<br />

43<br />

Average<br />

54<br />

46<br />

MRW<br />

12/10/93<br />

49<br />

3.0<br />

1.8<br />

60<br />

1.2<br />

40<br />

PRW<br />

11/16/94<br />

1<br />

3.5<br />

2.0<br />

57<br />

1.5<br />

43<br />

1/21/95<br />

66<br />

1.8<br />

1.1<br />

59<br />

0.7<br />

41<br />

1/31/95<br />

76<br />

2.0<br />

1.0<br />

51<br />

1.0<br />

49<br />

2/27/95<br />

103<br />

2.5<br />

1.5<br />

59<br />

1.0<br />

41<br />

4/26/95<br />

162<br />

3.0<br />

1.5<br />

51<br />

1.5<br />

49<br />

5/24/95<br />

190<br />

4.3<br />

2.5<br />

59<br />

1.8<br />

41<br />

8/2/95<br />

Average<br />

239<br />

3.1<br />

1.8<br />

57<br />

56<br />

1.3<br />

43<br />

44<br />

The effect <strong>of</strong> pretreatment and <strong>GAC</strong> treatment on the NOM was very similar<br />

for all four waters. The humic-nonhumic fractionation results for the raw water,<br />

<strong>GAC</strong> influent, and effluents from the pilot- and full-scale contactors for ORW,<br />

LGW, MRW, and PRW are shown in Figures 5.5,5.6,5.7, and 5.8, respectively. The<br />

bar charts represent the raw water and <strong>GAC</strong> influent. The difference in the results<br />

between these two sets <strong>of</strong> bars represents the impact <strong>of</strong> full-scale treatment systems<br />

currently in place at the four sites. The effect <strong>of</strong> pretreatment on the DOC and the<br />

humic-nonhumic distribution is shown on these figures. DOC removal <strong>by</strong> the<br />

various pretreatment processes ranged from 15 to 45 percent and resulted in a major<br />

shift from humic to nonhumic material. As evidenced <strong>by</strong> the first <strong>GAC</strong> effluent<br />

samples, the nonadsorbable fractions for ORW, LGW, MRW, and PRW were<br />

composed nearly entirely <strong>of</strong> the nonhumic fraction; the humic fraction was virtually<br />

completely removed. With time, the nonhumic fraction broke through rapidly,<br />

though the humic fraction continued to be better removed. The nonhumic fraction<br />

breakthrough reached a plateau level after 40,60,30, and 160 days for ORW, LGW,


NOM Characterization 53<br />

MRW, and PRW, respectively. Further increases in effluent DOC concentrations<br />

were a result <strong>of</strong> the humic fraction breaking through. The humic DOC fraction was<br />

still almost totally removed (below 0.1 mg/L) as <strong>of</strong> 60,50,4, and 80 days for ORW,<br />

LOW, MRW, and PRW, respectively. It is interesting to note that <strong>GAC</strong> treatment<br />

was in a filter adsorber mode for MRW, as compared to postfilter contactors for<br />

ORW, LOW, and PRW. The associated lower EBCT in the MRW filter adsorber<br />

may be related to MRW's significantly shorter period before humic breakthrough<br />

occurred compared to the other <strong>GAC</strong> applications.<br />

Biodegradability<br />

NOM was also characterized in terms <strong>of</strong> biodegradability (as shown <strong>by</strong><br />

BDOC levels) for all four waters. Figure 5.9 shows the effect <strong>of</strong> <strong>GAC</strong> on BDOC for<br />

MRW. The BDOC <strong>of</strong> the <strong>GAC</strong> influent varied from 0.5 to 1.3 mg/L, while the<br />

influent DOC varied from 2.5 to 4.5 mg/L for this preozonated water. BDOC was<br />

initially well removed <strong>by</strong> <strong>GAC</strong> and began to break through after 14 days <strong>of</strong><br />

operation. The measurement <strong>of</strong> relatively early BDOC breakthrough for MRW may<br />

again be attributed to the higher influent BDOC or the lower EBCT in the filter<br />

adsorber configuration, as well as to the fact that the water was ozonated prior to the<br />

<strong>GAC</strong> contractor. It was not possible to characterize the breakthrough behavior <strong>of</strong><br />

BDOC for ORW, LGW, and PRW. The BDOC values for ORW and PRW are shown<br />

in Table 5.7. For ORW the BDOC <strong>of</strong> the influent varied from 0.1 to 0.8 mg/L;<br />

however, it was <strong>of</strong>ten below the method detection limit <strong>of</strong> 0.1 mg/L. The effluent<br />

BDOC was below the detection limit for all cases. Similar results were observed for<br />

PRW. The influent BDOC varied from 0.4 to 0.6 mg/L, and the effluent BDOC was<br />

below the detection limit for six samples. For one sample, an effluent BDOC <strong>of</strong> 0.3<br />

mg/L was detected. No BDOC was detected for LGW in either the influent or the<br />

<strong>GAC</strong> effluent.<br />

Table 5.7 Summary <strong>of</strong> ORW and PRW field-scale BDOC results<br />

<strong>Water</strong><br />

Sample day<br />

Day <strong>of</strong> run<br />

<strong>GAC</strong> influent BDOC<br />

(mg/L)<br />

<strong>GAC</strong> effluent BDOC<br />

(mg/L)<br />

ORW<br />

8/9/93<br />

9/3/93<br />

10/6/93<br />

10/29/93<br />

1/27/94<br />

39<br />

64<br />

97<br />

120<br />

210<br />

0.80<br />

0.15<br />

BDL<br />

BDL<br />

0.12<br />

BDL<br />

BDL<br />

BDL<br />

BDL<br />

BDL<br />

PRW<br />

11/16/94<br />

1<br />

0.6<br />

BDL<br />

1/21/95<br />

66<br />

BDL<br />

BDL<br />

1/31/95<br />

76<br />

BDL<br />

BDL<br />

2/27/95<br />

103<br />

BDL<br />

BDL<br />

4/26/95<br />

162<br />

BDL<br />

BDL<br />

5/24/94<br />

8/2/95<br />

190<br />

239<br />

0.4<br />

0.4<br />

BDL<br />

0.3<br />

BDL = below detection limit <strong>of</strong> 0.1 mg/L


54 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

Bench-Scale NOM Characterization<br />

Approach<br />

NOM was characterized at bench scale with respect to humic-nonhumic<br />

fractionation, MS fractionation, biodegradability, and PY-GC-MS fractionation for<br />

ORW, LOW, MRW, PRW, SRPW, and FGW at four defined points on the<br />

breakthrough curve, as illustrated in Figure 5.10:<br />

1. <strong>GAC</strong> influent<br />

2. Point A, the nonadsorbable fraction<br />

3. Point B, intermediate (20-35 percent) breakthrough<br />

4. Point C, advanced (>50 percent) breakthrough<br />

These four points were chosen to represent critical points <strong>of</strong> a breakthrough curve.<br />

Comparison <strong>of</strong> the NOM compositions at these different points facilitates the<br />

assessment <strong>of</strong> NOM adsorbability.The raw water and <strong>GAC</strong> influent samples were<br />

taken from the field-scale plant at the same time. The DOC and the specific TTHM<br />

yields <strong>of</strong> the fractions were determined according to the methodology discussed in<br />

Chapter 2.<br />

Results and Discussion<br />

Humic-Nonhumic Fractionation<br />

Figure 5.11 shows the DOC <strong>of</strong> the unfractionated sample and the nonhumic<br />

and humic fractions <strong>of</strong> ORW. The bars represent the raw water and the <strong>GAC</strong><br />

influent, and the line graph indicates the breakthrough <strong>of</strong> the fractions plotted versus<br />

scaled operation time. As shown in the previous section <strong>of</strong> this chapter, conven<br />

tional pretreatment predominantly removed the humic fraction from the raw ORW,<br />

yielding a <strong>GAC</strong> influent that was 52 percent nonhumic and 48 percent humic in<br />

nature. Initially, the humic fraction was completely removed <strong>by</strong> <strong>GAC</strong> and the<br />

nonadsorbable fraction (point A) was entirely nonhumic. The nonhumic fraction<br />

broke through rapidly and completely, while the humic fraction was better removed<br />

and exhibited slower breakthrough. At point C, which characterizes the later portion<br />

<strong>of</strong> the breakthrough curve, the humic fraction was still being removed <strong>by</strong> more than<br />

50 percent.<br />

The specific TTHM yields <strong>of</strong> the unfractionated sample and the nonhumic<br />

and humic fractions for the raw water, <strong>GAC</strong> influent, and the three effluent samples<br />

for ORW are shown in Figure 5.1 2. The fractions and the unfractionated portion <strong>of</strong><br />

a given sample were chlorinated at constant Br~:DOC ratio to allow for comparison<br />

among them. However, the five separate samples cannot be compared to each other<br />

since the Br~:DOC ratio shifts, as indicated on the graph. The specific TTHM yields<br />

for the nonhumic and humic fractions were similar to each other and to the<br />

unfractionated sample for the raw water and at the <strong>GAC</strong> influent. It is difficult to<br />

compare the three effluent samples because the Br~ :DOC ratios were different. The<br />

specific TTHM yields for raw ORW were slightly higher than those for the <strong>GAC</strong><br />

influent; these values can be compared since the chlorination conditions (i.e.,<br />

Br:DOC ratios) were similar.


NOM Characterization 55<br />

Figures 5.13 and 5.14 show the DOC fractionation and specific TTHM<br />

yields for humic-nonhumic fractions <strong>of</strong> LGW. Similar graphs for MRW, PRW,<br />

SRPW, and FGW are shown in Figures 5.15 through 5.22.<br />

The raw water DOC for all surface water cases was more humic than<br />

nonhumic in nature, ranging from 52 percent humic for SRPW to 63 percent for<br />

MRW. The raw water DOC from FGW (a groundwater) had a much higher humic<br />

content, 75 percent. In general, the pretreatment process (summarized in Table 2.2,<br />

[p. 7] for each water) resulted in a reduction <strong>of</strong> DOC, which was mainly due to a<br />

removal or alteration (in the case <strong>of</strong> ozonation) <strong>of</strong> the hurnic fraction. In the case <strong>of</strong><br />

FGW, the removal was comparable for both fractions. This is summarized in Table<br />

5.8. For all waters, the nonadsorbable fraction was nearly entirely nonhumic in<br />

nature, indicating that the humic fraction was very well removed <strong>by</strong> <strong>GAC</strong>. With time<br />

the nonhumic fraction broke through rapidly, while the humic fraction broke<br />

through more slowly. Even for the last effluent sample, the effluent NOM was<br />

predominantly nonhumic in nature. These results are consistent with the field-scale<br />

behavior shown in the previous section (Figures 5.5 to 5.8).<br />

In terms <strong>of</strong> the specific TTHM yields (Figures 5.12 to 5.22), the humic<br />

fraction <strong>of</strong> five <strong>of</strong> the six waters had a slightly higher specific TTHM yield than did<br />

the nonhumic fraction for the raw and <strong>GAC</strong> influent waters. There were no apparent<br />

differences in <strong>DBP</strong> formation between the humic and nonhumic fractions in the<br />

<strong>GAC</strong> effluent at the advanced breakthrough points, although the results were<br />

confounded <strong>by</strong> the differing Br:DOC ratios. At the initial and intermediate<br />

breakthrough points, the humic fraction was too small to determine specific TTHM<br />

yields.<br />

Molecular Size Distribution<br />

Figure 5.23 shows the ORW DOC distribution <strong>of</strong> the 3,000 MS fractions for raw water, <strong>GAC</strong> influent, and three effluent<br />

samples. The dominant fraction in the raw water was the intermediate MS fraction,<br />

with the small and large MS fractions contributing approximately equal amounts <strong>of</strong><br />

DOC. After <strong>GAC</strong> treatment, the nonadsorbable fraction was composed entirely <strong>of</strong><br />


56 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

the latter portion <strong>of</strong> the breakthrough curve, the 500-3,000 MS fraction was still<br />

being removed <strong>by</strong> more than 50 percent.<br />

After conventional pretreatment, the >3,000 MS fraction was removed (or<br />

altered) appreciably in most cases and the other fractions were not significantly<br />

impacted. For all waters, the MS distribution in the <strong>GAC</strong> influent is summarized in<br />

Table 5.8.<br />

For ORW, the specific TTHM yields <strong>of</strong> the unfractionated sample and the<br />


NOM Characterization 57<br />

in the <strong>GAC</strong> influent. The fraction was initially well removed and rapidly broke<br />

through to the same concentration <strong>of</strong> that in the <strong>GAC</strong> influent.<br />

For LGW, MRW, PRW, and SRPW the <strong>GAC</strong> effluent specific TTHM<br />

yields <strong>of</strong> the


58 <strong>Removal</strong> <strong>of</strong><strong>DBP</strong> <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

Table 5.10 Summary <strong>of</strong> BDOC results for PRW at bench scale<br />

Sample Scaled operation time (days) BDOC (mg/L)<br />

<strong>GAC</strong> influent<br />

Nonadsorbable<br />

Intermediate breakthrough<br />

Advanced breakthrough<br />

NA<br />

20<br />

76<br />

190<br />

0.4<br />

BDL<br />

BDL<br />

0.3<br />

BDL = below detection limit <strong>of</strong> 0.1 mg/L<br />

NA = not applicable<br />

Table 5.11 PY-GC-MS classification <strong>of</strong> Ohio River water<br />

Percent based upon peak height<br />

Stage<br />

Aliphatic<br />

Aromatic<br />

Nitrogen<br />

containing<br />

Halogen<br />

substituted<br />

Unknown<br />

Raw<br />

<strong>GAC</strong> influent<br />

Nonadsorbable<br />

Intermediate breakthrough<br />

Advanced breakthrough<br />

34.5<br />

30.2<br />

76.5<br />

52.5<br />

47.8<br />

15.8<br />

1.6<br />

BDL<br />

1.0<br />

BDL<br />

35.9<br />

4.8<br />

3.5<br />

9.0<br />

18.3<br />

13.9<br />

52.0<br />

15.5<br />

36.6<br />

33.9<br />

BDL<br />

11.5<br />

4.4<br />

0.8<br />

BDL<br />

BDL = below detection limit<br />

Table 5.12 PY-GC-MS classification <strong>of</strong> Lake Gaillard water<br />

Percent based upon peak height<br />

Stage<br />

Aliphatic<br />

Aromatic<br />

Nitrogen<br />

containing<br />

Halogen<br />

substituted<br />

Unknown<br />

Raw<br />

<strong>GAC</strong> influent<br />

Nonadsorbabie<br />

Intermediate breakthrough<br />

Advanced breakthrough<br />

16.3<br />

12.5<br />

3.0<br />

2.1<br />

11.3<br />

70.0<br />

77.7<br />

85.4<br />

91.2<br />

73.2<br />

7.7<br />

2.0<br />

5.3<br />

0.5<br />

5.3<br />

BDL<br />

BDL<br />

6.3<br />

1.9<br />

1.6<br />

5.8<br />

NA<br />

NA<br />

4.2<br />

8.4<br />

BDL = below detection level<br />

NA = not applicable<br />

remained the dominant fragment at the intermediate and advanced breakthrough<br />

samples as well. In the first two breakthrough samples, the aromatic character was<br />

in the range <strong>of</strong> 85-90 percent <strong>of</strong> the total peak height. Typically, very simple<br />

fragmentation patterns were obtained at these early breakthrough points. In the final<br />

breakthrough sample, the fingerprint, while still predominantly aromatic (73<br />

percent), began to show the appearance <strong>of</strong> other fragments (i.e., acetic acid,<br />

propanoic acid, methyl furfural, and so on). A rather distinct feature <strong>of</strong> this water


NOM Characterization 59<br />

was that typical aliphatic, possibly polysaccharide-derived fragments appeared to<br />

be correlated with humic material.<br />

Based upon the results shown in Table 5.13, it is evident that the relative<br />

aliphatic nature <strong>of</strong> MRW's organic matrix remained unchanged through pretreatment<br />

and <strong>GAC</strong> adsorption. The aromatic fraction and nitrogen-containing frag<br />

ments were, however, preferentially removed <strong>by</strong> <strong>GAC</strong> adsorption. The occurrence<br />

<strong>of</strong> halogenated fragments increased after <strong>GAC</strong> adsorption. This could also be<br />

attributed to the removal <strong>of</strong> other fragments, which made the chromatograms more<br />

susceptible to detection <strong>of</strong> smaller fragments that were earlier masked <strong>by</strong> the large<br />

presence <strong>of</strong> other peaks. Inspection <strong>of</strong> the PY-GC-MS fingerprints illustrates that<br />

acetic acid was the major peak at each breakthrough point and that other major peaks<br />

were formic acid, benzaldehyde, and benzonitrile or benzoic acid (suspected<br />

markers <strong>of</strong> industrial discharge). It is suspected that one source <strong>of</strong> the halosubstituted<br />

fragments was chlorinated wastewater discharge; a strong presence <strong>of</strong><br />

such fragments may be observed at the breakthrough stages because the parent<br />

halogenated organic material has higher molecular weight and is not well adsorbed<br />

to <strong>GAC</strong> because <strong>of</strong> micropore exclusion.<br />

The general chemical nature <strong>of</strong> PRW was not significantly altered <strong>by</strong><br />

pretreatment prior to <strong>GAC</strong> (Table 5.14). The predominant identified pyrolysis<br />

fragments after <strong>GAC</strong> adsorption were acetic acid and only various halogenated<br />

fragments. The chemical signature <strong>of</strong> NOM at the second <strong>GAC</strong> breakthrough point<br />

was predominantly aliphatic and dominated <strong>by</strong> a strong acetic acid peak. Although<br />

Table 5.13 PY-GC-MS classification <strong>of</strong> Mississippi River water<br />

Percent based upon peak height<br />

Stage Aliphatic Aromatic<br />

Nitrogen<br />

containing<br />

Halogen<br />

substituted Unknown<br />

Raw<br />

<strong>GAC</strong> influent<br />

Nonadsorbable<br />

Intermediate breakthrough<br />

Advanced breakthrough<br />

37.7<br />

38.0<br />

39.0<br />

28.0<br />

33.0<br />

13.4<br />

12.6<br />

7.0<br />

3.03<br />

5.7<br />

40.7<br />

42.5<br />

6.3<br />

16.0<br />

21.0<br />

—<br />

—<br />

35.6<br />

18.8<br />

30.6<br />

8.3<br />

7<br />

12<br />

4<br />

9<br />

Table 5.14 PY-GC-MS classification <strong>of</strong> Passaic River water<br />

Percent based upon peak height<br />

Stage Aliphatic Aromatic<br />

Nitrogen<br />

containing<br />

Halogen<br />

substituted Unknown<br />

Raw 23.0<br />

<strong>GAC</strong> influent 37.4<br />

Nonadsorbable 10.4<br />

Intermediate breakthrough 54.8<br />

Advanced breakthrough 17.2<br />

3.9<br />

5.5<br />

12.0<br />

4.0<br />

1.6<br />

1.6<br />

5.5<br />

67.0<br />

52.2<br />

13.3<br />

43.7<br />

58.4<br />

4.3<br />

6.4<br />

74.7<br />

13.4


60 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

much weaker relative to acetic acid, the other three predominant peaks (in descend<br />

ing order) were trichlorobenzene, tetrachlorobenzene, and formic acid. At complete<br />

breakthrough, the chemical character <strong>of</strong> the organic matrix was very similar to that<br />

<strong>of</strong> the raw and <strong>GAC</strong> influent water. The halo-substituted signature was dominant.<br />

The strongest peak was for dichloroacetonitrile; the other predominant peaks (all<br />

roughly comparable) were for acetic acid, trichlorobenzene, tetrachlorobenzene,<br />

and pentachlorobenzene.<br />

The pyrolysis fingerprint <strong>of</strong> <strong>GAC</strong> influent water for SRPW (Table 5.15)<br />

showed some change in the characteristics <strong>of</strong> the aliphatic signature. While the<br />

chemical nature <strong>of</strong> this water remained predominantly aliphatic, this signature was<br />

simplified and composed primarily <strong>of</strong> a strong acetic acid peak and a somewhat<br />

smaller peak <strong>of</strong> propanoic acid. Whereas in the raw water strong peaks <strong>of</strong> substituted<br />

furfurals and furans were detected, these peaks were diminished relative to acetic<br />

acid. The slight increase in percent aromaticity appeared to be related to a lack <strong>of</strong><br />

removal <strong>of</strong> aromatic parent compounds compared to aliphatic material. There were<br />

no halogenated fragments detected in the <strong>GAC</strong> effluent. The nonadsorbable NOM<br />

could not be characterized <strong>by</strong> PY-GC-MS. The only peaks detected in the analysis<br />

<strong>of</strong> the second <strong>GAC</strong> breakthrough point were aliphatic fragments. Among these,<br />

acetic acid was the major peak, and a variety <strong>of</strong> carboxylic acid fragments were<br />

detected as minorpeaks. The chemical signature at complete breakthrough was very<br />

similar to the <strong>GAC</strong> influent, with a major acetic acid peak, a strong peak <strong>of</strong> pro<br />

panoic acid, and smaller furfural and formic acid fragments.<br />

The chemical classification <strong>of</strong> the PY-GC-MS data for FGW was essen<br />

tially unaltered <strong>by</strong> treatment, as shown in Table 5.16. Close to 200 fragments<br />

consisting primarily <strong>of</strong> substituted benzenes, naphthalenes, and phenols were<br />

identified in the treated water. In addition to a strong peak <strong>of</strong> phenol, major peaks<br />

<strong>of</strong>/?- and m-cresol, C2-phenol, and methyl naphthalene were observed. The pattern<br />

differences between the raw and treated water suggest that the organic material<br />

removed in coagulation and s<strong>of</strong>tening was resolved into a single, strong phenol<br />

fragment <strong>by</strong> pyrolysis, and with its removal more <strong>of</strong> the details <strong>of</strong> the matrix became<br />

visible. The chemical signature <strong>of</strong> the nonadsorbable NOM was dramatically<br />

different than either the raw or conventionally treated waters. As shown in Table<br />

5.16, the organic quality <strong>of</strong> the nonadsorbable fraction was predominantly aliphatic<br />

and was characterized <strong>by</strong> a very strong peak <strong>of</strong> acetic acid. The second <strong>GAC</strong><br />

breakthrough sample was also predominantly aliphatic, and its pyrolysis finger<br />

print was composed primarily <strong>of</strong> acetic acid and formic acid. At complete break<br />

through, the pyrolysis fingerprint was very similar to that <strong>of</strong> the influent.<br />

Summary________________________<br />

Samples <strong>of</strong> raw ORW, MRW, PRW, and SRPW showed very similar NOM<br />

fractionation results. These four waters were surface water. With respect to these<br />

four waters, LGW showed a similar humic and nonhumic DOC distribution, but the<br />

MS distribution showed much larger intermediate MS fractions. FGW showed the<br />

largest percentage <strong>of</strong> DOC in the intermediate MS range, and 75 percent <strong>of</strong> the<br />

DOC was found in the humic fraction. In general, specific TTHM yields <strong>of</strong> the raw<br />

water humic fractions and the


NOM Characterization 61<br />

Table 5.15 PY-GC-MS classification <strong>of</strong> Salt River Project water<br />

Stage<br />

Nitrogen Halogen<br />

Aliphatic Aromatic containing substituted<br />

Unknown<br />

Raw<br />

<strong>GAC</strong> influent<br />

Nonadsorbable<br />

Intermediate breakthrough<br />

Advanced breakthrough<br />

57.9<br />

68.6<br />

—<br />

100<br />

93.4<br />

9.1<br />

18.0<br />

—<br />

—<br />

0.6<br />

9.2<br />

5.1<br />

—<br />

—<br />

2.2<br />

13.2<br />

1.1<br />

—<br />

—<br />

—<br />

10.5<br />

7.2<br />

—<br />

—<br />

3.9<br />

Table 5.16 PY-GC-MS classification <strong>of</strong> Florida groundwater<br />

Stage<br />

Aliphatic<br />

Aromatic<br />

Nitrogen<br />

containing<br />

Halogen<br />

substituted<br />

Unknown<br />

Raw<br />

<strong>GAC</strong> influent<br />

Nonadsorbable<br />

Intermediate breakthrough<br />

Advanced breakthrough<br />

1.4<br />

1.7<br />

80.0<br />

64.4<br />

1.5<br />

97.0<br />

93.3<br />

—<br />

17.9<br />

93.5<br />

—<br />

1.7<br />

—<br />

17.7<br />

1.4<br />

—<br />

2.4<br />

20.0<br />

—<br />

0.2<br />

1.6<br />

0.9<br />

—<br />

—<br />

3.4<br />

For the field-scale analyses, the <strong>GAC</strong> influent (after pretreatment) was<br />

slightly more nonhumic than humic in nature, and the largest percentage <strong>of</strong> the DOC<br />

was found in the intermediate MS range. The influent DOC concentrations in two<br />

<strong>of</strong> the four waters tested in the field scale changed during the <strong>GAC</strong> runs; however,<br />

the NOM composition remained more nonhumic than humic. The humic fraction<br />

was initially well removed <strong>by</strong> <strong>GAC</strong> for all four waters (ORW, LGW, MRW, and<br />

PRW), and the nonadsorbable fraction was nonhumic in nature. The nonhumic<br />

fraction broke through rapidly, while the humic fraction exhibited slower break<br />

through behavior. The intermediate MS range was best removed <strong>by</strong> <strong>GAC</strong>, while the<br />


62 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

Table 5.17 Average DOC distribution <strong>of</strong> river waters with breakthrough: Nonhumic and<br />

humic fractions<br />

Nonhumic DOC (%)<br />

Humic DOC (%)<br />

Average<br />

Standard deviation<br />

Average<br />

Standard deviation<br />

<strong>GAC</strong> influent<br />

Nonadsorbable<br />

Intermediate breakthrough<br />

Advanced breakthrough<br />

55<br />

96<br />

87<br />

74<br />

4.8<br />

8.0<br />

15<br />

4<br />

45<br />

4<br />

13<br />

26<br />

4.8<br />

8.0<br />

15<br />

4<br />

Table 5.18 Average DOC distribution <strong>of</strong> river<br />

fractions<br />

waters with breakthrough: Molecular size<br />


NOM Characterization 63<br />

3.5<br />

3.0 -<br />

Unfractionated<br />

Nonhumic fraction<br />

Humic fraction<br />

c 2.0<br />

.o<br />

i= 1.5 -\<br />

o><br />

o<br />

3 1.0 -\<br />

o<br />

Q 0.5 -\<br />

0.0<br />

11/93 7/94 3/95 8/95<br />

Figure 5.1 Seasonal variation <strong>of</strong> nonhumic and humic fractions for Ohio River raw<br />

water<br />

Unfractionated<br />

Nonhumic fraction<br />

Humic fraction<br />

11/93 7/94 3/95 8/95<br />

Figure 5.2 Specific TTHM yields <strong>of</strong> nonhumic and humic fractions for Ohio River<br />

seasonal raw water samples


64 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

Unfractionated<br />

__ Nonhumic fraction<br />

I I Humic fraction<br />

11/94 2/95 5/95 8/95<br />

Figure 5.3 Seasonal variation <strong>of</strong> nonhumic and humic fractions for Passaic River raw<br />

water<br />

Unfractionated<br />

Nonhumic fraction<br />

Humic fraction<br />

11/94 2/95 5/95 8/95<br />

Figure 5.4 Specific TTHM yields <strong>of</strong> nonhumic and humic fractions for Passaic River<br />

seasonal raw water samples


NOM Characterization 65<br />

0.0<br />

Average Average 0 50 100 150 200 250<br />

raw <strong>GAC</strong><br />

influent Operation time (days)<br />

Figure 5.5 Full-scale nonhumic and humic fraction breakthrough for Ohio River<br />

water<br />

0.0<br />

Average Average<br />

raw <strong>GAC</strong><br />

influent<br />

Operation time (days)<br />

150<br />

Figure 5.6 Full-scale nonhumic and humic fraction breakthrough for Lake Gaillard<br />

water


66 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

Average Average<br />

raw <strong>GAC</strong><br />

influent<br />

10 20 30 40<br />

Operation time (days)<br />

50<br />

Figure 5.7 Full-scale nonhumic and humic fraction breakthrough for Mississippi River<br />

water<br />

Average Average<br />

raw <strong>GAC</strong><br />

influent<br />

50 100 150 200 250<br />

Operation time (days)<br />

Figure 5.8 Pilot-scale nonhumic and humic fraction breakthrough for Passaic River<br />

water


NOM Characterization 67<br />

• <strong>GAC</strong> Influent<br />

0 <strong>GAC</strong> Effluent<br />

o<br />

50%) breakthrough<br />

B. Intermediate (20-36%) breakthrough<br />

A. Nonadsorbable fraction<br />

40 80<br />

Scaled operation time (days)<br />

Figure 5.10 Example curve showing sampling points for NOM characterization at bench<br />

scale


68 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

2.5<br />

Unfractionated<br />

Nonhumic fraction<br />

Humic fraction<br />

o 1.5 H<br />

I ,.


NOM Characterization 69<br />

Raw <strong>GAC</strong><br />

Influent<br />

50 100 150<br />

Scaled operation time (days)<br />

Note: A, B, and C denote initial, intermediate, and advanced breakthrough points.<br />

Figure 5.13 Bench-scale nonhumic and humic fraction breakthrough for Lake Gaillard<br />

water<br />

Unfractionated<br />

Nonhumic fraction<br />

Humic fraction<br />

Raw <strong>GAC</strong><br />

Influent<br />

Note: A, B, and C denote initial, intermediate, and advanced breakthrough points.<br />

NA = not analyzed<br />

Figure 5.14 Bench-scale specific TTHM yields <strong>of</strong> humic and nonhumic fractions for<br />

Lake Gaillard water


70 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> GA C <strong>Adsorption</strong><br />

5.5<br />

5.0<br />

^ 4.5<br />

a. 4.0<br />

^ 3.5<br />

I 3.0<br />

§ 2 ' 5<br />

§ 2.0<br />

g 1.5<br />

1 • • ' i • • ' ' i • '<br />

Unfractionated<br />

Nonhumic fraction<br />

Humic fraction<br />

0.5<br />

0.0<br />

Raw <strong>GAC</strong><br />

Influent<br />

10 20 30 40 50<br />

Scaled operation time (days)<br />

Note: A, B, and C denote initial, intermediate, and advanced breakthrough points.<br />

Figure 5.15 Bench-scale nonhumic and humic fraction breakthrough for Mississippi<br />

River water<br />

Unfractionated<br />

Nonhumic fraction<br />

Humic fraction<br />

Raw <strong>GAC</strong><br />

Influent<br />

Note: A, B, and C denote initial, intermediate, and advanced breakthrough points.<br />

Figure 5.16 Bench-scale specific TTHM yields <strong>of</strong> humic and nonhumic fractions for<br />

Mississippi River water


NOM Characterization 71<br />

Raw <strong>GAC</strong><br />

Influent<br />

0 50 100 150 200<br />

Scaled operation time (days)<br />

Note: A, B, and C denote initial, intermediate, and advanced breakthrough points.<br />

Figure 5.17 Bench-scale nonhumic and humic fraction breakthrough for Passaic River<br />

water<br />

60<br />

o<br />

8<br />

o><br />

50-<br />

40-<br />

69<br />

211<br />

Unfractionated<br />

Nonhumic fraction<br />

Humic fraction<br />

54<br />

30 -<br />

Br":DOC = 62<br />

(ng/mg)<br />

•I 20<br />

8. 10<br />

V)<br />

Raw <strong>GAC</strong><br />

Influent<br />

Note: A, B, and C denote initial, intermediate, and advanced breakthrough points.<br />

NA = not analyzed<br />

Figure 5.18 Bench-scale specific TTHM yields <strong>of</strong> nonhumic and humic fractions for<br />

Passaic River water


72 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> GA C <strong>Adsorption</strong><br />

Unfractionated<br />

0 Nonhumic fraction<br />

A Humic fraction<br />

Raw <strong>GAC</strong><br />

Influent<br />

0 50 100 150<br />

Scaled operation time (days)<br />

Note: A, B, and C denote initial, intermediate, and advanced breakthrough points.<br />

Figure 5.19 Bench-scale nonhumic and humic fraction breakthrough for Salt River<br />

Project water<br />

Unfractionated<br />

Nonhumic fraction<br />

Humic fraction<br />

Raw <strong>GAC</strong> A<br />

Influent<br />

Note: A, B, and C denote initial, intermediate, and advanced breakthrough points.<br />

NA = not analyzed<br />

Figure 5.20 Bench-scale specific TTHM yields <strong>of</strong> nonhumic and humic fractions for Salt<br />

River Project water


NOM Characterization 73<br />

14<br />

12 -<br />

5<br />

!> 10 -<br />

Unfractionated<br />

° Nonhumic fraction<br />

A Humic fraction<br />

I 6 ~<br />

2 -<br />

0<br />

Raw <strong>GAC</strong><br />

Influent<br />

10 20<br />

Scaled operation time (days)<br />

30<br />

Note: A, B, and C denote initial, intermediate, and advanced breakthrough points.<br />

Figure 5.21 Bench-scale nonhumic and humic fraction breakthrough for Florida<br />

groundwater<br />

60<br />

Unfractionated<br />

Nonhumic fraction<br />

50-<br />

Humic fraction<br />

o><br />

I<br />

§<br />

40 -<br />

30 -<br />

2<br />

•I 20 H<br />

i<br />

8. 10 4<br />

CO<br />

Br":DOC = 21<br />

(ng/mg)<br />

102<br />

131<br />

130<br />

Raw <strong>GAC</strong><br />

Influent<br />

Note: A, B, and C denote initial, intermediate, and advanced breakthrough points.<br />

NA = not analyzed<br />

Figure 5.22 Bench-scale specific TTHM yields <strong>of</strong> nonhumic and humic fractions for<br />

Florida groundwater


74 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

Unfractionated<br />

< 500 MS fraction<br />

500-3000 MS fraction<br />

> 3000 MS fraction<br />

0.0<br />

Raw <strong>GAC</strong><br />

Influent<br />

50 100 150<br />

Scaled operation time (days)<br />

Note: A, B, and C denote initial, intermediate, and advanced breakthrough points.<br />

Figure 5.23 Molecular size fraction breakthrough for Ohio River water<br />

O<br />

g at<br />

X<br />

O)<br />

TJ<br />

0)<br />

&<br />

CO<br />

100 -r<br />

90 -<br />

80 -<br />

70 -<br />

60 -<br />

50 -<br />

40 -<br />

30 -<br />

20 -<br />

10 -<br />

Br":DOC = 257<br />

(H9/mg)<br />

452<br />

945<br />

Unfractionated<br />

< 500 MS fraction<br />

< 3000 MS fraction<br />

585<br />

0<br />

Raw <strong>GAC</strong><br />

Influent<br />

Note: A, B, and C denote initial, intermediate, and advanced breakthrough points.<br />

Figure 5.24 Specific TTHM yields <strong>of</strong> MS fractions for Ohio River water


I<br />

S-t<br />

f 5"<br />

3! s<br />

< C » ?!<br />

? I<br />

2 ® 32.<br />

><br />

3 ><br />

in p ro<br />

en P3<br />

KJ g a><br />

ro o)<br />

05 t a<br />

Specific yield (ug TTHM/mg DOC)<br />

01<br />

g<br />

^<br />

* 2<br />

O_ CD<br />

o<br />

§.<br />


76 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> GA C <strong>Adsorption</strong><br />

Unfractionated<br />

< 500 MS fraction<br />

500-3000 MS fraction<br />

> 3000 MS fraction<br />

Raw <strong>GAC</strong> 0 10 20 30 40<br />

Influent Scaled operation time (days)<br />

50<br />

Note: A, B, and C denote initial, intermediate, and advanced breakthrough points.<br />

Figure 5.27 Molecular size fraction breakthrough for Mississippi River water<br />

80<br />

70 - Br:DOC = 141<br />

(Hg/mg)<br />

Unfractionated<br />

< 500 MS fraction<br />

< 3000 MS fraction<br />

106<br />

Raw <strong>GAC</strong><br />

Influent<br />

B<br />

Note: A, B, and C denote initial, intermediate, and advanced breakthrough points.<br />

Figure 5.28 Specific TTHM yields <strong>of</strong> MS fractions for Mississippi River water


NOM Characterization 77<br />

4.0<br />

3.5 -<br />

§ 3.0 -<br />

Unfractionated<br />

< 500 MS fraction<br />

500-3000 MS fraction<br />

> 3000 MS fraction<br />

c 2.5 H<br />

* 2.0 -<br />

§ 1 -5 '<br />

o<br />

O 1.0 1<br />

0.5 -<br />

0.0<br />

Raw <strong>GAC</strong><br />

Influent<br />

50 100 150<br />

Scaled operation time (days)<br />

200<br />

Note: A, B, and C denote initial, intermediate, and advanced breakthrough points.<br />

Figure 5.29 Molecular size fraction breakthrough for Passaic River water<br />

Unfractionated<br />

< 500 MS fraction<br />

< 3000 MS fraction<br />

Raw <strong>GAC</strong><br />

Influent<br />

Note: A, B, and C denote initial, intermediate, and advanced breakthrough points.<br />

NA = not analyzed<br />

Figure 5.30 Specific TTHM yields <strong>of</strong> MS fractions for Passaic River water


78 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

Raw <strong>GAC</strong> 0 50 100 150<br />

Influent Scaled operation time (days)<br />

Note: A, B, and C denote initial, intermediate, and advanced breakthrough points.<br />

Figure 5.31 Molecular size fraction breakthrough for Salt River Project water<br />

o<br />

80<br />

Unfractionated<br />

< 500 MS fraction<br />

< 3000 MS fraction<br />

BrDOC = 196<br />

(Hg/mg) —i 200<br />

445<br />

Raw<br />

<strong>GAC</strong><br />

Influent<br />

Note: A, B, and C denote initial, intermediate, and advanced breakthrough points.<br />

NA = not analyzed<br />

Figure 5.32 Specific TTHM yields <strong>of</strong> MS fractions for Salt River Project water


NOM Characterization 79<br />

Unfractionated<br />

< 500 MS fraction<br />

500-3000 MS fraction<br />

> 3000 MS fraction<br />

I<br />

8<br />

8<br />

8<br />

o<br />

5 -<br />

Raw GAG<br />

Influent<br />

10 20<br />

Scaled operation time (days)<br />

30<br />

Note: ft,, B, and C denote initial, intermediate, and advanced breakthrough points.<br />

Figure 5.33 Molecular size fraction breakthrough for Florida groundwater<br />

120<br />

8 i°o<br />

Q<br />

Bf:DOC = 240<br />

(ng/mg) -^ 358<br />

Unfractionated<br />

< 500 MS fraction<br />

< 3000 MS fraction<br />

493<br />

9B5<br />

NA<br />

Raw <strong>GAC</strong><br />

Influent<br />

B<br />

Note: A, B, and C denote initial, intermediate, and advanced breakthrough points.<br />

NA = not analyzed<br />

Figure 5.34 Specific TTHM yields <strong>of</strong> MS fractions for Florida groundwater


80 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> GA C <strong>Adsorption</strong><br />

o<br />

a><br />

100<br />

80 -<br />

A unfractionated (XAD-8)<br />

o nonhumic<br />

a humic<br />

60 -<br />

o><br />

40 -<br />

£<br />

(0<br />

20-<br />

0 4<br />

•<br />

0 ° o •<br />

• unfractionated<br />

• < 500 MS<br />

• < 3000 MS<br />

100<br />

200 300<br />

BrDOC<br />

400 500<br />

Figure 5.35 Effect <strong>of</strong> BR~:DOC on DOC fractions for <strong>GAC</strong> influent for six waters


Chapter 6___________<br />

Design and Operational<br />

Considerations<br />

The design, operation, and performance <strong>of</strong> a <strong>GAC</strong> contactor depends on<br />

<strong>GAC</strong> type, the water to be treated, and the mode <strong>of</strong> contactor operation. The<br />

objective <strong>of</strong> process optimization is to meet effluent criteria at the lowest cost. The<br />

most important design variables are the EBCT and bed operation (Sontheimer et al.<br />

1988). Consequently, optimizing <strong>GAC</strong> performance involves maximizing <strong>GAC</strong> bed<br />

lifetime while minimizing the size <strong>of</strong> the contactor.<br />

The EBCT dictates the contactor size and is a function <strong>of</strong> the filter type and<br />

the water to be treated. Filter adsorbers typically have lower EBCTs and require<br />

more frequent backwashing to reduce head loss buildup than do <strong>GAC</strong> contactors<br />

operated in the postfilter adsorber mode. For a given water, the effluent criteria can<br />

be met <strong>by</strong> (1) a short EBCT (low capital cost) and frequent replacement or<br />

regeneration <strong>of</strong> <strong>GAC</strong> (high operations and maintenance [O&M] costs) or (2) a long<br />

EBCT (high capital cost) and a less frequent replacement or regeneration <strong>of</strong> <strong>GAC</strong><br />

(low O&M costs). For utilities where <strong>GAC</strong> is a viable treatment option, more than<br />

one contactor is usually operated. Depending on the treatment goal, the operation<br />

<strong>of</strong> the <strong>GAC</strong> adsorbers can occur either in series or in parallel (Sontheimer et al.<br />

1988). When very high removals are needed (e.g., greater than 90 percent), series<br />

operation is <strong>of</strong>ten best because it maximizes the use <strong>of</strong> the carbon in the first<br />

contactor in the series. An advantage <strong>of</strong> parallel operation is that the effluents <strong>of</strong> the<br />

contactors can be blended, which extends the life <strong>of</strong> a given contactor (Roberts and<br />

Summers 1982).<br />

Evaluation <strong>of</strong> various process configurations can lead to optimum <strong>GAC</strong><br />

performance for a given plant capacity. To determine the optimum process configu<br />

ration, representative process performance data need to be generated. Although<br />

pilot-scale tests can be run, these are costly and time-consuming. A viable alterna<br />

tive is the RSSCT. In Chapter 3, the use <strong>of</strong> the RSSCT and its verification with fulland<br />

pilot-scale data for four waters were presented. The investigation <strong>of</strong> the impact<br />

<strong>of</strong> design and operational considerations using the RSSCT is presented in this<br />

chapter. The effect <strong>of</strong> EBCT on adsorption behavior, the impact <strong>of</strong> blending<br />

effluents from multiple contactors in parallel, and the effect <strong>of</strong> backwashing<br />

on NOM adsorption are discussed. The cost considerations are presented in Chap<br />

ter 12.<br />

81


82 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> GA C <strong>Adsorption</strong><br />

Impact <strong>of</strong> EBCT on <strong>Adsorption</strong> Behavior<br />

The EBCT <strong>of</strong> a <strong>GAC</strong> contactor is a primary design factor; it determines the<br />

bed depth at a given hydraulic loading rate. The optimum selection <strong>of</strong> the EBCT <strong>of</strong><br />

a <strong>GAC</strong> contactor is an important parameter in determining the effectiveness <strong>of</strong> <strong>GAC</strong><br />

adsorption. Hong (1995) used the pore-surface diffusion model (Sontheimer, et al.<br />

1988) to simulate the breakthrough behavior for ORW. Based on the parameters<br />

from this simulation, the breakthrough behavior for ORW was simulated for EBCTs<br />

<strong>of</strong> 10, 15, and 20 minutes.<br />

As illustrated <strong>by</strong> the model simulations in Figure 6.1, increasing the EBCT<br />

increases the time it takes for the effluent to reach a given TOC concentration; thus,<br />

it decreases the reactivation frequency. However, it will also increase the bed depth<br />

and therefore the capital cost <strong>of</strong> the contactor. The increase in the EBCT may not<br />

be directly proportional to the increase in run time because the shape <strong>of</strong> the<br />

adsorption front may become "sharper" with longer EBCTs (Sontheimer et al.<br />

1988). This indirect proportionality is illustrated <strong>by</strong> the model simulations in<br />

Figures 6.1 and 6.2 and is attributable to the adsorption equilibrium relationship and<br />

mass transfer resistance. In Figure 6.2, the operation time is normalized <strong>by</strong> the<br />

EBCT to yield throughput measured in bed volumes (BV), where BV = operation<br />

time/EBCT. At a TOC effluent <strong>of</strong> 1.6 mg/L, there is very little difference in the<br />

throughput <strong>of</strong> the three different EBCTs; the bed life is directly proportional to the<br />

EBCT. One can also see this in Figure 6.1 <strong>by</strong> comparing the ratios <strong>of</strong> the operation<br />

times for this effluent criterion to the ratios <strong>of</strong> the respective EBCTs. At a TOC<br />

effluent <strong>of</strong> 0.8 mg/L, however, there is a 10 to 15 percent difference in the<br />

throughputs, with the shorter EBCTs yielding faster breakthroughs. The effect <strong>of</strong><br />

EBCT on NOM adsorption has not been extensively evaluated. Sontheimer et al.<br />

(1988) showed that for an ozonated water with a high amount <strong>of</strong> biodegradable<br />

NOM, more bioactivity occurs in columns with higher EBCT, which in turn<br />

increases the overall removal <strong>of</strong> TOC.<br />

EBCTs between 10 and 20 minutes are representative <strong>of</strong> <strong>GAC</strong> contactors<br />

used for NOM control in previous studies. An EBCT <strong>of</strong> 10 minutes represents the<br />

upper range for a filter adsorber and the lower range for a postfilter adsorber.<br />

Increases in EBCT beyond 20 minutes for a single contactor do not increase the run<br />

time significantly (Sontheimer et al. 1988).<br />

Objectives and Approach<br />

The objectives <strong>of</strong> the work presented in this section were to evaluate the<br />

effect <strong>of</strong> EBCT on NOM adsorption for a range <strong>of</strong> water sources. Four convention<br />

ally treated waters were evaluated: ORW, SRPW, PRW, and FGW. Each water was<br />

evaluated at EBCTs <strong>of</strong> 10, 15, and 20 minutes. Conventionally treated water was<br />

used, and the RSSCT operation followed the same protocol outlined for the<br />

verification RSSCTs.<br />

The RSSCTs were monitored for NOM removal <strong>by</strong> analyzing for TOC and<br />

UV254 . DBF formation was assessed <strong>by</strong> UFC-TOX, -TTHM, -HAA6, and -CH.<br />

The breakthrough behaviors with respect to optimization parameters can be<br />

compared based on effluent criteria or water quality goals. Ideally, these treatment<br />

goals should reflect current or proposed future regulatory agendas. However, only


Design and Operational Considerations 83<br />

two <strong>of</strong> the DBFs have current or proposed MCLs, and no regulatory criteria have<br />

been set for TOC or UV254 .<br />

Proposed TTHM and HAA5 MCLs for Stages 1 and 2 <strong>of</strong> the D-<strong>DBP</strong> Rule<br />

were used: 80 and 40 ug/L for TTHM and 60 and 30 ug/L for HAAS. (Although the<br />

authors had HAA6 data, HAA5 data were used in these calculations.) For the other<br />

two DBFs, no regulatory basis exists. For TOX, an effluent criterion <strong>of</strong> 70 ug/L was<br />

chosen, because this would allow comparison for all waters. The breakthrough<br />

behaviors <strong>of</strong> CH were not compared, since in most cases the effluent concentrations<br />

were extremely low.<br />

Treatment goals for TOC and UV254 were chosen somewhat pragmatically.<br />

Enhanced coagulation is not required to remove DBF precursors if TOC concentra<br />

tion is


84 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> GA C <strong>Adsorption</strong><br />

Table 6.1 Summary <strong>of</strong> bed lives for different effluent criteria_______<br />

Run time until criterion is met (days)<br />

Source<br />

EBCT<br />

TOC TOX TTHM HAA6<br />

1.6 mg/L 0.8 mg/L 0.01/cm 70 ug CI-/L 64 ug/L 32 ug/L 48 ug/L 24 ug/L<br />

ORW<br />

10<br />

15<br />

20<br />

—<br />

—<br />

—<br />

66<br />

80<br />

139<br />

62<br />

117<br />

131<br />

126<br />

—<br />

—<br />

—<br />

—<br />

—<br />

64<br />

87<br />

146<br />

— —<br />

— —<br />

— —<br />

PRW<br />

10<br />

15<br />

20<br />

58<br />

128<br />

135<br />

35<br />

62<br />

70<br />

33<br />

65<br />

69<br />

38<br />

89<br />

153<br />

113<br />

—<br />

208<br />

50<br />

100<br />

90<br />

— 42<br />

— 112<br />

— 117<br />

SRPW<br />

10<br />

15<br />

20<br />

—<br />

—<br />

—<br />

42<br />

83<br />

103<br />

47<br />

85<br />

128<br />

70<br />

120<br />

162<br />

—<br />

—<br />

—<br />

56<br />

103<br />

132<br />

_ _<br />

— —<br />

— —<br />

FGW<br />

10<br />

3.5<br />

—<br />

2<br />

4.5<br />

8.5<br />

8<br />

14 12<br />

15<br />

9.0<br />

4<br />

6<br />

9.5<br />

12<br />

12<br />

21 13<br />

20<br />

12<br />

8<br />

9<br />

10<br />

13<br />

19<br />

28 21<br />

— indicates effluent did not reach criterion<br />

systematically affect the normalized breakthrough <strong>of</strong> TTHM and HAA6 precursors<br />

for ORW.<br />

Figure 6.9 shows the effect <strong>of</strong> EBCT on effluent TOC concentrations for<br />

PRW. For an increase in EBCT from 10 to 20 minutes, the run time to a TOC<br />

breakthrough <strong>of</strong> 1.6 mg/L increased from 58 to 135 scaled operation days, which<br />

represents an increase in run time <strong>of</strong> 133 percent. The run time to a TOC<br />

breakthrough <strong>of</strong> 0.8 mg/L increased from 35 to 70 scaled operation days, which<br />

represents an increase in run time <strong>of</strong> 100 percent. The 15- and 20-minute-EBCT<br />

columns showed very similar behavior. When normalized <strong>by</strong> EBCT, the NOM<br />

adsorption behavior was found to be the same for all three EBCTs for PRW, as<br />

shown in Figure 6.10. All three EBCTs had the same impact on normalized <strong>DBP</strong><br />

precursor removal for PRW, as illustrated for UFC-TTHM in Figure 6.11.<br />

Figure 6.12 shows the effluent TOC concentration as a function <strong>of</strong> scaled<br />

operation time for SRPW. As expected, the 10-minute-EBCT contactor broke<br />

through earliest in time, and the 20-minute contactor broke through last. For an<br />

increase in EBCT from 10 to 20 minutes, the run time to a TOC breakthrough <strong>of</strong> 0.8<br />

mg/L increased from 41 to 104 scaled operation days, which represents an increase<br />

in run time <strong>of</strong> 154 percent. Similar to the results seen for ORW and PRW, EBCT did<br />

not systematically affect normalized NOM adsorption behavior for SRPW, as<br />

indicated in Figure 6.13, although after 6,000 BV the effluent from the 10-minute-<br />

EBCT contactor tended to exhibit breakthrough first. The same effect <strong>of</strong> EBCT<br />

was found for normalized <strong>DBP</strong> precursor removal, as illustrated in Figure 6.14 for<br />

UFC-TOX.


Design and Operational Considerations 85<br />

The EBCT effect on TOC is shown in Figure 6. 1 5 for FGW. The results are<br />

similar to those <strong>of</strong> the three surface waters discussed previously. An increase in run<br />

time <strong>of</strong> 243 percent was observed from a 10- to 20-minute EBCT for a TOC goal <strong>of</strong><br />

1 .6 mg/L, whereas the increase was 300 percent for a TOC goal <strong>of</strong> 0.8 mg/L. The<br />

EBCT did appear to affect normalized TOC adsorption behavior, as shown in Figure<br />

6. 1 6. In the first half <strong>of</strong> the curve, breakthrough occurred earliest in the 10-minute-<br />

EBCT contactor and latest in the 20-minute-EBCT contactor. The results are similar<br />

to those in Figure 6.2. EBCT did not have as clear <strong>of</strong> an impact on normalized DBF<br />

precursor adsorption, however, as illustrated in Figure 6.17 for UFC-TTHM.<br />

Impact <strong>of</strong> Blending ___________________<br />

In most <strong>GAC</strong> applications <strong>of</strong> any significant size, multiple contactors will<br />

be operated in a parallel configuration and the number <strong>of</strong> contactors will depend on<br />

the plant size. Parallel <strong>GAC</strong> contactors are operated in a staggered mode wherein<br />

each contactor is in operation for different lengths <strong>of</strong> time. This is illustrated in<br />

Figures 6.18, 6.19, and 6.20. In this mode <strong>of</strong> operation, one contactor at a time is<br />

taken <strong>of</strong>f-line when the blended effluent exceeds the target effluent concentration,<br />

and a column with fresh or reactivated <strong>GAC</strong> is then placed on-line. This will cause<br />

the blended effluent concentration to temporarily decrease, as illustrated in Figure<br />

6.20. The effluent from the contactor in operation the longest can have a higher<br />

breakthrough concentration than the target because it is blended with water from<br />

contactors that have effluent concentrations much lower than the target concentra<br />

tions (see time t in Figure 6.19). The effluents <strong>of</strong> parallel contactors are blended<br />

prior to disinfection. Thus, parallel operation in a multiple-contactor configuration<br />

will result in longer <strong>GAC</strong> bedlife, and the time between reactivations will be longer.<br />

Under ideal conditions, staged blending with multiple parallel contactors leads to<br />

near steady-state effluent concentration (Roberts and Summers 1982).<br />

The effluent concentration is controlled <strong>by</strong> the frequency <strong>of</strong> reactivation or<br />

the carbon use rate, i.e., how long a single adsorber is allowed to operate prior to<br />

reactivation. Because <strong>of</strong> blending <strong>of</strong> effluents from multiple contactors, the overall<br />

<strong>GAC</strong> plant performance will be better than predicted <strong>by</strong> single contactor results. The<br />

overall <strong>GAC</strong> plant performance for a system <strong>of</strong> n parallel contactors can be<br />

described <strong>by</strong> the following equation (Roberts and Summers 1982):<br />

c~ ifi •«*'"><br />

! = 5- = M_ ———— (6.0<br />

which, for large values <strong>of</strong> n, becomes<br />

n f. Jn<br />

f=I-J- = -Ifi (6.2)<br />

i=i n n 1=1<br />

where<br />

fj = normalized concentration in the effluent <strong>of</strong> the ith (<strong>of</strong> n)<br />

contactor


86 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

f = the average normalized concentration <strong>of</strong> the blended effluent<br />

CE = the average effluent concentration <strong>of</strong> the blended water<br />

C0 = the influent concentration<br />

t D = the reactivation time<br />

K<br />

Roberts and Summers (1982) indicated that this approximation is valid when the<br />

number <strong>of</strong> parallel contactors exceeds 10. Equation 6.2 can be related to the<br />

definition <strong>of</strong> mathematical integration. The following equation defines mathemati<br />

cal integration over time (t):<br />

Mm n<br />

jf(t)-dt = I f^t). At (6.3)<br />

where<br />

At = - (6.4)<br />

n<br />

from Equations 6.3 and 6.4, it follows that<br />

- hm I fi(t) = -Jf(t)dt (6.5)<br />

n n — >°° 1=1 t<br />

If Equations 6.2 and 6.5 are combined, the blended effluent breakthrough curve<br />

can be represented <strong>by</strong><br />

f = -Jf(t)dt (6.6)<br />

Equation 6.6 indicates that if one knows the mathematical description <strong>of</strong> a singlecontactor<br />

breakthrough curve, f. = f(t), the breakthrough curve for the blended<br />

effluent, f , can be derived <strong>by</strong> mathematical or numerical integration. Numerical<br />

integration, however, can be done without a mathematical equation being devel<br />

oped.<br />

A logistic equation <strong>of</strong> the following form was used to fit the experimental<br />

data from a single contactor:<br />

(6'7)<br />

where A, B, and C are fitting parameters and t is breakthrough time. Upon<br />

substituting Equation 6.7 for f(t) in Equation 6.6 and performing integration, one<br />

arrives at the following equation for the blended effluent breakthrough curve:<br />

f = A + -^-«]n(l+B.e"C") - — .ln(l + B) (6.8)<br />

Ot C


Design and Operational Considerations 87<br />

Equation 6.8 can be solved <strong>by</strong> trial and error for a target blended concentration to<br />

determine the breakthrough time to achieve the target.<br />

For mixtures like NOM, modeling has not been reported and is complicated<br />

<strong>by</strong> the impact <strong>of</strong> the Br:DOC ratio during chlorination on the formation <strong>of</strong> DBFs.<br />

Blending experiments are important because <strong>of</strong> the impact <strong>of</strong> changing the Br~:DOC<br />

ratio on DBF formation and speciation. At higher Br~:DOC ratios, more DBFs are<br />

formed and a higher percent <strong>of</strong> brominated DBFs occur. Because TOC in the <strong>GAC</strong><br />

effluent is initially low and Br is not removed <strong>by</strong> <strong>GAC</strong>, the Br":DOC ratio is<br />

initially high. This phenomenon leads to a higher percentage <strong>of</strong> brominated<br />

compounds compared to that in the influent. This may be problematic if the<br />

brominated DBFs turn out to be individually regulated in the future D-<strong>DBP</strong> Rule.<br />

Previously published work on DBF formation, however, has not considered the<br />

impact <strong>of</strong> blending <strong>of</strong> <strong>GAC</strong> effluents on this phenomenon. Prior to disinfection, the<br />

plant water will be made up <strong>of</strong> blended effluent from multiple <strong>GAC</strong> contactors at<br />

various stages <strong>of</strong> NOM breakthrough. This blended water may have a different<br />

distribution <strong>of</strong> DBF precursors compared to a single contactor. Blending is also<br />

impacted <strong>by</strong> different molecular size and hydrophobic-hydrophilic effluent compo<br />

sitions from different contactors, as well as <strong>by</strong> different Br" :DOC ratios.<br />

Objectives and Approach<br />

The objectives <strong>of</strong> the work presented in this section were to investigate the<br />

effect <strong>of</strong> blending effluent water on the breakthrough behavior <strong>of</strong> NOM and DBF<br />

precursors. The goal was to experimentally simulate blending from multiple parallel<br />

contactors and evaluate the use <strong>of</strong> the integrated logistic function to predict NOM<br />

removal after blending.<br />

To simulate the blended water scenario, an integral breakthrough experi<br />

ment with the RSSCT was performed. The effluent from the RSSCT was continu<br />

ously collected in a large reservoir. With time, the volume and NOM concentration<br />

in this composite reservoir increased and the water was sampled. The DBFs formed<br />

in the chlorinated samples represent the effluent in a blended system, and the time<br />

at which a potential MCL is exceeded can be used to calculate <strong>GAC</strong> usage rate. The<br />

results <strong>of</strong> the integral breakthrough experiment were predicted based on the<br />

integrated logistic function model, and the fitting parameters were determined<br />

based on a fit <strong>of</strong> the logistic function model on the data from a single contactor.<br />

Blending experiments were conducted for PRW (EBCT = 15 minutes),<br />

SRPW (EBCT = 10 minutes), and FGW (EBCT = 20 minutes), which represent a<br />

range <strong>of</strong> water qualities (see Table 2.1 [p. 6]). Five or six blended effluent samples<br />

were collected for each RSSCT. The samples were analyzed for TOC and UFC-<br />

TTHM. The single-contactor breakthrough behavior was compared to the blended<br />

water scenario using TOC and TTHM target concentrations discussed earlier in this<br />

chapter.<br />

Results and Discussion<br />

The impact <strong>of</strong> blending on effluent TOC concentrations for PRW is shown<br />

in Figure 6.21. The discrete effluent values <strong>of</strong> the single contactor are compared<br />

with those <strong>of</strong> the composite "blended" sample. The logistic model, Equation 6.7,<br />

was applied to the single-contactor breakthrough data, and a best-fit procedure


88 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> GA C <strong>Adsorption</strong><br />

yielded values for the parameters A, B and C (this model simulation is shown <strong>by</strong> the<br />

solid line). Using these parameter values, Equation 6.8 was used to predict the<br />

blended effluent concentration (the model prediction is shown <strong>by</strong> the dashed line).<br />

The impact <strong>of</strong> blending is summarized for PRW, SRPW, and FGW in Table 6.2.<br />

For PRW the operating time to a TOC breakthrough <strong>of</strong> 0.8 mg/L was<br />

extended from 63 to 116 scaled operation days <strong>by</strong> blending the effluent water. The<br />

numerical model showed very good agreement with the experimental blended<br />

effluent data. Figure 6.22 shows the results for <strong>DBP</strong> precursor removal for PRW as<br />

illustrated <strong>by</strong> UFC-TTHM. The run time to the TTHM goal <strong>of</strong> 32 ug/L was<br />

increased from 94 to 199 scaled operation days. The effect <strong>of</strong> blending on TTHM<br />

concentration was also well predicted <strong>by</strong> the model.<br />

The impact <strong>of</strong> blending water from multiple contactors is shown for effluent<br />

TOC concentrations for SRPW in Figure 6.23. The run time to a TOC concentration<br />

<strong>of</strong> 0.8 mg/L was extended from 42 to 104 scaled operation days. The blending model<br />

slightly underpredicted the experimental TOC data. The results <strong>of</strong> blending <strong>GAC</strong><br />

effluent are shown for UFC-TTHM for SRPW in Figure 6.24. The time to an effluent<br />

TTHM concentration <strong>of</strong> 32 ug/L was extended <strong>by</strong> blending from 56 to 281 scaled<br />

operation days. The blending model also underpredicted the TTHM concentration.<br />

The effluent TOC concentrations results from the blending experiments are<br />

shown in Figure 6.25 for FGW. As was the case with the previous two waters, the<br />

<strong>GAC</strong> run time was increased <strong>by</strong> blending the effluent from multiple <strong>GAC</strong> contactors.<br />

The time to a TOC concentration <strong>of</strong> 1.6 mg/L was extended from 13 to 28 scaled<br />

operation days. The time to a TOC concentration <strong>of</strong> 0.8 mg/L was extended from<br />

8 to 14 scaled operation days. The experimental blended data were slightly<br />

underpredicted <strong>by</strong> the numerical model. These results are mirrored in the results<br />

for TTHM precursors shown in Figure 6.26. The run time to the TTHM goal <strong>of</strong><br />

64 ug/L was extended from 15 to 25 scaled operation days. The run time to the<br />

TTHM goal <strong>of</strong> 32 ug/L was extended from 11 to 19 scaled operation days. Similar<br />

to the modeled results for TOC, the TTHM results were slightly underpredicted in<br />

the latter part <strong>of</strong> the breakthrough <strong>by</strong> the numerical model.<br />

NOM Desorption and the Impact <strong>of</strong> Backwashing<br />

Recently, evidence has been gathered that indicates that NOM does not<br />

completely desorb from the surface <strong>of</strong> <strong>GAC</strong> when concentration gradients are<br />

applied. Once adsorbed, some <strong>of</strong> the NOM will not desorb from <strong>GAC</strong> back into<br />

solution if the liquid phase concentration is reduced. Summers and Roberts (1988)<br />

and Summers et al. (1989, 1992) showed for four source waters that less than 6<br />

percent <strong>of</strong> the total adsorbed NOM, as measured in terms <strong>of</strong> TOC and UV254 ,<br />

desorbed when the influent concentration was reduced to the detection limit. Hong<br />

(1995) showed that less than 15 percent <strong>of</strong> previously adsorbed DOC was desorbed<br />

<strong>by</strong> a reverse in the concentration gradient for six natural waters.<br />

<strong>GAC</strong> filter adsorbers require frequent backwashing in order to reduce head<br />

loss accumulation. During backwashing, the <strong>GAC</strong> particles may become restratif ied,<br />

which can result in saturated particles moving from the top <strong>of</strong> the column (which has<br />

been exposed to high influent NOM concentrations) to the bottom <strong>of</strong> the column.<br />

If the liquid phase NOM concentration at the bottom <strong>of</strong> the column is low during this<br />

backwashing, the concentration gradient favors the desorption <strong>of</strong> NOM. Further, the


Design and Operational Considerations 89<br />

Table 6.2 Impact <strong>of</strong> blending on bed life___________________________<br />

__ Run time (days)_____________<br />

TOC TTHM<br />

Source<br />

configuration 1.6mg/L<br />

0.8 mg/L<br />

64 ug/L 32 ug/L<br />

PRW<br />

Single —<br />

Blended —<br />

63<br />

116<br />

— 94<br />

— 199<br />

SRPW<br />

Single —<br />

Blended —<br />

42<br />

104<br />

— 56<br />

— 281<br />

FGW<br />

Single 13<br />

Blended 28<br />

8<br />

14<br />

15 11<br />

25 19<br />

• indicates effluent did not reach criterion<br />

restratification <strong>of</strong> <strong>GAC</strong> particles may lead to a spreading out <strong>of</strong> the mass transfer<br />

zone, which can lead to earlier breakthrough (Sontheimer et al. 1988). However,<br />

Hong and Summers (1994) and Hong (1995) have shown that, in general, backwashing<br />

does not lead to significantly earlier NOM and DBF precursor breakthrough.<br />

The lack <strong>of</strong> desorption <strong>of</strong> NOM may have a major impact on the application<br />

<strong>of</strong> <strong>GAC</strong> for DBF precursor control. Concerns about the impact <strong>of</strong> backwashing on<br />

the mass transfer zone and premature breakthrough may not be relevant for NOM<br />

and DBF precursors, which would facilitate the use <strong>of</strong> filter adsorbers in situations<br />

where high removal efficiencies are not required.<br />

Objectives and Approach<br />

The objectives <strong>of</strong> the work presented in this section were to evaluate the<br />

impact <strong>of</strong> backwashing on the breakthrough behavior <strong>of</strong> NOM and DBF precursors<br />

and to determine the extent <strong>of</strong> NOM desorption from a column previously exposed<br />

to NOM.<br />

Desorption and backwashing experiments were run for PRW (EBCT =10<br />

minutes). Desorption experiments were run to evaluate NOM desorption as a result<br />

<strong>of</strong> a change in concentration gradient. An adsorption RSSCT with an influent TOC<br />

<strong>of</strong> 3 mg/L was run to 72 percent breakthrough. After completion <strong>of</strong> the RSSCT run,<br />

lab clean water (conductivity adjusted with KC1) containing less than 0.2 mg/L TOC<br />

was fed to the column at the same flow rate as during adsorption. The effluent was<br />

monitored for TOC and UV254 . The desorption experiment was conducted until the<br />

effluent TOC concentration was below 10 percent <strong>of</strong> the adsorption influent<br />

concentration, or 0.3 mg/L.<br />

To investigate the impact <strong>of</strong> backwashing, two identically designed RSSCTs<br />

(EBCT =10 minutes) were operated. A 10-minute EBCT was chosen because it<br />

represents the upper operation limit for a filter adsorber. One column was backwashedevery<br />

11 scaled operation days, while the other column was neverbackwashed.


90 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> GA C <strong>Adsorption</strong><br />

Results and Discussion<br />

The results for NOM desorption for PRW are shown in Figures 6.27 and<br />

6.28 for TOC and UV254, respectively. The run time for the desorption experiment<br />

was 18.5 scaled operation days. To determine the mass <strong>of</strong> NOM desorbed during<br />

that time, the breakthrough curve was integrated and compared to the integrated<br />

adsorption breakthrough curve. Eleven percent <strong>of</strong> the total adsorbed TOC and 5.8<br />

percent <strong>of</strong> the total removed UV254 were desorbed from the <strong>GAC</strong> <strong>by</strong> reversing the<br />

concentration gradient through a change in the influent TOC concentration <strong>of</strong> 3.0<br />

mg/L to 0.2 mg/L. Most <strong>of</strong> the NOM desorption for PRW occurred in the first 5<br />

scaled operation days, after which desorption was much slower. These results are<br />

similar to those reported <strong>by</strong> Hong (1995).<br />

Figures 6.29 and 6.30 show the impact <strong>of</strong> backwashing on NOM break<br />

through for PRW as measured <strong>by</strong> TOC and UV 254 , respectively. After backwashing,<br />

the effluent concentration initially increased <strong>by</strong> an average <strong>of</strong> 8 percent, and no<br />

trends were evident until the next backwashing. After 11 days the backwashed<br />

column broke through more rapidly than the nonbackwashed column. For the<br />

backwashed column, the total desorbed mass from solid phase to liquid phase as a<br />

result <strong>of</strong> backwashing was 13 percent <strong>of</strong> the adsorbed mass in the nonbackwashed<br />

column. The mass <strong>of</strong> TOC that desorbed because <strong>of</strong> backwashing was similar to the<br />

amount desorbed in the desorption experiment. Backwashing decreased the run<br />

time to 1.6 mg/L TOC from 58 to 48 days and to 0.8 mg/L from 34 to 24 days.<br />

A similar result is shown for UFC-TTHM in Figure 6.31. The backwashed<br />

column showed earlier breakthrough <strong>of</strong> TTHM precursors, 30 percent earlier at 32<br />

ug/L. However, after approximately 60 scaled operation days, the behavior <strong>of</strong> both<br />

columns was very similar.<br />

PRW breakthrough behavior showed a significant impact <strong>of</strong> backwashing.<br />

These results differ from those <strong>of</strong> Hong and Summers (1994) and Hong (1995), who<br />

reported that only one out <strong>of</strong> four waters studied showed minor evidence <strong>of</strong> early<br />

breakthrough due to backwashing. The results from the PRW desorption experi<br />

ment (Figures 6.27 amd 6.28) indicated that the mass <strong>of</strong> TOC desorbed was similar<br />

to that for other waters. PRW is a heavily industrially impacted surface water in<br />

which SOCs are known to be intermittently present. Other researchers (Sontheimer<br />

et al. 1988) have shown that backwashing leads to earlier breakthrough <strong>of</strong> SOCs.<br />

Thus, the nature <strong>of</strong> TOC may be an important factor in determining the impact <strong>of</strong><br />

backwashing.<br />

Summary_________________________<br />

Impact <strong>of</strong> EBCT on <strong>Adsorption</strong> Behavior<br />

Increasing the EBCT resulted in longer operating times to a specific<br />

breakthrough condition, as expected. For ORW, SRPW, and PRW, an increase in<br />

EBCT from 10 to 20 minutes resulted in a 100 to 154 percent increase in run time<br />

to a TOC breakthrough <strong>of</strong> 0.8 mg/L. Neither the NOM nor <strong>DBP</strong> precursor<br />

adsorption was found to be systematically affected <strong>by</strong> EBCT for these three waters<br />

when comparisons were based on throughput. For FGW, a 229 percent increase in


Design and Operational Considerations 91<br />

run time was observed to a TOC breakthrough <strong>of</strong> 1.6 mg/L, and a systematic impact<br />

<strong>of</strong> EBCT was found for TOC and to a lesser extent for the DBF precursors.<br />

A longer EBCT will result in longer run times to a given effluent criterion<br />

and yield lower <strong>GAC</strong> reactivation frequencies. The carbon usage rate (i.e., the<br />

amount <strong>of</strong> carbon that must be reactivated), however, may not be related to EBCT.<br />

Further, a longer EBCT will have higher costs associated with larger contactors. An<br />

economic analysis is needed to determine the optimum EBCT for a given system.<br />

Only the cost analysis can determine whether the greater capital expense associated<br />

with longer EBCTs compensates for a potential reduction in <strong>GAC</strong> reactivation<br />

costs. The cost analysis associated with these data is presented in Chapter 12.<br />

Impact <strong>of</strong> Blending<br />

The results presented in the blending section emphasize the advantage <strong>of</strong><br />

operating multiple contactors in parallel. The time to <strong>GAC</strong> performance goals can<br />

be significantly extended <strong>by</strong> blending the effluent from multiple contactors. For the<br />

three waters examined (PRW, SRPW, and FGW), blending increased the run time<br />

<strong>by</strong> an average <strong>of</strong> 150 percent for both TOC and UFC-TTHM. The integral logistic<br />

model was found to predict blended effluent TOC concentrations well and reason<br />

ably predict those for UFC-TTHM. It is important for one to consider the added<br />

benefits <strong>of</strong> blending when analyzing <strong>GAC</strong> cost with respect to a performance goal.<br />

The impact <strong>of</strong> these extended run times on <strong>GAC</strong> use rate and costs will be also be<br />

presented in Chapter 12.<br />

NOM Desorption and the Impact <strong>of</strong> Backwashing<br />

For PRW, NOM did not desorb easily from <strong>GAC</strong> as a function <strong>of</strong> a change<br />

in the concentration gradient. Only 11 percent <strong>of</strong> the TOC was found to desorb from<br />

a column that was 75 percent exhausted. For PRW, backwashing led to earlier NOM<br />

and <strong>DBP</strong> precursor breakthrough behavior. Other researchers, however, have found<br />

this not to be the case. It is postulated that the impact found in this study is due to<br />

the nature <strong>of</strong> the NOM in PRW and the possible presence <strong>of</strong> SOCs, the removal <strong>of</strong><br />

which has been demonstrated to be affected <strong>by</strong> backwashing.


92 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> GA C <strong>Adsorption</strong><br />

2.0<br />

TOC<br />

C0 = 2 mg/L<br />

10<br />

u><br />

1.5 -<br />

2 1.0 -<br />

I o<br />

O<br />

0.5 -<br />

EBCT (min)<br />

—— 10<br />

— -15<br />

........ 20<br />

0.0<br />

100 200 300<br />

Scaled operation time (days)<br />

400 500<br />

Figure 6.1 Model simulation <strong>of</strong> the effect <strong>of</strong> EBCT on TOC breakthrough for the Ohio<br />

River water<br />

2.0<br />

1.5 -<br />

TOC<br />

C0 = 2 mg/L<br />

20 ...--^<br />

10<br />

o<br />

1e 1.0 -<br />

§<br />

o<br />

0.5 -<br />

0.0<br />

10000 20000 30000<br />

Throughput, BV (bed volumes)<br />

EBCT (min)<br />

—— 10<br />

— - 15<br />

20<br />

40000<br />

Figure 6.2 Model simulation <strong>of</strong> the effect <strong>of</strong> EBCT on normalized TOC breakthrough for<br />

Ohio River water


s o<br />

s<br />

a.<br />

2<br />

5'<br />

2 5'<br />

•n<br />

•n<br />


94 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> GA C <strong>Adsorption</strong><br />

0.025<br />

0.020 -<br />

d><br />

o<br />

0.015 -<br />

TO<br />

-Q 0.010 -<br />

CO<br />

o<br />

0.005 -<br />

1 o<br />

EBCT(min) C0 (1/cm)<br />

0.000<br />

3000 6000 9000 12000<br />

Throughput, BV (bed volumes)<br />

15000 18000<br />

Figure 6.5 Effect <strong>of</strong> EBCT on normalized UV254 breakthrough for Ohio River water<br />

o<br />

5<br />

•o<br />

o<br />

O<br />

EBCT (min) C0 (ug CI7L)<br />

3000 6000 9000 12000<br />

15000 18000<br />

Throughput, BV (bed volumes)<br />

Figure 6.6 Effect <strong>of</strong> EBCT on normalized UFC-TOX breakthrough for Ohio River water


Design and Operational Considerations 95<br />

o><br />

O<br />

4-'<br />

1 8<br />

O EBCT (min) C0 (\ig/L)<br />

3000 6000 9000 12000<br />

15000 18000<br />

Throughput, BV (bed volumes)<br />

Figure 6.7 Effect <strong>of</strong> EBCT on normalized UFC-TTHM breakthrough for Ohio River<br />

water<br />

o><br />

•-<br />

I<br />

O O<br />

EBCT (min) C0 fog/L)<br />

3000 6000 9000 12000<br />

15000 18000<br />

Throughput, BV (bed volumes)<br />

Figure 6.8 Effect <strong>of</strong> EBCT on normalized UFC-HAA6 breakthrough for Ohio River<br />

water


s <br />


Design and Operational Considerations 97<br />

80<br />

70 -<br />

TTHM.<br />

60 -<br />

1<br />

o •JP<br />

50 -<br />

40 -<br />


a<br />

o<br />

-4»<br />

§<br />

t)<br />

to<br />

13<br />

I<br />

2<br />

O<br />

^<br />

8-<br />

o<br />

•3<br />

I<br />

<br />

m<br />

3}<br />


s a<br />

a. s<br />

3<br />

I a^I1<br />

2 <br />

m<br />

Concentration (mg/L)<br />

(C C<br />

(D<br />

O)<br />

O1<br />

Concentration (mg/L)<br />

o I-*<br />

a m CD<br />

o H<br />

O 3<br />

O<br />

i O<br />

m CD<br />

o<br />

o<br />

. _l . I I I . 1 . I . 1 , I<br />

SL N°<br />

(D a<br />

3 O<br />

o><br />

S<br />

o c<br />

CD<br />

T]<br />

O<br />

^<br />

a D)<br />

(Q<br />

O<br />

3<br />

a<br />

I<br />

Q)<br />

5<br />

o<br />

c<br />

(Q<br />

o^<br />

Tl<br />

O<br />

a<br />

Q><br />

(Q<br />

O<br />

Q.<br />

I


100 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

200<br />

_J<br />

a.<br />

o<br />

14>rf<br />

§<br />

500 1000 1500 2000 2500 3000<br />

3500<br />

Throughput, BV (bed volumes)<br />

Figure 6.17 Effect <strong>of</strong> EBCT on normalized UFC-TTHM breakthrough for Florida<br />

groundwater<br />

n<br />

Figure 6.18 Operation <strong>of</strong> multiple <strong>GAC</strong> contactors in parallel


Design and Operational Considerations 101<br />

t<br />

Time<br />

Figure 6.19 Individual breakthrough curves <strong>of</strong> multiple contactors operated in parallel<br />

c<br />

o<br />

start-up<br />

normal operation<br />

I o<br />

O<br />

plant effluent criterion<br />

Time<br />

Figure 6.20 Blended <strong>GAC</strong> effluent from multiple contactors operated in parallel


102 <strong>Removal</strong> <strong>of</strong><strong>DBP</strong> <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

o><br />

2.5<br />

2.0 -<br />

TOG<br />

EBCT= 15min<br />

Logistic Model<br />

I<br />

I o<br />

O<br />

1.5 -<br />

1.0 -<br />

0.5 -<br />

Predicted Blended Effluent<br />

C0 (mg/L)<br />

• Discrete effluent samples 3.0<br />

o Blended effluent samples 3.0<br />

0.0<br />

50 100 150 200<br />

Scaled operation time (days)<br />

250<br />

Figure 6.21 Effect <strong>of</strong> blending on TOC breakthrough for Passaic River water<br />

80<br />

70 -<br />

60 -<br />

* Discrete effluent samples 85<br />

0 Blended effluent samples 85<br />

TTHM<br />

EBCT=15min<br />

o<br />

1 *-»<br />

I o<br />

O<br />

Predicted Blended Effluent<br />

50 100 150<br />

200 250<br />

Scaled operation time (days)<br />

Figure 6.22 Effect <strong>of</strong> blending on UFC-TTHM breakthrough for Passaic River water


Design and Operational Considerations JOB<br />

1.6<br />

1.4<br />

1.2<br />

C0 (mg/L)<br />

• Discrete effluent samples 2.2<br />

o Blended effluent samples 2.2<br />

—— Model prediction<br />

TOO<br />

EBCT=10min<br />

o<br />

1<br />

Io<br />

o<br />

1.0<br />

0.8<br />

0.6<br />

Logistic Model<br />

0.4<br />

0.2<br />

0.0<br />

0 10 20 30 40 50 60 70 80 90 100<br />

Scaled operation time (days)<br />

Figure 6.23 Effect <strong>of</strong> blending on TOC breakthrough for Salt River Project water<br />

CD<br />

n.<br />

60<br />

50 -<br />

40 -<br />

3 30-<br />

£<br />

c0 (ug/L)<br />

• Discrete effluent samples 72<br />

0 Blended effluent samples 72<br />

Logistic Model<br />

TTHM<br />

EBCT = 10min<br />

I o<br />

O<br />

20 -<br />

10 -<br />

Predicted Blended Effluent<br />

10 20 30 40 50 60 70 80 90<br />

Scaled operation time (days)<br />

Figure 6.24 Effect <strong>of</strong> blending on UFC-TTHM breakthrough for Salt River Project water


104 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

o><br />

§<br />

o<br />

10<br />

9 -<br />

8 -<br />

7 -<br />

6 -<br />

5 -<br />

C0 (mg/L)<br />

• Discrete effluent samples 10.0<br />

° Blended effluent samples 10.0<br />

Logistic Model<br />

TOC<br />

EBCT = 20 min<br />

§<br />

4 -<br />

3 -<br />

2 -<br />

1 -<br />

Predicted Blended Effluent<br />

0<br />

—i—<br />

10<br />

—i—<br />

20 30<br />

—i—<br />

40 50<br />

Scaled operation time (days)<br />

Figure 6.25 Effect <strong>of</strong> blending on TOC breakthrough for Florida groundwater<br />

200<br />

150 -<br />

cn (ug/L)<br />

• Discrete effluent samples 238<br />

o Blended effluent samples 238<br />

— — Model prediction<br />

TTHM<br />

EBCT = 20 min<br />

o •a<br />

I<br />

8<br />

O<br />

Scaled operation time (days)<br />

Figure 6.26 Effect <strong>of</strong> blending on UFC-TTHM breakthrough for Florida groundwater


Design and Operational Considerations 105<br />

3.0<br />

2.5 -<br />

j TOC<br />

C0 (mg/L)<br />

<strong>Adsorption</strong> 3.0<br />

Desorption 0.2<br />

EBCT=10min<br />

co<br />

I<br />

o<br />

I o<br />

O<br />

2.0 -<br />

1.5 -<br />

1.0 -<br />

0.5 -<br />

0.0<br />

—i—<br />

10<br />

Scaled operation time (days)<br />

—i—<br />

15 20<br />

Figure 6.27 NOM desorption for Passaic River water: TOC levels<br />

o<br />

0.08<br />

0.07 -<br />

c<br />

0.06 -<br />

0.05 -<br />

UV.<br />

254 C0 (1/cm)<br />

<strong>Adsorption</strong> 0.072<br />

Desorption 0.001<br />

EBCT=10 min<br />

CO<br />

0.04 -<br />

o (A<br />

CO<br />

JB<br />

g<br />

0.03 -<br />

0.02 -<br />

I<br />

z><br />

0.01 -<br />

0.00<br />

T~<br />

5<br />

—i—<br />

10<br />

—i—<br />

15 20<br />

Scaled operation time (days)<br />

Figure 6.28 NOM desorption for Passaic River water: UV254 levels


106 <strong>Removal</strong> <strong>of</strong><strong>DBP</strong> <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

3.0<br />

2.5-<br />

TOC EBCT=10min<br />

I<br />

g<br />

S<br />

2.0 -<br />

1.5 -<br />

1.0 -<br />

0.5 -<br />

C0 (mg/L)<br />

o Nonbackwashed 3.0<br />

n Backwashed 3.0<br />

0.0<br />

0 20 40 60 80 100 120 140 160 180 200<br />

Scaled operation time (days)<br />

Figure 6.29 Impact <strong>of</strong> backwashing on TOC breakthrough for Passaic River water<br />

0.07<br />

0.06 H<br />

UV- 254 EBCT=10 min<br />

8<br />

S<br />

-e<br />

%<br />

x><br />

(0<br />

«<br />

o<br />

1<br />

0.05<br />

0.04<br />

0.03<br />

0.02<br />

0.01<br />

o Nonbackwashed 0.076<br />

o Backwashed 0.076<br />

0.00<br />

0 20 40 60 80 100 120 140 160 180 200<br />

Scaled operation time (days)<br />

Figure 6.30 Impact <strong>of</strong> backwashing on UV254 breakthrough for Passaic River water


Design and Operational Considerations 107<br />

O)<br />

.2<br />

1<br />

o Nonbackwashed<br />

n Backwashed<br />

20 40 60 80 100 120 140 160 180 200<br />

Scaled operation time (days)<br />

Figure 6.31 Impact <strong>of</strong> backwashing on UFC-TTHM breakthrough for Passaic River water


Chapter 7<br />

The Impact <strong>of</strong> Optimized Coagulation<br />

on NOM <strong>Removal</strong> and <strong>DBP</strong> Control <strong>by</strong><br />

<strong>GAC</strong> <strong>Adsorption</strong><br />

The capacity <strong>of</strong> <strong>GAC</strong> for NOM adsorption can be improved <strong>by</strong> a decrease<br />

in initial NOM concentration (Randtke and Jepsen 1982; Summers and Roberts<br />

1988; Sontheimer et al. 1988). <strong>GAC</strong>'s capacity for adsorption <strong>of</strong> NOM is also<br />

affected <strong>by</strong> pH (McCreary and Snoeyink 1980; Weber et al. 1983). These research<br />

ers demonstrated that as the initial pH was lowered, the adsorption capacity <strong>of</strong> <strong>GAC</strong><br />

was improved. This improvement is most <strong>of</strong>ten attributed to a decrease in the<br />

solubility <strong>of</strong> NOM at lower pH.<br />

Coagulation processes used as pretreatment to <strong>GAC</strong> can both reduce<br />

influent TOC concentrations and decrease the influent pH to the adsorber, thus<br />

leading to improved <strong>GAC</strong> performance. Coagulation also preferentially removes<br />

the large molecular size fraction and humic fraction <strong>of</strong> NOM (Semmens and Staples<br />

1986; Collins et al. 1986; Dryfuse et al. 1995). A study <strong>by</strong> Semmens et al. (1986)<br />

showed that coagulation pretreatment prior to <strong>GAC</strong> increased the run time <strong>of</strong> the<br />

contactor and that further improvements in <strong>GAC</strong> run time can be achieved at higher<br />

coagulant doses.<br />

Objectives and Approach________________<br />

The systematic evaluation <strong>of</strong> the impact <strong>of</strong> optimized coagulation pretreat<br />

ment on <strong>GAC</strong> performance for NOM removal and <strong>DBP</strong> control is discussed in this<br />

chapter. The <strong>GAC</strong> adsorption characteristics <strong>of</strong> two waters pretreated under<br />

conventional and optimized alum coagulation conditions were compared. Opti<br />

mized coagulation with respect to TOC is defined as the coagulant dose beyond<br />

which improvement in marginal TOC removal is not measurable. Changes in<br />

specific <strong>DBP</strong> yield in <strong>GAC</strong> effluent as a result <strong>of</strong> optimized coagulation pretreat<br />

ment were also investigated. Finally, to determine how the adsorbability <strong>of</strong> NOM<br />

characteristics was affected <strong>by</strong> coagulation pretreatment, a study <strong>of</strong> NOM charac<br />

teristics <strong>by</strong> two fractionatiori techniques, ultrafiltration and the use <strong>of</strong> XAD-8 resins,<br />

and PY-GC-MS was performed.<br />

709


110 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

Experimental Conditions<br />

Two waters were used in this study: ORW and SRPW. Raw water<br />

characteristics are summarized in Table 7.1.<br />

ORW was sampled at CWW after conventional treatment using 7.7 mg/L<br />

alum, 8.6 mg/L lime, pH 7.8, and rapid sand filtration. Conventionally treated water<br />

from this source was batch coagulated with an additional 80 mg/L <strong>of</strong> alum at pH 6.4.<br />

This dose was based on jar test results <strong>of</strong> Dryfuse et al. (1995). SRPW was a blend<br />

<strong>of</strong> the Verde River and Salt River (in Arizona). SRPW was conventionally treated<br />

<strong>by</strong> coagulation with alum, sedimentation, and sand filtration; the plant alum dose<br />

was 10.6 mg/L at pH 7.8. Based on jar test results, and as a result <strong>of</strong> the high<br />

alkalinity <strong>of</strong> the water, the pH <strong>of</strong> the water was first reduced to 6.9 with sulfuric acid,<br />

and an additional alum dose <strong>of</strong> 60 mg/L was then added. This yielded a pH <strong>of</strong> 6.5<br />

for optimized coagulation.<br />

Although high alum dosages were used, this study was designed to examine<br />

the impact on <strong>GAC</strong> performance <strong>of</strong> reductions in influent TOC concentrations after<br />

optimized coagulation. Lowering alum dosages for optimized coagulation similarly<br />

causes decreases in influent TOC concentration, but it was beyond the scope <strong>of</strong> this<br />

study to optimize <strong>GAC</strong> performance in terms <strong>of</strong> coagulant dosage.<br />

The batch coagulation process was similar for both ORW and SRPW. It was<br />

repeated three times on water contained in 55-gal (208-L) drums. A variable speed<br />

shaft mixer was operated at high speed prior to alum addition and for 4 minutes after<br />

alum addition. The water was then flocculated for 12 minutes at low speed. After<br />

settling, the water was passed through a 1.0-um cartridge filter. The treated water<br />

characteristics are summarized in Table 7.2. Also included in Table 7.2 are the Step<br />

1 enhanced coagulation requirements according to Stage 1 <strong>of</strong> the proposed D-<strong>DBP</strong><br />

Rule (USEPA 1994), based on raw water characteristics.<br />

Results and Discussion_________________<br />

The results presented in this section involve a comparison <strong>of</strong> the relative<br />

effects <strong>of</strong> optimized coagulation and conventional pretreatment on <strong>GAC</strong> break<br />

through curves for <strong>DBP</strong> surrogate parameters (TOC and UV254) and formed <strong>DBP</strong>s<br />

(TTHM, HAA6, TOX, and CH). All <strong>DBP</strong> formation was conducted under UFC<br />

protocol (see Appendix C). As a result <strong>of</strong> the unsteady-state nature <strong>of</strong> the break<br />

through process, it can be difficult to compare breakthrough curves beyond a<br />

qualitative assessment. It is possible, however, to use breakthrough characteristics<br />

to systematically compare breakthrough behavior. In this study, the concentration<br />

<strong>of</strong> the immediate breakthrough (Cimm), which is representative <strong>of</strong> the nonadsorbable<br />

fraction, and the time to initial breakthrough (tjm), which represents the time at<br />

which the adsorbable fraction begins to show breakthrough, are used. Furthermore,<br />

comparisons are made <strong>by</strong> establishing <strong>GAC</strong> run time criteria and comparing the<br />

difference in run time obtained after conventional treatment with that obtained after<br />

optimized coagulation. The <strong>GAC</strong> run time criterion for TOC is defined as the time<br />

for reaching an effluent concentration <strong>of</strong> 0.8 mg/L. For UV154 , the criterion is an<br />

effluent absorbance <strong>of</strong> 0.01/cm. The criterion for TTHM is guided <strong>by</strong> Stage 2 <strong>of</strong><br />

the D-<strong>DBP</strong> Rule and is 32 ug/L, a value 20 percent below the proposed MCL <strong>of</strong><br />

40 ug/L. For HAA6, the run time criterion is set at 24 ug/L, based on a target


The Impact <strong>of</strong> Optimized Coagulation 111<br />

Table 7.1 Raw water characteristics for Chapter 7 experiments<br />

<strong>Water</strong> TOC (mg/L) Alkalinity (mg/L as CaCO3)<br />

ORW<br />

SRPW<br />

2.2<br />

2.5<br />

40<br />

140<br />

Table 7.2 Treated water characteristics for Chapter 7 experiments<br />

<strong>Water</strong><br />

Alum<br />

dose<br />

(mg/L)<br />

Conventional<br />

treatment<br />

TOC<br />

(mg/L)<br />

PH<br />

TOC<br />

removal<br />

(%)<br />

Alum<br />

dose<br />

(mg/L)<br />

Optimized<br />

coagulation<br />

TOC<br />

(mg/L)<br />

PH<br />

TOC<br />

removal<br />

(%)<br />

Enhanced<br />

coagulation<br />

TOC removal<br />

requirement<br />

(%)<br />

ORW<br />

SRPW<br />

8<br />

11<br />

2.0<br />

2.2<br />

7.8<br />

7.8<br />

6<br />

9<br />

88<br />

71<br />

1.5<br />

1.7<br />

6.4<br />

6.5<br />

32<br />

31<br />

40<br />

20<br />

20 percent below the Stage 2 HAAS MCL, 30 ug/L. Finally, a value <strong>of</strong> 70 ug<br />

C1~/L is used for the TOX run time criterion.<br />

The <strong>GAC</strong> influent levels <strong>of</strong> TOC and UV254 and <strong>of</strong> the formed TTHM,<br />

HAA6, TOX, and CH for the influent waters are summarized in Table 7.3. The<br />

influent concentrations <strong>of</strong> CH were very low, and its formation was well controlled<br />

<strong>by</strong> <strong>GAC</strong>. Therefore, characterization <strong>of</strong> its breakthrough will not be discussed.<br />

<strong>GAC</strong> breakthrough behavior for ORW pretreated under conventional and<br />

optimized coagulation is shown in Figures 7.1 to 7.6 for TOC, UV 254 , and various<br />

DBFs. For each pair <strong>of</strong> breakthrough curves, the Cimm, tim , and respective run time<br />

criteria comparisons between the pretreatment types are summarized in Table 7.4.<br />

The run time criteria are also listed in Table 7.4. Although the Cimm was usually very<br />

low for the breakthrough curves <strong>of</strong> both pretreatment types, it is evident that<br />

optimized coagulation pretreatment dramatically improved <strong>GAC</strong> performance for<br />

the adsorption <strong>of</strong> DBF surrogates and DBFs. While the tint ranged from 30 to 50 days<br />

after conventional treatment, it was improved to 125 to 170 days after optimized<br />

coagulation. Furthermore, as defined <strong>by</strong> the criteria mentioned earlier, the <strong>GAC</strong> run<br />

time improved from a range <strong>of</strong> 80 to 125 days following conventional treatment to<br />

a range <strong>of</strong> 214 to 300 days following optimized coagulation. The comparison was<br />

not possible for HAA6 because effluent concentrations were very low. It is also clear<br />

that the effect <strong>of</strong> optimized coagulation on adsorption <strong>of</strong> TOC and UV254 was very<br />

similar to the effect on the DBFs analyzed. This is a good indication <strong>of</strong> the ability<br />

<strong>of</strong> TOC and UV254 to serve as surrogates for DBF precursor breakthrough.<br />

Figures 7.7 to 7.12 show the impact <strong>of</strong> optimized coagulation on <strong>GAC</strong><br />

breakthrough <strong>of</strong> TOC and UV254 , as well as various DBFs, for SRPW. The<br />

breakthrough behavior comparison parameters (Cimm , tint, and run time criteria) are<br />

summarized in Table 7.4. In a manner similar to the case <strong>of</strong> ORW, optimized<br />

coagulation greatly improved <strong>GAC</strong> performance for all parameters analyzed. The<br />

DBF surrogate parameters (TOC and UV254) and the formed DBFs analyzed


112 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

Table 7.3 Summary <strong>of</strong> DBF surrogate and DBF concentrations prior to <strong>GAC</strong> treatment<br />

<strong>Water</strong> CT<br />

Formation under UFC<br />

TOC UV254 TTHM HAA6 TOX<br />

(mg/L) (1/cm) (ng/L) (ug/L) (ug CI'/L)<br />

OC CT OC CT OC CT<br />

OC CT OC<br />

CT<br />

CH<br />

(M9/L)<br />

OC<br />

ORW 2.0<br />

SRPW 2.2<br />

1.5 0.04 0.03 86 59 26<br />

1.7 0.05 0.03 72 46 28<br />

19 170 92<br />

17 170 130<br />

3.9<br />

4.1<br />

1.4<br />

2.6<br />

CT = conventional treatment<br />

OC = optimized coagulation<br />

showed similar improvements in their breakthrough behavior with optimized<br />

coagulation. Under conventional pretreatment conditions, initial breakthrough<br />

occurred after 40 and 55 days for TOC and UV254 , respectively, as compared to<br />

between 25 and 45 days for the DBF precursors. After optimized coagulation, initial<br />

breakthrough occurred after 100 days for both TOC and UV254 , while that for the<br />

DBF precursors occurred after 90 to 110 days. <strong>GAC</strong> run time improved from a range<br />

<strong>of</strong> 83 to 120 days to a range <strong>of</strong> 155 to 300 days, based on the run time criteria.<br />

The results indicate that for the two waters examined, <strong>GAC</strong> performance<br />

after optimized coagulation was significantly improved compared to that after<br />

conventional treatment. On average, the percent improvement in tint for all param<br />

eters was greatest for ORW, at 254 percent, while that for SRPW was still very large,<br />

at 179 percent. Because <strong>GAC</strong> reactivation times are controlled <strong>by</strong> run time, the<br />

comparison based on run time criterion is very important. For all parameters, the<br />

average percent increase in run times were 148 and 129 percent for ORW and<br />

SRPW, respectively. These results demonstrate that optimized pretreatment prior<br />

to a <strong>GAC</strong> contactor may provide significant cost reductions associated with <strong>GAC</strong><br />

reactivation. Furthermore, <strong>GAC</strong> coupled with optimized coagulation may be a<br />

necessary step for many utilities to comply with future, more stringent <strong>DBP</strong><br />

regulations. However, the alum dosages used in this study were high, and it was<br />

beyond the scope <strong>of</strong> this project to optimize <strong>GAC</strong> performance at more moderate<br />

coagulant dosages.<br />

It should be noted that these results are based on single RSSCT runs. In<br />

practice, multiple <strong>GAC</strong> contactors are operated in a parallel, staggered mode, and<br />

the effluent water quality is blended prior to disinfection. As shown in Chapter 6,<br />

the impact <strong>of</strong> blending further improves <strong>GAC</strong> run time.<br />

In this study, optimized coagulation resulted in both a decrease in influent<br />

TOC and a decrease in influent pH for both waters. Other research has shown that<br />

both factors can contribute to an increase in the performance <strong>of</strong> <strong>GAC</strong> for the control<br />

<strong>of</strong> NOM. A survey <strong>of</strong> 26 <strong>GAC</strong> runs utilizing 18 different water sources yielded a<br />

logarithmic relationship <strong>of</strong> run time (as measured <strong>by</strong> bed volumes to 50 percent TOC<br />

breakthrough) as a function <strong>of</strong> influent TOC (Hooper 1996; Summers et al. 1994b).<br />

The relationship indicated that as influent TOC increased, run time decreased. A<br />

controlled study performed on a single water source (Hooper et al. 1995) showed<br />

that the impact <strong>of</strong> influent TOC on <strong>GAC</strong> run time was similar to that observed in the<br />

relationship just described. That study also investigated the impact <strong>of</strong> influent pH


OC<br />

(days)<br />

214<br />

245<br />

241<br />

*<br />

300<br />

155<br />

245<br />

300<br />

*<br />

180<br />

Change<br />

(%)<br />

+168<br />

+96<br />

+177<br />

*<br />

+150<br />

+148<br />

+87<br />

+188<br />

+191<br />

*<br />

+50<br />

+129<br />

3<br />

a.<br />

I a"<br />

s<br />

Table 7.4 Effect <strong>of</strong> coagulation pretreatment on <strong>GAC</strong> breakthrough characteristics<br />

Immediate breakthrough<br />

concentration, C lmm<br />

Time to initial<br />

breakthrough, tjnt<br />

Run time based on<br />

established criterion<br />

<strong>Water</strong><br />

Parameter<br />

Units<br />

Cone<br />

CT<br />

%lnf<br />

OC<br />

Cone<br />

%lnf<br />

CT<br />

(days)<br />

OC<br />

(days)<br />

Change<br />

(%)<br />

Criterion<br />

CT<br />

(days)<br />

ORW<br />

TOC<br />

uv254<br />

TTHM<br />

HAA6<br />

TOX<br />

Average<br />

mg/L<br />

1/cm<br />

M9/L<br />

H9/L<br />

pg cr/L<br />

0.2<br />

0.003<br />

3<br />

4<br />

5<br />

9.8<br />

6.8<br />

3.5<br />

15.4<br />

2.9<br />

8<br />

0.15<br />

BDL<br />

1<br />

4<br />

10<br />

10.1<br />

NA<br />

1.7<br />

21.1<br />

10.9<br />

11<br />

45<br />

50<br />

45<br />

30<br />

50<br />

125<br />

145<br />

150<br />

170<br />

150<br />

+178<br />

+190<br />

+233<br />

+467<br />

+200<br />

+254<br />

0.8<br />

0.01<br />

32<br />

24<br />

70<br />

80<br />

125<br />

87<br />

*<br />

120<br />

SRPW<br />

TOC<br />

UV254<br />

TTHM<br />

HAA6<br />

TOX<br />

Average<br />

mg/L<br />

1/cm<br />

H9/L<br />

M9/L<br />

ug CI-/L<br />

0.18<br />

0.001<br />

BDL<br />

2.5<br />

10<br />

8.1<br />

2.1<br />

NA<br />

8.8<br />

5.9<br />

6<br />

0.1<br />

0.001<br />

BDL<br />

2.5<br />

5<br />

6.0<br />

3.6<br />

NA<br />

15.0<br />

3.8<br />

6<br />

40<br />

55<br />

30<br />

45<br />

25<br />

100<br />

100<br />

90<br />

100<br />

110<br />

+150<br />

+82<br />

+200<br />

+122<br />

+340<br />

+179<br />

0.8<br />

0.01<br />

32<br />

24<br />

70<br />

83<br />

85<br />

103<br />

*<br />

120<br />

BDL = below detection limit<br />

Cone = concentration<br />

CT = Conventional treatment<br />

NA = not available<br />

%lnf = percent influent<br />

OC = optimized coagulation<br />

* Effluent concentration did not reach criterion


114 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

on <strong>GAC</strong> breakthrough under controlled conditions and found that as influent pH<br />

increased from 5.5 to 8.5, <strong>GAC</strong> run time decreased. For the two waters investigated<br />

in the present study, most <strong>of</strong> the improvement in <strong>GAC</strong> performance found after<br />

optimized coagulation was attributed to the decrease in influent TOC concentration<br />

(64 and 96 percent for ORW and SRPW, respectively).<br />

Effect <strong>of</strong> Coagulation Pretreatment<br />

on Specific DBF Yield <strong>of</strong> <strong>GAC</strong> Effluent_________<br />

Although the effectiveness <strong>of</strong> optimized coagulation pretreatment for the<br />

improved removal <strong>of</strong> DBF formation in <strong>GAC</strong> effluent has been shown, the<br />

comparisons were made with breakthrough curves plotted against scaled operation<br />

time. One can gain additional insight into the improvement <strong>of</strong> the RSSCT receiving<br />

optimized coagulated water (RSSCT-OC) <strong>by</strong> plotting DBF formation concentra<br />

tions against the corresponding TOC concentration for each sample and comparing<br />

these plots to those for the RSSCT receiving conventionally treated water (RSSCT-<br />

CT). These plots tend to have high linear regression correlations—as most showed<br />

an r2 value greater than 0.900. Two important features <strong>of</strong> these plots are the relative<br />

position and slope <strong>of</strong> each best-fit curve; significant changes in these parameters<br />

provide an indication <strong>of</strong> any changes in the specific DBF yield between RSSCT-CT<br />

and RSSCT-OC (Hooper 1996).<br />

Figure 7.13 shows the relationship between TTHM formation and TOC<br />

concentration for RSSCT-CT and RSSCT-OC for ORW. Both the slope and y-axis<br />

intercept were compared using regression analysis with categorical variables<br />

(Berthouex and Brown 1994). This method uses categorical variables to enable the<br />

entire data set (both pretreatments) to be fit to one model while still allowing the data<br />

sets to be distinguishable. The two categories were conventional pretreatment and<br />

optimized coagulation pretreatment. There are two null hypotheses:<br />

H Q : "The difference between the two slopes is zero."<br />

H0': "The difference between the two y-axis intercepts is zero."<br />

If the analysis shows that both null hypotheses should be accepted, then for the<br />

particular DBF examined, pretreatment <strong>by</strong> optimized coagulation did not signifi<br />

cantly affect the specific DBF yield after <strong>GAC</strong> treatment. If one or both hypotheses<br />

are rejected, then one or both <strong>of</strong> the following alternative hypotheses are accepted,<br />

and pretreatment significantly affected the specific DBF yield after <strong>GAC</strong> treatment:<br />

HA : "The difference between the two slopes is not zero."<br />

HA': "The difference between the two y-axis intercepts is not zero."<br />

TOC is<br />

A linear model for regression analysis <strong>of</strong> conventional pretreatment for<br />

V<strong>DBP</strong>= ao + a


The Impact <strong>of</strong> Optimized Coagulation 115<br />

for the DBF <strong>of</strong> interest, where y DBp and XTOC are the variables <strong>of</strong> the linear model<br />

and a0 and a { are constants. Similarly, for <strong>GAC</strong> effluent pretreated with optimized<br />

coagulation,<br />

where 60 and Bj are constants. A combined model can be created from these two<br />

models so that the model can be simultaneously fit to all the data. The combined<br />

model uses Z as a dummy variable: Z = 0 if the data are for the RSSCT-CT, and<br />

Z = 1 if the data are for the RSSCT-OC:<br />

y DBp = a0 + CC,XTOC + Z(y0 + y,xTOC) (7.3)<br />

60 = a0 + y0 (7.4)<br />

B^a, +Y, (7.5)<br />

where y0 and y, are constants. The combined model can be rearranged:<br />

y DBp = oc0 + alXTOC + Zy0 + y,ZxTOC (7.6)<br />

The regression is performed with independent variables XTOC , Z, and ZxTOC to<br />

estimate parameters oc0, (Xj , y0 , and yt . Once the model is fit to the data and estimates<br />

have been obtained for all four parameters, the model is simplified. Simplification<br />

is possible when y0 or y, is estimated as zero. When this occurs, the parameter is<br />

eliminated from the model, and the simplified version is fit to the data. For example,<br />

if y0 = 0, then B0 = aQ . This method leads to four possible outcomes, as shown in<br />

Figure 7.14. The four possible outcomes have been labeled type 1, 2, 3, and 4. A<br />

statistical s<strong>of</strong>tware package (SYSTAT Version 3.0 for Windows, Systat, Inc.,<br />

Evanston, 111.) was used for the analysis. The estimated parameter values for the<br />

four-parameter curve fit were associated with t statistics. A parameter was not<br />

significantly different than zero when the value <strong>of</strong> the t statistic was less than the<br />

critical t statistic, t*. The t* was determined from atable <strong>of</strong> critical t statistics, based<br />

on the sample size, at the 95 percent confidence level.<br />

The regression sum <strong>of</strong> squares for the complete model and the simplified<br />

version can be compared to ensure that elimination <strong>of</strong> one or more parameters from<br />

the model has not significantly affected the goodness <strong>of</strong> fit. To compare the<br />

regression sum <strong>of</strong> squares, the F statistic was computed and compared to the critical<br />

F statistic, F*, at the upper 5 percent point <strong>of</strong> the F distribution, for a 95 percent<br />

confidence level test <strong>of</strong> significance (Berthouex and Brown 1994).<br />

For the data shown in Figure 7.13, no significant change was observed in<br />

the y-axis intercepts, but a significant decrease in slope occurred between conven<br />

tional treatment and optimized coagulation pretreatment. Two conclusions can be<br />

drawn from these type 3 results. First, pretreatment <strong>by</strong> optimized coagulation<br />

reduced the specific TTHM yield after <strong>GAC</strong> treatment for ORW at a given effluent<br />

TOC concentration. Second, the increase in TTHM formation for a given increase<br />

in TOC concentration decreased after pretreatment <strong>by</strong> optimized coagulation.


116 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

The results obtained for the formation <strong>of</strong> TTHM, HAA6, and TOX for<br />

ORW and SRPW are summarized in Table 7.5, together with a comparison <strong>of</strong> F to<br />

F*. The analysis <strong>of</strong> HAA6 formation for ORW also yielded type 3 behavior, as<br />

shown in Figure 7.15. The analysis performed on TOX formation yielded a<br />

significant increase in y-axis intercept and a significant decrease in slope with <strong>GAC</strong><br />

treatment after optimized coagulation, resulting in a "crossing" pattern, as shown<br />

in Figure 7.16, which is a modified type 1 behavior. Therefore, at low TOC<br />

concentrations (below 0.5 mg/L), equivalent TOC concentrations showed a higher<br />

specific TOX yield in <strong>GAC</strong> effluent after optimized coagulation, while at TOC<br />

concentrations greater than 0.5 mg/L, a higher TOX yield was observed in <strong>GAC</strong><br />

effluent after conventional treatment. Overall, however, the increase in TOX<br />

formation for a given increase in TOC concentration in <strong>GAC</strong> effluent was lower for<br />

the water pretreated <strong>by</strong> optimized coagulation.<br />

For SRPW, the effects <strong>of</strong> treatment on the relationship between TOC and<br />

the formation <strong>of</strong> TTHM and HAA6, respectively, are is shown in Figures 7.17 and<br />

7.18. Statistical analysis yielded type 3 results for these DBFs, similar to ORW<br />

results for the same DBFs. Both the specific DBF yield at a given TOC and the<br />

increase in DBF formation for a given increase in TOC concentration decreased for<br />

these DBFs with optimized coagulation pretreatment.<br />

The comparison involving TOX data for SRPW (Figure 7.19) showed that<br />

there was a significant difference in neither the y-axis intercepts nor the slopes,<br />

which corresponded to type 4 behavior. Thus, the specific TOX yield in <strong>GAC</strong><br />

effluent was not significantly affected <strong>by</strong> optimized coagulation pretreatment for<br />

SRPW.<br />

The statistical analysis presented here allows for a direct comparison <strong>of</strong> the<br />

effect <strong>of</strong> optimized coagulation on the specific DBF yield <strong>of</strong> <strong>GAC</strong> effluent. In all<br />

cases, optimized coagulation yielded lower TOC concentrations in the <strong>GAC</strong><br />

effluent at any given time, which also yielded lower formed DBFs. In most cases,<br />

DBF formation was reduced at a given TOC concentration in <strong>GAC</strong> effluent after<br />

optimized coagulation pretreatment. In cases for which DBF formation was not<br />

Table 7.5 Effect <strong>of</strong> coagulation pretreatment on relationship between <strong>GAC</strong> effluent TOC<br />

and DBF formation<br />

Slope Y-axis intercept<br />

<strong>Water</strong><br />

Conventional Optimized<br />

treatment coagulation<br />

DBF (eg (B,)<br />

Conventional Optimized<br />

treatment coagulation<br />

K) (Bo) F<br />

F*<br />

Behavior<br />

type<br />

ORW<br />

TTHM<br />

HAA6<br />

TOX<br />

55.1<br />

11.5<br />

90.8<br />

> 38.2<br />

> 7.3<br />

> 31.0<br />

-11.6<br />

1.3<br />

-24.9<br />

= -11.6<br />

1.3<br />

< 6.3<br />

2.01<br />

0.01<br />

NA<br />

4.67<br />

4.67<br />

NA<br />

3<br />

3<br />

1 modified<br />

SRPW TTHM<br />

HAA6<br />

TOX<br />

31.9<br />

8.7<br />

56.9<br />

> 24.6<br />

> 6.9<br />

56.9<br />

-3.1<br />

1.2<br />

-0.6<br />

-3.1<br />

1.2<br />

-0.6<br />

0.30<br />

3.28<br />

3.59<br />

4.60<br />

4.60<br />

4.10<br />

3<br />

3<br />

4<br />

NA = not applicable


The Impact <strong>of</strong> Optimized Coagulation 117<br />

reduced, there was no significant difference in specific DBF yield. Therefore, based<br />

on this statistical analysis for the two waters examined, optimized coagulation<br />

pretreatment to <strong>GAC</strong> not only reduced the influent TOC concentration but also<br />

reduced DBF formation at a given TOC in the <strong>GAC</strong> effluent.<br />

Impact <strong>of</strong> Pretreatment on NOM<br />

Fraction Characteristics After <strong>GAC</strong>____________<br />

Humic-Nonhumic Fractionation<br />

Figure 7.20 is a composite graph that shows the humic and nonhumic<br />

makeup <strong>of</strong> ORW at various stages <strong>of</strong> treatment: raw, conventional treatment,<br />

optimized coagulation, and three <strong>GAC</strong> effluent samples for RSSCT-OC. The bar<br />

graphs compare the humic and nonhumic content <strong>of</strong> the raw water sample to the<br />

fractionation <strong>of</strong> water samples after conventional treatment and optimized coagu<br />

lation. Table 7.6 summarizes the humic and nonhumic makeup <strong>of</strong> the samples<br />

analyzed. The results show that the raw water sample was 55 percent humic and 45<br />

percent nonhumic. A small decrease in total DOC was observed between raw and<br />

conventionally treated water, mostly because <strong>of</strong> a decrease in the humic fraction.<br />

Optimized coagulation decreased the total DOC <strong>by</strong> 32 percent relative to the raw<br />

water, with substantial removal occurring in both fractions. The humic and nonhumic<br />

fractions composed 58 and 42 percent, respectively, <strong>of</strong> the total DOC <strong>of</strong> the<br />

optimized coagulated water.<br />

The right side <strong>of</strong> Figure 7.20 shows the humic and nonhumic makeup <strong>of</strong> the<br />

<strong>GAC</strong> effluent samples after optimized coagulation pretreatment. The nonadsorbable<br />

fraction <strong>of</strong> DOC was composed entirely <strong>of</strong> nonhumic organic matter, while the<br />

humic fraction was initially nearly completely removed and did not show break<br />

through until after 150 days <strong>of</strong> operation. This behavior is similar to what was<br />

presented for the conventionally pretreated water in Chapter 5. For both columns,<br />

the concentration <strong>of</strong> the nonhumic fraction in the <strong>GAC</strong> effluent gradually increased<br />

over the course <strong>of</strong> the <strong>GAC</strong> operation. The nonhumic fraction showed a steeper<br />

breakthrough behavior in the RSSCT-CT effluent than in the RSSCT-OC effluent.<br />

Table 7.6 Impact <strong>of</strong> coagulation treatment on humic-nonhumic and molecular size<br />

fractionation<br />

<strong>Water</strong><br />

source<br />

Treatment<br />

Humic<br />

Nonhumic<br />

Percent <strong>of</strong> total DOC<br />

3,000 MS<br />

ORW<br />

Raw<br />

Conventional treatment<br />

Optimized coagulation<br />

55<br />

52<br />

58<br />

45<br />

48<br />

42<br />

19<br />

28<br />

33<br />

60<br />

72<br />

60<br />

21<br />

0<br />

7<br />

SRPW<br />

Raw<br />

Conventional treatment<br />

Optimized coagulation<br />

53<br />

53<br />

65<br />

47<br />

47<br />

35<br />

20<br />

18<br />

20<br />

65<br />

76<br />

69<br />

15<br />

6<br />

11


118 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

The specific UFC-TTHM yields fortheunfractionated, humic, andnonhumic<br />

fractions <strong>of</strong> ORW are shown in Figure 7.21. Results shown are for raw water, both<br />

RSSCT-CT and RSSCT-OC influents, and RSSCT-OC effluent. The RSSCT<br />

effluent sample points (A, B, and C) correspond to those shown in Figure 7.20. As<br />

was stated previously, because <strong>of</strong> differences in the Br~:DOC ratio among different<br />

sample points, the specific UFC-TTHM yields are not comparable at different<br />

locations and times. For a given sample point, though, the results can be compared.<br />

The specific UFC-TTHM yields <strong>of</strong> the humic and nonhumic fraction <strong>of</strong> any given<br />

sample point or time were similar to each other and to the unfractionated sample.<br />

Thus, for ORW, both the humic and nonhumic fraction contributed equally as<br />

precursors to UFC-TTHM formation, and this was not affected <strong>by</strong> optimized<br />

coagulation or <strong>GAC</strong> treatment.<br />

Figure 7.22 shows the DOC results for humic-nonhumic fractionation in<br />

raw, conventionally treated, and optimized coagulated SRPW. RSSCT-OC effluent<br />

fractionation data are not presented as a result <strong>of</strong> analytical difficulties that affected<br />

all three samples. The raw SRPW DOC was 53 percent humic and 47 percent<br />

nonhumic, as shown in Table 7.6. No change in the relative composition <strong>of</strong> the<br />

SRPW was seen after conventional treatment. Optimized coagulation preferentially<br />

removed the nonhumic fraction, resulting in a 65 percent humic and 35 percent<br />

nonhumic makeup. The breakthrough behavior <strong>of</strong> the humic and nonhumic frac<br />

tions in the RSSCT-CT effluent was similar to that observed for ORW; this behavior<br />

was discussed in Chapter 5, where it was shown that the humic fraction was well<br />

adsorbed while the nonhumic fraction showed high levels <strong>of</strong> immediate break<br />

through.<br />

Molecular Size Fractionation<br />

The effect <strong>of</strong> treatment on the molecular size distribution <strong>of</strong> ORW as<br />

measured <strong>by</strong> DOC is shown in Figure 7.23. The molecular size distributions for all<br />

samples are summarized in Table 7.6. The raw sample <strong>of</strong> ORW was dominated <strong>by</strong><br />

the 500-3,000 MS fraction, with the remaining DOC approximately evenly divided<br />

between 3,000 MS fractions. With both conventional and optimized<br />

coagulation, virtually all the >3,000 MS fraction was removed. For conventional<br />

treatment, there was no removal <strong>of</strong> either the


The Impact <strong>of</strong> Optimized Coagulation 119<br />

The specific UFC-TTHM yields for ORW molecular size fractions are<br />

shown in Figure 7.24 for the raw water and RSSCT-OC influent and effluent<br />

samples. The specific yield <strong>of</strong> the


720 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

Table 7.7 PY-GC-MS classification <strong>of</strong> ORW: Enhanced coagulation<br />

Percent based upon peak height<br />

Stage<br />

Aliphatic<br />

Aromatic<br />

Nitrogen<br />

containing<br />

Halogen<br />

substituted<br />

Unknown<br />

Raw<br />

<strong>GAC</strong> influent<br />

Nonadsorbable<br />

Intermediate breakthrouogh<br />

Advanced breakthrough<br />

34.5<br />

2.5<br />

15.6<br />

55.6<br />

51.1<br />

15.8<br />

BDL<br />

BDL<br />

0.1<br />

BDL<br />

35.9<br />

16.3<br />

0.3<br />

2.2<br />

1.9<br />

13.9<br />

66.5<br />

81<br />

37.5<br />

46.2<br />

BDL<br />

14.7<br />

3.1<br />

4.6<br />

0.9<br />

BDL = below detection limit<br />

peaks <strong>of</strong> acetaldehyde, acetic acid, formic acid, and trichlorobenzene, was very<br />

similar in general character to that observed in the case <strong>of</strong> conventional pretreatment.<br />

As shown in Table 7.8, the pyrolysis fingerprint <strong>of</strong> the enhanced coagulated<br />

SRPW, was very similar to that produced <strong>by</strong> conventional coagulation (discussed<br />

in Chapter 5). The major peak was acetic acid, with minor peaks <strong>of</strong> furfural,<br />

propanoic acid, and methyl furfural. The chemical signature <strong>of</strong> the nonadsorbable<br />

NOM was similar to that <strong>of</strong> the influent, with a predominant acetic acid fragment<br />

produced. The organic matrix <strong>of</strong> the second breakthrough point was characterized<br />

<strong>by</strong> a strong acetic acid and methyl furfural and also a number <strong>of</strong> peaks that could not<br />

be identified. The complete breakthrough sample was more similar to the raw water.<br />

This fingerprint contained strong acetic acid, methyl furfural, propanoic acid, and<br />

furfural.<br />

Summary_________________________<br />

Optimized coagulation improved <strong>GAC</strong> run times <strong>by</strong> 129 to 148 percent over<br />

conventional pretreatment for the two waters studied, as measured <strong>by</strong> breakthrough<br />

behavior <strong>of</strong> NOM surrogates and DBF formation. Although the optimized coagu<br />

lation alum doses used were high, the findings indicate the potential benefits <strong>of</strong><br />

increasing coagulant doses prior to <strong>GAC</strong> adsorption. In addition, optimized coagu<br />

lation pretreatment significantly decreased specific DBF yield in the <strong>GAC</strong> effluent.<br />

The adsorbability <strong>of</strong> the NOM fractions as measured <strong>by</strong> XAD-8 fractionation and<br />

ultrafiltration was not affected <strong>by</strong> coagulation conditions.<br />

Because pretreatment conditions did not affect NOM characteristics and<br />

adsorbability, the significant improvement in <strong>GAC</strong> performance with optimized<br />

coagulation may be attributed to lower influent concentrations and lower influent<br />

pH values. It is difficult to separate these effects in a comparison <strong>of</strong> <strong>GAC</strong><br />

performance under two coagulation conditions for these waters.<br />

For the two waters examined, optimized coagulation removed a fraction <strong>of</strong><br />

the TOC present after conventional treatment. For ORW, both humic and nonhumic<br />

fractions were affected about equally, while for SRPW the nonhumic fraction was<br />

preferentially removed. Other research has shown that the humic fraction is usually<br />

preferentially removed over the nonhumic fraction <strong>by</strong> coagulation (Semmens and<br />

Staples 1986; Collins et al. 1986; Dryfuse et al. 1995). MS fractionation studies


The Impact <strong>of</strong> Optimized Coagulation 121<br />

Table 7.8 PY-GC-MS classification <strong>of</strong> SRPW: Enhanced coagulation<br />

Percent based upon peak height<br />

Stage<br />

Aliphatic<br />

Aromatic<br />

Nitrogen<br />

containing<br />

Halogen<br />

substituted<br />

Unknown<br />

Raw<br />

<strong>GAC</strong> influent<br />

Nonadsorbable<br />

Intermediate breakthrough<br />

Advanced breakthrough<br />

57.9<br />

66.9<br />

87.1<br />

49.2<br />

59.7<br />

9.1<br />

25.3<br />

1.8<br />

6.6<br />

17.0<br />

9.2<br />

BDL<br />

5.7<br />

4.3<br />

2.6<br />

13.2<br />

BDL<br />

4.1<br />

10.4<br />

BDL<br />

10.5<br />

7.8<br />

1.2<br />

29.5<br />

20.7<br />

BDL = below detection limit<br />

showed that the 500-3,000 MS fraction was reduced after optimized coagulation for<br />

both waters, while the


722 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

2.25<br />

Conventional treatment influent<br />

Optimized coagulation influent<br />

0 Conventional treatment effluent<br />

• Optimized coagulation effluent<br />

50 100 150 200<br />

Scaled operation time (days)<br />

250 300<br />

Figure 7.1 Effect <strong>of</strong> coagulation pretreatment on TOC breakthrough for Ohio River<br />

water<br />

Conventional treatment influent 0.044<br />

Optimized coagulation influent 0.031<br />

° Conventional treatment effluent<br />

• Optimized coagulation effluent<br />

0.000<br />

100 150 200<br />

Scaled operation time (days)<br />

Figure 7.2 Effect <strong>of</strong> coagulation pretreatment on UV254 breakthrough for Ohio River<br />

water


The Impact <strong>of</strong> Optimized Coagulation 123<br />

Conventional treatment influent 86.1<br />

Optimized coagulation influent 58.8<br />

0 Conventional treatment effluent<br />

• Optimized coagulation effluent<br />

50 100 150 200<br />

250 300<br />

Scaled operation time (days)<br />

Figure 7.3 Effect <strong>of</strong> coagulation pretreatment on TTHM breakthrough for Ohio River<br />

water<br />

Conventional treatment influent 25.5<br />

Optimized coagulation influent 18.8<br />

0 Conventional treatment effluent<br />

* Optimized coagulation effluent<br />

50 100 150 200<br />

250 300<br />

Scaled operation time (days)<br />

Figure 7.4 Effect <strong>of</strong> coagulation pretreatment on HAA6 breakthrough for Ohio River<br />

water


124 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

O<br />

o><br />

120<br />

100 -<br />

80 -<br />

60<br />

.0<br />

I c 40 H<br />

8<br />

§<br />

O 20 -\<br />

TOX<br />

Conventional treatment influent<br />

Optimized coagulation influent<br />

0 Conventional treatment effluent<br />

• Optimized coagulation effluent<br />

C0 (ug CI-/L).<br />

170<br />

92<br />

50 100 150 200<br />

250 300<br />

Scaled operation time (days)<br />

Figure 7.5 Effect <strong>of</strong> coagulation pretreatment on TOX breakthrough for Ohio River<br />

water<br />

o.u -<br />

•<br />

4.5 -<br />

4.0 -<br />

.<br />

3.5 -<br />

^j<br />

°> 3.0 -<br />

I 2.5-<br />

| 2.0-<br />

0)<br />

c 1-5 -<br />

o<br />

0 1.0-<br />

0.5 -<br />

CH<br />

1 .,,.,..,. 1 .... 1 ...,,,,.,<br />

Conventional treatment influent 3.9<br />

Optimized coagulation influent 1.4<br />

0 Conventional treatment effluent<br />

• Optimized coagulation effluent<br />

C0 (M9/L) -<br />

' >> -<br />

-<br />

-<br />

-<br />

0.0 -<br />

.,., I ..,, I ,.,,,,,,.,.,,,,.,.,<br />

() 50 100 150 200 250 31DO<br />

Scaled operation time (days)<br />

Figure 7.6 Effect <strong>of</strong> coagulation pretreatment on CH breakthrough for Ohio River water


The Impact <strong>of</strong> Optimized Coagulation 125<br />

2.25<br />

2.00 -<br />

1.75 -<br />

1.50 -<br />

TOC<br />

Conventional treatment influent<br />

Optimized coagulation influent<br />

Conventional treatment effluent<br />

Optimized coagulation effluent<br />

C0 (mg/L)<br />

2.2<br />

1.7<br />

50 100 150 200<br />

250 300<br />

Scaled operation time (days)<br />

Figure 7.7 Effect <strong>of</strong> coagulation pretreatment on TOC breakthrough for Salt River<br />

Project water<br />

Conventional treatment influent 0.047<br />

Optimized coagulation influent 0.028<br />

Conventional treatment effluent<br />

Optimized coagulation effluent<br />

0.000<br />

50 100 150 200<br />

Scaled operation time (days)<br />

250 300<br />

Figure 7.8 Effect <strong>of</strong> coagulation pretreatment on UV254 breakthrough for Salt River<br />

Project water


126 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

90<br />

80<br />

70<br />

60<br />

TTHM C0 (ug/L)<br />

Conventional treatment influent 71.5<br />

Optimized coagulation influent 45.8<br />

Conventional treatment effluent<br />

Optimized coagulation effluent<br />

50<br />

*•/<br />

I<br />

0)<br />

o<br />

§O<br />

40<br />

30<br />

20<br />

10<br />

0<br />

100 150 200<br />

Scaled operation time (days)<br />

250 300<br />

Figure 7.9 Effect <strong>of</strong> coagulation pretreatment on TTHM breakthrough for Salt River<br />

Project water<br />

25<br />

20 -<br />

15 -<br />

HAA8<br />

CQ (ug/p<br />

Conventional treatment influent 28.3<br />

Optimized coagulation influent 16.7<br />

° Conventional treatment effluent<br />

* Optimized coagulation effluent<br />

I<br />

10 -<br />

o<br />

O<br />

—i—<br />

50<br />

100 150 200<br />

Scaled operation time (days)<br />

250 300<br />

Figure 7.10 Effect <strong>of</strong> coagulation pretreatment on HAA6 breakthrough for Salt River<br />

Project water


The Impact <strong>of</strong> Optimized Coagulation 127<br />

140<br />

120 -<br />

100-<br />

TOX<br />

Conventional treatment influent<br />

Optimized coagulation influent<br />

° Conventional treatment effluent<br />

* Optimized coagulation effluent<br />

C0 (ug CI-/L)<br />

170<br />

130<br />

50 100 150 200<br />

250 300<br />

Scaled operation time (days)<br />

Figure 7.11 Effect <strong>of</strong> coagulation pretreatment on TOX breakthrough for Salt River<br />

Project water<br />

5.0<br />

4.5 -<br />

4.0 -<br />

3.5 -<br />

g> 3.0 -<br />

CH C0 (ug/L)<br />

Conventional treatment influent 4.1<br />

Optimized coagulation influent 2.6<br />

0 Conventional treatment effluent<br />

• Optimized coagulation effluent<br />

C •) r I<br />

o t.O<br />

£ 2.0 H<br />

8 15 ,<br />

C '-5 n<br />

O<br />

0 1.0 H<br />

50 100 150 200<br />

250 300<br />

Scaled operation time (days)<br />

Figure 7.12 Effect <strong>of</strong> coagulation pretreatment on CH breakthrough for Salt River<br />

Project water


128 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> GA C <strong>Adsorption</strong><br />

100<br />

s<br />

o<br />

I 8<br />


The Impact <strong>of</strong> Optimized Coagulation 129<br />

O)<br />

20<br />

16 -<br />

HAA6<br />

00=1.3<br />

a, = 11.5<br />

Pi » 7.3<br />

i——'——i——>——i——'——i———"~<br />

—e-- Conventional treatment (OQ, a,)<br />

—•—Optimized coagulation ((30 , p,)<br />

8<br />

, 8H<br />

4 -<br />

0.0 0.2 0.4 0.6 0.8 1.0 1.2<br />

1.4<br />

TOC concentration (mg/L)<br />

Figure 7.15 Effect <strong>of</strong> coagulation pretreatment on the relationship between TOC and<br />

HAA6 formation during <strong>GAC</strong> breakthrough for Ohio River water<br />

iuu -<br />

~ 8° "<br />

3"<br />

b<br />

I 60-<br />

?<br />

* 1 ' 1 ' 1 ' 1 • 1 ' 1 '<br />

TOX o Conventional treatment (a0 , a.,) .<br />

• Optimized coagulation (p0, p,)<br />

a0 = -24.9<br />

o<br />

a, = 90.8<br />

0^31.0 °<br />

•<br />

| 4°o<br />

O 20 -<br />

• o<br />

• *<br />

n -<br />

• o<br />

0<br />

0.0 0.2 0.4 0.6 0.8 1.0<br />

1.2 1.4<br />

TOC concentration (mg/L)<br />

Figure 7.16 Effect <strong>of</strong> coagulation pretreatment on the relationship between TOC and<br />

TOX formation during <strong>GAC</strong> breakthrough for Ohio River water


130 <strong>Removal</strong> <strong>of</strong><strong>DBP</strong> <strong>Precursors</strong> <strong>by</strong> GA C <strong>Adsorption</strong><br />

50 - ——— i ——— | ——— i ——— i ——— > ——— i —— i ——— i — • — i • i<br />

1<br />

I<br />

§<br />

3<br />

40 -<br />

30 -<br />

20 -<br />

10 -<br />

TTHM o Conventional treatment (OQ, a,) .<br />

• Optimized coagulation (P0, p,)<br />

oto = P 0 = -3.05<br />

a, = 31.9 °<br />

P, = 24.6 °<br />

0 •<br />

0 »<br />

o<br />

0 •<br />

•<br />

n -<br />

0.0 0.2 0.4 0.6 0.8 1.0<br />

TOC concentration (mg/L)<br />

1.2 1.4<br />

Figure 7.17 Effect <strong>of</strong> coagulation pretreatment on the relationship beween TOC and<br />

TTHM formation during <strong>GAC</strong> breakthrough for Salt River Project water<br />

«u<br />

15 -<br />

s<br />

?<br />

| 10-<br />

c<br />

8 co<br />

o 5-<br />

HAA6 o Conventional treatment (OQ, a,)<br />

• Optimized coagulation (P0, p,)<br />

OQ = p0 = 1 .2<br />

a, = 8.7<br />

P, = 6-9 ° .<br />

o<br />

m •<br />

o<br />

QQ ^^^<br />

o •<br />

0 0<br />

•<br />

•<br />

-<br />

n -<br />

0.0 0.2 0.4 0.6 0.8 1.0<br />

1.2 1.4<br />

TOC concentration (mg/L)<br />

Figure 7.18 Effect <strong>of</strong> coagulation pretreatment on the relationship between TOC and<br />

HAA6 formation during <strong>GAC</strong> breakthrough for Salt River Project water


The Impact <strong>of</strong> Optimized Coagulation 131<br />

100 -<br />

o Conventional treatment (OQ, a.,)<br />

• Optimized coagulation (P0 , p,)<br />

o<br />

at<br />

80 -<br />

Og = Po = -0.6<br />

a, = p, = 56.9<br />

o 60<br />

|<br />

| 40<br />

o<br />

O<br />

20 -J<br />

0.0 0.2 0.4 0.6 0.8 1.0 1.2<br />

TOC concentration (mg/L)<br />

1.4<br />

Figure 7.19 Effect <strong>of</strong> coagulation pretreatment on the relationship between TOC and<br />

TOX formation during <strong>GAC</strong> breakthrough for Salt River Project water<br />

Raw Conventional Optimized 0 50 100 150 200 250<br />

treatment coagulation<br />

Scaled operation time (days)<br />

Note: A, B, and C denote initial, intermediate, and advanced breakthrough points.<br />

Figure 7.20 Effect <strong>of</strong> coagulation pretreatment on humic and nonhumic fraction<br />

breakthrough for Ohio River water


132 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> GA C <strong>Adsorption</strong><br />

Unfractionated<br />

Nonhumic fraction<br />

Humic fraction<br />

Raw Conventional Optimized<br />

treatment coagulation<br />

(<strong>GAC</strong> influent)<br />

<strong>GAC</strong> effluent samples<br />

Note: A, B, and C denote initial, intermediate, and advanced breakthrough points. NA = not analyzed<br />

Figure 7.21 Effect <strong>of</strong> coagulation pretreatment on the specific UFC-TTHM yields <strong>of</strong><br />

humic-nonhumic fractions for Ohio River water<br />

2.5<br />

1 • i • ' ' ' i ' ' > ' i ' •<br />

Unfractionated<br />

Nonhumic fraction<br />

Humic fraction<br />

Raw Conventional Optimized 0<br />

treatment coagulation<br />

50 100 150 200 250 300<br />

Scaled operation time (days)<br />

Note: Effluent sample fractionation not performed.<br />

Figure 7.22 Effect <strong>of</strong> coagulation pretreatment on humic and nonhumic fraction<br />

breakthrough for Salt River Project water


The Impact <strong>of</strong> Optimized Coagulation 133<br />

en<br />

2.5<br />

Unfractionated<br />

3000 MS fraction<br />

1.5 -<br />

8<br />

1.0 H<br />

0.5 -<br />

0.0<br />

Raw Conventional Optimized Q<br />

• treatment coagulation<br />

100 150 200 250<br />

Scaled operation time (days)<br />

Note: A, B, and C denote initial, intermediate, and advanced breakthrough points.<br />

Figure 7.23 Effect <strong>of</strong> coagulation pretreatment on molecular size fraction breakthrough<br />

for Ohio River water<br />

o><br />

100<br />

80 -<br />

452<br />

IHH Unfractionated<br />

< 500 MS fraction<br />

I I < 3000 MS fraction<br />

t<br />

o><br />

2 40 -<br />

o<br />

60-<br />

20-<br />

Bf:DOC 257<br />

(ng/mg)<br />

427<br />

Raw Conventional Optimized A<br />

treatment coagulation _._ „ . .<br />

(<strong>GAC</strong> influent) <strong>GAC</strong> effluent samples<br />

Note: A, B, and C denote initial, intermediate, and advanced breakthrough points.<br />

Figure 7.24 Effect <strong>of</strong> coagulation pretreatment on the specific UFC-TTHM yields <strong>of</strong><br />

molecular size fractions for Ohio River water


134 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

2.5<br />

~ 2.0-<br />

Unfractionated<br />

3000 MS fraction<br />

1.5 -<br />

O<br />

0.5 -<br />

0.0<br />

Raw Conventional Optimized 0<br />

treatment coagulation<br />

50 100 150 200 250 300<br />

Scaled operation time (days)<br />

Note: A, B, and C denote initial, intermediate, and advanced breakthrough points.<br />

Figure 7.25 Effect <strong>of</strong> coagulation pretreatment on molecular size fraction breakthrough<br />

for Salt River Project water<br />

Unfractionated<br />

__ < 500 MS fraction<br />

I I < 3000 MS fraction<br />

Raw Conventional Optimized<br />

treatment coagulation<br />

(<strong>GAC</strong> influent)<br />

A B<br />

<strong>GAC</strong> effluent samples<br />

Note: A, B, and C denote initial, intermediate, and advanced breakthrough points.<br />

Figure 7.26 Effect <strong>of</strong> coagulation pretreatment on the specific UFC-TTHM yields <strong>of</strong><br />

molecular size fractions for Salt River Project water


Chapter 8___________<br />

The Impact <strong>of</strong> Ozonation and<br />

Biological Filtration on NOM<br />

<strong>Removal</strong> and DBF Control <strong>by</strong> GAG<br />

The adsorption capacity <strong>of</strong> <strong>GAC</strong> is affected <strong>by</strong> the influent concentration<br />

and chemical nature <strong>of</strong> the water being treated. For natural waters, in general, lower<br />

initial concentrations yield longer <strong>GAC</strong> adsorption capacities and therefore longer<br />

run times to a given effluent concentration. Changes in composition jand organic<br />

matter structure, however, can also affect adsorption capacity.<br />

One <strong>GAC</strong> pretreatment option that may potentially improve adsorption<br />

behavior is ozonation with biological filtration. Ozonation <strong>of</strong> the NOM causes a<br />

shift to lower-molecular-weight compounds that have lower UV absorbance and are<br />

more easily biodegradable. Microorganisms attached to biologically active filters<br />

can then utilize some <strong>of</strong> these biodegradable compounds, reducing the overall NOM<br />

concentration as well as biostabilizing the water. Typically, the adsorbability <strong>of</strong><br />

NOM decreases with ozonation because <strong>of</strong> the creation <strong>of</strong> more polar, hydrophilic<br />

compounds. Sontheimer et al. (1988) showed that biological filtration can compen<br />

sate for the negative effect <strong>of</strong> ozonation on adsorption <strong>by</strong> removing these weakly<br />

adsorbing compounds. Summers and Crittenden (1989) showed that ozonation and<br />

biological filtration reduced the organic concentration in filtered ORW and in<br />

creased the run time <strong>of</strong> a <strong>GAC</strong> contactor. The amount <strong>of</strong> reduction in NOM resulting<br />

from ozonation and biological filtration is a function <strong>of</strong> the ozone dose and the water<br />

being treated. Miltner and Summers (1992) reported a 20 to 40 percent TOC<br />

removal for ORW after ozonation and biological filtration, while Joselyn and<br />

Summers (1992) reported only a 15 percent TOC removal in settled ORW.<br />

Metz et al. (1993) compared <strong>GAC</strong> performance <strong>of</strong> ozonated, biologically<br />

filtered, ozonated and biologically filtered, and conventionally treated ORW. The<br />

case <strong>of</strong> ozonated and biologically filtered water produced the lowest DBF yields<br />

after <strong>GAC</strong> treatment for a given run time.<br />

Objectives and Approach________________<br />

The objective <strong>of</strong> the work presented in this chapter was to evaluate the<br />

impact <strong>of</strong> ozonation and subsequent biological filtration on NOM removal and DBF<br />

control <strong>by</strong> <strong>GAC</strong>. Conventionally treated water was ozonated and passed through an<br />

acclimated bi<strong>of</strong>ilter before being treated <strong>by</strong> <strong>GAC</strong>, as discussed in Chapter 2. <strong>GAC</strong><br />

135


136 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

performance was monitored in terms <strong>of</strong> TOC and UV,54 , and DBF removal was<br />

assessed in terms <strong>of</strong> UFC-TOX, -TTHM, -HAA6, and -CH. The impact <strong>of</strong> ozonation<br />

and biological filtration on NOM characteristics was investigated in terms <strong>of</strong> the<br />

effects on humic and nonhumic composition, molecular size fractions, biodegradability,<br />

and PY-GC-MS results.<br />

Experimental Conditions________________<br />

The impact <strong>of</strong> ozonation and biological filtration as pretreatment before<br />

<strong>GAC</strong> was investigated for ORW, PRW, and FGW. Conventionally treated water<br />

was passed through a bench-scale ozonation system with a contact time <strong>of</strong> 7 minutes.<br />

Details <strong>of</strong> the system are presented in Chapter 2. Table 8.1 summarizes the<br />

experimental conditions for all three waters. The ozone dose reported is the ratio <strong>of</strong><br />

transferred ozone to DOC in mg/mg. The BDOC after ozonation was approximately<br />

10 percent <strong>of</strong> the TOC for both PRW and FGW, and it was 20 percent <strong>of</strong> the TOC<br />

for ORW.<br />

After ozonation the water was passed through a biologically active sand<br />

filter. The sand filter was designed to remove the quickly biodegradable fraction and<br />

had an EBCT <strong>of</strong> 5 minutes. The biologically active sand was acclimated with the<br />

water to be treated prior to the run, as discussed in Chapter 2. The TOC removals<br />

after biological filtration were 33, 26, and 17 percent for ORW, FGW, and PRW,<br />

respectively. After biological filtration, the water was filtered through a 1-um<br />

cartridge and treated <strong>by</strong> <strong>GAC</strong>.<br />

Impact <strong>of</strong> Pretreatment on <strong>GAC</strong> Breakthrough<br />

Behavior: Results and Discussion___________<br />

The impact <strong>of</strong> ozonation and biological filtration on <strong>GAC</strong> breakthrough<br />

behavior was compared for ORW, PRW, and FGW. Table 8.2 summarizes the effect<br />

<strong>of</strong> ozonation and biological filtration on <strong>GAC</strong> influent. The <strong>GAC</strong> influent TOC<br />

concentrations were not significantly affected <strong>by</strong> ozonation alone. After biological<br />

filtration, however, they decreased 17 to 33 percent. The breakthrough behavior was<br />

compared using the concentration <strong>of</strong> immediate breakthrough, C imm , and the time<br />

to initial breakthrough, tint, as well as run times to the effluent criteria described in<br />

Chapter 6. The terms Cimm and tint represent the nonadsorbable concentration and<br />

the time at which the adsorbable fraction begins to break through, respectively.<br />

Table 8.3 summarizes the comparisons between Cimm , tjnt , and run times to effluent<br />

criteria for all three waters. The results for CH are not reported because concentra<br />

tions were <strong>of</strong>ten at or below the detection limit.<br />

Figure 8.1 shows the impact <strong>of</strong> ozonation and biological filtration on<br />

normalized TOC breakthrough for ORW. Because the influent concentrations were<br />

different, the normalized concentration was used to facilitate comparisons <strong>of</strong><br />

breakthrough behavior. Ozonaticn and biological filtration reduced the influent<br />

TOC concentration from 2.1 to 1.4 mg/L. Based on the lower influent concentration,<br />

the column treating the ozonated and biologically filtered water would be expected<br />

to break through later. However, for ORW the ozonated and biologically filtered<br />

water initially broke through more quickly. After 40 percent breakthrough, both


The Impact <strong>of</strong> Ozonation and Biological Filtration 137<br />

Table 8.1 Summary <strong>of</strong> experimental conditions for ozonation and biological filtration<br />

<strong>Water</strong><br />

After<br />

conventional<br />

TOC<br />

(mg/L)<br />

UV254<br />

(1/cm)<br />

Ozone dose<br />

(mg Cymg TOC)<br />

TOC<br />

(mg/L)<br />

After<br />

ozonation<br />

uv 254<br />

(1/cm)<br />

BDOC<br />

(mg/L)<br />

TOC<br />

(mg/L)<br />

After<br />

biological filtration<br />

UV254<br />

(1/cm)<br />

BDOC<br />

(mg/L)<br />

ORW.<br />

PRW<br />

FGW<br />

2.1<br />

3.0<br />

10.0<br />

0.048<br />

0.076<br />

0.29<br />

1.04<br />

1.58<br />

1.03<br />

2.0<br />

2.8<br />

9.8<br />

0.015<br />

0.041<br />

0.11<br />

0.4<br />

0.3<br />

0.8<br />

1.4<br />

2.5<br />

7.4<br />

0.016<br />

0.038<br />

0.10<br />

BDL<br />

BDL<br />

BDL<br />

BDL = below detection limit<br />

Table 8.2 Summary <strong>of</strong> <strong>GAC</strong> influent characteristics after ozonation and biological<br />

filtration<br />

TOC<br />

(mg/L)<br />

UV254<br />

(1/cm)<br />

TOX<br />

(M9 CL-/L)<br />

TTHM<br />

(H9/L)<br />

HAA6<br />

(P9/L)<br />

CH<br />

(M9/L)<br />

<strong>Water</strong><br />

Conv.<br />

Oj/bio<br />

Conv.<br />

O3/bio<br />

Conv.<br />

Oj/bio<br />

Conv.<br />

03/bio<br />

Conv.<br />

O^bio<br />

Conv.<br />

O3/bio<br />

ORW<br />

PRW<br />

FGW<br />

2.1<br />

3.0<br />

10.0<br />

1.4<br />

2.5<br />

7.4<br />

0.048<br />

0.076<br />

0.29<br />

0.016<br />

0.038<br />

0.10<br />

170<br />

277<br />

1,046<br />

56<br />

135<br />

645<br />

86<br />

73<br />

238<br />

38<br />

57<br />

205<br />

26<br />

70<br />

142<br />

13<br />

31<br />

102<br />

3.9<br />

15<br />

7.8<br />

0.5<br />

11<br />

9.5<br />

Conv. = conventional treatment<br />

O 3/bio = ozonation and biological filtration<br />

columns performed similarly. This behavior indicates that after ozonation and<br />

biological filtration, a higher percentage <strong>of</strong> the TOC was composed <strong>of</strong> more polar<br />

and less adsorbable compounds, which were created during ozonation but were not<br />

removed <strong>by</strong> biological filtration.<br />

Figure 8.2 compares the absolute effluent TOC concentrations from the<br />

columns treating conventionally treated ORW and ozonated and biologically<br />

filtered ORW. The effluent TOC concentrations <strong>of</strong> both columns were similar up<br />

to 60 scaled operation days, after which the column treating the ozonated and<br />

biologically filtered water performed better. The run time to an effluent TOC<br />

criterion <strong>of</strong> 0.8 mg/L was increased from 80 to 105 days <strong>by</strong> pretreatment. As<br />

indicated in Table 8.3, the nonadsorbable TOC concentration remained the same,<br />

but as a fraction <strong>of</strong> total TOC it increased from 9.8 to 14 percent. The time to initial<br />

breakthrough was reduced from 45 to 35 days. Figure 8.3 shows the breakthrough<br />

behavior <strong>of</strong> UV254 . The nonadsorbable fraction was lower for the ozonated and<br />

biologically filtered water. Between 30 and 90 scaled operation days, both columns<br />

behaved similarly. After 90 days, the column treating ozonated and biologically<br />

filtered water performed better. The tint decreased from 50 to 10 days for UV254 . It<br />

is recognized, however, that comparisons between conventional treatment and<br />

ozonation with biological filtration are complicated <strong>by</strong> ozonation's destruction <strong>of</strong><br />

a water's capacity for adsorbing UV light at 254 nm.


120<br />

62<br />

65<br />

90<br />

95<br />

115<br />

9<br />

7.5<br />

1<br />

10<br />

9<br />

5<br />

30<br />

15<br />

112<br />

40<br />

49<br />

48<br />

105<br />

>150<br />

17<br />

13<br />

12<br />

6<br />

27<br />

15<br />

>40<br />

21<br />

31<br />

>20<br />

>25<br />

>70<br />

—<br />

-67<br />

-35<br />

-25<br />

-47<br />

11<br />

>30<br />

89<br />

73<br />

1,100<br />

-40<br />

200<br />

200<br />

>100<br />

40<br />

~<br />

i ciuie o.«a ouuiiinai y ui in ipai*i wi u'^wiiauwn anu uivsiwyivai mil auuii uu vmo pel i ui mam<br />

<strong>Water</strong> Parameter<br />

ORW TOC<br />

• uv254<br />

TOX<br />

TTHM<br />

HAA6<br />

Units<br />

mg/L<br />

1/cm<br />

M9 CI-/L<br />

Mg/L<br />

Mg/L<br />

Cone.<br />

0.2<br />

0.003<br />

3<br />

4<br />

5<br />

Immediate breakthrough<br />

concentration, C lmm<br />

Conv.<br />

%lnf<br />

9.8<br />

6.8<br />

3.5<br />

15<br />

2.9<br />

Cone.<br />

0.2<br />

BDL<br />

10<br />

4<br />

5<br />

Cybio<br />

%lnf<br />

14<br />

—<br />

18<br />

11<br />

39<br />

Conv.<br />

(days)<br />

45<br />

50<br />

45<br />

30<br />

50<br />

Time to initial<br />

breakthrough, t<br />

O-/bio<br />

(days)<br />

35<br />

10<br />

85<br />

—<br />

—<br />

•Int<br />

Change<br />

-22<br />

-80<br />

89<br />

—<br />

— .<br />

rC<br />

Criterion<br />

0.8<br />

0.01<br />

70<br />

32<br />

24<br />

Run time based on<br />

established criterion<br />

Conv.<br />

(days)<br />

80<br />

125<br />

120<br />

88<br />

>150<br />

O3/bio Change<br />

(days) (%)<br />

105<br />

>150<br />

>150<br />

>150<br />

>150<br />

Oo<br />

<strong>of</strong><strong>DBP</strong> <strong>Removal</strong><br />

"0<br />

s<br />

3o<br />

•i"<br />

d<br />

0<br />

1 o<br />

•3<br />

PRW TOC<br />

UV254<br />

TOX<br />

TTHM<br />

HAA6<br />

mg/L<br />

1/cm<br />

Mg CI-/L<br />

Mg/L<br />

Mg/L<br />

0.2<br />

0.003<br />

10<br />

5<br />

7.5<br />

6.7<br />

3.5<br />

3.6<br />

6.8<br />

11<br />

0.3<br />

0.005<br />

10<br />

5<br />

7.5<br />

12<br />

13<br />

7.4<br />

8.8<br />

24<br />

50<br />

45<br />

55<br />

55<br />

75<br />

20<br />

20<br />

40<br />

20<br />

20<br />

-60<br />

-56<br />

-27<br />

-64<br />

-73<br />

1.6<br />

0.8<br />

0.01<br />

70<br />

32<br />

24<br />

FGW TOC<br />

UV254<br />

TOX<br />

TTHM<br />

HAA6<br />

mg/L<br />

1/cm<br />

Mg CI-/L<br />

Mg/L<br />

Mg/L<br />

0.6<br />

0.01<br />

25<br />


The Impact <strong>of</strong>Ozonation and Biological Filtration 139<br />

Figures 8.4,8.5, and 8.6 compare the breakthrough behavior <strong>of</strong> UFC-TOX,<br />

-TTHM, and -HAA6 for ORW. The effluent concentrations in both columns were<br />

similar initially for all three DBFs; however, the ozonated and biologically filtered<br />

column performed significantly better than the column to which conventionally<br />

treated water was applied. The nonadsorbable concentration <strong>of</strong> TTHM and HAA6<br />

precursors remained the same, while that for TOX precursors increased from 3 to<br />

10 ug/L after ozonation and biological filtration. The time to initial breakthrough<br />

increased for TOX precursors, decreased for TTHM precursors, and remained the<br />

same for HAA6 precursors. The effluent concentrations <strong>of</strong> the column treating<br />

ozonated and biologically filtered water did not reach the effluent criteria used in<br />

this study; in fact, they had not reached 50 percent <strong>of</strong> the effluent criteria after 150<br />

days.<br />

Figure 8.7 compares normalized TOC breakthrough behavior for PRW.<br />

Ozonation and biological filtration resulted in a reduction <strong>of</strong> influent TOC from 3.0<br />

to 2.5 mg/L. TOC in the column treating ozonated and biologically filtered water<br />

broke through more rapidly than in the column treating conventionally treated<br />

water. Although the influent TOC concentration was decreased, ozonation and<br />

biological filtration affected the nature <strong>of</strong> the NOM, which could have resulted in<br />

earlier breakthrough. After approximately 65 percent breakthrough, the normalized<br />

effluents from both columns were similar. Figure 8.8 compares the effluent TOC<br />

concentrations on an absolute basis. The column treating ozonated and biologically<br />

filtered water broke through earlier. However, after 80 scaled operation days, the<br />

columns showed similar behavior. The TOC tint decreased from 50 to 20 days as a<br />

result <strong>of</strong> ozonation and biological filtration. The nonadsorbable concentration<br />

increased <strong>by</strong> 50 percent with ozonation and biological filtration. As with ORW, this<br />

adsorption behavior indicates that ozonation created polar, less adsorbable com<br />

pounds that were not entirely removed <strong>by</strong> biological filtration. These compounds<br />

led to earlier breakthrough. The run time to an effluent TOC criterion <strong>of</strong> 1.6 mg/L<br />

decreased from 120 to 112 scaled operation days with ozonation and biological<br />

filtration. The run time to aTOC criterion <strong>of</strong> 0.8 mg/L decreased from 62 to 50 scaled<br />

operation days. The UV254 results for PRW effluent are shown in Figure 8.9. The<br />

column treating ozonated and biologically filtered water broke through earlier.<br />

However, after 70 scaled operation days, the performance <strong>of</strong> this column was better<br />

than the column representing conventional treatment. The run time to an effluent<br />

criterion <strong>of</strong> 0.01/cm decreased from 65 to 49 scaled operation days, although this<br />

was likely significantly affected <strong>by</strong> the destruction <strong>of</strong> UV-absorbing compounds<br />

during ozonation prior to <strong>GAC</strong> treatment.<br />

The impact <strong>of</strong> ozonation and biological filtration on DBF control for PRW<br />

is shown in Figures 8.10, 8.11, and 8.12 for UFC-TOX, -TTHM and -HAA6,<br />

respectively. The behavior <strong>of</strong> the three formed DBFs was similar to that <strong>of</strong> UV254 .<br />

The column treating ozonated biologically filtered water initially broke through<br />

earlier than the column treating conventionally treated water, but after 80 to 100<br />

days the column treating ozonated and biologically filtered water showed better<br />

DBF removal. The nonadsorbable concentrations <strong>of</strong> TOX, TTHM, and HAA6<br />

precursors remained the same after ozonation and biological filtration but increased<br />

as a fraction <strong>of</strong> the total influent concentration. The run time for an effluent TOX<br />

criterion <strong>of</strong> 70 ug C1~/L decreased from 90 to 4.8 scaled operation days; the run time<br />

for a TTHM effluent criterion <strong>of</strong> 32 ug/L increased from 95 to 105 scaled operation<br />

days, and that for HAA6 increased from 115 to > 150 scaled operation days.


140 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

Figure 8.13 shows the impact <strong>of</strong> ozonation and biological filtration on<br />

normalized TOC breakthrough for FGW. The influent <strong>GAC</strong> concentration de<br />

creased from 10.0 to 7.4 mg/L after pretreatment, and the performance <strong>of</strong> the column<br />

treating ozonated and biologically filtered water was better than that <strong>of</strong> the column<br />

representing conventional treatment. The same impact is shown in Figure 8.14,<br />

which compares the effluent TOC concentrations. The nonadsorbable TOC concen<br />

tration after ozonation and biological filtration decreased and the tint doubled. The<br />

run time to an effluent TOC concentration <strong>of</strong> 1.6 mg/L increased from 9 to 17 scaled<br />

operation days. The run time to an effluent TOC concentration <strong>of</strong> 0.8 mg/L<br />

increased from 7.5 to 13 scaled operation days. The results for UV254 breakthrough<br />

in Figure 8.15 were similar to TOC results. The column treating ozonated and<br />

biologically filtered water lowered UV254 very well. The nonadsorbable matter<br />

measured <strong>by</strong> UV254 decreased <strong>by</strong> 50 percent.<br />

For FGW, ozonation and biological filtration prior to <strong>GAC</strong> resulted in<br />

better removal <strong>of</strong> TOX, TTHM, and HAA6 precursors <strong>by</strong> <strong>GAC</strong>, as shown in Figures<br />

8.16, 8.17, and 8.18, respectively. The nonadsorbable concentrations increased for<br />

TOX and HAA6 precursors. The run times to the given effluent criteria for TTHM<br />

and HAA6 increased as a result <strong>of</strong> ozonation and biological filtration <strong>by</strong> 300 and 40<br />

percent, respectively, while that for TOX decreased <strong>by</strong> 40 percent. For FGW, the<br />

TOX effluent criterion is very low compared to the influent TOX concentration.<br />

Also, fewer data points were taken for TOX than for the other formed DBFs.<br />

Effect <strong>of</strong> Ozonation and Biological Filtration on<br />

Specific DBF Yield <strong>of</strong> <strong>GAC</strong>________________<br />

The results from the preceding section indicate that the effect <strong>of</strong> using<br />

ozonation and biological filtration as <strong>GAC</strong> pretreatment may be dependent on the<br />

water source. It is difficult to directly compare the <strong>GAC</strong> control <strong>of</strong> DBF formation<br />

in ozonated and biologically filtered water to that in conventionally treated water<br />

because the influent concentrations were different. To facilitate the comparison,<br />

DBF formation can be plotted as a function <strong>of</strong> TOC concentration. As discussed in<br />

Chapter 7, this method allows the DBF formations to be compared at the same TOC<br />

values.<br />

The plots <strong>of</strong> DBF formation versus TOC tended to be highly linear. A bestfit<br />

curve was applied to the DBFs formed and TOC concentration data <strong>of</strong> both<br />

columns. The slope and the y-axis intercept <strong>of</strong> the curves from the ozonated and<br />

biologically filtered water and conventionally treated water were compared using<br />

regression analysis with categorical variables. The method for comparison was<br />

discussed in detail in Chapter 7. The four parameters, a0 , a,, 80, and 6,, were<br />

estimated and used to compare the specific yields for the two treatments. The t-test<br />

was used to determine whether the parameters were significantly different than<br />

zero. The results were categorized into five possible outcomes— types 1, 1 modi<br />

fied, 2, 3, and 4—as discussed in Chapter 7 (see Figure 7.14, p. 128).<br />

Figure 8.19 shows the effect <strong>of</strong> biotreatment <strong>of</strong> ORW on TOX formation.<br />

These results were classified as type 3 behavior. The intercepts <strong>of</strong> the two curves<br />

were the same while the slopes were different. Ozonation and biological filtration<br />

decreased the TOX formed at a given TOC. The slope represents the specific DBF<br />

yield. The results for specific TOX, TTHM, and HAA6 yields for ORW, PRW, and


The Impact <strong>of</strong> Ozonation and Biological Filtration 141<br />

Table 8.4 Summary <strong>of</strong> the effect <strong>of</strong> ozonation and biotreatment on the relationship<br />

between TOC and DBF formation<br />

Slope Y-axis intercept<br />

<strong>Water</strong><br />

Conventional O3/bio Conventional Oj/bio<br />

treatment, filtered, treatment, filtered,<br />

Parameter a, I3n<br />

«0 B 0<br />

F<br />

F*<br />

Behavior<br />

Type<br />

ORW<br />

TOX<br />

TTHM<br />

HAA6<br />

86<br />

59<br />

11<br />

><br />

><br />

><br />

22<br />

19<br />

5.5<br />

1.0<br />

-15<br />

2.0<br />

1.0<br />

< -2.0<br />

2.0<br />

2.44<br />

*<br />

3.91<br />

5.99<br />

1c<br />

4.60<br />

3<br />

1 modified<br />

3<br />

PRW<br />

TOX<br />

TTHM<br />

HAA6<br />

67<br />

35<br />

18<br />

><br />

><br />

><br />

55<br />

28<br />

9.3<br />

-11<br />

-11<br />

-1.8<br />

. = -11<br />

= -11<br />

< 2.1<br />

0.43<br />

0.07<br />

*<br />

4.49<br />

4.49<br />

*<br />

3<br />

3<br />

1 modified<br />

FGW<br />

TOX<br />

TTHM<br />

HAA6<br />

80<br />

19<br />

10<br />

><br />

><br />

><br />

22<br />

12<br />

3.1<br />

-1.0<br />

12<br />

-7.0<br />

= -1.0<br />

= 12<br />

< 9.6<br />

*<br />

0.37<br />

*<br />

•<br />

4.67<br />

*<br />

3<br />

3<br />

1 modified<br />

* Full model was not simplified.<br />

FGW are summarized in Table 8.4. The comparisons between F and the critical F<br />

value, F*, are also summarized in Table 8.4. Figures 8.20 and 8.21 show the TTHM<br />

and HAA6 results, respectively. TTHM results showed type 1 behavior; the slopes<br />

and intercepts <strong>of</strong> the two curves were different. HAA6 results showed type 3<br />

behavior, indicating that after ozone and biological pretreatment the DBF formation<br />

was lower for a given TOC concentration above 0.4 mg/L. In addition, the increase<br />

in specific DBF yield for a given increase in TOC was less for the ozonated and<br />

biologically filtered water than for the conventionally treated water.<br />

The PRW results for TOX and TTHM formation indicate Type 3 behavior<br />

and are shown in Figures 8.22 and 8.23, respectively. The HAA6 results for PRW,<br />

shown in Figure 8.24, show modified type 1 behavior, i.e., the two curves intersect.<br />

Thus, at aTOC concentration below 0.5 mg/L, the H AA6 formation for the ozonated<br />

and biologically filtered water could be higher than for the conventionally treated<br />

water; however, there were no data in this range. .<br />

The results for TOX, TTHM, and HAA6 formation for FGW are shown in<br />

Figures 8.25,8.26, and 8.27, respectively. Similar to the results for PRW, the TOX<br />

and TTHM results showed type 3 behavior. The TTHM results <strong>of</strong> the ozonated and<br />

biologically filtered water in Figure 8.26 showed a lot <strong>of</strong> scatter, though, making it<br />

very difficult to classify the data. For a given TOC concentration, however, the<br />

specific yield was lower after ozonation and biological filtration. The HAA6<br />

formation data shown in Figure 8.27 exhibits modified type 1 behavior, again with<br />

a lot <strong>of</strong> scatter in the ozonated and biologically filtered data.<br />

In all cases, ozonation and biological filtration decreased the DBF forma<br />

tion for a given TOC concentration. Thus, the specific DBF yields were lower. Also,<br />

for a given increase in TOC concentration, the corresponding increase in specific<br />

DBF yield was lower after ozonation and biological filtration than for conventional


742 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> GA C <strong>Adsorption</strong><br />

treatment. For all nine cases, the slopes <strong>of</strong> the two curves were significantly<br />

different, while the intercepts were significantly different in only three <strong>of</strong> the nine<br />

cases. Therefore, ozonation and biological filtration followed <strong>by</strong> <strong>GAC</strong> seemed to<br />

preferentially remove DBF precursors as compared to <strong>GAC</strong> alone.<br />

Impact <strong>of</strong> Ozonation and Biological Filtration<br />

on NOM Fraction Characteristics After <strong>GAC</strong>______<br />

<strong>GAC</strong> influent and three effluent samples were fractionated into humic and<br />

nonhumic components and molecular weight components as described in Chapter<br />

5. Three effluent samples were collected to represent the nonadsorbable fraction,<br />

intermediate breakthrough, and advanced breakthrough, as shown in Figure 5.10<br />

(p. 67). The fractions were characterized in terms <strong>of</strong> TOC and UV254 , and the DBF<br />

formation was assessed <strong>by</strong> TTHM formation under UFC at constant Br~:DOC<br />

ratios. The fractionation results for the raw and conventionally treated water were<br />

discussed in detail in Chapter 5.<br />

Humic-Nonhumic Fractionation<br />

Figure 8.28 shows the humic and nonhumic fractionation results for ORW.<br />

The bar graph region shows the raw water, the conventionally treated <strong>GAC</strong> influent,<br />

and the ozonated and biologically filtered <strong>GAC</strong> influent. The line graph region<br />

shows DOC composition in the <strong>GAC</strong> effluent from the column treating ozonated<br />

and biologically filtered water. Table 8.5 summarizes the DOC composition <strong>of</strong> the<br />

raw water and influent to the <strong>GAC</strong> columns. Conventional treatment <strong>of</strong> ORW<br />

reduced the DOC slightly, mainly as a result <strong>of</strong> a decrease in the humic fraction.<br />

Ozonation and biological filtration decreased the TOC <strong>by</strong> 33 percent, and the DOC<br />

removal was due to the removal <strong>of</strong> both fractions. The resulting <strong>GAC</strong> influent was<br />

59 percent nonhumic in nature. Other researchers (Koechling et al. 1996) have<br />

shown that ozonation caused a shift in DOC, creating a more nonhumic NOM;<br />

subsequent biological filtration reduced the nonhumic fraction. For this study,<br />

NOM was not characterized after ozonation but before biological filtration. Thus,<br />

it is not possible to determine whether the same effect occurred in this work. <strong>GAC</strong><br />

treatment affected the DOC composition in a similar manner as was described in<br />

Chapter 5. The humic fraction was initially very well removed, and the nonadsorbable<br />

fraction was entirely nonhumic in nature. The humic fraction continued to be well<br />

removed. Even at advanced breakthrough, the DOC was still 73 percent nonhumic.<br />

The specific TTHM yields for the humic and nonhumic fractions are shown<br />

in Figure 8.29 for ORW. The Br:DOC ratio was adjusted to the same value for each<br />

unfractionated and fractionated sample group so that the TTHM formations from<br />

these samples could be compared. However, the raw, conventionally treated,<br />

ozonated and biologically filtered, and <strong>GAC</strong> effluent samples cannot be compared<br />

to each other since the Br:DOC ratio shifts, as indicated on the graph. The humic<br />

fraction <strong>of</strong> the ozonated and biologically filtered water had a slightly lower specific<br />

yield than the unfractionated water and nonhumic fraction.<br />

The humic-nonhumic DOC fraction results for PRW and FGW are shown<br />

in Figures 8.30 and 8.31, respectively. Ozonation and biological filtration decreased<br />

the total DOC concentration <strong>of</strong> the <strong>GAC</strong> influent <strong>by</strong> 10 and 25 percent for PRW and


The Impact <strong>of</strong> Ozonation and Biological Filtration 143<br />

Table 8.5 Impact <strong>of</strong> ozonation and biotreatment on humic-nonhumic and MS<br />

fractionation results<br />

Percent <strong>of</strong> total DOC<br />

<strong>Water</strong><br />

Treatment<br />

Nonhumic<br />

Humic<br />

3,000 MS<br />

ORW<br />

Raw<br />

Conventional<br />

Ozone-biological<br />

45<br />

48<br />

59<br />

55<br />

52<br />

.41<br />

19<br />

28<br />

21<br />

60<br />

72<br />

65<br />

21<br />

0<br />

14<br />

PRW<br />

Raw<br />

Conventional<br />

Ozone-biological<br />

44<br />

48<br />

70<br />

56<br />

52<br />

30<br />

16<br />

21<br />

20<br />

63<br />

79<br />

64<br />

21<br />

0<br />

16<br />

FGW<br />

Raw<br />

Conventional<br />

Ozone-biological<br />

25<br />

25<br />

54<br />

75<br />

75<br />

46<br />

8<br />

7<br />

11<br />

90<br />

76<br />

84<br />

2<br />

10<br />

5<br />

FGW, respectively. This removal was due to a removal <strong>of</strong> the humic fraction. The<br />

nonhumic concentration <strong>of</strong> both waters increased after ozonation and biological<br />

filtration. The DOC composition <strong>of</strong> both waters after <strong>GAC</strong> treatment was similar to<br />

that for ORW. The nonadsorbable fraction was entirely nonhumic in nature. The<br />

humic fraction was very well removed in both waters, while the nonhumic fraction<br />

broke through quickly.<br />

The specific TTHM yields for the humic-nonhumic fractions for PRW and<br />

FGW are shown in Figures 8.32 and 8.33. For both PRW and FGW, the specific<br />

yields for the humic and unfractionated samples were similar for the raw and<br />

conventionally treated water, while those for the nonhumic fraction were lower.<br />

Ozonation and biological filtration did not result in a significantly lower specific<br />

yield for either <strong>of</strong> the two waters. There was no systematic effect <strong>of</strong> ozonation and<br />

biological filtration on specific yields <strong>of</strong> the humic-nonhumic fractions after <strong>GAC</strong><br />

treatment for either PRW or FGW.<br />

Molecular Size Fractionation<br />

Figure 8.34 shows the effect <strong>of</strong> the ozonation and biological filtration<br />

pretreatment on MS fractions for ORW. The fractionation results are summarized<br />

in Table 8.5. The dominant fraction in raw ORW was the intermediate fraction<br />

(500-3,000 MS). Conventional treatment caused a slight reduction in DOC as a<br />

result <strong>of</strong> the removal <strong>of</strong> the >3,000 MS fraction. The decrease in DOC after<br />

ozonation and biological filtration was mainly attributed to removal in the small and<br />

intermediate MS fractions. The effect <strong>of</strong> <strong>GAC</strong> treatment on MS fractions was<br />

similar to that described in Chapter 5. The small MS fraction quickly yielded<br />

complete breakthrough, while the intermediate fraction was better removed.


144 <strong>Removal</strong> <strong>of</strong><strong>DBP</strong> <strong>Precursors</strong> <strong>by</strong> GA C <strong>Adsorption</strong><br />

The effect <strong>of</strong> pretreatment on the specific TTHM yields <strong>of</strong> MS fractions for<br />

ORW is shown in Figure 8.35. The specific yields <strong>of</strong> the <strong>GAC</strong> influent and effluent<br />

showed no systematic difference among the MS fractions.<br />

The MS fractionation DOC results for PRW and FGW are shown in Figures<br />

8.36 and 8.37, respectively. The results for PRW and FGW were similar to those for<br />

ORW. The dominant fraction in the raw, conventionally treated, and ozonated and<br />

biologically filtered water was the intermediate MS fraction. For PRW, DOC<br />

removal <strong>by</strong> conventional treatment was largely due to the total removal <strong>of</strong> the large<br />

MS fraction, while for FGW it was mainly the intermediate MS fraction that was<br />

removed. As with ORW, the >3,000 MS fraction <strong>of</strong> PRW increased in the ozonated<br />

and biologically filtered water relative to conventional treatment. The effect <strong>of</strong><br />

<strong>GAC</strong> treatment on the MS fractions <strong>of</strong> PRW and FGW was similar to that for ORW.<br />

The small MS fraction broke through quickly, while the intermediate MS fraction<br />

was much better removed.<br />

Figure 8.38 shows the specific TTHM yields for MS fractions <strong>of</strong> PRW. The<br />

specific yields <strong>of</strong> the


The Impact <strong>of</strong> Ozonation and Biological Filtration 145<br />

Table 8.6 PY-GC-MS classification <strong>of</strong> ORW: Ozonation and biotreatment<br />

Percent based upon peak height<br />

Stage<br />

Aliphatic<br />

Nitrogen Halogen<br />

Aromatic containing substituted<br />

Unknown<br />

Raw<br />

<strong>GAC</strong> influent<br />

Nonadsorbable<br />

Intermediate breakthrough<br />

Advanced breakthrough<br />

34.5<br />

57.9<br />

50.3<br />

55.2<br />

48.1<br />

15.8 35.9 13.9<br />

0.4 6.1 30.3<br />

0.8 7.1 37.2<br />

BDL 3.2 38.3<br />

1.4 8.0 28.7<br />

BDL<br />

5.2<br />

4.6<br />

3.3<br />

13.9<br />

BDL = below detection limit<br />

Table 8.7 PY-GC-MS classification<br />

<strong>of</strong> PRW: Ozonation and biotreatment<br />

Percent based upon peak height<br />

Stage<br />

Aliphatic<br />

Nitrogen Halogen<br />

Aromatic containing substituted<br />

Unknown<br />

Raw<br />

<strong>GAC</strong> influent<br />

Nonadsorbable<br />

Intermediate breakthrough<br />

Advanced breakthrough<br />

23.0<br />

38.3<br />

BDL<br />

58.6<br />

9.04<br />

3.9 12.0 67.0<br />

5.5 8.7 40.0<br />

BDL BDL BDL<br />

BDL 1.1 • 36.6<br />

0.9 3.75 78.0<br />

4.3<br />

6.8<br />

BDL<br />

3.75<br />

8.2<br />

BDL = below detection limit<br />

Table 8.8 PY-GC-MS<br />

classification<br />

<strong>of</strong> FGW: Ozonation and biotreatment<br />

Percent based upon peak height<br />

Stage<br />

Aliphatic<br />

Nitrogen Halogen<br />

Aromatic containing substituted<br />

Unknown<br />

Raw<br />

<strong>GAC</strong> influent<br />

Nonadsorbable<br />

Intermediate breakthrough<br />

Advanced breakthrough<br />

1.4<br />

7.6<br />

94.3<br />

54.4<br />

56.7<br />

97.0 BDL BDL<br />

83.0 2.6 2.7<br />

BDL 2.3 3.0<br />

1.0 5.1 39.5<br />

14.5 8.6 10.5<br />

1.6<br />

4.1<br />

BDL<br />

BDL<br />

9.7<br />

BDL = below detection limit


146 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> GA C <strong>Adsorption</strong><br />

fragments (dibromochloromethane>tribromomethane>dichloroacetonitrile). The<br />

organic matrix at complete <strong>GAC</strong> breakthrough remained predominantly aliphatic,<br />

although the aromatic fraction showed a small increase in comparison to the second<br />

breakthrough point. The pyrolysis fingerprint was characterized <strong>by</strong> a strong acetic<br />

acid peak and minor peaks <strong>of</strong> formic and propanoic acid.<br />

Summary_________________________<br />

The impact <strong>of</strong> ozonation and biological filtration on <strong>GAC</strong> breakthrough<br />

behavior was water dependent. For ORW and PRW, ozonation likely created polar,<br />

less adsorbable compounds that were not removed entirely <strong>by</strong> biological filtration.<br />

The presence <strong>of</strong> these compounds in the <strong>GAC</strong> influent caused poorer performance<br />

in the initial part <strong>of</strong> breakthrough. During the latter stages <strong>of</strong> breakthrough, the <strong>GAC</strong><br />

column treating ozonated and biologically filtered water showed better perfor<br />

mance, which is likely due to the overall lower influent concentration. For ORW and<br />

PRW, ozonation and biological filtration increased the nonadsorbable fraction (0.7<br />

percent and 37 percent, respectively) and decreased the time to initial breakthrough<br />

<strong>of</strong> the adsorbable fraction. For ORW, <strong>GAC</strong> performance based on the effluent<br />

criteria improved after ozonation and biological filtration. The improvement<br />

occurred after 70 days for the NOM surrogates, TOC and UV, 54 , and after 30 to 50<br />

days for the DBF precursors. For PRW, ozonation and biological filtration de<br />

creased the run time to effluent criteria for TOC, UV254 , and TOX and increased the<br />

run time to the effluent criteria for TTHM and HAA6. For FGW, ozonation and<br />

biological filtration resulted in a significant improvement in <strong>GAC</strong> performance; the<br />

run times to initial breakthrough and to the effluent criteria increased. Thus,<br />

biological filtration <strong>of</strong> FGW was effective in removing or altering the polar, less<br />

adsorbable compounds created <strong>by</strong> ozonation.<br />

For the three waters studied, ozonation and biological filtration resulted in<br />

a shift in NOM composition. Ozonation and biological filtration yielded NOM that<br />

was more nonhumic than humic in nature, especially for ORW and PRW, where the<br />

nonhumic fractions were 59 and 70 percent, respectively. This shift to a greater<br />

percentage <strong>of</strong> the less adsorbable compounds explains the earlier breakthrough after<br />

ozonation and biological filtration <strong>of</strong> ORW and PRW.<br />

Ozonation and biological filtration also resulted in a decrease in the TOC<br />

concentration <strong>of</strong> the <strong>GAC</strong> influent. If a decrease in TOC is not accompanied <strong>by</strong> a<br />

change in the chemical makeup <strong>of</strong> the water, the reduced influent concentration<br />

should increase the run time to a given effluent criterion. An increase in the<br />

nonhumic content <strong>of</strong> NOM will result in a decrease in adsorbability <strong>of</strong> the NOM.<br />

Thus, it is postulated that the change in NOM composition decreased the benefit<br />

associated with a decrease in influent TOC concentration. This effect was not<br />

observed for FGW, which was highly humic in nature prior to treatment. Even after<br />

ozonation and biological filtration, 46 percent <strong>of</strong> the DOC was still humic for FGW.<br />

Thus, decreasing the influent concentration outweighed the change in humicnonhumic<br />

composition.<br />

Ozonation and biological filtration resulted in a decrease in the intermedi<br />

ate MS fraction. Ozonation has been shown to cause a shift toward the small MS<br />

fraction (Koechling et al. 1996). These small MS compounds can then easily be<br />

degraded. Thus, ozonation likely caused a decrease in the intermediate and large MS


The Impact <strong>of</strong> Ozonation and Biological Filtration 147<br />

fraction, while biological filtration removed a substantial amount <strong>of</strong> the small MS<br />

fractions. The intermediate MS fraction was shown in Chapter 5 to be better<br />

adsorbed than the small and large MS fractions. Thus, decreasing this fraction and<br />

increasing the more poorly adsorbed large MS fractions could contribute to the<br />

earlier <strong>GAC</strong> breakthrough seen after ozonation and biological filtration.


148 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

-a<br />

o<br />

o.<br />

o<br />

I<br />

o o<br />

o<br />

1.0 -<br />

0.9 -<br />

0.8 -<br />

0.7 -<br />

0.6 -<br />

0.5 -<br />

0.4 -<br />

TOC C0 (mg/L)<br />

• Conventional treatment 2.1<br />

0 Ozonation + biotreatment 1.4<br />

EBCT= 15 min<br />

0.3 -<br />

0.2 -<br />

0.1 -:<br />

0.0<br />

50 100<br />

150 200<br />

Scaled operation time (days)<br />

Figure 8.1 Effect <strong>of</strong> ozonation and biotreatment on normalized TOC breakthrough for<br />

Ohio River water<br />

1.6<br />

1.2-<br />

TOC C0 (mg/L)<br />

• Conventional treatment 2.1<br />

0 Ozonation + biotreatment 1.4<br />

EBCT = 15min<br />

I<br />

o •s<br />

I<br />

Io<br />

O<br />

0.8<br />

0.4 -<br />

0.0<br />

50 100 150<br />

Scaled operation time (days)<br />

200<br />

Figure 8.2 Effect <strong>of</strong> ozonation and biotreatment on TOC breakthrough for Ohio River<br />

water


The Impact <strong>of</strong> Ozonation and Biological Filtration 149<br />

o<br />

0.020<br />

0.018 -<br />

Conventional treatment 0.048<br />

Ozonation + biotreatment 0.016<br />


150 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

80<br />

70-<br />

60 -<br />

TTHM C0 (M9/L) EBCT = 15min<br />

* Conventional treatment 86<br />

0 Ozonation + biotreatment 38<br />

50 -<br />

I<br />

O<br />

40<br />

30 -<br />

20 -<br />

10 -<br />

—i—<br />

50 100 150 200<br />

Scaled operation time (days)<br />

Figure 8.5 Effect <strong>of</strong> ozonation and biotreatment on UFC-TTHM breakthrough for Ohio<br />

River water<br />

30<br />

HAA6 (H9/L) EBCT= 15 min<br />

25 -<br />

• Conventional treatment 26<br />

0 Ozonation + biotreatment 13<br />

20 -<br />

c<br />

o<br />

2•s<br />

8<br />

o<br />

O<br />

15 -<br />

10 -<br />

5 -<br />

—i—<br />

50 100 150 200<br />

Scaled operation time (days)<br />

Figure 8.6 Effect <strong>of</strong> ozonation and biotreatment on UFC-HAA6 breakthrough for Ohio<br />

River water


The Impact <strong>of</strong> Ozonation and Biological Filtration 151<br />

5<br />

1<br />

§o<br />

8<br />

I<br />

1.0<br />

0.9<br />

0.8<br />

0.7<br />

0.6<br />

0.5<br />

0.4<br />

0.3<br />

TOC C0 (mg/L)<br />

• Conventional treatment 3.0<br />

o Ozonation + biotreatment 2.5<br />

EBCT=15min<br />

0.2<br />

0.1<br />

0.0<br />

50 100 150 200<br />

Scaled operation time (days)<br />

250<br />

Figure 8.7 Effect <strong>of</strong> ozonation and biotreatment on normalized TOC breakthrough for<br />

Passaic River water<br />

2.8<br />

2.4 -<br />

2.0 -<br />

TOC C0 (mg/L)<br />

• Conventional treatment 3.0<br />

0 Ozonation + biotreatment 2.5<br />

i '' •^<br />

EBCT=15min<br />

o<br />

1 §<br />


752 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

8<br />

ro<br />

-e<br />

o 10<br />

.a<br />

0.04 -<br />

0.03 -<br />

0.02 -<br />

(Q<br />

0.05<br />

• Conventional treatment 0.072<br />

0 Ozonation + biotreatment 0.038<br />

0.01 -<br />

0.00<br />

50 100 150<br />

Scaled operation time (days)<br />

200 250<br />

Figure 8.9 Effect <strong>of</strong> ozonation and biotreatment on UV254 breakthrough for Passaic<br />

River water<br />

180<br />

160 -<br />

140 -<br />

TOX cr/L)<br />

• Conventional treatment 277<br />

0 Ozonatfon + biotreatment 135<br />

EBCT = 15 min<br />

o<br />

o><br />

o<br />

•£<br />

8<br />

o<br />

O<br />

120 -<br />

100 -<br />

80 -<br />

60 -<br />

40 -<br />

20 -<br />

50 100 150<br />

200 250<br />

Scaled operation time (days)<br />

Figure 8.10 Effect <strong>of</strong> ozonation and biotreatment on UFC-TOX breakthrough for Passaic<br />

River water


The Impact <strong>of</strong> Ozonation and Biological Filtration 153<br />

80<br />

TTHM C0 (M9/L)<br />

70 -<br />

• Conventional treatment 73<br />

0 Ozonation + biotreatment 57<br />

60 -<br />

EBCT=15min<br />

50<br />

Io<br />

8<br />

o<br />

O<br />

40 -\<br />

30<br />

20<br />

10 A<br />

—i— 50<br />

—i—<br />

100 150<br />

Scaled operation time (days)<br />

200 250<br />

Figure 8.11 Effect <strong>of</strong> ozonation and biotreatment on UFC-TTHM breakthrough for<br />

Passaic River water<br />

50<br />

40 -<br />

HAA6 C0 (M9/L)<br />

* Conventional treatment 70<br />

0 Ozonation + biotreatment 31<br />

EBCT= 15 min<br />

o><br />

o<br />

8<br />

o<br />

O<br />

30 -<br />

20 -<br />

10 -<br />

50 100 150<br />

200 250<br />

Scaled operation time (days)<br />

Figure 8.12 Effect <strong>of</strong> ozonation and biotreatment on UFC-HAA6 breakthrough for<br />

Passaic River water


754 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> GA C <strong>Adsorption</strong><br />

o<br />

1.0<br />

0.9<br />

0.8<br />

0.7<br />

0.6<br />

TOC C0 (mg/L)<br />

• Conventional treatment 10.0<br />

0 Ozonation + biotreatment 7.4<br />

EBCT=15min<br />

I<br />

a<br />

0.5<br />

0.4<br />

0.3<br />

o<br />

0.2<br />

0.1<br />

0.0<br />

10 20<br />

Scaled operation time (days)<br />

30 40<br />

Figure 8.13 Effect <strong>of</strong> ozonation and biotreatment on normalized TOC breakthrough for<br />

Florida groundwater<br />

o><br />

10<br />

9 -<br />

8 -<br />

7 -<br />

6 -<br />

TOC C0 (mg/L)<br />

• Conventional treatment 10.0<br />

0 Ozonation + biotreatment 7.4<br />

EBCT=15min<br />

g<br />

5 -<br />

•<br />

4 -<br />

I o<br />

O<br />

3 -<br />

2 -<br />

1<br />

—i—<br />

10<br />

—i—<br />

20<br />

30 40<br />

Scaled operation time (days)<br />

Figure 8.14 Effect <strong>of</strong> ozonation and biotreatment on TOC breakthrough for Florida<br />

groundwater


The Impact <strong>of</strong> Ozonation and Biological Filtration 155<br />

0.30<br />

UV. 254<br />

u<br />

?= 0.20 -<br />

8<br />

OJ<br />

€<br />

o<br />

in<br />

.Q<br />

(0<br />

0.25 - • Conventional treatment 0.29<br />

0 Ozonation + biotreatment 0.10<br />

0.15 -<br />

IZD<br />

0.00 -^<br />

Scaled operation time (days)<br />

Figure 8.15 Effect <strong>of</strong> ozonation and biotreatment on UV254 breakthrough for Florida<br />

groundwater<br />

800 -T<br />

TOX C0 (M9CIVL)<br />

700 -<br />

• Conventional treatment 1046<br />

0 Ozonation + biotreatment 645<br />

600 -<br />

EBCT = 15min<br />

o<br />

1<br />

g<br />

2<br />

*-•<br />

I o<br />

O<br />

500 -<br />

400 -<br />

300 -<br />

200 -<br />

100 -<br />

10 20<br />

30 40<br />

Scaled operation time (days)<br />

Figure 8.16 Effect <strong>of</strong> ozonation and biotreatment on UFC-TOX breakthrough for Florida<br />

groundwater


756 <strong>Removal</strong> <strong>of</strong><strong>DBP</strong> <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

200<br />

180<br />

160<br />

TTHM C0 (M9/L) EBCT=15min<br />

Conventional treatment 238<br />

Ozonation + biotreatment 205<br />

140<br />

o<br />

^=<br />

2<br />

I o<br />

O<br />

120<br />

100<br />

80<br />

60<br />

40<br />

20<br />

0<br />

Scaled operation time (days)<br />

Figure 8.17 Effect <strong>of</strong> ozonation and biotreatment on UFC-TTHM breakthrough for<br />

Florida groundwater<br />

100<br />

90<br />

80<br />

HAA6 c0 (ng/L) EBCT=15min<br />

• Conventional treatment 142<br />

° Ozonation + biotreatment 102<br />

O)<br />

o<br />

I<br />

70<br />

60<br />

50<br />

40<br />

30<br />

20<br />

10<br />

0<br />

—i—<br />

10 20<br />

Scaled operation time (days)<br />

—r—<br />

30 40<br />

Figure 8.18 Effect <strong>of</strong> ozonation and biotreatment on UFC-HAA6 breakthrough for<br />

Florida groundwater


The Impact <strong>of</strong> Ozonation and Biological Filtration 157<br />

100<br />

80 -<br />

Oo-P0 =1.0<br />

a, = 80<br />

'—-•-- Conventional treatment (a,,, a,)<br />

Ozonation/biotreatment (P 0 , p,)<br />

O 60 -<br />

O)<br />

o<br />

8<br />

o<br />

O<br />

0.0 0.2 0.4 0.6 0.8 1.0<br />

1.2 1.4<br />

TOC concentration (mg/L)<br />

Figure 8.19 Effect <strong>of</strong> ozonation and biotreatment on the relationship between TOC and<br />

TOX formation during <strong>GAC</strong> breakthrough for Ohio River water<br />

100<br />

0.2 0.4 0.6 0.8 1.0 1.2<br />

TOC concentration (mg/L)<br />

1.4<br />

Figure 8.20 Effect <strong>of</strong> ozonation and biotreatment on the relationship between TOC and<br />

TTHM formation during <strong>GAC</strong> breakthrough for Ohio River water


158 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> GA C <strong>Adsorption</strong><br />

^u -<br />

16 -<br />

i '<br />

1 12 -<br />

o<br />

s<br />

Concent<br />

OB<br />

4 -<br />

HAA6 • Conventional treatment (a0 , a,) .<br />

o Ozone/biotreatment (P0 , p,)<br />

a, = 11 •<br />

Pl = 5.5<br />

* 0 00 0<br />

o °<br />

* ° 0<br />

n -<br />

0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4<br />

TOC concentration (mg/L)<br />

Figure 8.21 Effect <strong>of</strong> ozonation and biotreatment on the relationship between TOC and<br />

HAA6 formation during <strong>GAC</strong> breakthrough for Ohio River water<br />

o<br />

a<br />

o<br />

8<br />

o<br />

200<br />

175<br />

150 -<br />

125 -<br />

100 -<br />

75 -<br />

50<br />

25<br />

TOX<br />

=Po = -<br />

a, =67<br />

Conventional treatment (a0 , a,)<br />

Ozone/biotreatment (p0 , p,)<br />

0.0 0.5 1.0 1.5 2.0<br />

2.5 3.0<br />

TOC concentration (mg/L)<br />

Figure 8.22 Effect <strong>of</strong> ozonation and biotreatment on the relationship between TOC and<br />

TOX formation during <strong>GAC</strong> breakthrough for Passaic River water


The Impact <strong>of</strong> Ozonation and Biological Filtration 159<br />

s<br />

c<br />

o<br />

8<br />

o<br />

O<br />

100<br />

80 -<br />

60 -<br />

40<br />

20 -<br />

TTHM<br />

a, = 35<br />

Conventional treatment (a0 , a,)<br />

Ozone/biotreatment (P0 , p,)<br />

0.0 0.5 1.0 1.5 2.0<br />

2.5 3.0<br />

TOC concentration (mg/L)<br />

Figure 8.23 Effect <strong>of</strong> ozonation and biotreatment on the relationship between TOC and<br />

TTHM formation during <strong>GAC</strong> breakthrough for Passaic River water<br />

60<br />

50 -<br />

HAA6<br />

00 =-1.8<br />

' Conventional treatment (a0 , a,)<br />

" Ozone/biotreatment (p0 , p,)<br />

^- 40 -<br />

^<br />

§ 30 H<br />

a, = 18<br />

Pi = 9-3<br />

20 -<br />

10 -<br />

0.0 0.5 1.0 1.5 2.0<br />

2.5 3.0<br />

TOC concentration (mg/L)<br />

Figure 8.24 Effect <strong>of</strong> ozonation and biotreatment on the relationship between TOC and<br />

HAA6 formation during <strong>GAC</strong> breakthrough for Passaic River water


160 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

£<br />

b<br />

0)<br />

»uu -<br />

800 -<br />

700 -<br />

600 -<br />

500 -<br />

400 -<br />

' i ' i • i • i • i • i • i • i • i •<br />

TOX ~*~~ Conventional treatment (o^, a,)<br />

a, = 80<br />

— e — Ozone/biotreatment (P0 , p,)<br />

Pi=22 .//^ '.<br />

/•<br />

9<br />

c o1<br />

Io<br />

O<br />

300 -<br />

200 -<br />

100 -<br />

/<br />

s<br />

/^ o<br />

s' o ^^__^—<br />

S 00 ^ —— — —<br />

0 T ——— ' ——— | ——— i ——— 1 ——— ' ——— 1 ——— ' ——— | ——— ' ——— 1 ——— ' ——— 1 ——— • ——— 1 ——— ' ——— 1 ——— ' ——— | ——— ' ——— 1<br />

01234567891<br />

TOC concentration (mg/L)<br />

Figure 8.25 Effect <strong>of</strong> ozonation and biotreatment on the relationship between TOC and<br />

TOX formation during <strong>GAC</strong> breakthrough for Florida groundwater<br />

Conventional treatment (OQ, a,)<br />

Ozone/biotreatment (P0 , p,)<br />

o><br />

o •-C<br />

I0)<br />

o<br />

o<br />

234567<br />

TOC concentration (mg/L)<br />

Figure 8.26 Effect <strong>of</strong> ozonation and biotreatment on the relationship between TOC and<br />

TTHM formation during <strong>GAC</strong> breakthrough for Florida groundwater


The Impact <strong>of</strong> Ozonation and Biological Filtration 161<br />

100<br />

HAA6<br />

Conventional treatment (a,,, a,)<br />

Ozone/biotreatment (p0, p,)<br />

I<br />

8<br />

8<br />

60 -<br />

40<br />

20 -<br />

Pi =3.1<br />

234567<br />

TOO concentration (mg/L)<br />

9 10<br />

Figure 8.27 Effect <strong>of</strong> ozonation and biotreatment on the relationship between TOC and<br />

HAA6 formation during <strong>GAC</strong> breakthrough for Florida groundwater<br />

2.5<br />

" Unfractionated<br />

0 Nonhumic fraction<br />

A Humic fraction<br />

o 1.5 J<br />

1<br />

8<br />

8<br />

1.0 -<br />

80-<br />

2.0-<br />

0.5-<br />

0.0<br />

Raw Conventional OJBio „ ... .... ^_.<br />

treatment y 0 50 100 150<br />

Scaled operation time (days)<br />

Note: A, B, and C denote initial, intermediate, and advanced breakthrough points. O3/Bio = ozonation and<br />

biological filtration.<br />

Figure 8.28 Effect <strong>of</strong> ozonation and biotreatment on humic-nonhumic fraction<br />

breakthrough for Ohio River water


762 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> GA C <strong>Adsorption</strong><br />

Unfractionated<br />

Nonhumic fraction<br />

Humic fraction<br />

CO<br />

Raw Conventional cyBio A<br />

treatment<br />

Note: A, B, and C denote initial, intermediate, and advanced breakthrough points. NA = not analyzed.<br />

O 3/Bio = ozonation and biological filtration.<br />

Figure 8.29 Effect <strong>of</strong> ozonation and biotreatment on humic-nonhumic fraction specific<br />

yields for Ohio River water<br />

Unfractionated<br />

0 Nonhumic fraction<br />

A Humic fraction<br />

Raw Conventional CyBio Q<br />

treatment<br />

50 100<br />

Scaled operation time (days)<br />

150<br />

Note: A, B, and C denote initial, intermediate, and advanced breakthrough points. O/Bio = ozonation and<br />

biological filtration.<br />

Figure 8.30 Effect <strong>of</strong> ozonation and biotreatment on humic-nonhumic fraction<br />

breakthrough for Passaic River water


The Impact <strong>of</strong>Ozonation and Biological Filtration 163<br />

i<br />

o<br />

14<br />

12 -<br />

10-<br />

8 -<br />

m Unfractionated<br />

0 Nonhumic fraction<br />

A Humic fraction<br />

8 6^<br />

8<br />

2-<br />

0<br />

Raw Conventional Oj/Bio 0<br />

treatment<br />

10 20 30<br />

Scaled operation time (days)<br />

40<br />

Note: A, B, and C denote initial, intermediate, and advanced breakthrough points. O3/Bio = ozonation and<br />

biological filtration.<br />

Figure 8.31 Effect <strong>of</strong> ozonation and biotreatment on humic-nonhumic fractionation for<br />

Florida groundwater<br />

Unfractionated<br />

Nonhumic fraction<br />

Humic fraction<br />

Raw Conventional<br />

treatment<br />

Note: A, B, and C denote initial, intermediate, and advanced breakthrough points. NA = not analyzed.<br />

O/Bio = ozonation and biological filtration.<br />

Figure 8.32 Effect <strong>of</strong> ozonation and biotreatment on humic-nonhumic fraction specific<br />

yields for Passaic River water


164 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

Unfractionated<br />

Nonhumic fraction<br />

Humic fraction<br />

Br:DOC 21<br />

(Hg/mg) ,—,<br />

Raw Conventional (v/Bio A<br />

treatment<br />

Note: A, B, and C denote initial, intermediate, and advanced breakthrough points. NA = not analyzed.<br />

O3/Bio = ozonation and biological filtration.<br />

Figure 8.33 Effect <strong>of</strong> ozonation and biotreatment on humic-nonhumic fraction specific<br />

yields for Florida groundwater<br />

2.5<br />

O)<br />

Unfractionated<br />

< 500 MS fraction<br />

500-3000 MS fraction<br />

> 3000 MS fraction<br />

1.5 -<br />

8<br />

8<br />

1.0 -<br />

0.5-<br />

0.0<br />

Raw Conventional Oj/Bio rj<br />

treatment<br />

50 100 150<br />

Scaled operation time (days)<br />

Note: A, B, and C denote initial, intermediate, and advanced breakthrough points. 0,/Bio = ozonation and<br />

biological filtration.<br />

Figure 8.34 Effect <strong>of</strong> ozonation and biotreatment on MS fraction breakthrough for Ohio<br />

River water


The Impact <strong>of</strong> Ozonation and Biological Filtration 165<br />

Unfractionated<br />

< 500 MS fraction<br />

< 3000 MS fraction<br />

O)<br />

I<br />

2<br />

•I<br />

0 Q.<br />

to<br />

Raw Conventional cyBio<br />

treatment<br />

Note: A, B, and C denote initial, intermediate, and advanced breakthrough points. CyBio = ozonation and<br />

biological filtration.<br />

Figure 8.35 Effect <strong>of</strong> ozonation and biotreatment on specific yields for MS fractions for<br />

Ohio River water<br />

4.0<br />

3.5<br />

I"<br />

| 2.0<br />

8<br />

I 1 -5<br />

O 1.0<br />

Unfractionated<br />

0 < 500 MS fraction<br />

* 500-3000 MS fraction<br />

0 > 3000 MS fraction<br />

B<br />

0.5<br />

0.0<br />

Raw Conventional O/Bio o<br />

treatment<br />

50 100 150<br />

Scaled operation time (days)<br />

Note: A, B, and C denote initial, intermediate, and advanced breakthrough points. 0,/Bio = ozonation and<br />

biological filtration.<br />

Figure 8.36 Effect <strong>of</strong> ozonation and biotreatment on MS fraction breakthrough for<br />

Passaic River water


166 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

15<br />

10 -<br />

Unfractionated<br />

0 < 500 MS fraction<br />

A 500-3000 MS fraction<br />

0 > 3000 MS fraction<br />

I<br />

8 5<br />

8<br />

Raw Conventional cyBio<br />

treatment<br />

Scaled operation time (days)<br />

Note: A, B, and C denote initial, intermediate, and advanced breakthrough points. O3/Bio = ozonation and<br />

biological filtration.<br />

Figure 8.37 Effect <strong>of</strong> ozonation and biotreatment on MS fraction breakthrough for<br />

Florida groundwater<br />

Unfractionated<br />

< 500 MS fraction<br />

< 3000 MS fraction<br />

Raw Conventional O/Bio<br />

treatment<br />

Note: A, B, and C denote initial, intermediate, and advanced breakthrough points. CyBio = ozonation and<br />

biological filtration.<br />

Figure 8.38 Effect <strong>of</strong> ozonation and biotreatment on specific yields for MS fractions for<br />

Passaic River water


Chapter 9<br />

Impact <strong>of</strong> Treatment on <strong>DBP</strong><br />

Speciation<br />

Changes in the ratio <strong>of</strong> Br to TOC (Br:TOC) as a result <strong>of</strong> preferential<br />

removal <strong>of</strong> TOC <strong>by</strong> a treatment process can strongly affect <strong>DBP</strong> speciation upon<br />

chlorination. <strong>DBP</strong> speciation is important because the proposed MCLs for TTHM<br />

and HAAS as part <strong>of</strong> the D-<strong>DBP</strong> Rule are mass-based levels. The atomic weight <strong>of</strong><br />

the bromine atom is 2.25 times greater than that <strong>of</strong> the chlorine atom. Thus, an<br />

increase in bromine incorporation into <strong>DBP</strong> speciation, given the same molar yield,<br />

will lead to higher TTHM or HAAS mass concentrations. These increases in mass<br />

concentrations may not necessarily be reflected in changes in molar TTHM and<br />

HAAS concentrations. <strong>DBP</strong> speciation is also significant for health related issues<br />

because some <strong>of</strong> the more brominated species may be <strong>of</strong> a greater health concern<br />

than chlorinated species (Krasner et al. 1992). It is also important to consider that<br />

not all nine HAA species are measured <strong>by</strong> HAAS or HAA6. In fact, all four species<br />

not accounted for <strong>by</strong> HAAS are bromo-substituted species; therefore, increases in<br />

the concentrations <strong>of</strong> these species due to shifts in speciation after treatment are not<br />

accounted for <strong>by</strong> measurements <strong>of</strong> HAAS. The formed concentrations <strong>of</strong> HAA<br />

species not accounted for <strong>by</strong> HAAS or HAA6 measurements can constitute a<br />

significant fraction <strong>of</strong> total HAA formation at moderate to high Br~:TOC ratios<br />

(Cowman and Singer 1994).<br />

Results and Discussion_________________<br />

<strong>GAC</strong> Effluent <strong>DBP</strong> Speciation<br />

One can observe effect <strong>of</strong> <strong>GAC</strong> treatment on <strong>DBP</strong> speciation <strong>by</strong> plotting the<br />

breakthrough <strong>of</strong> each THM compound on one graph, as shown in Figure 9.1 for<br />

ORW after conventional pretreatment. After 65 days <strong>of</strong> scaled operation, concen<br />

trations in the chlorinated <strong>GAC</strong> effluent were highest for brom<strong>of</strong>orm and<br />

dibromochloromethane and lowest for bromodichloromethane and chlor<strong>of</strong>orm.<br />

This can be contrasted with THMs measured in the influent upon chlorination,<br />

where brom<strong>of</strong>orm was lowest and bromodichloromethane was highest. As <strong>GAC</strong><br />

breakthrough continued, the brom<strong>of</strong>orm concentration following chlorination <strong>of</strong><br />

the <strong>GAC</strong> effluent reached a peak, and then the bromodichloromethane concentra<br />

tion increased above brom<strong>of</strong>orm. Figure 9.2 shows THM species breakthrough in<br />

ORW after <strong>GAC</strong> treatment preceded <strong>by</strong> optimized coagulation.<br />

167


168 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

An example <strong>of</strong> the breakthrough <strong>of</strong> HAA species in chlorinated <strong>GAC</strong><br />

effluent following conventional treatment <strong>of</strong> SRPW is shown in Figure 9. 3. Influent<br />

DBF formation was highest for dichloroacetic acid (DCAA) and trichloroacetic<br />

acid (TCAA), both nonbrominated species. During <strong>GAC</strong> breakthrough, two brominated<br />

species, dibromoacetic acid (DBAA) and bromochloroacetic acid (BCAA),<br />

showed concentrations that were higher than both DCAA and TCAA. The figure<br />

also shows results for monochloracetic acid (MCAA) and monobromoacetic acid<br />

(MBAA). Results observed for HAA species breakthrough in SRPW after opti<br />

mized coagulation are shown in Figure 9.4.<br />

The effect <strong>of</strong> <strong>GAC</strong> treatment on the breakthrough <strong>of</strong> THM precursors for<br />

PRW is shown in Figure 9.5. At 75 scaled operation days, dibromochloromethane<br />

concentrations were highest. At the end <strong>of</strong> the run, 190 scaled operation days,<br />

chlor<strong>of</strong>orm had the highest concentration. This corresponds to the influent speciation,<br />

where chlor<strong>of</strong>orm also had the highest concentration. The concentration <strong>of</strong><br />

brom<strong>of</strong>orm peaked at 75 scaled operation days, after which it decreased. Thus, after<br />

75 days, the Br~:TOC ratio decreased as a result <strong>of</strong> TOC breakthrough to an extent<br />

that the formation <strong>of</strong> more chlorinated species was favored. The effect <strong>of</strong> ozonation<br />

and biological filtration on THM species for PRW is shown in Figure 9.6. A shift<br />

in speciation occurred in the <strong>GAC</strong> influent, with dibromochloromethane concentra<br />

tions being the highest throughout the breakthrough period. After <strong>GAC</strong> treatment,<br />

the more brominated compounds broke through to a greater extent. At the end <strong>of</strong> the<br />

run at 1 50 scaled operation days, dibromochloromethane had the highest concentra<br />

tion in the effluent, while chlor<strong>of</strong>orm concentrations were only slightly higher than<br />

those <strong>of</strong> brom<strong>of</strong>orm.<br />

Breakthrough <strong>of</strong> HAA species after conventional treatment is shown in<br />

Figure 9.7 for FGW. The highest <strong>GAC</strong> influent concentrations were for DCAA and<br />

TCAA. These two species also had the highest concentrations at the end <strong>of</strong> the run.<br />

DBAA peaked at 1 6 scaled operation days, after which its concentration in the <strong>GAC</strong><br />

effluent decreased. The effect <strong>of</strong> ozonation and biological filtration on HAA species<br />

breakthrough is shown in Figure 9.8 for FGW. After pretreatment, the speciation in<br />

the <strong>GAC</strong> influent shifted. Large concentrations <strong>of</strong> MCAA and DBAA were<br />

detected, while the concentrations <strong>of</strong> DCAA and TCAA decreased. This is also<br />

reflected in the <strong>GAC</strong> effluent, where DBAA had the highest concentration at the end<br />

<strong>of</strong> the run.<br />

Bromine Incorporation Factors<br />

These speciation results after <strong>GAC</strong> treatment are similar to those reported<br />

<strong>by</strong> Summers et al. (1994b). It is difficult, however, to directly compare shifts in<br />

speciation as breakthrough occurs in chlorinated <strong>GAC</strong> effluent for different<br />

pretreatment scenarios, such as conventional pretreatment, optimized coagulation,<br />

or ozonation and biological filtration. To assist in making these comparisons, the<br />

bromine incorporation factor, n, can be calculated on a molar basis for THM<br />

speciation (Gould et al. 1983) using Equation 9.1 :<br />

„ = (9.1)<br />

TTHM


Impact <strong>of</strong> Treatment on DBF Speciation 169<br />

where THM-Br is defined as:<br />

THM-Br = Ii«CHCl,_.Br.<br />

j=0 311<br />

(9-2)<br />

Both THM-Br and TTHM have units <strong>of</strong> u.mol/L. The n value can range from 0 to<br />

3 depending on the extent <strong>of</strong> bromine incorporation; n = 0 if only chlor<strong>of</strong>orm is<br />

formed, while n = 3 if only brom<strong>of</strong>orm is present.<br />

A similar calculation, Equation 9.3, can be made to calculate the bromine<br />

incorporation factor for HAA speciation, n' (Shukairy and Summers 1993):<br />

n =<br />

HAA - Br<br />

HAA<br />

(9.3)<br />

where HAA-Br is defined as the sum <strong>of</strong> the bromo-substituted HAA species<br />

analyzed. Within HAA6, not all bromo-substituted HAA species are included.<br />

Therefore, HAA-Br is defined <strong>by</strong> Equation 9.4:<br />

HAA-Br = (1 • MBAA) + MBAA) + (1 • BCAA) + (2 • DBAA) (9.4)<br />

where MBAA is monobromoacetic acid, BCAA is bromochloroacetic acid, and<br />

DBAA is dibromoacetic acid. Again, the units used for HAA-Br and HAA are<br />

umol/L. The n' value can range from 0 to 2 depending on the extent <strong>of</strong> bromine<br />

incorporation.<br />

Optimized Coagulation<br />

Table 9.1 lists TOC, bromide, Br:TOC, n, and n' for ORW and SRPW after<br />

conventional treatment and optimized coagulation. For ORW, optimized coagula<br />

tion increased the bromine incorporation for both THMs and HAAs as compared to<br />

conventional treatment. This increase was primarily caused <strong>by</strong> an increase in<br />

Br:TOC with optimized treatment, since TOC was reduced while the bromide<br />

concentration was assumed to be unaffected. The Br":TOC ratio increased<br />

38 percent with optimized coagulation for ORW, and n and n' increased 40 and<br />

Table 9.1 Effect <strong>of</strong> optimized coagulation on DBF speciation<br />

ORW SRPW<br />

Parameter<br />

Conventional Optimized<br />

treatment coagulation<br />

Change Conventional Optimized Change<br />

(%) treatment coagulation (%)<br />

Influent TOC (mg/L)<br />

Bromide (ug/L)<br />

Br:TOC (ug/mg)<br />

n<br />

n'<br />

2.0<br />

132<br />

64.7<br />

0.92<br />

0.61<br />

1.5<br />

132*<br />

89.2<br />

1.28<br />

0.76<br />

-28<br />

NA<br />

38<br />

40<br />

24<br />

2.2<br />

89<br />

39.9<br />

0.75<br />

0.40<br />

1.7<br />

89*<br />

52.9<br />

0.98<br />

0.52<br />

-25<br />

NA<br />

33<br />

30<br />

31<br />

NA = not applicable<br />

*Not measured; it was assumed that <strong>GAC</strong> treatment did not affect bromide concentrations.


170 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> GA C <strong>Adsorption</strong><br />

24 percent, respectively. For SRPW, the 33 percent increase in Br:TOC was matched<br />

<strong>by</strong> 30 and 31 percent increases in n and n'. Thus, for these two waters, the change in<br />

BrrTOC with optimized coagulation was matched <strong>by</strong> similar changes in n and n'.<br />

Ozonation and Biological Filtration<br />

The effect <strong>of</strong> ozonation and biological filtration on <strong>GAC</strong> influent speciation<br />

is shown in Table 9.2. For ORW, ozonation and biological filtration decreased the<br />

Br:TOC ratio <strong>by</strong> 12 percent, while for PRW and FGW, the Br :TOC ratio increased<br />

<strong>by</strong> 20 and 31 percent, respectively. The bromine incorporation factors for both<br />

THMs and HAAs increased for all three waters after ozonation and biological<br />

filtration. For ORW, Br:TOC decreased <strong>by</strong> 12 percent, while 60 and 34 percent<br />

increases in n and n', respectively, were observed. For PRW, there was a 20 percent<br />

increase in the Br~:TOC, while n and n' increased <strong>by</strong> 177 and 135 percent,<br />

respectively. For FGW, n and n' values increased <strong>by</strong> 124 and 117 percent,<br />

respectively, after a 31 percent increase in the Br:TOC ratio. As discussed earlier,<br />

increases in n and n' values after ozonation and biological filtration were larger than<br />

after optimized coagulation for similar increases in the Br~:TOC ratio. An increase<br />

in the Br:TOC ratio is known to cause the shift to more brominated species. After<br />

ozonation and biological filtration, the characteristics <strong>of</strong> the organic matter changed<br />

as discussed in Chapter 8. This change in NOM characteristics could have caused<br />

the larger shift in speciation observed after pretreatment <strong>by</strong> ozonation and biological<br />

filtration. The results for ORW, where a decrease in the Br:TOC ratio resulted in<br />

an increase in n and n', would seem to support this.<br />

It is difficult to determine the effect <strong>of</strong> pretreatment on <strong>GAC</strong> effluent n and<br />

n' when these values are plotted as a function <strong>of</strong> scaled operation time because, at<br />

any given time, effluent TOC concentrations in the RSSCT effluent for the two<br />

pretreatments were not equal. Therefore, n and n' have been plotted against the<br />

measured TOC for each sample. In this manner, the behavior <strong>of</strong> n and n' with<br />

increasing TOC can be compared between the two types <strong>of</strong> pretreatment conditions<br />

for each water. Figure 9.9 shows the effect <strong>of</strong> pretreatment on n values for ORW.<br />

At TOC concentrations <strong>of</strong> 0.4 mg/L and higher, n decreased for both pretreatment<br />

types. Although at TOC concentrations below 0.8 mg/L the n values <strong>of</strong> the effluent<br />

samples from RSSCT-OC were slightly lower than those from RSSCT-CT, above<br />

0.8 mg/L there was little difference between the two pretreatments. At TOC<br />

concentrations below 0.4 mg/L, n increased with TOC for both RSSCTs. This trend<br />

would not have been expected based on the high Br:TOC ratio at very low TOC<br />

concentrations. It may be caused in part, however, <strong>by</strong> analytical difficulties at low<br />

TOC concentrations, as well as <strong>by</strong> the very low levels <strong>of</strong> substrate (TOC) available<br />

for reaction. A similar relationship illustrating n as a function <strong>of</strong> TOC for SRPW<br />

showed the same overall trends shown for ORW, as well as little to no effect <strong>of</strong><br />

pretreatment on <strong>GAC</strong> effluent n values (Figure 9.10).<br />

Figure 9.11 shows the effect <strong>of</strong> ozonation and biological filtration on <strong>GAC</strong><br />

effluent n values for PRW. The results seen in Figures 9.5 and 9.6 are again evident<br />

here. The n values after ozonation and biological filtration were larger, indicating<br />

a more brominated THM distribution. At TOC concentrations greater than 0.7<br />

mg/L, n decreased with increasing TOC. At TOC concentrations less than 0.7<br />

mg/L, n increased. Similar results were seen for FGW, shown in Figure 9.12. The<br />

n values after conventional treatment were in general lower than after ozonation and


FGW<br />

Ozonation/<br />

biological<br />

filtration<br />

7.4<br />

268<br />

36.2<br />

0.75<br />

0.26<br />

Change<br />

%<br />

-26<br />

_g<br />

31<br />

124<br />

117<br />

13<br />

O<br />

n 5<br />

s<br />

o<br />

re<br />

S'<br />

Table 9.2 Effect <strong>of</strong> ozonation and biological filtration on DBF speciation<br />

ORW<br />

PRW<br />

Conventional<br />

Parameter<br />

treatment<br />

Influent TOC (mg/L) 2.0<br />

Bromide (ug/L)<br />

132<br />

Br:TOC (ug/mg)<br />

64.7<br />

n 0.92<br />

n'<br />

0.61<br />

Ozonation/<br />

biological<br />

filtration<br />

1.4<br />

80<br />

57.1<br />

1.46<br />

0.82<br />

Change<br />

%<br />

-33<br />

-39<br />

-12<br />

60<br />

34<br />

Conventional<br />

treatment<br />

3.0<br />

80<br />

26.7<br />

0.47<br />

0.24<br />

Ozonation/<br />

biological<br />

filtration<br />

2.5<br />

80<br />

32.0<br />

1.29<br />

0.56<br />

Change<br />

%<br />

-17<br />

0<br />

20<br />

177<br />

135<br />

Conventional<br />

treatment<br />

10.0<br />

276<br />

27.6<br />

0.34<br />

0.12


772 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

biological filtration. Ozonation and biological filtration increased the bromine<br />

incorporation factor. Similar results were reported in Shukairy et al. (1994) and<br />

Koechlingetal. (1996).<br />

The effect <strong>of</strong> pretreatment on <strong>GAC</strong> effluent n' values is shown in Figure<br />

9.13 for ORW. Conventionally pretreated <strong>GAC</strong> effluent sample n' values were<br />

slightly higher than those for <strong>GAC</strong> effluent samples after optimized coagulation.<br />

The comparison is confounded however, <strong>by</strong> the fact that three <strong>of</strong> the nine HAA<br />

species are not measured when HAA6 is being determined; shifts in speciation to<br />

these species are not incorporated <strong>by</strong> n'. Below a TOC concentration <strong>of</strong> 1.0 mg/L,<br />

a direct relationship between n' and TOC was found. Again, possible shifts to the<br />

nonanalyzed HAA species, which are all brominated, at low TOC concentrations<br />

could account for this relationship. For SRPW, pretreatment did not seem to have<br />

a large effect on <strong>GAC</strong> effluent n' values, as shown in Figure 9.14.<br />

The effect <strong>of</strong> ozonation and biological filtration on <strong>GAC</strong> effluent n' values<br />

for PRW is shown in Figure 9.15. At TOC concentrations below 1.6 mg/L, n' values<br />

for both pretreatments were similar. At TOC concentrations greater than 1.6 mg/L<br />

n' values were greater after ozonation and biological filtration. For FGW, shown in<br />

Figure 9.16, there was little effect <strong>of</strong> ozonation and biological filtration on <strong>GAC</strong><br />

effluent n' values. Other researchers (Shukairy et al. 1994; Koechling et al. 1996)<br />

have shown that n' values increased after ozonation and biological filtration.<br />

Summary_________________________<br />

It is evident that treatment <strong>by</strong> <strong>GAC</strong> after conventional pretreatment in<br />

creases the Br~:TOC ratio, as well as the Br~:Q 2 ratio when chlorinating takes place<br />

at a fixed Cl ,:TOC ratio, and therefore shifts DBF speciation to the more bromosubstituted<br />

species. Although an increase in bromo-substitution as measured <strong>by</strong> the<br />

bromine incorporation factors n and n' was observed with optimized coagulation as<br />

compared to conventional treatment, the percent increases in n and n' were well<br />

predicted <strong>by</strong> the increase in Br~:TOC ratio after optimized coagulation. This<br />

indicates that DBF speciation was not affected <strong>by</strong> changes in the chemistry <strong>of</strong> NOM<br />

after optimized coagulation, but that the increased bromo-substitution was mainly<br />

a result <strong>of</strong> the change in Br~:TOC resulting from the decrease in TOC after<br />

optimized coagulation.<br />

Ozonation with biological filtration resulted in significant increases in<br />

bromine incorporation for both THMs and HAAs that were not reflected <strong>by</strong> the<br />

changes in the Br~:TOC ratio between conventional treatment and ozonation with<br />

biological filtration. In fact, ORW showed a decrease in Br~:TOC but still showed<br />

increases in both n and n'. PRW and FGW yielded increases in n and n' <strong>by</strong> factors<br />

<strong>of</strong> 8 and 4, respectively, which were greater than the increases seen inBr~:TOC.This<br />

behavior is a strong indication that ozonation and biological filtration changed the<br />

character <strong>of</strong> NOM precursors to THMs and HAAs.<br />

The results also showed that differences in pretreatment to <strong>GAC</strong>, when<br />

normalized <strong>by</strong> the effluent TOC concentration, did not have a significant impact on<br />

effluent <strong>DBP</strong> speciation. For pretreatment with optimized coagulation and with<br />

ozonation and biological filtration, most n and n' curves showed an initial increase,<br />

which may have been caused <strong>by</strong> low substrate or analytical difficulties, followed <strong>by</strong><br />

a decrease as TOC concentrations increased and yielded lower Br~:TOC ratios.


Impact <strong>of</strong> Treatment on <strong>DBP</strong> Speciation 173<br />

Scaled operation time (days)<br />

175<br />

Figure 9.1<br />

treatment<br />

<strong>GAC</strong> breakthrough <strong>of</strong> THM species in Ohio River water after conventional<br />

1<br />

o<br />

1<br />

V<br />

u<br />

o<br />

O<br />

5 -<br />

50 100 150 200<br />

250 300<br />

Scaled operation time (days)<br />

Figure 9.2 <strong>GAC</strong> breakthrough <strong>of</strong> THM species in Ohio River water after optimized<br />

coagulation


174 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> GA C <strong>Adsorption</strong><br />

HAA species<br />

o MCAA<br />

a DCM<br />

• TCAA<br />

^5 *S<br />

2 -<br />

1 -<br />

25 50 75<br />

Scaled operation time (days)<br />

100 125<br />

Figure 9.3 <strong>GAC</strong> breakthrough <strong>of</strong> THM species in Salt River Project water after<br />

conventional treatment<br />

o<br />

100 150 200<br />

Scaled operation time (days)<br />

300<br />

Figure 9.4 <strong>GAC</strong> breakthrough <strong>of</strong> HAA species in Salt River Project water optimized<br />

coagulation


Impact <strong>of</strong> Treatment on DBF Speciation 175<br />

i<br />

•S3 o<br />

1<br />

8 O<br />

50 100<br />

Scaled operation time (days)<br />

200<br />

Figure 9.5 <strong>GAC</strong> breakthrough <strong>of</strong> THM species in Passaic River water after conventional<br />

treatment<br />

g<br />

"•4-1<br />

S<br />

O<br />

100<br />

Scaled operation time (days)<br />

150 200<br />

Figure 9.6 <strong>GAC</strong> breakthrough <strong>of</strong> THM species in Passaic River water after ozonation<br />

and biotreatment


776 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

i<br />

o MCAA<br />

Q DCAA<br />

TCAA<br />

o<br />

8<br />

o<br />

O<br />

Scaled operation time (days)<br />

Figure 9.7 <strong>GAC</strong> breakthrough <strong>of</strong> HAA species in Florida groundwater after<br />

conventional treatment<br />

2<br />

o<br />

1<br />

8<br />

10 20 30<br />

Scaled operation time (days)<br />

40<br />

Figure 9.8 <strong>GAC</strong> breakthrough <strong>of</strong> HAA species in Florida groundwater after ozonation<br />

and biotreatment


Impact <strong>of</strong> Treatment on DBF Speciation 177<br />

3.0<br />

c 2.5 H I"<br />

o Conventional treatment<br />

• Optimized coagulation<br />

? 2'°1<br />

o ••sa<br />

S 1.5 -<br />

0> 1.0 -<br />

2<br />

m 0.5 -<br />

0.0<br />

0.0 0.5<br />

1.0 1.5<br />

TOO (mg/L)<br />

2.0 2.5<br />

Wore: Data represent <strong>GAC</strong> effluent following pretreatment.<br />

Figure 9.9 Effect <strong>of</strong> treatment on THM bromine incorporation factor (n) for ORW<br />

2.5<br />

L.- 2.0 -<br />

o Conventional treatment<br />

• Optimized coagulation pretreatment<br />

3.<br />

1.5 H<br />

5<br />

I<br />

§ 1.0 -<br />

Q)<br />

1<br />

8 0.5 -<br />

m<br />

0.0<br />

0.0<br />

0.5 1.0 1.5<br />

TOC (mg/L)<br />

2.0 2.5<br />

Wore: Data represent <strong>GAC</strong> effluent following pretreatment.<br />

Figure 9.10 Effect <strong>of</strong> treatment on THM bromine incorporation factor (n) for SRPW


178 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

2.5<br />

o Conventional treatment<br />

• Ozonation and biotreatment<br />

2.0 -<br />

I C J<br />

1.5 1<br />

0)<br />

1.0 -<br />

m 0.5 -<br />

0.0<br />

0.0 0.5 1.0<br />

1.5 2.0<br />

TOC (mg/L)<br />

2.5 3.0 3.5<br />

Note: Data represent <strong>GAC</strong> effluent following pretreatment.<br />

Figure 9.11 Impact <strong>of</strong> pretreatment on THM bromine incorporation factor (n) for<br />

Passaic River water<br />

5.0<br />

4.5 -<br />

4.0 -<br />

o Conventional treatment<br />

• Ozonation and biotreatment<br />

3.5 -<br />

3.0 -<br />

2.5 -<br />

2.0-<br />

8.<br />

m<br />

1.5 -<br />

1.0 -<br />

0.5 -<br />

0.0 T-<br />

2<br />

T~<br />

6<br />

TOC (mg/L)<br />

Note: Data represent <strong>GAC</strong> effluent following pretreatment.<br />

Figure 9.12 Impact <strong>of</strong> pretreatment on THM bromine incorporation factor (n) for Florida<br />

groundwater


Impact <strong>of</strong> Treatment on <strong>DBP</strong> Speciation 179<br />

1.6<br />

1.4 -<br />

1.2 -<br />

1.0 -<br />

n 1<br />

o Conventional treatment<br />

• Optimized coagulation pretreatment<br />

0.8-<br />

c 0.6 -<br />

o><br />

E 0.4 -<br />

8<br />

CQ<br />

0.2 H<br />

0.0<br />

0.0 0.5<br />

1.0 1.5<br />

TOC (mg/L)<br />

2.0 2.5<br />

Note: Data represent <strong>GAC</strong> effluent following pretreatment.<br />

Figure 9.13 Effect <strong>of</strong> treatment on HAA bromine incorporation factor (n') for ORW<br />

1.6<br />

1.4 H<br />

n'<br />

o Conventional treatment<br />

• Optimized coagulation pretreatment<br />

g 1.0 H<br />

1 2 0.8 H<br />

c 0.6 -<br />

0)<br />

I 0.4 1<br />

8<br />

m 0.2 H<br />

0.0<br />

0.0<br />

0.5 1.0 1.5<br />

TOC (mg/L)<br />

2.0 2.5<br />

Note: Data represent <strong>GAC</strong> effluent following pretreatment.<br />

Figure 9.14 Effect <strong>of</strong> treatment on HAA bromine incorporation factor (n') for Salt River<br />

Project water


180 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> GA C <strong>Adsorption</strong><br />

s<br />

o<br />

0.8<br />

0.6 -<br />

o Conventional treatment<br />

• Ozonation and biotreatment<br />

0.4 -<br />


Chapter 10<br />

The Impact <strong>of</strong> NOM Preloading on<br />

SOC <strong>Adsorption</strong> Behavior<br />

Even when designed and operated to remove NOM and control DBF<br />

formation, <strong>GAC</strong> can adsorb other compounds, such as SOCs, although at dimin<br />

ished capacities. Several researchers have evaluated the impact <strong>of</strong> prior exposure to<br />

NOM (preloading) on <strong>GAC</strong>'s adsorption capacity for SOCs (Sontheimeret al. 1988;<br />

Summers et al. 1989; Cummings 1992). In the referenced research, the extent and<br />

kinetics <strong>of</strong> the loss <strong>of</strong> <strong>GAC</strong> adsorption capacity (fouling) have been investigated for<br />

a wide range <strong>of</strong> <strong>GAC</strong>s and SOCs. Unfortunately, only a few sources <strong>of</strong> NOM have<br />

been examined, and most <strong>of</strong> them have been groundwaters. This chapter describes<br />

the impact <strong>of</strong> NOM preloading on the adsorption <strong>of</strong> SOCs for all six source waters<br />

used in this study.<br />

Objectives and Approach________________<br />

The objective <strong>of</strong> the work presented in this chapter was, for a range <strong>of</strong> water<br />

sources, to assess the loss <strong>of</strong> <strong>GAC</strong> adsorption capacity for a specific SOC as a result<br />

<strong>of</strong> prior exposure to NOM. The <strong>GAC</strong> was collected from the RSSCT columns at the<br />

end <strong>of</strong> the NOM adsorption runs for all six sources: LOW, MRW, ORW, PRW,<br />

SRPW, and FGW. Preloaded <strong>GAC</strong> was also collected from the pilot plant and fullscale<br />

adsorbers treating LGW and ORW, respectively, at the end <strong>of</strong> the operating<br />

life. Preloaded <strong>GAC</strong> was sampled from the bottom and top <strong>of</strong> each column so that<br />

the impact <strong>of</strong> bed depth could be examined. After the <strong>GAC</strong> was dry, and after<br />

grinding in the case <strong>of</strong> the pilot and full-scale <strong>GAC</strong>s exposed to LGW and ORW,<br />

bottle-point batch isotherms were conducted, as described in Chapter 2, to assess<br />

the residual <strong>GAC</strong> adsorption capacity for DCE. The isotherm results were then<br />

compared to determine the influence <strong>of</strong> NOM source, <strong>GAC</strong> bed depth, time <strong>of</strong><br />

preloading, and scale <strong>of</strong> preloading (RSSCT versus field scale) on adsorption <strong>of</strong><br />

DCE.<br />

A summary <strong>of</strong> the operating conditions under which the <strong>GAC</strong> was preloaded<br />

is presented in Table 10.1. TheEBCT <strong>of</strong> the pilot- or full-scale contactors was'15<br />

minutes, and that <strong>of</strong> the RSSCTs was the equivalent <strong>of</strong> 15 minutes at the pilot or full<br />

scale. The exception was the RSSCT for MRW (because <strong>of</strong> pilot test design<br />

constraints for the utility), which was designed to simulate a pilot-scale EBCT <strong>of</strong> 6.2<br />

minutes. The influent TOC concentrations ranged from 1.5 to 10.0 mg/L, and the<br />

columns were operated until 60 to 80 percent TOC breakthrough was attained. The<br />

breakthrough curves associated with these studies are shown in previous chapters.<br />

181


182 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

Table 10.1 Summary <strong>of</strong> operating conditions for preloaded <strong>GAC</strong><br />

Source<br />

Influent DOC<br />

Scale (mg/L)<br />

EBCT*<br />

(minutes)<br />

Operation time<br />

(days)<br />

Throughput<br />

(bed volumes)<br />

ORW<br />

Field<br />

RSSCT<br />

2.1<br />

2.0<br />

15.0<br />

1.19(15)<br />

210<br />

150<br />

20,160<br />

14,400<br />

LGW<br />

Field '<br />

RSSCT<br />

1.5<br />

1.5<br />

15.0<br />

1.88 (15)<br />

127<br />

135<br />

12,190<br />

12,960<br />

MRW<br />

RSSCT<br />

2.7<br />

0.66 (6.2)<br />

47<br />

10,830<br />

PRW<br />

RSSCT<br />

3.0<br />

1.59(15)<br />

205<br />

19,500<br />

SRPW<br />

RSSCT<br />

2.2<br />

1.59(15)<br />

150<br />

14,500<br />

FGW<br />

RSSCT<br />

10.0<br />

2.85(15)<br />

31<br />

3,000<br />

* Values in parentheses represent equivalent large-scale EBCT.<br />

A single-solute isotherm was conducted to establish the adsorption capacity<br />

<strong>of</strong> the <strong>GAC</strong> for DCE without any competition from other cosolutes or preadsorbed<br />

compounds. In addition, a coadsorption isotherm was run with fresh <strong>GAC</strong> and DCE<br />

in the presence <strong>of</strong> PRW to determine the adsorption competition resulting from the<br />

background NOM.<br />

Results and Discussion_________________<br />

The results <strong>of</strong> the single-solute and coadsorption isotherms are shown in<br />

Figure 10.1. The results <strong>of</strong> these two isotherms, as well as the other 16 isotherms<br />

conducted for this study, were modeled using the Freundlich equation:<br />

q = KFC" (10.1)<br />

where q = the adsorbed or solid phase concentration<br />

C = liquid phase concentration<br />

Kp n = the Freundlich model coefficients<br />

Application <strong>of</strong> the Freundlich equation to the isotherm data is summarized in Table<br />

10.2. Nonlinear regression <strong>of</strong> the Freundlich equation to the data resulted in r2 values<br />

above 0.90 for all isotherms and 0.98 or above for 13 <strong>of</strong> the 18 isotherms. Theq, 00<br />

value represents the solid phase concentration at a DCE liquid phase concentration<br />

<strong>of</strong> 100 ug/L and is used to compare the isotherm results. The Freundlich coefficient<br />

KF was not used for comparison purposes because it represents the solid phase<br />

concentration at a DCE liquid phase concentration <strong>of</strong> 1 ug/L, which is not in the<br />

range <strong>of</strong> values tested in this research.


The Impact <strong>of</strong>NOM Preloading on SOC <strong>Adsorption</strong> 183<br />

Table 10.2 Comparison <strong>of</strong> <strong>GAC</strong> residual capacity for DCE<br />

Source<br />

Scale<br />

Location<br />

c<br />

(M9/L)<br />

KF<br />

n<br />

r2<br />

9100<br />

(ug DCE/g <strong>GAC</strong>)<br />

ORW<br />

Field<br />

RSSCT<br />

Top<br />

Bottom<br />

Top<br />

Bottom<br />

507<br />

507<br />

507<br />

507<br />

32<br />

63<br />

29<br />

95<br />

0.73<br />

0.65<br />

0.73<br />

0.63<br />

0.98<br />

0.97<br />

0.99<br />

0.99<br />

923<br />

1,237<br />

836<br />

1,730<br />

LGW<br />

Field<br />

RSSCT<br />

Top<br />

Bottom<br />

Top<br />

Bottom<br />

507<br />

507<br />

507<br />

507<br />

73<br />

69<br />

31<br />

92<br />

0.69<br />

0.68<br />

0.75<br />

0.57<br />

0.99<br />

0.98<br />

0.96<br />

0.99<br />

1,751<br />

1,581<br />

947<br />

1,393<br />

MRW<br />

RSSCT<br />

Top<br />

Bottom<br />

450<br />

450<br />

29<br />

66<br />

0.77<br />

0.57<br />

0.98<br />

0.95<br />

1,006<br />

911<br />

PRW<br />

RSSCT<br />

Top<br />

Bottom<br />

450<br />

450<br />

92<br />

108<br />

0.42<br />

0.47<br />

0.90<br />

0.94<br />

636<br />

932<br />

SRPW<br />

RSSCT<br />

Top<br />

Bottom<br />

450<br />

450<br />

103<br />

148<br />

0.55<br />

0.50<br />

0.99<br />

0.99<br />

1,284<br />

1,470<br />

FGW<br />

RSSCT<br />

Top<br />

Bottom<br />

450<br />

450<br />

121<br />

126<br />

0.58<br />

0.67<br />

0.99<br />

0.99<br />

1,749<br />

2,757<br />

As shown <strong>by</strong> the results in Figure 10.1 and Table 10.2, the presence <strong>of</strong> the<br />

background organic matter in PRW decreased the adsorption capacity for DCE <strong>by</strong><br />

37 percent. Other researchers have shown that in some cases the background NOM<br />

will decrease the adsorption capacity for SOCs in a coadsorption isotherm (Sontheimer<br />

et al. 1988; Cummings 1992), while in other cases the background NOM has no<br />

negative impact (Sontheimer et al. 1988; Summers et al. 1989). These differences<br />

are likely due to the strength <strong>of</strong> adsorption <strong>of</strong> the target compound relative to that<br />

<strong>of</strong> components in the background NOM.<br />

The effect <strong>of</strong> operating the RSSCT column with PRW for the equivalent <strong>of</strong><br />

205 days on the residual DCE adsorption capacity is also shown in Figure 10.1. The<br />

adsorption capacities for DCE were reduced <strong>by</strong> 84 and 77 percent for the <strong>GAC</strong> taken<br />

from the top and bottom <strong>of</strong> the RSSCT column, respectively. The residual adsorp<br />

tion capacity <strong>of</strong> the <strong>GAC</strong> from the top <strong>of</strong> the column was lower than that from the<br />

bottom because more <strong>of</strong> the "fouling" substances are removed from the background<br />

NOM at the top <strong>of</strong> the column (Summers et al. 1989).<br />

The isotherm results for <strong>GAC</strong> preloaded with ORW and LGW are shown<br />

in Figures 10.2 and 10.3, respectively. In addition to results from <strong>GAC</strong> preloaded<br />

in the RSSCT, the results from <strong>GAC</strong> preloaded in the field columns are shown. The<br />

average DCE capacity reductions were 70 and 65 percent because <strong>of</strong> preloading <strong>by</strong><br />

ORW and LGW, respectively. The <strong>GAC</strong> from the top <strong>of</strong> the LGW field column<br />

yielded unusually high values. With the exception <strong>of</strong> the pilot-scale <strong>GAC</strong> from


184 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> GA C <strong>Adsorption</strong><br />

LOW preloading, the <strong>GAC</strong> from the top <strong>of</strong> the column generally had a lower<br />

adsorption capacity for these two waters (similar to the <strong>GAC</strong> preloaded with PRW).<br />

The preloaded <strong>GAC</strong> from the RSSCT yielded results that were very similar to those<br />

for the <strong>GAC</strong> preloaded in the field. Cummings (1992) also found this for <strong>GAC</strong> from<br />

a Florida groundwater.<br />

Figures 10.4,10.5, and 10.6 show the preloaded isotherm results for MRW,<br />

SRPW, and FGW, respectively. The <strong>GAC</strong> in all three cases was collected from the<br />

RSSCT. For SRPW and FGW, lower capacities were found for <strong>GAC</strong> from the top<br />

<strong>of</strong> the column resulting in a depth effect. The <strong>GAC</strong> preloaded with MRW was taken<br />

from a short column (EBCT = 6.2 minutes); thus, the <strong>GAC</strong> from the bottom <strong>of</strong> this<br />

column would be more equivalent to <strong>GAC</strong> from the middle <strong>of</strong> the other columns,<br />

which had EBCTs equivalent to 15 minutes. Thus, the lack <strong>of</strong> a depth effect for the<br />

<strong>GAC</strong> preloaded with MRW is understandable.<br />

To assess the impact <strong>of</strong> the time <strong>of</strong> preloading and NOM source water on<br />

the residual adsorption capacity for DCE, the relative q, 00 values were plotted as a<br />

function <strong>of</strong> throughput for the <strong>GAC</strong>s from the RSSCTs. The results are shown in<br />

Figure 10.7. The q loo values were normalized to that <strong>of</strong> the single-solute isotherm,<br />

q ss , 00, and are shown for <strong>GAC</strong> from the top <strong>of</strong> the column, as are average q )00 values<br />

<strong>of</strong> <strong>GAC</strong> from the top and bottom. For both types <strong>of</strong> q |00 values, a trend <strong>of</strong> decreasing<br />

value with increasing throughput was found. Thus, for a given contactor, longer<br />

operation times with the background NOM would yield lower residual adsorption<br />

capacities for the target SOC. Similar results have been found <strong>by</strong> others (Sontheimer<br />

etal. 1988; Summers etal. 1989; Cummings 1992). Of particular interest, this trend<br />

was independent <strong>of</strong> NOM source, which has not been shown <strong>by</strong> other researchers.<br />

The column run time or throughput was found to be dependent on influent<br />

concentration, as discussed in the other chapters.<br />

The trend shows a quick decrease in DCE capacity followed <strong>by</strong> a slow<br />

gradual decrease with time. A second-order polynomial was used to fit the preloaded<br />

isotherm results, and the value from the coadsorption isotherm was plotted with a<br />

throughput value <strong>of</strong> 700 bed volumes, which is the equivalent <strong>of</strong> the length <strong>of</strong> time<br />

the isotherm was run in terms <strong>of</strong> a 15 minute-EBCT contactor. For the average q, 00<br />

values, the results from MRW (throughput = 10,830; see Table 10.1) were lower than<br />

expected because <strong>of</strong> the short EBCT <strong>of</strong> that column. When the q, 00 values from the<br />

<strong>GAC</strong> at the top <strong>of</strong> the column are compared, the results for MRW are similar to the<br />

trend found for the other waters. No relationship between the relative adsorption<br />

capacity and operation time or mass <strong>of</strong> TOC removed was found.<br />

Summary_________________________<br />

Exposure <strong>of</strong> <strong>GAC</strong> to the background water during the removal <strong>of</strong> NOM and<br />

<strong>DBP</strong> precursors led to a loss <strong>of</strong> the <strong>GAC</strong> residual adsorption capacity for DCE. The<br />

loss in capacity was higher for preloaded <strong>GAC</strong> from the top <strong>of</strong> the column. The<br />

results from preloaded <strong>GAC</strong> from the RSSCT were similar to those found for <strong>GAC</strong><br />

from the pilot- and full-scale contactors. The residual capacity was found to<br />

decrease with operation time, expressed as throughput, and was not directly<br />

dependent on the source water.


The Impact <strong>of</strong>NOM Preloading on SOC <strong>Adsorption</strong> 185<br />

o 10000 -<br />

s<br />

o><br />

PRW<br />

A Single solute<br />

T Coadsorption<br />

0 Preloaded- top<br />

D Preloaded- bottom<br />

I 1000 -<br />

§<br />


186 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

O 10000 -<br />

LGW<br />

Single solute<br />

c 1000 -<br />

1<br />

8<br />

8<br />

• F|ekJ- top<br />

• Field- bottom<br />

0 RSSCT- top<br />

0 RSSCT- bottom<br />

Q. 100<br />

•o<br />

=5 1<br />

CO<br />

i i i i 111<br />

10<br />

1 ' I<br />

100<br />

Liquid phase concentration, C (ng DCE/L)<br />

1000<br />

Figure 10.3 Impact <strong>of</strong> preloading <strong>of</strong> Lake Gaillard water at pilot and small scale on DCE<br />

adsorption capacity<br />

o 10000 -<br />

o<br />

MRW<br />

O<br />

o<br />

I<br />

§ 1000<br />

I<br />

§8<br />

0)<br />

0 RSSCT- top<br />

D RSSCT- bottom<br />

o. 100<br />

"o<br />

CO<br />

10 100<br />

Liquid phase concentration, C (ng DCE/L)<br />

1000<br />

Figure 10.4 Impact <strong>of</strong> preloading <strong>of</strong> Mississippi River water at small scale on DCE<br />

adsorption capacity


The Impact <strong>of</strong>NOM Preloading on SOC <strong>Adsorption</strong> 187<br />

O 10000<br />

I u?<br />

O<br />

Q<br />

O)<br />

SRPW<br />

Single solute<br />

1000 -<br />

I<br />

8<br />

m CO<br />

O<br />

(O<br />

100<br />

10 .100<br />

Liquid phase concentration, C (ng DCE/L)<br />

1000<br />

0 RSSCT- top<br />

D RSSCT- bottom<br />

Figure 10.5 Impact <strong>of</strong> preloading <strong>of</strong> Salt River Project water at small scale on DCE<br />

adsorption capacity<br />

O 10000 d<br />

<<br />

O<br />

FGW<br />

O<br />

Q<br />

S cr<br />

c<br />

.2<br />

1000 -<br />

8<br />

8<br />

o><br />

CD<br />

Q.<br />

100<br />

0 RSSCT- top<br />

D RSSCT- bottom<br />

10 100<br />

Liquid phase concentration, C (ng DCE/L)<br />

1000<br />

Figure 10.6 Impact <strong>of</strong> preloading <strong>of</strong> Florida groundwater at small scale on DCE<br />

adsorption capacity


188 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

Single solute<br />

T Coadsorption<br />

- Preloaded-top<br />

— Preloaded- average<br />

0.0<br />

5000 10000<br />

Throughput, BV (bed volumes)<br />

20000<br />

Figure 10.7 Correlation between solid phase DCE concentration and throughput in bed<br />

volumes for six sites


Chapter 11<br />

Relationship Among NOM<br />

Characteristics, <strong>Removal</strong> <strong>of</strong> DBF<br />

<strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> Treatment, and<br />

<strong>DBP</strong> Formation<br />

NOM is ubiquitous in source waters and is composed <strong>of</strong> a broad spectrum<br />

<strong>of</strong> organic compounds, most <strong>of</strong> which cannot be measured directly, but rather must<br />

be quantified <strong>by</strong> surrogate measures. A subset <strong>of</strong> NOM is composed <strong>of</strong> precursor<br />

compounds to DBFs; these DBFs form upon the addition <strong>of</strong> a disinfectant, such as<br />

free chlorine. Consequently, it is <strong>of</strong> great interest to understand the relationship<br />

among NOM characteristics, removal <strong>of</strong> <strong>DBP</strong> precursors <strong>by</strong> treatment, and <strong>DBP</strong><br />

formation so that the drinking water community can better understand the potential<br />

for successful application <strong>of</strong> treatment processes to reduce <strong>DBP</strong>s based upon an<br />

understanding <strong>of</strong> source water characteristics. A study sponsored <strong>by</strong> AWWARF<br />

and completed in 1992 (Owen et. al. 1992) began to address the relationship among<br />

these variables across abroad range <strong>of</strong> source waters. As an outgrowth <strong>of</strong> that effort,<br />

this project was developed to specifically address the removal <strong>of</strong> <strong>DBP</strong> precursors<br />

<strong>by</strong> <strong>GAC</strong> adsorption. During the conduct <strong>of</strong> this work, NOM characterization was<br />

performed to assist in elucidating the interrelationship among the characteristics <strong>of</strong><br />

<strong>DBP</strong> precursors, their removal <strong>by</strong> <strong>GAC</strong> adsorption, and ultimately the resulting<br />

formation <strong>of</strong> <strong>DBP</strong>s.<br />

Overview <strong>of</strong> Characterization Techniques_______<br />

The characterization work performed in this study included humic-nonhumic<br />

fractionation, molecular size fractionation, and pyrolysis GC-MS. The <strong>DBP</strong> precur<br />

sor analyses included surrogates commonly used in the water industry (TOC and<br />

UV254), and the <strong>DBP</strong>s analyzed included a host <strong>of</strong> compounds <strong>of</strong> regulatory interest,<br />

as well as those that assist in understanding the broad spectrum <strong>of</strong> halogenated <strong>DBP</strong>s<br />

that may be formed upon chlorination (e.g., TOX). Therefore, there is a significant<br />

amount <strong>of</strong> data collected in this work that may provide meaningful insight into<br />

source water characteristics and successful treatment <strong>by</strong> <strong>GAC</strong>. To account for the<br />

preferential removal <strong>of</strong> NOM compared to bromide, various fractions (humicnonhumic<br />

and molecular size fractions) were equilibrated with respect to Br~:TOC<br />

ratio in this research to minimize the impact <strong>of</strong> variations in that ratio.<br />

189


790 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

As discussed in Chapter 2, pyrolysis GC-MS is a relatively new analytical<br />

technique that directly measures the fragments <strong>of</strong> compounds that have been<br />

pyrolyzed. An evaluation <strong>of</strong> the pyrolyzed fragments lends insight into the nature<br />

and composition <strong>of</strong> the parent compounds. When coupled with DBF analyses<br />

resulting from the chlorination <strong>of</strong> the parent material, the pyrolysis GC-MS<br />

technique has the potential to evaluate "markers" that may indicate parent material<br />

that is responsible for specific DBF formation.<br />

Detailed discussions <strong>of</strong> the effect <strong>of</strong> pretreatment and <strong>GAC</strong> adsorption on<br />

humic-nonhumic fractionation and molecular size fractionation are presented in<br />

Chapters 5 and 7. A summary <strong>of</strong> these results will be presented here. In addition, the<br />

results <strong>of</strong> pyrolysis GC-MS analyses on various samples <strong>of</strong> raw and treated water<br />

will be discussed in this chapter. These results, together with data on DBF formation<br />

in treated and untreated waters, provide insight into the relationship among NOM,<br />

DBF precursor removal <strong>by</strong> <strong>GAC</strong> treatment, and DBF formation.<br />

Molecular Size Fractionation___________________<br />

The intermediate MS fraction (500 to 3,000) dominated the MS distribution<br />

in the raw water samples analyzed during this research. Most waters were equally<br />

distributed between the largest (MS >3,000) and the smallest (MS 3,000 MS fraction was removed and the other<br />

fractions were not effectively impacted. After <strong>GAC</strong> treatment, the nonadsorbable<br />

fraction was composed entirely <strong>of</strong>


NOM Characteristics, <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong>, and DBF Formation 191<br />

Ozonation and biptreatment resulted in a decrease in the intermediate MS<br />

fraction. Ozonation has been shown to cause a shift toward the small MS fraction<br />

(Koechling et al. 1996). These small MS compounds are then easily degraded. Thus,<br />

ozonation likely caused a decrease in the intermediate and large MS fractions, while<br />

biotreatment removed the more substantial small MS fraction. The intermediate MS<br />

fractions was shown (in Chapter 5) to be better adsorbed than the small and large<br />

MS fraction. Thus, decreasing this fraction and increasing the more poorly adsorbed,<br />

large MS fractions could contribute to the earlier <strong>GAC</strong> breakthrough seen after<br />

ozonation and biotreatment.<br />

Humic-Nonhumic Fractionation______________<br />

DOC fractionation for the six raw waters examined during this research<br />

indicated humic fractions in the range <strong>of</strong> 53 to 75 percent (a slight dominance over<br />

the nonhumic fraction). The relative percent composition <strong>of</strong> each fraction was<br />

dependent on the water source. FGW was the most humic in nature (75 percent <strong>of</strong><br />

the DOC was found in the humic fraction), while SRPW was the least humic in<br />

nature (53 percent). The effect <strong>of</strong> pretreatment and <strong>GAC</strong> treatment on the NOM was<br />

very similar for all six waters. DOC removal <strong>by</strong> the various pretreatment processes<br />

ranged from 15 to 45 percent and resulted in a major shift from humic to nonhumic<br />

material (the nonhumic fraction <strong>of</strong> the treated waters ranged from 56 to 67 percent<br />

<strong>of</strong> the total DOC). As evidenced <strong>by</strong> the <strong>GAC</strong> effluent samples collected during the<br />

initial portion <strong>of</strong> the breakthrough curves, the nonadsorbable fractions were<br />

composed nearly entirely <strong>of</strong> the nonhumic fraction; the humic fraction was virtually<br />

completely removed. With time, the nonhumic fraction broke through rapidly, while<br />

the humic fraction continued to be better removed. The nonhumic fraction break<br />

through reached a plateau, while further increases in effluent DOC concentrations<br />

were observed as a result <strong>of</strong> the breakthrough <strong>of</strong> the humic fraction.<br />

The specific TTHM yield <strong>of</strong> the humic fraction <strong>of</strong> either the raw or<br />

conventionally treated water was only slightly higher than that <strong>of</strong> the nonhumic<br />

fraction. After <strong>GAC</strong> treatment, no systematic differences could be found. This is<br />

significant in that most researchers have historically considered the humic fraction<br />

to be more problematic from a standpoint <strong>of</strong> DBF yield; consequently, more<br />

attention has been focused on removal <strong>of</strong> the humic fraction. Recent work (Owen<br />

et. al., 1992) has recognized the importance <strong>of</strong> the nonhumic fraction in DBF<br />

formation, but attempts to quantify the comparison <strong>of</strong> DBF yields between humic<br />

and nonhumic fractions was confounded <strong>by</strong> varying Br~:TOC ratios. Consequently,<br />

this research represents the first work indicating comparable yields between humic<br />

and nonhumic fractions under equilibrated Br":TOC ratios.<br />

For the two waters examined (ORW and SRPW), optimized coagulation<br />

removed a fraction <strong>of</strong> the TOC present after conventional treatment. For ORW, both<br />

humic and nonhumic fractions were affected about equally, while for SRPW the<br />

nonhumic fraction was preferentially removed. Other research has shown that<br />

coagulation usually preferentially removes the humic fraction over the nonhumic<br />

fraction (Semmens and Staples 1986; Collins et al. 1986; Dryfuse et al. 1995).<br />

For the three waters studied (ORW, PRW, and FGW), ozonation and<br />

biotreatment resulted in a shift in NOM composition. Ozonation and biotreatment<br />

yielded NOM that was more nonhumic than humic in nature, especially for ORW


792 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

and PRW, for which the nonhumic fractions were 59 and 70 percent, respectively.<br />

This shift to more <strong>of</strong> the less adsorbable compounds explains the earlier break<br />

through for ORW and PRW after ozonation and biotreatment. This type <strong>of</strong><br />

pretreatment resulted in a decrease in the TOC concentration <strong>of</strong> the <strong>GAC</strong> influent.<br />

If the decrease in TOC is not accompanied <strong>by</strong> a change in the chemical makeup <strong>of</strong><br />

the water, the reduced influent concentration should increase the run time before a<br />

given effluent criterion is reached. An increase in the nonhumic content <strong>of</strong> NOM<br />

will result in a decrease in adsorbability <strong>of</strong> the NOM. Thus, it is postulated that the<br />

change in NOM makeup decreased the benefit associated with a decrease in influent<br />

concentration. This effect, however, was not observed for FGW. FGW was highly<br />

humic in nature prior to treatment. Even after ozonation and biotreatment, 46<br />

percent <strong>of</strong> the DOC was still humic in nature. Thus, decreasing the influent<br />

concentration outweighed the change in humic-nonhumic composition.<br />

The Role <strong>of</strong> TOC and UV254________________<br />

The comparisons <strong>of</strong> relative breakthrough curves for TOC, UV., 54 , and DBF<br />

precursors for ORW, MRW, PRW, SRPW, and FGW are shown in Figures 11.1 to<br />

11.5. The results shown are for the RSSCT. TOC and UV254 were used to<br />

characterize NOM breakthrough, and TOX, TTHM, HAA6, and CH precursor<br />

breakthrough were assessed under UFC. For ORW, HAA6 precursors had the<br />

earliest breakthrough <strong>of</strong> the measured <strong>DBP</strong> precursors; CH precursors broke<br />

through latest, and the other two parameters were found in between. For the<br />

remaining four waters, TTHM precursors broke through earliest, and the relative<br />

breakthrough patterns for <strong>DBP</strong> precursors were similar to those for ORW.<br />

For ORW, TOC can be used as a conservative indicator <strong>of</strong> the breakthrough<br />

<strong>of</strong> <strong>DBP</strong> precursors after 35 percent breakthrough. In the early portions <strong>of</strong> the<br />

breakthrough curve, HAA6 precursor breakthrough was more rapid than TOC<br />

breakthrough. For PRW, SRPW, and FGW, TOC broke through earliest and thus<br />

can also be used as a conservative indicator <strong>of</strong> <strong>DBP</strong> precursor breakthrough. An<br />

exception to this was observed for MRW, for which a <strong>DBP</strong> precursor broke through<br />

earlier than TOC (Figure 11.2). One possible reason for this anomaly could be the<br />

short EBCT used during the testing <strong>of</strong> this water compared to others.<br />

Pyrolysis GC-MS_____________________<br />

Six source waters and associated influents and effluents <strong>of</strong> <strong>GAC</strong> have been<br />

analyzed <strong>by</strong> PY-GC-MS. The following is a summary <strong>of</strong> observations related to PY-<br />

GC-MS analysis. These analyses showed in general that the humic fraction <strong>of</strong> NOM<br />

is not necessarily aromatic in nature as is commonly believed. Only two <strong>of</strong> the six<br />

waters analyzed (LGW and FGW) indicated a more aromatic nature <strong>of</strong> the humic<br />

fractions. In many cases in this study, the humic and nonhumic fractions had similar<br />

percentages and showed similar chemical signatures.


NOM Characteristics, <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong>, and DBF Formation 193<br />

Raw <strong>Water</strong> Characteristics<br />

Three <strong>of</strong> the six surface waters evaluated <strong>by</strong> PY-GC-MS were found to be<br />

industrially impacted: MRW, ORW, and PRW. Although each water possessed<br />

distinct features, there were three features as characterized <strong>by</strong> PY-GC-MS that were<br />

common to these waters. In each <strong>of</strong> these raw waters, benzonitrile and benzoic acid<br />

were detected at relatively significant levels. These pyrolysis fragments are thought<br />

to be markers <strong>of</strong> industrial discharge and may be derived from aromatic polyamides<br />

or phthalic-based polymers. Each <strong>of</strong> these waters showed relatively strong halogenated<br />

signatures. These results suggest that, in general, industrially impacted waters<br />

had much higher levels <strong>of</strong> halogenated organic material than nonindustrially<br />

impacted waters and that halo-substituted fragments in combination with benzonitrile<br />

and benzoic acid may be good markers for industrial and wastewater discharge. The<br />

pyrolysis fingerprints <strong>of</strong> all <strong>of</strong> these waters lacked a biological signature, such as<br />

fragments associated with biopolymers.<br />

The other three source waters—LGW, SRPW, and FGW—were not ex<br />

pected to show signs <strong>of</strong> industrial impact. LGW represented a high-quality surface<br />

water. The proposed industrial markers <strong>of</strong> benzonitrile and benzoic acid were not<br />

detected in any <strong>of</strong> these waters. Although minor peaks <strong>of</strong> halo-substituted fragments<br />

were detected in these samples, only in SRPW were they identified in the raw water.<br />

In the two surface water samples, LGW and SRPW, biological signatures were<br />

evident based on the variety <strong>of</strong> fragments identified and the presence <strong>of</strong> fragments<br />

typical <strong>of</strong> biopolymeric parent material (e.g., polysaccharides, proteins).<br />

Seasonal samples indicated similar organic characteristics; however, the<br />

relative proportions <strong>of</strong> different markers varied from season to season. All <strong>of</strong> the<br />

seasonal samples were characterized <strong>by</strong> very strong halogenated signatures. For the<br />

most part, the chemical signatures were dominated <strong>by</strong> markers <strong>of</strong> industrial or<br />

wastewater discharge and there was a general absence <strong>of</strong> biological markers. Halosubstituted<br />

fragments were the more prominent fragments in the winter samples,<br />

and benzonitrile became a major industrial marker in the spring and summer<br />

samples.<br />

Effect <strong>of</strong> Coagulation<br />

The effect <strong>of</strong> enhanced or optimized coagulation on <strong>GAC</strong> performance and<br />

<strong>DBP</strong> yield was studied for two waters: ORW and SRPW. For ORW, enhanced<br />

coagulation increased TOC removal to 31 percent (compared to 5 percent with<br />

conventional treatment). The residual organic matrix <strong>of</strong> the sample pretreated <strong>by</strong><br />

enhanced coagulation was predominantly halogenated, as was the organic matrix <strong>of</strong><br />

the conventionally pretreated sample. There were, however, a number <strong>of</strong> notable<br />

differences between the chemical fingerprints <strong>of</strong> the two pretreatment types. The<br />

enhanced coagulated water appeared to have very little acetic acid and acetaldehyde<br />

present, a slightly greater percentage <strong>of</strong> brominated fragments, and mostly earlyeluting<br />

fragments in contrast to the conventionally treated sample. The organic<br />

quality <strong>of</strong> the <strong>GAC</strong> breakthrough samples in the enhanced coagulation experiments<br />

also showed differences with respect to conventional pretreatment. Enhanced<br />

coagulation pretreatment produced a matrix in the <strong>GAC</strong> breakthrough samples<br />

having a different chemical signature, as evidenced <strong>by</strong> the strong presence <strong>of</strong><br />

acetaldehyde and lesser relative evidence <strong>of</strong> acetic acid and formic acid. Significant


194 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

improvement in <strong>GAC</strong> performance with optimized coagulation pretreatment was<br />

observed and could be partially attributed to quality differences revealed <strong>by</strong> PY-<br />

GC-MS.<br />

In general, the nonadsorbable fraction <strong>of</strong> the NOM was very difficult to<br />

analyze <strong>by</strong> PY-GC-MS for a number <strong>of</strong> reasons. The organic material present at this<br />

early breakthrough point is typically characterized <strong>by</strong> low molecular weights and<br />

thus is poorly suited to a method tailored to nonvolatile, macromolecular material.<br />

These samples had very low TOC values, which required concentrating large<br />

volumes (10-20 L) <strong>of</strong> water to obtain sufficient organic material to analyze. This<br />

also resulted in concentrating large quantities <strong>of</strong> inorganic salts, which can possibly<br />

result in a variety <strong>of</strong> interferences. Despite these difficulties, some very general<br />

trends are visible among the data for the nonadsorbable fraction. For most samples,<br />

the nonadsorbable NOM was characterized <strong>by</strong> a major peak <strong>of</strong> acetic acid. A notable<br />

exception to this was LGW, which was primarily aromatic but showed a shift in<br />

fragment dominance from phenol top-cresol. In a number <strong>of</strong> instances, a significant<br />

halogenated signature emerged. It is thought that PY-GC-MS is particularly<br />

sensitive to halogenated macromolecular material, especially in the absence <strong>of</strong> other<br />

macromolecular compounds.<br />

Effect <strong>of</strong> Ozone and Bi<strong>of</strong>iltration<br />

The effect <strong>of</strong> ozone and bi<strong>of</strong>iltration on <strong>GAC</strong> performance and <strong>DBP</strong> yield<br />

was studied for three waters: ORW, PRW, and FGW. For ORW, the chemical<br />

characteristics <strong>of</strong> the organic matrix pretreated <strong>by</strong> ozonation and bi<strong>of</strong>iltration were<br />

remarkably altered in comparison to conventional treatment or enhanced coagula<br />

tion. This pretreatment strategy produced a pyrolysis fingerprint bearing a much<br />

reduced halogenated character. The major fragments were acetaldehyde and acetic<br />

acid, and the predominant chemical class was aliphatic. These same characteristics<br />

were carried throughout the various <strong>GAC</strong> breakthrough samples such that the<br />

halogenated fraction was never the predominant fraction <strong>of</strong> pyrolysis fragments.<br />

For PRW, the chemical classification <strong>of</strong> PY-GC-MS data for ozonation and<br />

bi<strong>of</strong>iltration pretreatment, however, was similar to that <strong>of</strong> conventional pretreat<br />

ment. Although there was a decrease in the number and proportion <strong>of</strong> halosubstituted<br />

pyrolysis fragments with ozonation and biotreatment, the group <strong>of</strong> major<br />

peaks (dichloroacetonitrile, acetic acid, trichlorobenzene) remained unchanged.<br />

The relative increase in the nonhumic fraction may be associated with the relative<br />

decrease in the halogenated class <strong>of</strong> fragments. Although the TOC removals<br />

produced <strong>by</strong> ozonation and bi<strong>of</strong>iltration were similar for ORW and PRW, the<br />

behavior <strong>of</strong> the halogenated fraction differed. While in the case <strong>of</strong> both waters the<br />

halo-substituted signature was diminished and simplified at the first two break<br />

through points, a single halogenated fragment, dichloroacetonitrile, composed most<br />

<strong>of</strong> the halo-substituted character <strong>of</strong> the PRW samples; this peak was the major peak<br />

at complete breakthrough.<br />

For FGW, the chemical nature remained primarily aromatic, but in com<br />

parison to conventional treatment it had slightly greater aliphatic character. With<br />

<strong>GAC</strong> adsorption, the chemical signature <strong>of</strong> this water became aliphatic. Yet, even<br />

at complete breakthrough, the organic quality <strong>of</strong> the <strong>GAC</strong> effluent produced <strong>by</strong><br />

ozonation and biotreatment was greatly modified in comparison to the influent<br />

quality and in contrast to that observed in the case <strong>of</strong> conventional pretreatment; in


NOM Characteristics, <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong>, and DBF Formation 195<br />

the latter case, at complete breakthrough, the water was very similar to the <strong>GAC</strong><br />

influent and highly aromatic. This observation suggests that the humic material <strong>of</strong><br />

FGW broke through less readily than in the case <strong>of</strong> conventional pretreatment; this<br />

may correspond to the reported significant improvement in <strong>GAC</strong> performance for<br />

ozonation and pretreatment <strong>of</strong> FGW.


n S:<br />

^<br />

o<br />

30 3


O<br />

^<br />

&<br />

s<br />

2<br />

^ t^<br />

3<br />

S<br />

§<br />

"O<br />

S<br />

O<br />

3<br />

a.<br />

I<br />

s 5'<br />

33 -Tl<br />

If<br />

,3 Si -*<br />

to' t*<br />

o»•*<br />

< O<br />

5 o<br />

£* 3<br />

5-6<br />

0)<br />

55'<br />

o<br />

3<br />

O<br />

Normalized concentration, C/C0<br />

o><br />

2.<br />

F*<br />

1'<br />

o-<br />

s»<br />

3<br />

5<br />

O<br />

z<br />

o 2<br />

CQ<br />

(As*<br />

o OJ<br />

3<br />

s-<br />

(/><br />

to<br />

33 2!<br />

(0 c<br />

I-<br />

O<br />

•o<br />

D><br />

55'<br />

o<br />

o<br />

2.<br />

o><br />

CD<br />

o-<br />

0)<br />

o<br />

(Q 3"<br />

(A<br />

—h<br />

O<br />

B><br />

Q.<br />

C<br />

Tl<br />

O<br />

603<br />

~O<br />

to<br />

-4i<br />

O<br />

0)<br />

w<br />

(0<br />

fi> o'<br />

Normalized concentration, C/C0<br />

lll! *-


Oo<br />

50<br />

1<br />

O<br />

g<br />

5<br />

to<br />

Co<br />

c<br />

1<br />

2<br />

.O-<br />

•<br />

3'<br />

(Q 11<br />

o 2*<br />

(D<br />

QJ<br />

|<br />

I<br />

(O 3-<br />

(A<br />

^*<br />

O<br />

Q)<br />

a<br />

c<br />

Tl<br />

O<br />

6<br />

DO<br />

T3<br />

(0<br />

—h<br />

O<br />

o•^<br />

a<br />

o>


Chapter 12___________<br />

Evaluation <strong>of</strong> Design Criteria<br />

and Costs<br />

The operation <strong>of</strong> a <strong>GAC</strong> process in a treatment facility is primarily<br />

controlled <strong>by</strong> the breakthrough characteristics <strong>of</strong> the contaminant <strong>of</strong> concern and the<br />

relevant treated water quality goal. NOM breakthrough, in terms <strong>of</strong> DBF precursor<br />

compounds and chlorinated DBFs, has been discussed in previous chapters <strong>of</strong> this<br />

report, as have the effects <strong>of</strong> pretreatment and operating characteristics (i.e., EBCT,<br />

blending, and backwashing) on <strong>GAC</strong> adsorption.<br />

This chapter presents design criteria and conceptual treatment costs for<br />

implementing <strong>GAC</strong> at four <strong>of</strong> the participating utilities treating FGW, ORW, PRW,<br />

and SRPW, respectively. Conceptual cost estimates for pretreatment to optimize<br />

<strong>GAC</strong> contactor operation are also presented in this chapter. The chapter concludes<br />

with a discussion <strong>of</strong> cost implications related to the benefits <strong>of</strong> pretreatment in<br />

reducing overall facility costs for implementing <strong>GAC</strong> treatment.<br />

<strong>GAC</strong> Design Parameters_________________<br />

A <strong>GAC</strong> treatment process includes several facility components, including<br />

<strong>GAC</strong> contactors, media backwashing, storage for fresh and spent carbon, and<br />

pumping to the <strong>GAC</strong> contactors. Depending on the size <strong>of</strong> the treatment plant,<br />

carbon reactivation also may be required. In full-scale applications, reactivation is<br />

achieved <strong>by</strong> thermal methods, typically with multiple-hearth incinerators. Conse<br />

quently, on-site reactivation requires facilities for transporting spent and reacti<br />

vated carbon, carbon dewatering, and <strong>of</strong>f-gas treatment.<br />

The capital cost for the <strong>GAC</strong> process is a function <strong>of</strong> the treatment plant<br />

flow and the relative size <strong>of</strong> the contactor (i.e., EBCT). Operation and maintenance<br />

costs for the <strong>GAC</strong> facility also depends on the plant production rate and the rate at<br />

which spent carbon needs to be replaced or reactivated (i.e., carbon usage rate).<br />

Because the CUR determines the amount <strong>of</strong> carbon that must be reactivated, this<br />

parameter also determines the capital cost for the <strong>GAC</strong> reactivation facilities.<br />

Pretreatment (e.g., optimized coagulation; ozone and bi<strong>of</strong>iltration) can<br />

reduce the CUR, there<strong>by</strong> reducing the capital and O&M costs for the <strong>GAC</strong><br />

reactivation facility, as well as O&M costs for the <strong>GAC</strong> facility. The capital and<br />

O&M costs for pretreatment are functions <strong>of</strong> plant production rate and the design<br />

criteria for pretreatment.<br />

199


200 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> GA C <strong>Adsorption</strong><br />

Empty Bed Contact Time<br />

The EBCT is a measure <strong>of</strong> the volume <strong>of</strong> a carbon contactor, expressed in<br />

terms <strong>of</strong> the retention time in an empty contactor at design flow. Consequently, the<br />

EBCT is in direct proportion to the volume <strong>of</strong> carbon in a contactor, and longer<br />

EBCTs translate to larger contactors. Short EBCTs can be advantageous for welladsorbing<br />

specific micropollutants to reduce competition effects from other organics<br />

that may adsorb deeper in the <strong>GAC</strong>. Longer EBCTs are necessary to capture<br />

organic material that is more poorly adsorbed, which may nevertheless be problem<br />

atic. Because <strong>of</strong> the heterogeneous nature <strong>of</strong> NOM and the potential for DBF<br />

precursor material to be composed <strong>of</strong> a range <strong>of</strong> organics with varying adsorption<br />

characteristics, research for and application <strong>of</strong> <strong>GAC</strong> adsorption for DBF precursor<br />

material have typically focused on EBCTs between 10 and 20 minutes. As discussed<br />

in Chapter 6, experimental studies as a part <strong>of</strong> this research were limited to three<br />

EBCTs: 10 minutes, 15 minutes, and 20 minutes.<br />

The selection <strong>of</strong> EBCT may also be dependent upon the desired frequency<br />

<strong>of</strong> reactivation. For specific source water conditions, longer EBCTs will typically<br />

result in longer operation cycles and less frequent reactivation. Consequently, a<br />

system may increase EBCT to reduce the reactivation frequency and replacement<br />

<strong>of</strong> carbon in a contactor. The reactivation frequency, however, should not be<br />

confused with CUR. The CUR measures the capacity <strong>of</strong> carbon to adsorb organics<br />

from a water; consequently, it is related to both the EBCT and reactivation<br />

frequency as described in the following section. Therefore, although increasing the<br />

EBCT may decrease the reactivation frequency, it is entirely possible that the CUR<br />

will not change and that consequently the amount <strong>of</strong> carbon to be reactivated can<br />

remain constant. On the other hand, increasing the EBCT may decrease the CUR,<br />

thus permitting a reduction in reactivation frequency proportionally greater than the<br />

ratio <strong>of</strong> the EBCTs. Under these circumstances, cost trade-<strong>of</strong>fs can be considered<br />

between increasing EBCT and reducing the size <strong>of</strong> the reactivation facilities.<br />

Carbon Usage Rate<br />

As stated previously, the CUR represents the amount <strong>of</strong> <strong>GAC</strong> needed to<br />

treat a unit volume <strong>of</strong> water satisfying the finished water quality goal. CUR can be<br />

determined using the following formula which is based upon the length <strong>of</strong> time<br />

along the breakthrough curve for which the finished water quality is at or below the<br />

target level:<br />

CUR = EBCT. 28. 1,000,000<br />

T. 1,440 .7.48<br />

where CUR = carbon usage rate (Ib/mil gal)<br />

EBCT = empty bed contact time (minutes)<br />

28 = density <strong>of</strong> <strong>GAC</strong> (Ib/ft3)(= 0.45 g/cm 3 )<br />

1,000,000 = conversion factor (1,000,000 gal = 1 mil gal [3,785 m 3])<br />

T = breakthrough time (days)<br />

1,440 = conversion factor (1,400 minutes = 1 day)<br />

7.48 = conversion factor (7.48 gal = 1 ft3 [0.028 m 3])


Evaluation <strong>of</strong> Design Criteria and Costs 201<br />

Relevant breakthrough times for a given source water can be estimated from<br />

the results <strong>of</strong> a single-contactor RSSCT experiment. To represent a realistic mode<br />

<strong>of</strong> operation, however, the single-contactor results can be transformed into a<br />

blended effluent breakthrough curve. Mathematically simulated blended effluent<br />

breakthrough curves can be used to estimate CUR values that are more representa<br />

tive <strong>of</strong> those found in full-scale applications.<br />

Effect <strong>of</strong> Pretreatment<br />

<strong>Removal</strong> <strong>of</strong> contaminants prior to <strong>GAC</strong> treatment is expected to result in<br />

longer contactor operation prior to reactivation. Pretreatment <strong>of</strong> the source water to<br />

remove TOC and DBF precursors is also expected to result in lower CUR values<br />

compared to water that has not been pretreated. It should be noted, however, that the<br />

breakthrough curves developed during this study were based upon settled or filtered<br />

water from the participating utilities. For this reason, all the waters used for the<br />

RSSCT evaluation were pretreated using the existing treatment scheme (conven<br />

tional coagulation, flocculation, sedimentation, and filtration). Further optimiza<br />

tion for the removal <strong>of</strong> TOC and DBF precursors was studied <strong>by</strong> additional treatment<br />

<strong>of</strong> the settled or filtered water through the application <strong>of</strong> coagulant or <strong>by</strong> treat<br />

ment <strong>of</strong> the water with ozone and biotreatment, as discussed in Chapters 7 and 8,<br />

respectively.<br />

From a cost standpoint, the benefit <strong>of</strong> pretreatment is dependent upon<br />

whether the cost <strong>of</strong> the pretreatment is <strong>of</strong>fset <strong>by</strong> the reduction in <strong>GAC</strong> facility costs.<br />

In this chapter, the <strong>GAC</strong> facility costs with conventional pretreatment alone are<br />

compared with the costs <strong>of</strong> a <strong>GAC</strong> facility plus optimized coagulation or ozone with<br />

biotreatment, as practiced in this research effort, to determine whether pretreatment<br />

lowers overall facility costs for implementing <strong>GAC</strong> treatment.<br />

Determination <strong>of</strong> Design Criteria____________<br />

Breakthrough characteristics for TOC and chlorinated DBFs for various<br />

source waters were used to determine the CUR estimates for the participating<br />

utilities. Three different EBCT values (10, 15, and 20 minutes) were evaluated<br />

during the testing under pretreatment conditions currently used at each utility. The<br />

effect <strong>of</strong> pretreatment was evaluated for selected utilities for one EBCT value (15<br />

minutes). The following text presents a discussion <strong>of</strong> the water quality goals<br />

selected for the development <strong>of</strong> the design criteria and an estimate <strong>of</strong> CURs for<br />

various <strong>GAC</strong> treatment scenarios at the participating utilities.<br />

<strong>Water</strong> Quality Goals<br />

<strong>Water</strong> quality goals were selected to reflect the regulatory agenda <strong>of</strong> the<br />

USEPA. Proposed THM and HAA5 MCLs for Stage 2 <strong>of</strong> the D-<strong>DBP</strong> Rule were<br />

used as goals for the chlorinated DBFs. The Stage 2 goals were selected for two<br />

reasons: (1) the treatability studies in the ICR are based, in part, upon determining<br />

the design criteria and associated costs for systems to meet potential Stage 2<br />

requirements using <strong>GAC</strong> treatment; and (2) the Stage 1 MCLs are sufficiently high


202 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> GA C <strong>Adsorption</strong><br />

that operating times to breakthrough could not reliably be determined for three <strong>of</strong><br />

the four waters tested. The Stage 2 proposed MCLs for TTHM and HAAS are<br />

40 ug/L and 30 ug/L, respectively. It is recognized, however, that these MCLs are<br />

"placeholders" subject to renegotiation in the future.<br />

The selected goal for TOC is 1.0 mg/L. Again, a value <strong>of</strong> 2.0 mg/L has more<br />

regulatory relevance for the Stage 1 D-<strong>DBP</strong> Rule, in that enhanced coagulation, a<br />

treatment technique to remove DBF precursors in conventional treatment, is not<br />

required if the treated water TOC prior to continuous disinfection is less than 2.0<br />

mg/L. Similar to the case with DBFs, however, a treatment goal <strong>of</strong> 2.0 mg/L did not<br />

permit reliable estimates <strong>of</strong> operating times for three <strong>of</strong> the four waters evaluated.<br />

A 20 percent factor <strong>of</strong> safety was used to determine the maximum allow<br />

able values <strong>of</strong> these parameters. This safety factor, typical <strong>of</strong> what is used in<br />

practice, allows for operational variation and was endorsed <strong>by</strong> the water industry as<br />

an acceptable margin <strong>of</strong> safety during the regulatory negotiations for the D-<strong>DBP</strong><br />

Rule. As a result, the following were selected as the target parameter values after<br />

<strong>GAC</strong> treatment:<br />

• TOC = 0.8 mg/L<br />

• TTHM =32 ug/L<br />

• HAA5 =24 ug/L<br />

Single Contactor Breakthrough Curves<br />

Table 12.1 summarizes the single-contactor breakthrough times for reach<br />

ing the target TOC, THM, and HAA5 concentrations for the participating utilities.<br />

A logistic equation <strong>of</strong> the following form was used to fit the experimental data:<br />

f"(t) - ——— -r (12.2)<br />

"<br />

where A, B, and C are fitting parameters and t is breakthrough time.<br />

Equation 12.2 can be rearranged to show breakthrough time (t) as a function<br />

<strong>of</strong> effluent concentration (f(t)), which can be used to determine the breakthrough<br />

time in (days) for a given effluent concentration. The restructured equation is as<br />

follows:<br />

t = -_ «in<br />

C<br />

1 ( A<br />

-1 (12.3)<br />

B lf(t)<br />

Equation 12.3 was used to estimate the value <strong>of</strong>t, shown in the "scaled days"<br />

column in Table 12.1, for the single-contactor breakthrough curves. Corresponding<br />

estimates <strong>of</strong> CUR (using Equation 12.1) are also shown in Table 12.1.


Evaluation <strong>of</strong> Design Criteria and Costs 203<br />

Table 12.1 Estimates <strong>of</strong> CURs for the participating utilities<br />

Utility name<br />

Pretreatment<br />

type<br />

Controlling<br />

parameter and<br />

influent concentration<br />

EBCT<br />

(minutes)<br />

Scaled days<br />

Single<br />

Blended<br />

CUR (Ib/mil gal)<br />

Single<br />

Blended<br />

Cincinnati <strong>Water</strong><br />

Works<br />

Conventional<br />

coagulation<br />

TOC: 2.0 mg/L<br />

THM: 86 ug/L<br />

10<br />

15<br />

20<br />

10<br />

15<br />

20<br />

66<br />

80<br />

139<br />

64<br />

87<br />

146<br />

149<br />

169<br />

328<br />

248<br />

183<br />

876<br />

393<br />

490<br />

375<br />

404<br />

447<br />

357<br />

175<br />

231<br />

158<br />

105<br />

213<br />

59<br />

Optimized<br />

coagulation<br />

TOC: 1 .5 mg/L<br />

THM: 58 ug/L<br />

15<br />

15<br />

214<br />

241<br />

348<br />

1,046<br />

182<br />

162<br />

112<br />

37<br />

Ozone and<br />

biotreatment<br />

TOC: 1 .4 mg/L<br />

15<br />

104<br />

505<br />

377<br />

77<br />

City <strong>of</strong> Phoenix<br />

Conventional<br />

coagulation<br />

TOC: 2.3 mg/L<br />

THM: 71 ug/L<br />

10<br />

15<br />

20<br />

10<br />

15<br />

42<br />

83<br />

103<br />

56<br />

103<br />

104<br />

178<br />

208<br />

281<br />

363<br />

615<br />

470<br />

503<br />

463<br />

378<br />

250<br />

219<br />

251<br />

93<br />

107<br />

Optimized<br />

coagulation<br />

TOC: 1 .7 mg/L<br />

15<br />

155<br />

382<br />

251<br />

102<br />

Passaic Valley<br />

<strong>Water</strong><br />

Commission<br />

Conventional<br />

coagulation<br />

TOC: 3.2 mg/L<br />

10<br />

15 '<br />

20<br />

34<br />

63<br />

68<br />

63<br />

116<br />

128<br />

756<br />

616<br />

762<br />

415<br />

335<br />

408<br />

THM: 85 ug/L<br />

10<br />

15<br />

20<br />

50<br />

94<br />

94<br />

99<br />

199<br />

188<br />

525<br />

417<br />

553<br />

264<br />

196<br />

277<br />

HAAS: 57 ug/L<br />

10<br />

15<br />

144<br />

143<br />

394<br />

401<br />

181<br />

272<br />

66<br />

97<br />

Ozone and<br />

biotreatment<br />

TOC: 2.5 mg/L<br />

THM: 57 ug/L<br />

15<br />

15<br />

45<br />

172<br />

87<br />

487<br />

874<br />

226<br />

450<br />

80<br />

Florida Cities<br />

Conventional<br />

coagulation<br />

TOC: 7.4 mg/L<br />

15<br />

20<br />

4<br />

8<br />

7<br />

14<br />

10,056<br />

6,695<br />

5,410<br />

3,774<br />

THM: 238 ug/L<br />

10<br />

15<br />

20<br />

4<br />

6<br />

11<br />

8<br />

12<br />

19<br />

5,999<br />

6,003<br />

4,605<br />

3,284<br />

3,293<br />

2,712<br />

HAAS: 128 ug/L<br />

10<br />

15<br />

20<br />

12<br />

15<br />

21<br />

21<br />

24<br />

36<br />

2,125<br />

2,685<br />

2,476<br />

1,243<br />

1,593<br />

1,439<br />

Ozone and<br />

biotreatment<br />

TOC: 10 mg/L<br />

15<br />

13<br />

21<br />

3,062<br />

1,826


204 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> GA C <strong>Adsorption</strong><br />

Effect <strong>of</strong> Blending From Multiple Contactors<br />

As discussed earlier, blending <strong>of</strong> effluent from multiple parallel <strong>GAC</strong><br />

contactors decreases the CUR values for satisfying a given water quality target. The<br />

mathematical approach described in Chapter 6 was used to transform the singlecontactor<br />

breakthrough curves (Equation 12.2) into blended effluent curves. As<br />

discussed in Chapter 6, blended effluent breakthrough curve can be represented <strong>by</strong><br />

f - -Jf(t) dt (12.4)<br />

Substituting Equation 12.2 for f(t) in the preceding equation and performing<br />

integration, one obtains the following equation for the blended effluent break<br />

through curve results:<br />

f = A + -^- • ln(l + B • e~C * l ) - — • ln(l + B) (12.5)<br />

Equation 12.5 is transcendental (no direct solution for t); at very small values <strong>of</strong> t,<br />

which are not representative <strong>of</strong> <strong>GAC</strong> applications, it will predict an increase in the<br />

blended effluent concentrations (resulting from the presence <strong>of</strong> t in the denomina<br />

tor). This equation can be solved <strong>by</strong> trial and error for a target blended concentration<br />

to determine the breakthrough time to achieve the target. Breakthrough times and<br />

CURs for the blended effluent curve, calculated using these equations, are summa<br />

rized in Table 12.1.<br />

It should be noted that the underlying assumption in the derivation <strong>of</strong> the<br />

blended effluent equation is that the multiple contactors will be placed on-line (with<br />

fresh <strong>GAC</strong>) at equal intervals. If the frequency <strong>of</strong> placing a new contactor on-line<br />

is held constant, the actual concentration from the set <strong>of</strong> parallel contactors will<br />

fluctuate around the value <strong>of</strong> the target concentration with a mean value equal to the<br />

target concentration. The amplitude <strong>of</strong> fluctuation will depend on the number <strong>of</strong><br />

contactors.<br />

Cost Curve Development________________<br />

A computer program generated <strong>by</strong> Culp-Wesner-Culp (CWC Engineering<br />

S<strong>of</strong>tware, San Clemente, CA) was used to develop costs for <strong>GAC</strong> and pretreatment<br />

processes. The process components for the <strong>GAC</strong> and pretreatment options, along<br />

with the design parameters and the appropriate assumptions required for executing<br />

the CWC s<strong>of</strong>tware, are outlined in the following subsections.<br />

Process Components for <strong>GAC</strong><br />

Figure 12.1 shows a schematic <strong>of</strong> the assumed components for the <strong>GAC</strong><br />

process. These components include an influent pumping station, <strong>GAC</strong> contactors,<br />

backwash pumping station, carbon storage, and multiple-hearth incineration<br />

with <strong>of</strong>f-gas treatment for on-site reactivation. It was assumed that in a retr<strong>of</strong>it


Evaluation <strong>of</strong> Design Criteria and Costs 205<br />

application, plant filtered water will be pumped into the <strong>GAC</strong> adsorbers, which can<br />

be operated as either a gravity-flow or pressure-driven process.<br />

On-site reactivation <strong>of</strong> spent <strong>GAC</strong> was assumed for total carbon usage rates<br />

<strong>of</strong> 4,000 Ib/d (1,815 kg/d) or higher. A stand<strong>by</strong> furnace was assumed for reactivation<br />

facilities as a result <strong>of</strong> the significant maintenance requirement. Storage facilities<br />

for spent, reactivated, and makeup carbon were included. For smaller facilities<br />

(requiring less than 4,000 Ib <strong>GAC</strong>/d), <strong>of</strong>f-site disposal and replacement <strong>of</strong> spent<br />

<strong>GAC</strong> were assumed. Adequate storage facilities for the spent <strong>GAC</strong> were provided.<br />

Process Components for Pretreatment<br />

Two types <strong>of</strong> pretreatment were considered during this study: (1) optimized<br />

coagulation and (2) ozonation with biological filtration. Optimized coagulation was<br />

achieved either <strong>by</strong> an increase in the amount <strong>of</strong> alum addition or <strong>by</strong> a combination<br />

<strong>of</strong> lower coagulation pH at elevated alum dosage. Figure 12.2 shows a schematic <strong>of</strong><br />

the process components for the optimized coagulation option. Acid storage and<br />

pumping facilities were included for the optimized coagulation method that requires<br />

lowering the pH. Both methods for achieving optimized coagulation require<br />

additional alum storage, chemical feed facilities, and residuals-handling facilities.<br />

Sludge-drying beds were assumed for the residuals-handling process, although it is<br />

recognized that these may not be feasible in facilities with site constraints and<br />

climatic limitations. Nevertheless, sludge-drying beds can be land-intensive, and<br />

costs were included for land purchase, there<strong>by</strong> increasing the cost <strong>of</strong> this disposal<br />

option. For both optimized coagulation methods, caustic storage and pumping<br />

facilities were required to adjust the low coagulation pH to an acceptable distribu<br />

tion system pH.<br />

Figure 12.3 presents a schematic <strong>of</strong> the assumed process components for<br />

the ozonation and bi<strong>of</strong>iltration process. Settled water was assumed to be pumped<br />

into the ozone contactor. On-site generation <strong>of</strong> ozone from ambient air was assumed<br />

during the cost computations. No additional costs were attributed to the filtration<br />

process.<br />

Design Parameters<br />

Conceptual costs were calculated using the CWC model for eight distinct<br />

flow rates ranging from 0.7 to 270 mgd (2.65 to 1,022 ML/d). The selected flow rates<br />

represent USEPA flow categories 5 through 12. As shown in Table 12.2, these flow<br />

categories represent service area populations <strong>of</strong> 3,000 to greater than 1,000,000.<br />

A total dynamic head <strong>of</strong> 30 ft (9.1 m) was assumed for pumping filtered<br />

water to the <strong>GAC</strong> adsorbers. EBCTs used for developing the cost curves were 10,<br />

15, and 20 minutes. For carbon amounts less than 80,000 Ib (36,300 kg), standard<br />

pressure contactors were used. For larger carbon installations, gravity contactors<br />

were used. The minimum number <strong>of</strong> contactors was four for both types <strong>of</strong><br />

contactors, and the maximum number <strong>of</strong> pressure contactors was six. The maximum<br />

number <strong>of</strong> gravity contactors was determined <strong>by</strong> the number <strong>of</strong> contactors with a<br />

surface area less than 1,950 ft2 (180 m2). A hydraulic loading rate <strong>of</strong> 5 gpm/ft2 (12.2<br />

m/h) was assumed.<br />

Off-site reactivation was assumed for plants with average design flow less<br />

than 2.1 mgd (8 ML/d). Off-site reactivation was also assumed for spent carbon


206 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> GA C <strong>Adsorption</strong><br />

Table 12.2 Treatment plant flows considered for cost curves<br />

Category<br />

Population range<br />

Median population<br />

Average flow (mgd)<br />

5<br />

6<br />

7<br />

8<br />

9<br />

10<br />

11<br />

12<br />

3,301-10,000<br />

10,001-25,000<br />

25,001-50,000<br />

50,001-75,000<br />

75,001-100,000<br />

100,001-500,000<br />

500,001-1,000,000<br />

Greater than 1 ,000,000<br />

5,500<br />

15,500<br />

35,500<br />

60,000<br />

88,100<br />

175,000<br />

730,000<br />

1,550,000<br />

0.7<br />

2.1<br />

5<br />

8.8<br />

13<br />

27<br />

120<br />

270<br />

loading rates less than 4,000 Ib/d (1,815 kg/d) based upon available furnace sizes.<br />

For higher loadings, on-site carbon reactivation was used. CURs used for develop<br />

ing cost curves were 100, 300, and 500 Ib/mil gal (12, 36, and 60 g/m 3).<br />

The design <strong>of</strong> facilities for optimized coagulation was based upon the<br />

following chemical dosages:<br />

• 65 mg/L <strong>of</strong> additional alum for the alum-only option<br />

• 45 mg/L <strong>of</strong> additional alum and 15 mg/L <strong>of</strong> additional sulfuric acid<br />

for the alum-and-acid option<br />

• 20 mg/L <strong>of</strong> caustic<br />

These dosages were selected to approximate the range <strong>of</strong> dosages used in the<br />

research. It is recognized that these dosages are relatively large, and it is uncertain<br />

whether smaller dosages could have achieved the same effect as that described in<br />

Chapter 7. Such an optimization was beyond the scope <strong>of</strong> this research.<br />

In the alum-only option, the selected alum dosage is expected to lower the<br />

coagulation pH to approximately 7.0 for source waters with alkalinity <strong>of</strong> approxi<br />

mately 120 mg/L as CaCO 3 . For the alum-and-acid option, the combination <strong>of</strong> alum<br />

and acid dosages is expected to effect a similar pH depression. A 20-mg/L dosage<br />

<strong>of</strong> caustic is expected to raise the coagulation pH to a distribution system pH <strong>of</strong> 8.0<br />

for the same source. The design parameters for additional sludge-drying beds were<br />

based upon a loading rate <strong>of</strong> 12.5 lb/ft2 (61 kg/m2) with a filling cycle <strong>of</strong> 100 days.<br />

The design <strong>of</strong> ozonation facilities included in-plant pumping to provide an<br />

additional 30 ft <strong>of</strong> total dynamic head. The ozone contactor was assumed to have a<br />

contact time <strong>of</strong> 10 minutes. Costs were calculated for three different applied ozone<br />

dosages: 3, 5, and 10 mg/L.<br />

Cost Factors<br />

July 1995 cost indices were used in developing the cost curves (Engineer<br />

ing News Record 1995). Table 12.3 summarizes the relevant cost indices. Various<br />

cost factors were applied to the construction costs predicted <strong>by</strong> the CWC model to<br />

determine the overall capital cost <strong>of</strong> the facilities. These cost factors are summarized<br />

in Table 12.4. Table 12.5 summarizes the unit costs for various building and<br />

operating materials.


Evaluation <strong>of</strong> Design Criteria and Costs 207<br />

Table 12.3 Indices used in developing cost curves<br />

Cost Index<br />

ENR building cost<br />

ENR skilled labor<br />

ENR material prices<br />

Housing cost ($/ft2)<br />

Index value<br />

3,114<br />

4,945<br />

1,996<br />

125<br />

ENR = Engineering News Record<br />

Table 12.4 Factors applied to construction cost estimates<br />

Cost parameter Factor<br />

Sitework (%)<br />

Contractor's overhead and pr<strong>of</strong>it (%)<br />

Contingencies (%)<br />

Engineering and design (%)<br />

Mobilization and bonding (%)<br />

Legal and administration (%)<br />

Interest rate (%)<br />

Number <strong>of</strong> years<br />

Land cost ($/acre)<br />

15<br />

12<br />

15<br />

15<br />

3<br />

10<br />

7<br />

30<br />

25,000<br />

Source: Engineering News Record 1995<br />

Table 12.5 Unit costs for building and operating materials<br />

Cost parameter Unit cost<br />

Electricity ($/kW»h)<br />

Labor ($/hour)<br />

Diesel fuel ($/gal)<br />

Natural gas ($/ft3 )<br />

Building energy use (kW«h/ft2/year)<br />

Carbon attrition at 10 percent ($/lb)<br />

<strong>GAC</strong> ($/ton)<br />

Alum—liquid ($/ton)<br />

Sulfuric acid ($/ton)<br />

Sodium hydroxide—50% ($/ton)<br />

0.12<br />

30<br />

1.25<br />

0.006<br />

102.6<br />

0.9<br />

1,600<br />

230<br />

100<br />

350<br />

Source: Engineering News Record 1995


208 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> GA C <strong>Adsorption</strong><br />

Cost Curves<br />

Figure 12.4 shows capital and O&M costs for the <strong>GAC</strong> facilities for EBCTs<br />

<strong>of</strong> 10,15, and 20 minutes. An exponential model was used to fit the costs presented<br />

in this and the subsequent figures. The fitting parameters for the exponential model<br />

are shown on the figures. The capital costs shown on Figure 12.4 do not include<br />

carbon reactivation. The O&M cost for the <strong>GAC</strong> facilities is primarily associated<br />

with the pumping costs. The capital and O&M costs for on-site and <strong>of</strong>f-site<br />

reactivation are presented on Figures 12.5 and 12.6, respectively. These costs were<br />

generated for carbon usage rates <strong>of</strong> 100, 300, and 500 Ib/mil gal (12, 36, and 60<br />

g/m 3). It is evident from the cost curves that a majority <strong>of</strong> the capital cost is<br />

associated with the carbon contactors. In addition, the O&M cost for carbon<br />

reactivation is less than that for facility pumping (included in the <strong>GAC</strong> facility<br />

O&M) at lower CURs.<br />

Figure 12.7 shows capital and O&M costs for the optimized coagulation<br />

facilities. Costs for both methods for achieving optimized coagulation are illus<br />

trated on this figure. Figure 12.8 presents capital and O&M costs for the ozonation<br />

facilities for applied ozone dosages <strong>of</strong> 3, 5, and 10 mg/L.<br />

Verification <strong>of</strong> Costs<br />

Capital and O&M costs from the CWC model were compared with actual<br />

costs incurred during the construction and operation <strong>of</strong> the <strong>GAC</strong> facilities for the<br />

California <strong>Water</strong> Treatment Plant (WTP) at Cincinnati, Ohio. The California WTP<br />

consists <strong>of</strong> presedimentation, coagulation with alum, rapid sand filtration, <strong>GAC</strong><br />

adsorption, and disinfection with chlorine. The <strong>GAC</strong> adsorption facilities consist <strong>of</strong><br />

12 gravity contactors with an EBCT <strong>of</strong> 15 minutes and a hydraulic loading rate <strong>of</strong><br />

5 gpm/ft2 (12.2 m/h) at design flow. Spent carbon is reactivated <strong>by</strong> an on-site<br />

reactivation system. This system includes storage facilities for spent, reactivated,<br />

and makeup carbon and the reactivating furnace. The reactivating furnace consists<br />

<strong>of</strong> two multiple furnaces (both in service) designed to operate at a spent carbon rate<br />

<strong>of</strong> 80,000 Ib/d (36,300 kg/d). Results <strong>of</strong> the comparison <strong>of</strong> capital and O&M costs<br />

between the CWC model and the California WTP are shown in Table 12.6. From<br />

this table, it is evident that the CWC model was able to accurately predict the O&M<br />

costs for the Cincinnati facilities. The difference in the capital costs can be attributed<br />

to additional facilities, beyond the construction <strong>of</strong> the <strong>GAC</strong> contactor and reactiva<br />

tion facilities alone, that were included in the bid estimate. The $60 million bid<br />

estimate for the California WTP also included other processes, such as improve<br />

ments in the clearwell, addition <strong>of</strong> new maintenance area, modifications to the<br />

chemical feed facilities, and modifications to the pumping stations.<br />

Capital and O&M costs based on cost curves for optimized coagulation<br />

facilities were compared with cost estimates for the South WTP at Tempe, Ariz.<br />

These latter costs were generated during a preliminary engineering cost estimate<br />

(Malcolm Pirnie, Inc. 1994). The South WTP consists <strong>of</strong> presedimentation, coagu<br />

lation with alum, filtration, and disinfection with chlorine. Bench-scale studies<br />

indicated that in order for the WTP to comply with Stage 1 <strong>of</strong> the D-<strong>DBP</strong> Rule, the<br />

alum dosage needed (1) to be increased from 25 to 30 mg/L with pH lowered to 7.3,<br />

or (2) to be increased to 60 mg/L without any pH adjustment. Sulfuric acid was used<br />

at a concentration <strong>of</strong> 15 mg/L to adjust the pH to 7.3. Caustic soda was used at a


Evaluation <strong>of</strong> Design Criteria and Costs 209<br />

Table 12.6 Comparison <strong>of</strong> costs for the GAG processes_______________<br />

___ Cost California WTP* CWC modelf__________<br />

Capital costs (million $) 60 51<br />

O&M costs (cents/1,000 gal) 10 10<br />

* Design flow for California WTP is 175 mgd; average flow for California WTP is 120 mgd; the CUR for California WTP is 228<br />

Ib/mil gal for design flow and 333 Ib/mil gal for average flow,<br />

t A CUR <strong>of</strong> 300 Ib/mil gal and EBCT <strong>of</strong> 15 minutes were used in the CWC model.<br />

Table 12.7 Comparison <strong>of</strong> costs for optimized coagulation _______________<br />

Capital cost (million $) O&M Cost (cents/1,000 gal)<br />

Treatment South WTP* Cost curves South WTP* Cost curves<br />

Alum only 0.93 1.71 9.9 11.3<br />

Alum at low pH 0.95 1.57 8.5 9.8<br />

* Source: Malcolm Pirnie, Inc. (1994). Additional chemical dosages were adjusted to the selected dosages for this study.<br />

Design flow for South WTP is 40 mgd. Average flow for South WTP is 25 mgd.<br />

concentration <strong>of</strong> 20 mg/L to raise the coagulation pH to a distribution system pH <strong>of</strong><br />

8.0. At average plant flow, the increase <strong>of</strong> alum dose to 60 mg/L generated 3,960<br />

Ib/d (1,800 kg/d) additional solids, whereas the increase <strong>of</strong> alum dose to 30 mg/L<br />

with pH lowered to 7.3 generated 990 Ib/d (450 kg/d) additional solids. The<br />

additional sludge production was assumed to be accommodated <strong>by</strong> new drying beds<br />

constructed next to the existing drying beds. Results <strong>of</strong> the comparison <strong>of</strong> capital and<br />

O&M costs predicted <strong>by</strong> the cost curves and the South WTP are shown in Table 12.7.<br />

This table indicates that the cost curves overpredicted both capital and O&M costs<br />

for the South WTP. It should be noted, though, that the costs generated for the South<br />

WTP were very site specific and assumed use <strong>of</strong> some available facilities that were<br />

not actually in use. For example, at this facility, two sodium chlorite tanks that had<br />

been abandoned were converted to alum tanks to minimize the costs to the utility.<br />

The costs curves, on the other hand, assumed the construction <strong>of</strong> all new facilities.<br />

Capital and O&M costs based on cost curves for ozonation facilities were<br />

compared with costs presented in the document entitled Technology and Costs for<br />

the Control <strong>of</strong> Disinfectants and Disinfection By-Products (USEPA 1993), which<br />

was peer reviewed and validated during the D-<strong>DBP</strong> Rule regulatory negotiation.<br />

Results from this comparison are shown in Table 12.8. It is evident from the table<br />

that cost curves very accurately predicted the capital cost from the report. However,<br />

the CWC model appeared to overpredict the O&M cost.


210 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> GA C <strong>Adsorption</strong><br />

Table 12.8 Comparison <strong>of</strong> ozone facilities costs<br />

Capital cost (million $) O&M cost (cents/1,000 gal)<br />

Flow (mgd) USEPA report* Cost curves USEPA report Cost curves<br />

11<br />

210<br />

2.3<br />

19<br />

2.5<br />

21.6<br />

3.9<br />

3.2<br />

8.8<br />

8.8<br />

Note: The ozone concentration was 5 mg/L<br />

" * Source: US ERA (1993)<br />

<strong>GAC</strong> Treatment Costs for Participating Utilities____<br />

The costs curves presented in the previous section were used to determine<br />

the cost <strong>of</strong> retr<strong>of</strong>itting the participating utilities with <strong>GAC</strong> treatment process. Costs<br />

were developed for <strong>GAC</strong> treatment with or without optimized pretreatment at these<br />

facilities. A summary <strong>of</strong> all the costs for the participating utilities is included in<br />

Table 12.9. The following is a discussion <strong>of</strong> the additional treatment cost for each<br />

<strong>of</strong> the participating utilities.<br />

Cincinnati <strong>Water</strong> Works<br />

Cincinnati <strong>Water</strong>works treats Ohio River water in a 175-mgd (660-ML/d)<br />

WTP. The average production rate at this facility is 120 mgd (450 ML/d). The<br />

treatment facility at this utility consists <strong>of</strong> conventional treatment followed <strong>by</strong> <strong>GAC</strong><br />

adsorption in a 15-minute-EBCT contactor (at design flow). <strong>GAC</strong> is reactivated onsite.<br />

The RSSCT results with Ohio River water under conventional pretreatment<br />

suggested that the CUR estimates were not significantly affected <strong>by</strong> the EBCTs (see<br />

Table 12.1). Consequently, although the reactivation frequency was lower as EBCT<br />

increased, the total amount <strong>of</strong> carbon to be reactivated was similar. Therefore, there<br />

is little savings in reactivation capital for longeer EBCTs, and O&M costs and the<br />

overall treatment costs were lowest for the 10-minute EBCT, directly because <strong>of</strong><br />

the lower capital costs associated with this EBCT. It is important to recognize that<br />

the extensive studies performed at Cincinnati to arrive at the design criteria for the<br />

full-scale <strong>GAC</strong> facility were based upon considerations other than the removal <strong>of</strong><br />

DBF precursor material. The <strong>GAC</strong> process was installed to provide removal <strong>of</strong><br />

SOCs, which can intermittently be measured in the Ohio River as a result <strong>of</strong> spills.<br />

Consequently, the optimization for DBF precursor removal in this research cannot<br />

be compared directly to the constraints at the full-scale facility.<br />

Both optimized coagulation and ozone with biotreatment were studied as<br />

pretreatment for this source water. Estimated CUR values were significantly<br />

lowered <strong>by</strong> both pretreatment schemes—<strong>by</strong> one-half and two-thirds for optimized<br />

coagulation and ozone with biotreatment, respectively. The overall treatment costs<br />

with the pretreatment, however, were found to be higher because <strong>of</strong> the costs<br />

associated with the pretreatment. Therefore, the reduction in costs for <strong>GAC</strong><br />

treatment, primarily in capital and O&M for reactivation, were not sufficient to<br />

<strong>of</strong>fset the increase in costs associated with the pretreatment. It is important to


Evaluation <strong>of</strong> Design Criteria and Costs 211<br />

Table 12.9 Estimates <strong>of</strong> <strong>GAC</strong> treatment costs for the participating utilities<br />

Utility name<br />

Controlling<br />

Pretreatment parameter and<br />

type influent concentration<br />

EBCT<br />

(minutes)<br />

Treatment cost (cents/1 ,000 gal)<br />

Pretreatment <strong>GAC</strong> Total<br />

Cincinnati <strong>Water</strong><br />

Works<br />

Conventional<br />

coagulation<br />

TOC: 2.0 mg/L<br />

THM: 86 ug/L<br />

10<br />

15<br />

20<br />

10<br />

15<br />

20<br />

17.7<br />

20.7<br />

21.9<br />

16.5<br />

20.4<br />

20.3<br />

17.7<br />

20.7<br />

21.9<br />

16.5<br />

20.4<br />

20.3<br />

Optimized<br />

coagulation<br />

TOC: 1 .5 mg/L<br />

THM: 58 ug/L<br />

15<br />

15<br />

11.2 18.6<br />

11.2 17.4<br />

29.8<br />

28.6<br />

Ozone<br />

and biotreatment<br />

TOC: 1 .4 mg/L<br />

15<br />

10.8 18.0<br />

28.8<br />

City <strong>of</strong> Phoenix<br />

Conventional<br />

coagulation<br />

TOC: 2.3 mg/L<br />

THM: 71 ug/L<br />

10<br />

15<br />

20<br />

10<br />

15<br />

21.4<br />

23.4<br />

27.0<br />

18.8<br />

21.6<br />

21.4<br />

23.4<br />

27.0<br />

18.8<br />

21.6<br />

Optimized<br />

coagulation<br />

TOC: 1 .7 mg/L<br />

15<br />

11.6 21.5<br />

33.1<br />

Passaic Valley<br />

<strong>Water</strong><br />

Commission<br />

Conventional<br />

coagulation<br />

TOC: 3.2 mg/L<br />

THM: 85 ug/L<br />

HAA5: 57 ug/L<br />

10<br />

15<br />

20<br />

10<br />

15<br />

20<br />

10<br />

15<br />

28.6<br />

30.2<br />

35.7<br />

25.5<br />

27.6<br />

33.0<br />

17.5<br />

25.9<br />

28.6<br />

30.2<br />

35.7<br />

25.5<br />

27.6<br />

33.0<br />

17.5<br />

25.9<br />

Ozone and<br />

biotreatment<br />

TOC: 2.5 mg/L<br />

THM: 57 ug/L<br />

15<br />

15<br />

13.2 32.7<br />

13.2 25.6<br />

45.9<br />

38.8<br />

Florida Cities<br />

Conventional<br />

coagulation<br />

and s<strong>of</strong>tening<br />

TOC: 7.4 mg/L<br />

THM: 238 ug/L<br />

HAAS: 1 28 ug/L<br />

15<br />

20<br />

10<br />

15<br />

20<br />

10<br />

15<br />

20<br />

228.3<br />

212.5<br />

152.9<br />

194.7<br />

195.7<br />

120.6<br />

167.8<br />

175.5<br />

228.3<br />

212.5<br />

152.9<br />

194.7<br />

195.7<br />

120.6<br />

167.8<br />

175.5<br />

Ozone<br />

and biotreatment<br />

TOC: 10 mg/L<br />

15<br />

57.3 171.5<br />

228.8


2/2 <strong>Removal</strong> <strong>of</strong><strong>DBP</strong> <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

recognize that a relatively high coagulant dosage was used in this research.<br />

Consequently, if CURs could be similarly improved at lower coagulant dosages,<br />

the cost-effectiveness <strong>of</strong> optimized coagulation pretreatment may improve. Further,<br />

it should be noted that the benefit <strong>of</strong> ozone with biotreatment was measured only in<br />

terms <strong>of</strong> costs associated with DBF reduction in subsequent treatment <strong>by</strong> <strong>GAC</strong>.<br />

Other benefits <strong>of</strong> ozone with biotreatment, such as disinfection and improvement in<br />

biostability, could not be factored into this conceptual cost analysis.<br />

The treatment costs were also observed to be lower for satisfying the DBF<br />

goals rather than satisfying the TOC goal. Because the TOC goal is not related to<br />

any specific regulatory constraint at this time, meeting the Stage 2 DBF goals may<br />

be a more relevant comparison for <strong>GAC</strong> costs.<br />

City <strong>of</strong> Phoenix<br />

The City <strong>of</strong> Phoenix operates five different surface water treatment facili<br />

ties, four <strong>of</strong> which use water from the Salt River Project (SRP). The average design<br />

capacity <strong>of</strong> SRP plants is 130 mgd (490 ML/d), and the average annual water<br />

production rate at these facilities is 75 mgd (280 ML/d). RSSCTs were performed<br />

for 10-, 15-, and 20-minute EBCTs using filtered water (see Table 12.1). Conven<br />

tional treatment at these utilities includes coagulation for particulate removal<br />

followed <strong>by</strong> dual media filtration.<br />

Similar to ORW, CUR estimates for the blended effluent curve using Salt<br />

River Project water were found to be relatively unaffected <strong>by</strong> EBCT. Consequently,<br />

the 10-minute EBCT exhibited the lowest conceptual cost.<br />

Optimized coagulation was found to reduce the CUR for the Salt River<br />

Project water <strong>by</strong> approximately 50 percent. The overall treatment cost with<br />

optimized coagulation, however, again was found to be higher because <strong>of</strong> the<br />

relatively high coagulation cost under the selected conditions. As shown on Table<br />

12.9, the treatment cost estimates are slightly lower if only the THM target is to be<br />

satisfied.<br />

Passaic Valley <strong>Water</strong> Commission<br />

The Passaic Valley <strong>Water</strong> Commission treats Passaic River water in a<br />

110-mgd (415-ML/d) capacity WTP. The average production rate at this utility is<br />

50 mgd (190 ML/d). Three alternative EBCT values were studied using this source<br />

water. The CUR estimate was found to be lower at a 15-minute EBCT, than at either<br />

a 10- or 20-minute EBCT (see Table 12.1). The overall treatment costs, however,<br />

were relatively unaffected <strong>by</strong> the differences in the CUR estimate because the<br />

majority <strong>of</strong> the capital costs were associated with the carbon contactors. The lowest<br />

overall <strong>GAC</strong> treatment costs at this utility were found to be associated with a 10-<br />

minute EBCT.<br />

Ozonation followed <strong>by</strong> biological filtration was studied as a pretreatment<br />

for the <strong>GAC</strong> process for this source water. No improvement in CUR was observed<br />

as a result <strong>of</strong> this pretreatment scheme. As a result <strong>of</strong> the additional cost <strong>of</strong><br />

ozonation, the pretreatment-<strong>GAC</strong> process trains were found to be the most expen<br />

sive. As with the Cincinnati and Phoenix data, the treatment costs for satisfying the<br />

<strong>DBP</strong> goals were lower compared to that for TOC.


Evaluation <strong>of</strong> Design Criteria and Costs 213<br />

Florida Cities <strong>Water</strong> Company<br />

A groundwater sample from Florida Cities <strong>Water</strong> Company was used in the<br />

RSSCT experiments to determine the breakthrough characteristics. This water<br />

sample was relatively high in NOM, as indicated <strong>by</strong> the high TOC and DBF<br />

formation potentials. The participating utility for this source water is a small facility<br />

with a design capacity <strong>of</strong> 1 mgd (3.78 ML/d) and an average production rate <strong>of</strong> 0.7<br />

mgd (2.65 ML/d). The CUR estimates for this source water were much higher than<br />

the CUR ranges evaluated to prepare the cost curves. For this reason, the costs<br />

shown for this utility were extrapolated and may not be realistic.<br />

As a result <strong>of</strong> the small size <strong>of</strong> the treatment facility and the high NOM<br />

content in the source water, the estimated overall cost <strong>of</strong> <strong>GAC</strong> treatment is very high.<br />

Ozonation and bi<strong>of</strong>iltration <strong>of</strong> this water prior to <strong>GAC</strong> treatment, however, resulted<br />

in a similar overall treatment cost. The treatment costs were found to be significantly<br />

lower if the process is controlled <strong>by</strong> effluent DBF concentration rather than the<br />

effluent TOC concentration.<br />

Implications for Process Optimization_________<br />

The results presented in the previous section yield several implications<br />

associated with process optimization. The following text discusses these implica<br />

tions in terms <strong>of</strong> optimizing EBCT and the impact <strong>of</strong> pretreatment on <strong>GAC</strong><br />

efficiency for the removal <strong>of</strong> DBF precursors.<br />

Empty Bed Contact Time<br />

The cost <strong>of</strong> <strong>GAC</strong> treatment process is primarily controlled <strong>by</strong> the size <strong>of</strong> the<br />

contactors (a function <strong>of</strong> EBCT) and the reactivation process (a function <strong>of</strong> the<br />

CUR). Of these two components, the bulk <strong>of</strong> the costs, in terms <strong>of</strong> both capital and<br />

unit costs (after amortizing the capital investment), is in the contactors. Conse<br />

quently, reducing the size <strong>of</strong> the contactors has the greatest impact on the overall cost<br />

reduction <strong>of</strong> the <strong>GAC</strong> facility.<br />

The experimental observation <strong>of</strong> NOM breakthrough curves, as indicated<br />

<strong>by</strong> the breakthrough <strong>of</strong> TOC and DBF precursors, suggested that a longer EBCT<br />

does not necessarily improve the CUR. Therefore, the size <strong>of</strong> the reactivation<br />

process does not decrease as EBCT increases, there<strong>by</strong> negating any potential cost<br />

benefit. Because the CUR is relatively unaffected <strong>by</strong> EBCT in the 10 to 20 minute<br />

range evaluated in this study for these waters, it is most economical to minimize the<br />

EBCT. It is recognized, however, that a utility may place a practical minimum value<br />

on the duration <strong>of</strong> contactor operation prior to reactivation; the duration <strong>of</strong> contactor<br />

operation will decrease as the EBCT decreases for a constant CUR. Therefore, the<br />

EBCT should not be reduced to such a degree that the period <strong>of</strong> contactor operation<br />

is so short as to result in other O&M related constraints.<br />

Further, it is recognized that as EBCT continues to decrease, eventually the<br />

CUR will begin to increase as premature breakthrough <strong>of</strong> difficult-to-adsorb NOM<br />

and DBF precursors occurs. This was not found for the waters in this research at an<br />

EBCT <strong>of</strong> 10 minutes; at some value below 10 minutes, though, it would be expected<br />

to occur. Therefore, it is recommended that utilities consider evaluating EBCTs


214 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

lower than 10 minutes to determine the EBCT at which the CUR begins to increase<br />

as a result <strong>of</strong> premature breakthrough. As long as the period <strong>of</strong> contactor operation<br />

is acceptable at this minimum EBCT before CUR begins increasing, this EBCT will<br />

yield the most economical facility.<br />

Finally, because the majority <strong>of</strong> the cost is associated with the contactors,<br />

it still may be advantageous from a cost standpoint to allow the CUR to increase a<br />

certain amount as EBCT decreases. At some point, the CUR will increase to an<br />

extent that the capital and O&M costs associated with the reactivation furnaces will<br />

<strong>of</strong>fset the cost savings <strong>of</strong> reducing the EBCT. This is the true optimization point<br />

from a cost standpoint, but it was not reached at an EBCT as low as 10 minutes for<br />

the waters used in this research.<br />

Pretreatment Optimization<br />

Pretreatment is expected to increase the length <strong>of</strong> operation <strong>of</strong> a <strong>GAC</strong><br />

facility as it reduces the CUR. The overall cost <strong>of</strong> precursor removal via enhanced<br />

pretreatment and <strong>GAC</strong> adsorption, however, must be considered to determine the<br />

benefit <strong>of</strong> the pretreatment. The following paragraphs discuss implications <strong>of</strong><br />

optimized coagulation and ozone with biotreatment in terms <strong>of</strong> cost-effectively<br />

improving <strong>GAC</strong> performance for the removal <strong>of</strong> DBF precursors.<br />

Optimized Coagulation<br />

The experimental studies conducted during this research focused on rela<br />

tively extreme conditions for optimizing precursor removal via coagulation; in<br />

addition, the "optimized" coagulation conditions used in this study should not be<br />

confused with the regulatory requirements <strong>of</strong> "enhanced" coagulation in Stage 1 <strong>of</strong><br />

the D-<strong>DBP</strong> Rule. Although a 50 percent reduction in CUR was achieved under the<br />

selected pretreatment conditions, the overall costs were higher as a result <strong>of</strong> the cost<br />

associated with the high dosages used to define "optimized" coagulation conditions.<br />

If the 50 percent reduction in CUR yielded under the optimized coagulation<br />

conditions used in this study could be achieved using lower dosages, an overall<br />

system cost reduction could be achieved if the cost <strong>of</strong> additional coagulation could<br />

be reduced to less than approximately 5 cents/1,000 gal (approximately 10 cents per<br />

1,000 gal [3,785 L] was the estimated cost for the treatment in this study).<br />

Alternatively, the CUR could be held constant <strong>by</strong> pretreatment, and the<br />

EBCT could be decreased. Although this effort was beyond the scope <strong>of</strong> work for<br />

this research, lowering EBCT at a constant CUR may have a greater potential to<br />

reduce the overall system costs because the majority <strong>of</strong> the cost is associated with<br />

the contactors. It is recommended that utilities, when evaluating the benefits <strong>of</strong><br />

pretreatment, consider both (1) reducing CUR while holding EBCT constant and (2)<br />

reducing EBCT while reducing CUR. Finally, it is important for a utility to find the<br />

breakpoint at which only marginal reductions in CUR and EBCT are achieved for<br />

large increases in costs associated with coagulation. At this point, it is not necessary<br />

to further enhance the pretreatment processes to improve <strong>GAC</strong> performance.<br />

Ozone With Biotreatment<br />

Pretreatment using ozone with biotreatment was found to be effective in<br />

lowering the CUR values for removal <strong>of</strong> <strong>DBP</strong> precursors with postfilter contactors<br />

in two <strong>of</strong> the three waters evaluated: ORW and FGW. Further, the reduction in


Evaluation <strong>of</strong> Design Criteria and Costs 215<br />

reactivation costs associated with this reduction in CUR was sufficient to justify the<br />

cost <strong>of</strong> ozone with biotreatment for FGW. It must also be recognized that other<br />

benefits associated with ozone and biotreatment, such as biostability <strong>of</strong> the treated<br />

water and disinfection benefits, were not factored into this analysis. The costbenefit<br />

analysis <strong>of</strong> ozone with biotreatment considered only the improvemed <strong>GAC</strong><br />

adsorption for the removal <strong>of</strong> DBF precursor compounds.<br />

Cost comparisons at FGW implied that the higher the CUR, the better the<br />

opportunity for any pretreatment to become cost-effective if the percentage reduc<br />

tion in CUR remains constant. That is, if a 50 percent reduction in CUR can be<br />

achieved for two different waters, one with an initial CUR <strong>of</strong> 100 Ib/mil gal (12<br />

kg/ML) and the other with an initial CUR <strong>of</strong> 500 Ib/mil gal (60 kg/ML) the cost<br />

benefit will be greater for the latter. Therefore, optimizing pretreatment provides<br />

better potential for lowering overall system cost as the CUR increases. Similarly, if<br />

an equivalent reduction in EBCT can be achieved as the CUR is lowered, then<br />

potentially even greater benefits can be achieved <strong>by</strong> pretreatment.


276 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> GA C <strong>Adsorption</strong><br />

Spent<br />

Carbon Storage<br />

Off-Site<br />

Reactivation<br />

Filtered <strong>Water</strong>-<br />

Spent/Reactvated/Makeup<br />

Carbon Storage<br />

Plant<br />

Pumping Station<br />

Multiple<br />

Hearth Furnace<br />

<strong>GAC</strong> Adsorber<br />

Backwash<br />

Pumping Station<br />

Figure 12.1 Schematic <strong>of</strong> process components for <strong>GAC</strong><br />

Acid Storage<br />

Acid<br />

Pumping Station<br />

Raw <strong>Water</strong><br />

Flocculaton/Sedimentatjon Basin<br />

Fitters<br />

-\ I<br />

Alum Storage<br />

Solar Drying Beds<br />

Alum<br />

! Pumping Station<br />

I Caustic Storage<br />

Caustic<br />

Pumping Station<br />

Figure 12.2 Schematic <strong>of</strong> process components for optimized coagulation


Evaluation <strong>of</strong> Design Criteria and Costs 217<br />

Bi<strong>of</strong>Stratton<br />

Settled <strong>Water</strong><br />

Plant Pumping Station Ozone Contactor<br />

Ozone Generator<br />

Figure 12.3 Process schematic for ozonation and bi<strong>of</strong>iltration treatment


218 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

80<br />

70 ^<br />

I 6M<br />

5 50 ^<br />

I «<br />

"5. 30 -<br />

O<br />

20 -<br />

y=-338.4-exp(-0.0008894-x)+340.1<br />

y=-320.8-exp(-0.0007332-x)+322.4<br />

y=-208.0-exp(-0.0008516-x}+209.3<br />

• EBCT = 10 min<br />

• EBCT =15 min<br />

+ EBCT = 20 min<br />

Capital Cost<br />

10 -<br />

0 -<br />

70<br />

3 10 30 100 300 1000<br />

Plant Capacity (mgd)<br />

O&M Cost<br />

co<br />

_o<br />

§><br />

8<br />

in<br />

o<br />

O<br />

60 -<br />

50 -<br />

40 -<br />

30 -<br />

20 -<br />

y=79.16-exp(-0.5031<br />

y=74.75-exp(-0.512-x)+5.315<br />

y=41.23'exp(-0.7425->)+5.64<br />

• EBCT =10 min<br />

• EBCT = 15 min<br />

• EBCT = 20 min<br />

10 -<br />

3 10 30 100<br />

Plant Capacity (mgd)<br />

300 1000<br />

Figure 12.4 Cost curves for <strong>GAC</strong> facilities


Evaluation <strong>of</strong> Design Criteria and Costs 219<br />

25<br />

20 -<br />

CUR = 100lb/milgal<br />

CUR = 300 Ib/ma gal<br />

CUR = 500 Ib/mil gal<br />

Capital Cost<br />

(0<br />

O<br />

= 15 -\<br />

y=-98.18-exp(-0.0005836-xJ+103.4<br />

y=-552.2-exp(-0.00006584-^+557.1<br />

Tn<br />

o<br />

O<br />

I<br />

Q.<br />

ra<br />

O<br />

10 -<br />

5 -<br />

y=0.0084-x+5.682<br />

in<br />

jO<br />

15<br />

O)<br />

25<br />

20 -<br />

15 -<br />

10 30 100 300<br />

Plant Capacity (mgd)<br />

O&M Cost<br />

• CUR =100 Ib/mil gal<br />

• CUR = 300 Ib/mil gal<br />

• CUR = 500 Ib/mil gal<br />

y=18.46-exp(-0.06274o$+8.425<br />

1000<br />

10 -<br />

y=14.53-exp(-0.06216<br />

w<br />

o<br />

O<br />

5 -<br />

0 -<br />

y=-0.0032-x*-2.584<br />

10 30 100 300<br />

Plant Capacity (mgd)<br />

1000<br />

Figure 12.5 Cost curves for on-site reactivation


220 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

• CUR = 100 Ib/mil gal<br />

• CUR = 300 Ib/ma gal<br />

• CUR = 500 Ib/mil gal<br />

Capital Cost<br />

2 -<br />

(O<br />

5 H<br />

Q.<br />

CO<br />

O<br />

0 -<br />

y=-162.4-exp(-0.0004874-^+162.5<br />

i.149-exp(0.03111-x)-5.977<br />

y=29.09-exp(0.01386-x)-29.06<br />

~l———————————————I—<br />

1 3 10<br />

Plant Capacity (mgd)<br />

30<br />

O&M Cost<br />

7 -<br />

in<br />

£"co<br />

o><br />

6-<br />

y=7.28<br />

36<br />

• CUR = 100 Ib/mil gal<br />

• CUR = 300 Ib/mil gal<br />

• CUR = 500 Ib/mil gal<br />

4<br />

u<br />

3 3 H<br />

08<br />

° 2-\<br />

y=1.46<br />

-•—•-<br />

3 10<br />

Plant Capacity (mgd)<br />

30<br />

Figure 12.6 Cost curves for <strong>of</strong>f-site reactivation


Evaluation <strong>of</strong> Design Criteria and Costs 221<br />

10<br />

8 -<br />

Alum-only<br />

Alum-at low pH<br />

Capital Cost<br />

ID<br />

i 6<br />

y=-15.21 -exp(-0.003189->)-H 5.50 -<br />

y=-12.68-exp(-0.003523-x)+12.96<br />

(A<br />

O<br />

To<br />

'a.<br />

CO<br />

O<br />

4 -<br />

2 -<br />

0 -<br />

10 30 100 300 1000<br />

Plant Capacity (mgd)<br />

20<br />

18 -<br />

O&M Cost<br />

Alum-only<br />

Alum-at low pH<br />

W<br />

O<br />

1 16<br />

O)<br />

y=14.23-exp(-0.8034-x)+11.33<br />

^ 14 -<br />

y=14.30-exp(-0.7244-x)+9.782<br />

I<br />

9f 12 -<br />

O<br />

5<br />

o 10 -<br />

8<br />

1 3 10 30 100<br />

Plant Capacity (mgd)<br />

300 1000<br />

Figure 12.7 Cost curves for optimized coagulation


co<br />

en<br />

O<br />

o> •o<br />

O<br />

o<br />

w<br />

§<br />

CO<br />

8-<br />

o<br />

•3<br />

5'<br />

(Q<br />

JO<br />

CO<br />

a<br />

o<br />

I<br />

en<br />

08M Cost (cents/1,000 gallons)<br />

o (ft<br />

to<br />

o<br />

o o<br />

Capital Cost ($-millions)<br />

ro<br />

o<br />

ro<br />

en<br />

O Q)<br />

T3<br />

01<br />

O<br />

CA><br />

O<br />

M en<br />

O<br />

0)<br />

T><br />

01<br />

O.<br />

*< ?v<br />

O<br />

S<br />

O<br />

Q><br />

3<br />

co _>.<br />

B 8<br />

CO<br />

Q.<br />

CO<br />

o<br />

o<br />

O<br />

o<br />

0) o


Chapter 13<br />

Conclusions and Recommendations<br />

This chapter summarizes the conclusions and recommendations reached<br />

based upon the work conducted in this research effort.<br />

Verification <strong>of</strong> RSSCT Results_____________<br />

• A comparison <strong>of</strong> breakthrough curves from full-scale application,<br />

pilot-scale testing, and RSSCTs indicated that the rapid test<br />

procedure can be successfully used to predict NOM breakthrough<br />

behavior in terms <strong>of</strong> TOC, UV254, and several chlorinated DBFs.<br />

The importance <strong>of</strong> representative sample collection, however, was<br />

highlighted during this study. In order to use RSSCTs for the<br />

prediction <strong>of</strong> breakthrough characteristics as a function <strong>of</strong> seasonal<br />

change, several RSSCTs should be performed with batches <strong>of</strong><br />

influent waters that represent the seasons <strong>of</strong> interest.<br />

• Most raw and conventionally treated waters evaluated during this<br />

study contained small amounts <strong>of</strong> biodegradable organic matter. For<br />

this reason, the biodegradation occurring in the full-scale and pilotscale<br />

applications did not result in any significant differences with<br />

the RSSCT applications. Further, breakthrough behavior between<br />

the RSSCT and the pilot or full scale could be compared on the<br />

basis <strong>of</strong> <strong>GAC</strong> effluent concentration as a proportion <strong>of</strong> <strong>GAC</strong> influent<br />

concentration (C/C0), there<strong>by</strong> allowing a comparison <strong>of</strong> the relative<br />

adsorbability <strong>of</strong> a constituent rather than the absolute concentration<br />

if the source water varied significantly at full or pilot-scale during<br />

the period <strong>of</strong> comparison. <strong>GAC</strong> treatment following ozonation,<br />

however, is expected to include a significant biodegradation<br />

component. For this type <strong>of</strong> pretreatment, RSSCTs should be<br />

performed on-site to evaluate breakthrough characteristics. The time<br />

required to ship samples to a laboratory for RSSCTs may<br />

significantly alter the NOM characteristics and hence the adsorption<br />

behavior.<br />

• A statistical method has been developed during this study to<br />

compare breakthrough curves developed from RSSCTs with the<br />

curves from full- and pilot-scale tests. The statistical test was based<br />

upon the paired t-test procedure. The qualitative inferences were<br />

adequately matched <strong>by</strong> the statistical tests, although one must take<br />

223


224 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

certain precautions when interpreting values from statistical tests in<br />

the absence <strong>of</strong> visual comparison <strong>of</strong> the breakthrough curves.<br />

• The formation <strong>of</strong> DBFs was evaluated under two sets <strong>of</strong> chlorination<br />

conditions: SDS conditions and UFC. SDS conditions were used to<br />

conduct experiments to simulate the impact <strong>of</strong> distribution system<br />

conditions, specific to a given utility, on DBF formation.<br />

Alternatively, the UFC were used to study fundamental adsorption<br />

behavior as a function <strong>of</strong> source water variation and were therefore<br />

applied uniformly across all waters for comparison. Because SDS<br />

conditions reflected the conditions at a particular water utility, these<br />

results were significantly different, in some cases, relative to the<br />

UFC. The adsorption behavior, however, was found to be predicted<br />

similarly <strong>by</strong> either condition type when the effluent concentrations<br />

were reported on a normalized basis with respect to the influent<br />

concentrations.<br />

Impact <strong>of</strong> <strong>GAC</strong> Treatment on<br />

NOM Characteristics and DBF Yield_________<br />

• For the raw source waters analyzed during this study, humic<br />

fractions slightly dominated the NOM content as characterized <strong>by</strong><br />

DOC and UV,54 . The specific DBF yield <strong>of</strong> the humic fraction was<br />

also observed to be slightly higher than that <strong>of</strong> the nonhumic<br />

fraction under equalized Br:DOC ratios. Conventional treatment<br />

removed a greater amount <strong>of</strong> the humic fraction, resulting in a<br />

predominantly nonhumic NOM in the <strong>GAC</strong> influent water.<br />

Additional optimized coagulation <strong>of</strong> the conventionally coagulated<br />

water removed both humic and nonhumic fractions in approximately<br />

equal proportions. Ozonation followed <strong>by</strong> biotreatment <strong>of</strong> the<br />

conventionally treated source waters resulted in a relatively more<br />

predominant nonhumic fraction.<br />

• In the adsorption experiments, the initial samples <strong>of</strong> the<br />

nonadsorbable fraction <strong>of</strong> NOM were observed to be almost<br />

completely nonhumic in nature. The nonhumic fraction broke<br />

through rather quickly and reached a plateau after a short period <strong>of</strong><br />

operation, while the humic fraction was removed more efficiently<br />

during <strong>GAC</strong> adsorption for a significant portion <strong>of</strong> the breakthrough<br />

curve. The humic fraction broke through toward the end <strong>of</strong> the<br />

adsorption experiments, when C/C0 exceeded 0.5 to 0.6 for TOC.<br />

• Molecular size fractionation <strong>of</strong> the source water samples indicated<br />

that the majority <strong>of</strong> the DOC was in the intermediate range (500-<br />

3,000 MS) with the remaining fractions approximately evenly<br />

distributed between 3,000 MS fractions. Conventional<br />

treatment was found to effectively remove the >3,000 MS fraction,<br />

whereas the other fractions remained relatively unaffected.<br />

Optimized coagulation <strong>of</strong> the conventionally treated water removed<br />

DOC mostly from the intermediate MS range, whereas ozonation


Conclusions and Recommendations 225<br />

followed <strong>by</strong> biotreatment resulted in a shift toward smaller MS<br />

fractions.<br />

• The initial samples from <strong>GAC</strong> adsorption experiments indicated that<br />

the nonadsorbable fraction was almost entirely composed <strong>of</strong> the<br />


226 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> GA C <strong>Adsorption</strong><br />

• Increasing EBCT resulted in longer run times for the <strong>GAC</strong><br />

contactor. The CUR, however, was not significantly affected <strong>by</strong><br />

increasing EBCT between 10 and 20 minutes, implying that the total<br />

amount <strong>of</strong> carbon used (and, hence, carbon requiring reactivation) is<br />

equivalent in this EBCT range. Breakthrough curves in which the<br />

effluent concentrations were plotted as a function <strong>of</strong> number <strong>of</strong> bed<br />

volumes processed at the time <strong>of</strong> sample collection indicated no<br />

significant differences with respect to EBCT values.<br />

• Additional optimized coagulation <strong>of</strong> the conventionally treated<br />

water resulted in significant removal <strong>of</strong> NOM and increased <strong>GAC</strong><br />

run times <strong>by</strong> 129 to 148 percent. It should be noted, however, that<br />

the optimized coagulation dosages examined during this study were<br />

significantly greater than those required for enhanced coagulation as<br />

a part <strong>of</strong> the D-<strong>DBP</strong> Rule.<br />

• Ozonation followed <strong>by</strong> biotreatment resulted in a shift <strong>of</strong> the NOM<br />

toward more nonhumic fractions and smaller MS fractions. The<br />

benefit <strong>of</strong> this pretreatment scheme in lowering the influent NOM<br />

concentration was somewhat <strong>of</strong>fset <strong>by</strong> the shift toward the<br />

nonadsorbable fraction. The ultimate effect <strong>of</strong> this pretreatment<br />

scheme was that the CUR values were not significantly lowered. In<br />

two <strong>of</strong> the source waters (i.e., FGW and ORW), this pretreatment<br />

scheme resulted in lower CUR values. It should be noted, however,<br />

that the benefits <strong>of</strong> ozone with biotreatment were evaluated only in<br />

terms <strong>of</strong> their impact on the removal <strong>of</strong> DBF precursors <strong>by</strong> <strong>GAC</strong><br />

adsorption. An evaluation <strong>of</strong> other benefits, such as disinfection and<br />

biological stability, was not a part <strong>of</strong> this work.<br />

Estimation <strong>of</strong> Design Criteria and<br />

Conceptual Costs for <strong>GAC</strong> Treatment__________<br />

• A mathematical procedure was developed during this study to<br />

predict breakthrough curves in a multiple-contactor <strong>GAC</strong> adsorption<br />

system based on data from a single-contactor breakthrough curve.<br />

Multiple-contactor operation is much more representative <strong>of</strong> fullscale<br />

applications. Verification <strong>of</strong> the procedure using actual<br />

blending experiments indicated the success <strong>of</strong> this approach. The<br />

blended effluent curves resulted in significantly longer periods <strong>of</strong><br />

operation prior to an effluent criterion being reached, there<strong>by</strong><br />

resulting in significantly lower CURs compared to single contactors.<br />

• Conceptual treatment cost estimates for the <strong>GAC</strong> process indicated<br />

that minimizing the EBCT lowered the unit costs. This was<br />

attributed to the fact that the CUR did not change as EBCT<br />

decreased, and therefore capital costs reduction associated with the<br />

contactors were reduced at lower EBCTs without consequent cost<br />

increases associated with carbon reactivation. The lowest EBCT<br />

evaluated in this study was 10 minutes.


Conclusions and Recommendations 227<br />

Extrapolations based upon these results indicate that even lower<br />

EBCTs (less than 10 minutes) could further reduce unit costs for<br />

<strong>GAC</strong> treatment. At some lower value <strong>of</strong> EBCT, however, CUR<br />

would begin to increase as a result <strong>of</strong> premature breakthrough, and<br />

consequent increases in reactivation costs would <strong>of</strong>fset the reduction<br />

in capital cost yielded <strong>by</strong> the lower EBCT. It is also recognized that<br />

a practical limitation may exist in operational complexity resulting<br />

from the increase in reactivation frequency that accompanies a<br />

lower EBCT.<br />

Optimized coagulation resulted in an approximately 50 percent<br />

reduction in CUR, there<strong>by</strong> reducing the cost <strong>of</strong> the <strong>GAC</strong> facility.<br />

The overall system cost (i.e., the cost <strong>of</strong> pretreatment and <strong>GAC</strong>),<br />

however, was not reduced the high cost <strong>of</strong> the relatively high<br />

coagulant dosages used in this study outweighed the facility cost<br />

reduction. If a similar reduction in CUR could be achieved with a<br />

smaller increase in coagulant dosage, optimized coagulation<br />

potentially could lower the overall system cost. A unit cost lower<br />

than 5 cents per 1,000 gal (3,785 L) (including sludge handling and<br />

disposal) for optimized coagulation was the breakpoint at which the<br />

system cost was reduced in this study.<br />

Ozonation followed <strong>by</strong> biotreatment resulted in lower CUR<br />

estimates for two <strong>of</strong> the three waters evaluated during this study.<br />

The overall cost <strong>of</strong> treatment was found to be lower for one <strong>of</strong> the<br />

source waters. Again, lowering EBCT while maintaining a constant<br />

CUR may result in overall lower system costs when pretreatment is<br />

being applied.<br />

The extent to which pretreatment can reduce overall system costs is<br />

partially dependent upon the CUR. As the CUR for a given water<br />

increases, the cost benefits for a specific percent reduction in CUR<br />

<strong>by</strong> pretreatment also increase. Because <strong>of</strong> the relatively large<br />

proportion <strong>of</strong> unit costs associated with the capital cost <strong>of</strong> the <strong>GAC</strong><br />

contactors, a more efficient method <strong>of</strong> reducing cost <strong>by</strong> pretreatment<br />

may be to reduce EBCT while maintaining the CUR constant. Such<br />

an optimization study was beyond the scope <strong>of</strong> this research.


Appendix A<br />

RSSCT Design<br />

The scaling equations that were used to design the RSSCT were developed<br />

based on similitude and dimensional analysis (Crittenden et al. 1986, 1987, 1989,<br />

1991; Hineline et al. 1987; Sontheimer et al. 1988). The scaling relationship is a<br />

function <strong>of</strong> the carbon particle sizes used in the small-scale (RSSCT) and large-scale<br />

(pilot- or full-scale) columns. A problem with RSSCT design is the possible<br />

dependence <strong>of</strong> the intraparticle diffusivity on <strong>GAC</strong> particle size. The design equa<br />

tions were originally developed with the assumption <strong>of</strong> constant diffusivity (CD).<br />

However, a number <strong>of</strong> small column tests and batch kinetic tests have shown that<br />

the diffusion coefficient decreases proportionally with decreasing particle size, and<br />

a proportional diffusivity (PD) design approach has been developed (Crittenden et<br />

al. 1991; Sontheimer et al. 1988; Benz 1989). The proportional diffusivity design<br />

was used for all RSSCTs conducted as part <strong>of</strong> this project.<br />

The critical RSSCT design parameters are the EBCT and the hydraulic<br />

loading rate, or superficial velocity (v). Also <strong>of</strong> interest is the operation time (t). The<br />

scaling factor (SF) can be defined as the ratio <strong>of</strong> particle diameters (d) <strong>of</strong> the large<br />

particle column (LC) to that <strong>of</strong> small particle column (SC):<br />

SF = dLC/dsc (A.D<br />

The dependency <strong>of</strong> the intraparticle diffusion coefficient (D) on particle size can be<br />

expressed as<br />

Dsc = (dsc/dLC)X DLC = SF"XDLC (A-2)<br />

where X is the diffusivity factor. The following three equations express the<br />

relationships for EBCT, v, and t:<br />

EBCTSC/EBCTLC = (dsc/dLC)2-x = SFX-2 (A.3)<br />

lsAc = (dsc/dLC)2~X = SF*-2 (A.5)<br />

When similitude prevails and the intraparticle diffusion coefficient is not a function<br />

<strong>of</strong> carbon particle size, then the CD design equations with X = 0 are used.<br />

For several studies, especially with NOM as measured <strong>by</strong> TOC and UV254 ,<br />

a linear relationship between D and d has been found (e.g., as shown <strong>by</strong> Benz 1989).<br />

229


230 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

This dependency, with X = 1 , leads to the proportional diffusivity design approach<br />

in which the EBCT and operation time <strong>of</strong> the RSSCT are defined as<br />

EBCTSC = (dsc/d LC) EBCTLC = EBCTLC/SF<br />

(A.6)<br />

The similitude velocity in the RSSCT, v*sc , is directly related to the velocity in the<br />

large-scale column <strong>by</strong> the following equation:<br />

The design <strong>of</strong> an RSSCT based on Equations A.6 and A.8 will yield an RSSCT<br />

column with the same length (1) as the large- or full-scale column.<br />

lsc = v*sc • EBCTSC = (V LC • SF) • (EBCTLC/SF)<br />

(A.9)<br />

lsc = V LC ' EBCTLC = ILC (A' 10)<br />

RSSCTs designed with such long lengths will likely produce excessive head loss<br />

and are difficult to operate at the bench scale.<br />

To shorten the RSSCT bed length, the dominance <strong>of</strong> internal mass transfer<br />

over external mass transfer is utilized. The design velocity <strong>of</strong> the RSSCT, vsc , can<br />

be set to a lower value than that <strong>of</strong> the similitude velocity, as long as it is above the<br />

minimum velocity at which the RSSCT can be operated such that internal mass<br />

transfer still dominates. This minimum velocity is defined with the minimum<br />

Reynolds number, ReSCmjn , which ranges from 0.02 to 0.13 depending on the<br />

molecular weight <strong>of</strong> the compound (Crittenden et al. 1 986). The lack <strong>of</strong> impact <strong>of</strong><br />

Resc on the breakthrough curve has been previously shown, where reducing the<br />

Reynolds number from 5 in the large-scale column, d LC =1.1 mm, to 0.05 in the<br />

RSSCT, dsc = 0.1 1 mm, had very little impact on the modeled breakthrough curve<br />

(Summers et al. 1994b).<br />

Using a value <strong>of</strong> Resc in the RSSCT that is lower than that in the full-scale<br />

system allows a lower vsc in the RSSCT to be used. As shown in Equation A.8, lower<br />

vsc values will result in shorter columns and lower flow rates, both <strong>of</strong> which are<br />

desirable for bench-scale operation. The design equation for the RSSCT hydraulic<br />

loading thus becomes<br />

V = SFvLc(Resc.mi,/Re Lc> (A- 11 )<br />

Values <strong>of</strong> Resc min in the range <strong>of</strong> 1 .0 to 0.5 have been successfully used and are<br />

recommended.<br />

Equations A.6 and A.I 1 were used for the RSSCT design in this study.<br />

These equations are based on proportional diffusivity and have been found to yield<br />

results that are comparable to those <strong>of</strong> pilot- and full-scale systems. Details on the<br />

design, construction, operation, and sampling <strong>of</strong> RSSCTs and pilot plants are<br />

provided in the guidance manual for <strong>GAC</strong> treatment studies for compliance with the<br />

ICR requirements (USEPA 1996).


Appendix B<br />

A Statistical Method for the<br />

Comparison <strong>of</strong> RSSCT and<br />

Field-Scale <strong>GAC</strong><br />

Breakthrough Curves<br />

Many studies performed on a variety <strong>of</strong> waters have shown that the RSSCT<br />

is a good predictor <strong>of</strong> field-scale <strong>GAC</strong> performance for control <strong>of</strong> NOM and DBF<br />

precursors when representative samples are collected for the RSSCT influent<br />

(Summers et. al. 1995). Comparisons are usually made qualitatively, using descrip<br />

tors such as "well-predicted," "well-simulated," and "good predictability." A quan<br />

titative method <strong>of</strong> comparing RSSCT and field-scale breakthrough patterns using<br />

statistical analysis has not been rigorously developed. Such a method would be used<br />

on a case-<strong>by</strong>-case basis to quantitatively determine whether the RSSCT is a good<br />

predictor <strong>of</strong> field-scale <strong>GAC</strong> performance. A statistical method <strong>of</strong> breakthrough<br />

curve comparison can strengthen the foundation upon which the RSSCT is com<br />

pared to field-scale data.<br />

Paired t-Test Procedure__________________<br />

To systematically and quantitatively evaluate the RSSCT prediction <strong>of</strong><br />

field-scale <strong>GAC</strong> performance, the paired t-test was utilized. A typical use <strong>of</strong> the<br />

paired t-test is to compare two analytical methods. Several samples that are uniquely<br />

matched and related, or paired, are split and analyzed <strong>by</strong> each method to obtain a<br />

series <strong>of</strong> paired measurements (Berthouex and Brown 1994). The average differ<br />

ence between the two methods is calculated, and the paired t-test is used to determine<br />

whether, at some confidence level, the average difference between paired points is<br />

significantly different from zero. The paired t-test assumes that the series <strong>of</strong><br />

differences are independent and normally distributed (Daniel 1987).<br />

To be used in a paired t-test, <strong>GAC</strong> breakthrough data sets must be in the form<br />

<strong>of</strong> a time series <strong>of</strong> normalized effluent concentrations (effluent values reported as<br />

fractions <strong>of</strong> influent values, C/C0). Since sample points <strong>of</strong> the RSSCT and fieldscale<br />

adsorber are not normally paired in time, one data set is interpolated to obtain<br />

values corresponding to the sample times <strong>of</strong> the second data set. Linear interpolation<br />

is performed after the RSSCT breakthrough data have been scaled to equivalent<br />

field-scale time. To minimize error, the data set chosen for interpolation is that with<br />

231


252 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

the highest average density <strong>of</strong> sample points present within the time range that both<br />

sets share. This data set is termed the interpolated data set. Therefore, the location<br />

in time <strong>of</strong> each sample point within the data set with a lower density <strong>of</strong> points (the<br />

reference data set) is used to obtain a point on the interpolated data set. Sample<br />

points from either data set lying outside <strong>of</strong> the range that both share are discarded,<br />

except for two points located at each end <strong>of</strong> the interpolated data set. These two<br />

points must be retained in order to perform the first and last interpolations. Once two<br />

sets <strong>of</strong> paired sample points are generated, the difference between each pair, or the<br />

residual, is calculated. The average difference across the entire data set is an<br />

indicator <strong>of</strong> how well the RSSCT predicts field-scale performance.<br />

The null hypothesis, HQ, is defined as follows:<br />

HQ :<br />

"The difference between the two breakthrough curves is zero,"<br />

or udo = 0.<br />

where udo is the hypothesized population mean difference.<br />

The alternate hypothesis, HA , is defined as:<br />

H A :<br />

"The difference between the two breakthrough curves is not zero," or<br />

The test statistic, t, is calculated <strong>by</strong> Equation B.I, assuming the differences are a<br />

random sample <strong>of</strong> a normally distributed population <strong>of</strong> differences (Daniel 1987):<br />

where<br />

« - * (B.1)<br />

( S SA)<br />

_<br />

d = the sample mean difference<br />

s^ = the standard deviation <strong>of</strong> the difference<br />

n = the number <strong>of</strong> observations<br />

A level <strong>of</strong> significance, a, is chosen for acceptance or rejection <strong>of</strong> the null<br />

hypothesis. The test statistic is compared to the critical test statistic, t* (obtained<br />

from a table <strong>of</strong> probabilities <strong>of</strong> student's t distribution), to determine whether, at the<br />

(1 - a ) x 100 percent confidence level, the null hypothesis is accepted.<br />

Important assumptions <strong>of</strong> the paired t-test are that the series <strong>of</strong> differences<br />

are independent and that their distribution is normal. An empirical test <strong>of</strong> serial<br />

dependence is the autocorrelation function (ACF), available in statistical s<strong>of</strong>tware<br />

packages, which measures the correlation between differences at certain units apart<br />

in the series (West and Hepworth 1991).<br />

To test whether a series <strong>of</strong> differences fulfills a normal distribution, a<br />

probability plot is generated. The plotting position <strong>of</strong> the ranked differences is<br />

calculated using Equation B.2 (Berthouex and Brown 1994):<br />


Appendix B 233<br />

where p is the probability and i is the rank. Based on their calculated probability, the<br />

differences are plotted against the position they would occupy in a normal distribu<br />

tion. A straight line indicates a normal distribution. Deviations from a straight line<br />

are an indication that the distribution is not normal, but results <strong>of</strong> the paired t-test<br />

procedure are still valid for deviations that are not drastic (Daniel 1987). This<br />

procedure is not quantitative and was not used to distinguish non-normal distribu<br />

tions in this study.<br />

As many sample points as possible should be used for comparisons <strong>of</strong><br />

breakthrough curves <strong>by</strong> the paired t-test procedure. When the method is applied to<br />

small sample sizes, especially those containing fewer than 10 points, results may be<br />

negatively affected <strong>by</strong> the presence <strong>of</strong> one or two outliers in the original break<br />

through curves, even when the breakthrough curves show similar overall trends.<br />

Results obtained for small sample sizes should be accepted with caution. A<br />

statistical method that eliminates outliers prior to the test is recommended for small<br />

sample sizes.<br />

Application_____________________________<br />

The paired t-test method has been applied to data obtained from field-scale<br />

<strong>GAC</strong> columns and RSSCTs for the four waters presented in Chapter 3. The waters<br />

studies were Ohio River water, Lake Gaillard water, Mississippi River water, and<br />

Passaic River water. For ORW, MRW, and PRW, the breakthrough curve compari<br />

son method was applied to TOC, UV254 , TTHM, HAA6, and TOX. For LOW, the<br />

breakthrough curve comparison method was applied to TOC, TTHM, and HAAS.<br />

For this study, a was chosen as 0.05 to yield a 95 percent confidence level.<br />

Figure B.I shows the RSSCT and field-scale <strong>GAC</strong> breakthrough behavior<br />

for TOC obtained using ORW. The RSSCT data set contained fewer points within<br />

the range that both data sets occupy and was used as the reference data set. The<br />

RSSCT and interpolated field-scale <strong>GAC</strong> breakthroughcurves are shown in Figure<br />

B.2. The comparison <strong>of</strong> the paired points showed that d = 0.0167, s^- = 0.0595, and<br />

n = 34, which yields t = 1.637. The critical test statistic, t*, for a = 0.05, obtained<br />

from a table <strong>of</strong> probabilities <strong>of</strong> t distributions with 33 degrees <strong>of</strong> freedom, is 2.036.<br />

Thus, t is less than t*, and the null hypothesis (i.e., no difference between the<br />

breakthrough curves) can be accepted once the assumption that the series <strong>of</strong><br />

differences is independent is verified. The results <strong>of</strong> the ACF analysis are presented<br />

in Figure B .3, which is a plot <strong>of</strong> autocorrelations against each lag, or autocorrelogram.<br />

A significant autocorrelation would be represented as a bar extending beyond the<br />

plotted parentheses, which represent a 95 percent confidence level for a serial<br />

autocorrelation <strong>of</strong> the data at a particular lag. The probability plot, which qualita<br />

tively assesses the normality <strong>of</strong> the distribution, is shown in Figure B.4. Because the<br />

ACF analysis showed that there were no serial correlations, and since t is less than<br />

t*, the null hypothesis was accepted: TOC breakthrough behavior for ORW was not<br />

significantly different between the RSSCT and field-scale adsorber at the 95 percent<br />

confidence level. A summary <strong>of</strong> the results <strong>of</strong> this method applied to all RSSCT<br />

verification studies is shown in Table B.I.<br />

The comparison between field-scale and RSSCT UV254 breakthrough for<br />

PRW (Figure B.5) yielded a t value <strong>of</strong> 1.947. The critical t statistic, t*, value was


234 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> GA C <strong>Adsorption</strong><br />

Table B.1 Summary <strong>of</strong> field-scale and RSSCT breakthrough curve comparisons<br />

<strong>Water</strong><br />

Parameter<br />

n<br />

Average<br />

difference,<br />

d<br />

Independent<br />

series <strong>of</strong><br />

differences?<br />

Significant<br />

difference in<br />

d at a = 0.05?<br />

Variance,<br />

s^ 2<br />

ORW<br />

TOC<br />

uv254<br />

TTHM<br />

HAA6<br />

TOX<br />

34<br />

13<br />

8<br />

8<br />

5<br />

0.017<br />

0.057<br />

0.037<br />

0.101<br />

0.094<br />

yes<br />

yes<br />

yes<br />

yes<br />

yes<br />

no<br />

yes<br />

no<br />

yes<br />

no<br />

0.0035<br />

0.0045<br />

0.0055<br />

0.013<br />

0.010<br />

LGW<br />

TOC<br />

TTHM<br />

HAAS<br />

7<br />

7<br />

7<br />

0.010<br />

0.030<br />

0.051<br />

yes<br />

yes<br />

yes<br />

no<br />

no<br />

no<br />

0.0070<br />

0.017<br />

0.010<br />

MRW<br />

TOC<br />

uv254<br />

TTHM<br />

HAA6<br />

22<br />

11<br />

9<br />

9<br />

0.075<br />

0.149<br />

0.072<br />

0.066<br />

no<br />

yes<br />

yes<br />

no<br />

NA*<br />

yes<br />

no<br />

NA<br />

0.011<br />

0.019<br />

0.071<br />

0.026<br />

TOX<br />

5<br />

0.058<br />

yes<br />

no<br />

0.0070<br />

PRW<br />

TOC<br />

uv254<br />

TTHM<br />

HAA6<br />

TOX<br />

32<br />

31<br />

9<br />

10<br />

10<br />

0.008<br />

0.014<br />

0.038<br />

0.048<br />

0.026<br />

no<br />

no<br />

yes<br />

yes<br />

yes<br />

NA<br />

NA<br />

no<br />

no<br />

no<br />

0.0054<br />

0.0017<br />

0.0083<br />

0.0060<br />

0.0038<br />

NA = not applicable; series <strong>of</strong> differences is independent.<br />

2.042. Since t is less than t*, the null hypothesis was accepted: The UV254<br />

breakthrough curves for PRW were not significantly different.<br />

Figure B.6 shows the field-scale and RSSCT TTHM breakthrough for<br />

LGW. For these breakthrough curves, the results showed a t value <strong>of</strong> 0.617, which<br />

is less than the t* value <strong>of</strong> 2.447. The RSSCT and field-scale TTHM breakthrough<br />

curves for LGW were not significantly different at the 95 percent confidence level.<br />

The results <strong>of</strong> the statistical method <strong>of</strong> comparing breakthrough curves for<br />

these three preceding examples do not contradict a qualitative assessment <strong>of</strong> how<br />

well the RSSCT predicted field-scale <strong>GAC</strong> breakthrough. However, this is not<br />

always the case, as illustrated <strong>by</strong> the following example. Figure B.7 shows fieldscale<br />

and RSSCT H AA6 breakthrough curves obtained for MRW. It is clear that the<br />

field-scale curve had a high immediate breakthrough as compared to the RSSCT,<br />

which showed a very low immediate breakthrough. The RSSCT data then showed<br />

a steeper breakthrough than did the field-scale data, and at about 19 scaled operation<br />

days the two curves crossed. Qualitatively, the RSSCT did not provide a good fit <strong>of</strong><br />

field-scale breakthrough. The results <strong>of</strong> the statistical method yielded a t value <strong>of</strong><br />

1.234, which is less than the t* value <strong>of</strong> 2.306. Thus, according to the method, the<br />

two curves were not significantly different. These results are caused <strong>by</strong> the addition


Appendix B 235<br />

<strong>of</strong> large positive and negative differences, which led to a mean value <strong>of</strong> the<br />

differences near zero. A similar effect was seen when TTHM breakthrough curves<br />

were compared for MRW, as shown in Figure B.8. For these cases, the calculated<br />

variance, s^2 , may be a useful indicator <strong>of</strong> the occurrence <strong>of</strong> this effect. Figure B.9<br />

shows the variance for all RSSCT and field-scale breakthrough curve comparisons.<br />

The highest variances seen were those generated from MRW HAA6 and TTHM<br />

breakthrough curve comparisons (Figures B.7 and B.8). A high variance was also<br />

associated with MRW UV254 breakthrough (Figure B.I7), which showed the same<br />

breakthrough effect seen for HAA6 and TTHM breakthrough but for which the null<br />

hypothesis was rejected. A high variance was also found for LOW TTHM break<br />

through (Figure B.6) and was due to scatter in the data. All remaining breakthrough<br />

curve comparisons are shown in Figures B. 10 through B .22, and the results obtained<br />

are summarized in Table B.I.<br />

Summary_________________________<br />

A statistical method has been developed, based on paired t-test analysis, to<br />

quantitatively compare RSSCT and field-scale <strong>GAC</strong> breakthrough curves. The<br />

procedure has been applied to four waters for a range <strong>of</strong> parameters, and the results<br />

are summarized in Table B.I. Of the 18 cases studied, 14 yielded an independent<br />

series <strong>of</strong> differences. Of these, 11 were not significantly different at the a = 0.05<br />

confidence level. Assumptions <strong>of</strong> independent and normally distributed differences<br />

should be verified. Certain breakthrough patterns that contain large positive and<br />

negative differences and that qualitatively do not show a good fit may be incorrectly<br />

assessed <strong>by</strong> this test. However, a high variance associated with the average<br />

difference can be used as an indicator that the results <strong>of</strong> the test should be accepted<br />

with caution. A variance greater than 0.015 is recommended as the criterion based<br />

on these results.<br />

The drawback <strong>of</strong> any statistical method to quantitatively compare break<br />

through curves is that the procedure results in acceptance or rejection <strong>of</strong> the fit. A<br />

statistical method's "acceptance" <strong>of</strong> the null hypothesis that there is no significant<br />

difference between two breakthrough curves does not mean the RSSCT was<br />

successful. On the other hand, when the same method determines that the two curves<br />

are significantly different, it does not mean that the RSSCT results are unusable.<br />

Either <strong>of</strong> the two conclusions reached using a statistical method must be combined<br />

with sound engineering judgment and common sense before a final evaluation <strong>of</strong> the<br />

data is made. When it is expected that the RSSCT breakthrough data "well predict"<br />

field-scale breakthrough data, it is useful to have a method available that provides<br />

a statistical basis for the comparison, beyond a simple qualitative statement based<br />

on visual inspection. However, certain factors can affect the comparison made using<br />

the statistical method, leading to a result that contradicts that based on sound<br />

engineering judgment. It is extremely important that when a statistical method is<br />

used for breakthrough curve comparisons, it should not be used blindly, without a<br />

thorough study <strong>of</strong> the data available.


Ov<br />

SO<br />

O<br />

•5<br />

I<br />

i<br />

i<br />

3<br />

2<br />

O<br />

•5<br />

5'<br />

s<br />

(Q C<br />

(D<br />

o<br />

(O<br />

i<br />

CO<br />

Normalized concentration, C/C0<br />

Normalized concentration, C/C0<br />

•<br />

(D a<br />

— h<br />

(A<br />

O<br />

SL<br />

(D<br />

D)<br />

5<br />

0.<br />

37<br />

(/><br />

CO<br />

O<br />

O<br />

O<br />

CT<br />

CD<br />

o<br />

c<br />

(Q 3"<br />

—t,<br />

O<br />

O 3; o'<br />

i<br />

o SL<br />

(D<br />

0)<br />

a<br />

CO<br />

CO<br />

O<br />

O<br />

o<br />

a-<br />

3<br />

a><br />

5<br />

o<br />

in<br />

o<br />

O 3; 5'<br />

(D<br />

QJ<br />

^^<br />

(D<br />

30<br />


Appendix B 237<br />

LAG 1<br />

CORK<br />

1<br />

2<br />

3456789<br />

15<br />

10<br />

11<br />

12<br />

13<br />

14<br />

.324<br />

.059<br />

-.082<br />

-.115<br />

.079<br />

-.203<br />

-.112<br />

-.100<br />

-.031<br />

.050<br />

-.055<br />

.074<br />

.036<br />

.121<br />

.108<br />

SB -1.0 -.8 -. 6 -.4 -.2 .0 .2 4 .6 .8 1.0<br />

+____+ ——._+——— + ——— + ——— + ——— + —— _+——— + ——— + ——— +<br />

.171<br />

( IXXXXXXX)<br />

.189<br />

( IX<br />

)<br />

.189<br />

( XXI<br />

)))))<br />

.190<br />

( XXI<br />

.192<br />

( IX<br />

.193<br />

( XXXXX I<br />

.199<br />

( XXI<br />

.201<br />

XXI<br />

))))))))<br />

.203<br />

( 1<br />

.203<br />

IX<br />

.203<br />

( x|<br />

.204<br />

IX<br />

.204<br />

( 1<br />

.205<br />

( IXXX<br />

.207<br />

( IXX<br />

Note: LAG = units apart in series; CORR = correlation; SE = standard error<br />

Figure B.3 Autocorrelogram <strong>of</strong> TOC residuals for Ohio river water<br />

2-<br />

2<br />

1-<br />

0 -<br />

-1 -<br />

-2 -<br />

-3 -i t i i i i i i i i t , i i i i . i- ^ i i i r<br />

-0.10 -0.05 0.00 0.05 0.10 0.15<br />

Residual<br />

Figure B.4 Probability polot <strong>of</strong> TOC residuals for Ohio River water


§ "0<br />

|5-<br />

2<br />

•3<br />

»-*<br />

I<br />

pa<br />

b><br />

•<br />

o_<br />

D><br />

<br />

i__.__i<br />

5 "H<br />

n>


a. 5'<br />

Co<br />


Tl<br />

N)<br />

p<br />

bo<br />

i • i<br />

C<br />

ro<br />

00<br />


jjj<br />

I<br />

I Co<br />

(Q<br />

(Q<br />

P<br />

ro<br />

•<br />

o_<br />

a<br />

(0<br />

a<br />

—K<br />

o<br />

to<br />

en<br />

p<br />

o<br />

Normalized concentration, C/C0<br />

p p<br />

K) "*•<br />

i__.__i<br />

p p<br />

b) bo<br />

i__,__i<br />

1<br />

CD<br />

•D<br />

O_<br />

S<br />

bo o<br />

tfi<br />

O<br />

£.<br />

(D<br />

n><br />

a.<br />

3]<br />

(/><br />

0)<br />

O<br />

C/3<br />

8<br />

CD<br />

•§<br />

I O'<br />

O<br />

SL<br />

(D<br />

D><br />

Q.<br />

33<br />

(A<br />

(/><br />

O<br />

(D Q. in<br />

o °<br />

o<br />

I a1<br />

&><br />

|<br />

o<br />

(Q<br />

O<br />

O =r o'<br />

3)<br />

I<br />

I<br />

3* tfl<br />

(D<br />

"S<br />

Q)<br />

•52-<br />

o -<br />

o<br />

Ul<br />

o<br />

(D<br />

0<br />

5<br />

O c<br />

(Q<br />

O<br />

^<br />

O 5' 3;<br />

i' 33<br />

i<br />

sf.<br />

3<br />

CD<br />

"3<br />

0)<br />

o<br />

ro -<br />

in<br />

in<br />

o<br />

O) =<br />

CO T<br />

O K<br />

w


242 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> GA C <strong>Adsorption</strong><br />

1.0<br />

TOX<br />

0°<br />

0.8 -<br />

g 0.6 -<br />

§ 0.4 H<br />

8<br />

•o<br />

a<br />

0.2-<br />

o<br />

0.0<br />

—i—<br />

25 50 75 100<br />

Scaled operation time (days)<br />

• Full-scale effluent<br />

0 RSSCT effluent<br />

125 150<br />

Figure B.13 Interpolated full-scale and RSSCT TOX breakthrough for Ohio River water<br />

1.0<br />

TOC<br />

0.8 -<br />

8<br />

o 0.6 -\<br />

0.4 -I<br />

8<br />

•a<br />

a> N<br />

ra 0.2 H<br />

0.0<br />

25 50 75 100<br />

* Pilot-scale effluent<br />

0 RSSCT effluent<br />

125 150<br />

Scaled operation time (days)<br />

Figure B.14 Interpolated full-scale and RSSCT and pilot-scale TOC breakthrough for<br />

Lake Gaillard water


Appendix B 243<br />

0.0<br />

25 50 75 100<br />

• Pilot-scale effluent<br />

0 RSSCT effluent<br />

125 150<br />

Scaled operation time (days)<br />

Figure B.15 Interpolated pilot-scale and RSSCT HAAS breakthrough for Lake Gaillard<br />

water<br />

• Pilot-scale effluent<br />

0 RSSCT effluent<br />

Scaled operation time (days)<br />

Figure B.16 Interpolated pilot-scale and RSSCT TOC breakthrough for Mississippi River<br />

water


244 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> GA C <strong>Adsorption</strong><br />

1.0<br />

UV. 254<br />

10<br />

0°<br />

0.8 -<br />

O<br />

o 0.6 H<br />

I<br />

§ 0.4 H<br />

8<br />

I 0.2 H<br />

0.0<br />

—I—<br />

20<br />

—i—<br />

30<br />

• Pilot-scale effluent<br />

0 RSSCT effluent<br />

—r~<br />

40 50<br />

Scaled operation time (days)<br />

Figure B.17 Interpolated RSSCT and pilot-scale UV254 breakthrough for Mississippi River<br />

water<br />

8<br />

1.0<br />

'g<br />

I 0.6 H<br />

8<br />

TOX<br />

i(0<br />

o 0.2-<br />

0.0<br />

10 20 30<br />

Scaled operation time (days)<br />

• Pilot-scale effluent<br />

o RSSCT effluent<br />

40 50<br />

Figure B.18 Interpolated pilot-scale and RSSCT TOX breakthrough for Mississippi River<br />

water


Appendix B 245<br />

0.0<br />

100 125<br />

9 Pilot-scale effluent<br />

0 RSSCT effluent<br />

Scaled operation time (days)<br />

Figure B.19 Interpolated pilot-scale and RSSCT TOC breakthrough for Passaic River<br />

water<br />

1.0<br />

TTHM<br />

o<br />

o<br />

0.8 -<br />

0.6 -<br />

1 8§<br />

o<br />

"•^<br />

0.4 -<br />

8<br />

1 0.2 -^<br />

o<br />

0.0<br />

25 50 75 100 125 150<br />

• Pilot-scale effluent<br />

0 RSSCT effluent<br />

175 200<br />

Scaled operation time (days)<br />

Figure B.20 Interpolated pilot-scale and RSSCT TTHM breakthrough for Passaic River<br />

water


0\<br />

1 §<br />

i<br />

s-<br />

G<br />

5 3!<br />

SB (5<br />

«S<br />

03<br />

ro<br />

•3 I<br />

o<br />

Q.<br />

•g<br />

I w<br />

o SL<br />

(D<br />

D)<br />

3<br />

Q.<br />

V)<br />

o<br />

iat<br />

D)<br />

I<br />

oc<br />

(O<br />

Normalized concentration, C/C0<br />

o -T<br />

"0<br />

B><br />

0)<br />

03<br />

CO<br />

o'<br />

3D 1'


Appendix C__________<br />

Development <strong>of</strong> a New Test<br />

for Disinfection By-Product<br />

Formation Assessment:<br />

Uniform Formation Conditions<br />

Until recently, two approaches have been used to assess the formation <strong>of</strong><br />

chlorination disinfection <strong>by</strong>-products in drinking water: the formation potential test<br />

and the simulated distribution system test. Under the FP test, waters are chlorinated<br />

at high doses and long incubation times to maximize DBF formation and to<br />

indirectly assess the presence <strong>of</strong> DBF precursors (Stevens and Symons 1977).<br />

Typically, these conditions are a 7-day incubation period, a temperature <strong>of</strong> 25°C, a<br />

pH <strong>of</strong> 7, and a chlorine residual <strong>of</strong> 3 to 5 mg/L at the end <strong>of</strong> the incubation period<br />

(APHA et al. 1992). These high chlorine doses and long incubation times lead to the<br />

formation <strong>of</strong> more chloro-substituted than bromo-substituted DBFs (Symons et al.<br />

1993; Shukairy et al. 1994) compared to the speciation <strong>of</strong> DBFs formed under the<br />

conditions common to a treatment plant and distribution system. Moreover, a utility<br />

cannot directly use the FP test results to evaluate how changes in a given treatment<br />

will affect compliance with DBF regulations.<br />

To more accurately represent DBF formation in a specific distribution<br />

system, utilities can perform SDS tests, where samples are chlorinated under sitespecific<br />

conditions <strong>of</strong> time, temperature, pH, and chlorine dose or residual that<br />

reflect the conditions found in the distribution system. This test was initiated <strong>by</strong> J.<br />

DeMarco and A. Stevens at the USEPA for five field investigations in 1978 (Miller<br />

et al. 1982), and the results from SDS tests have been shown to represent DBF<br />

formation found in distribution systems well (Koch et al. 1991). A limitation <strong>of</strong> the<br />

SDS test is that chlorination conditions vary among distribution systems, compli<br />

cating DBF formation comparisons among different waters.<br />

A new test has been developed in an effort to provide a means for direct DBF<br />

formation comparisons among different waters, and it utilizes chlorination under<br />

constant conditions that are representative <strong>of</strong> average distribution system condi<br />

tions. This approach has been termed the uniform formation conditions test for DBF<br />

formation (Summers et al. 1996). The UFC test can also be used to study the effect<br />

<strong>of</strong> treatment conditions on subsequent DBF formation.<br />

247


248 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

Objectives<br />

The objectives <strong>of</strong> this appendix are (1) to describe the UFC test conditions<br />

and the rationale for their selection, (2) to present the results <strong>of</strong> chlorine demand<br />

studies performed on various waters, and (3) to present the results <strong>of</strong> sensitivity<br />

analyses performed on each UFC parameter for three waters.<br />

Uniform Formation Conditions_____________<br />

The chlorination conditions selected for the UFC test are as follows: 24±1<br />

hours, 20.0+1.0°C, pH 8.0±0.2, and 1.0±0.4 mg/L free chlorine residual after 24<br />

hours, as outlined in Table C. 1.<br />

A complete description <strong>of</strong> the UFC test procedure can be found in Summers<br />

et al. (1996). The process <strong>of</strong> selecting the conditions for the UFC test began with the<br />

distribution <strong>of</strong> a comprehensive survey to utilities and researchers who perform the<br />

SDS and FP tests. A range <strong>of</strong> conditions for the SDS test were obtained from data<br />

generated <strong>by</strong> the survey. Together with data made available from the AWWA <strong>Water</strong><br />

Industry Database (WIDE), based on information gathered from 318 large utilities<br />

that indicated free chlorine as the only residual disinfectant, a preliminary UFC set<br />

was developed. An important difference with respect to previous approaches was<br />

the adoption <strong>of</strong> a target chlorine residual rather than the use <strong>of</strong> a chlorine dose based<br />

on TOC. After discussions with several researchers and utility representatives,<br />

proposed conditions for the UFC test were agreed upon, as shown in Table C.I.<br />

Incubation Time<br />

Incubation times <strong>of</strong> 1 to 3. days were initially considered. Data from the<br />

WIDE showed that the average maximum time in the distribution system is 3.0 days,<br />

and that the average mean time is 1.3 days. Initially, an incubation time <strong>of</strong> 3 days<br />

was proposed, representing average maximum conditions in the distribution sys<br />

tem. From a practical viewpoint, however, a 1-day incubation time yields faster<br />

results and allows for more tests to be conducted. In a 5-day work week, four tests<br />

could be conducted if the incubation time were 1 day, whereas a 3-day incubation<br />

time would allow for only three tests to be conducted per week. Additionally, if a<br />

chlorine demand study were required for a given sample, then the complete test<br />

could be completed in 2 days with a 1-day incubation time, as opposed to 6 days if<br />

a 3-day incubation time were used. An investigation <strong>of</strong> DBF formation kinetics<br />

(described later) showed that the increase in DBF formation from 1 to 3 days<br />

Table C.1 UFC test conditions<br />

Parameter Value<br />

Incubation time (hours) 24±1<br />

Incubation temperature (°C) 20.0±1.0<br />

pH 8.0±0.2<br />

24-hour chlorine residual (mg/L) 1.0±0.4


Appendix C 249<br />

averaged 22 percent for two waters examined and three DBFs analyzed. Ultimately,<br />

an incubation time <strong>of</strong> 1 day was chosen for the UFC test.<br />

Incubation Temperature<br />

Initially, a temperature <strong>of</strong> 22±2°C was proposed, allowing labs to conduct<br />

the test at ambient temperatures, without an incubator. After further discussion, it<br />

was determined that temperature extremes commonly found in labs were greater<br />

than those acceptable for the UFC test and that, as such, many labs would require<br />

a temperature-controlled incubator anyway. In view <strong>of</strong> this information, a lower<br />

temperature with a narrower range <strong>of</strong> acceptable deviation was proposed: 20.0± 1.0°C.<br />

pH Value<br />

The pH value for the test has not changed throughout the discussions. A pH<br />

<strong>of</strong> 8.0 was chosen to reflect the impact <strong>of</strong> the Lead and Copper Rule on distribution<br />

system pH. Both samples and the chlorine dosing solution were buffered at pH 8.0<br />

with a borate buffer. Approximately 2 mL/L pH 8.0 borate buffer (1 .OM boric acid<br />

and 0.26M sodium hydroxide) was added to each sample before dosing. The dosing<br />

solution was buffered to pH 8.0 <strong>by</strong> addition <strong>of</strong> pH 6.7 borate buffer (1 .OM boric acid<br />

and 0.1 \M sodium hydroxide) based on the method described <strong>by</strong> Koch et al. (1987).<br />

Chlorine Residual<br />

A target chlorine residual after 24 hours <strong>of</strong> 1.0 mg/L free chlorine was<br />

chosen. This value is supported <strong>by</strong> the WIDE data, which reported an average mean<br />

residual <strong>of</strong> 0.90 mg/L. The average minimum and maximum chlorine residuals<br />

supplied <strong>by</strong> the WIDE are 0.39 and 1.57 mg/L, respectively. The acceptable range<br />

for the UFC test was decreased from ±0.5 to ±0.4 mg/L, after concerns were<br />

expressed that low residuals after 24 hours would not lead to detectable residuals<br />

after 3 to 5 days. The minimum acceptable chlorine residual, 0.7 mg/L, was thought<br />

still to yield a detectable residual after 3 to 5 days in most waters. Some concern was<br />

also expressed that for waters with very high TOC concentrations, a residual would<br />

not be detectable after 5 days. To address these issues, chlorine demand studies were<br />

performed, the results are presented in the next section.<br />

Results and Discussion_________________<br />

Chlorine Demand<br />

One <strong>of</strong> the primary objectives in the establishment <strong>of</strong> the UFC test is that<br />

the chlorination conditions be representative <strong>of</strong> average distribution system condi<br />

tions. As stated earlier, the UFC test requires a 1.0-mg/L chlorine residual after 24<br />

hours. Another concern is the maintenance <strong>of</strong> a residual for periods that reflect<br />

maximum times in a distribution system. Therefore, chlorine decay kinetics were<br />

investigated for a range <strong>of</strong> waters to determine whether the selected chlorine<br />

residual <strong>of</strong> 1.0 mg/L after 1 day would still yield detectable residuals after 3 to 5<br />

days, as would be necessary in a distribution system. Figure C.I shows chlorine


250 <strong>Removal</strong> <strong>of</strong><strong>DBP</strong> <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

decay kinetics for several raw waters with moderate to high dissolved organic<br />

carbon concentrations (Dugan et al. 1995). For a chlorine dose that yielded a 24-hour<br />

chlorine residual near 1.0 mg/L, a chlorine residual was detected on day 3 and day<br />

5 for both Little Miami River (Ohio) water (DOC = 3.3 mg/L) and Winton Lake<br />

(Cincinnati, Ohio) water (DOC = 5.2 mg/L). Harsha Lake (Cincinnati, Ohio) water<br />

(DOC = 6.1 mg/L) showed no detectable residual after n day 3, and an extreme<br />

example, Hillsborough River (Florida) water (DOC = 16.4 mg/L), showed no<br />

detectable residual after day 2.<br />

The chlorine demand <strong>of</strong> all four raw waters should be significantly lowered<br />

<strong>by</strong> treatment. Chlorine decay studies performed on finished waters in which a<br />

1.0-mg/L residual was measured after 1 day have consistently yielded detectable<br />

chlorine residuals after 3 to 5 days. Results for three finished waters, Lake Gaillard<br />

water, Ohio River water, and Salt River Project water, with TOC concentrations<br />

ranging from 1.4 to 2.2 mg/L, are shown in Figure C.2. All three waters yielded<br />

detectable residuals on days 3 and 5 when the chlorine residual after 1 day was near<br />

1.0 mg/L.<br />

Because the UFC test accepts a range <strong>of</strong> 0.3 mg/L above and below the<br />

target 1.0-mg/L 24-hour chlorine residual, the lower limit <strong>of</strong> the target residual was<br />

investigated to demonstrate that, after a 24-hour residual near 0.7 mg/L, a chlorine<br />

residual would still be detectable after 3 and 5 days. Figure C.3 shows a comparison<br />

<strong>of</strong> two chlorine doses on SRPW that yielded two residuals after 24 hours (1.0 and<br />

0.6 mg/L). A detectable residual was measured for both waters on day 5. As shown<br />

in Figure C.4, similar results were found with Ohio River water for three chlorine<br />

doses yielding 24-hour residuals as low as 0.6 mg/L. A chlorine residual was<br />

detectable after 3 to 5 days for all three doses. For waters with higher TOC<br />

concentrations, however, the 24-hour residual needed to be higher to yield a 5-day<br />

detectable residual, as illustrated in Figure C.5 for Manatee Lake (Brandenton, Fla.)<br />

water (MLW). Dosed to yield the target 24-hour residual <strong>of</strong> 1.0 mg/L, this water<br />

yielded a detectable residual on day 3 but not on day 5. Increasing the dosage to yield<br />

a 1.5 mg/L residual after 24 hours resulted in a 0.5 mg/L 5-day residual.<br />

Sensitivity Analysis<br />

An investigation <strong>of</strong> the sensitivity <strong>of</strong> DBF formation to small and large<br />

deviations from UFC was performed on three waters. For the sensitivity analysis,<br />

one parameter was varied over a range <strong>of</strong> interest while the remaining parameters<br />

were held constant at the proposed UFC. This procedure was repeated for all<br />

parameters for three waters: ORW, SRPW, and MLW. All three waters were<br />

sampled after conventional treatment, which included alum coagulation, sedimen<br />

tation, and filtration. ORW is an industrially impacted source water with low TOC<br />

concentration, medium bromide concentration, and medium alkalinity. SRPW is a<br />

water with moderate TOC concentration, high bromide concentration, and high<br />

alkalinity. MLW is a surface water with high TOC concentration, low alkalinity, and<br />

high bromide concentration. General water quality parameters are summarized in<br />

Table C.2.<br />

The range <strong>of</strong> parameter values examined in the sensitivity analysis is shown<br />

in Table C.3. The values were selected to represent the range commonly found in<br />

distribution systems.


Appendix C 251<br />

Table C.2 <strong>Water</strong> quality <strong>of</strong> waters used in sensitivity analysis<br />

<strong>Water</strong><br />

Parameter<br />

ORW<br />

SRPW<br />

MLW<br />

Treatment<br />

TOC concentration (mg/L)<br />

Bromide concentration (ug/L)<br />

Alkalinity (mg/L as CaCO 3)<br />

Total hardness (mg/L as CaCO 3)<br />

PH<br />

Conventional<br />

1.3<br />

35<br />

.60<br />

• 128<br />

8.2<br />

Conventional<br />

.2.2<br />

89<br />

113<br />

176<br />

8.1<br />

Conventional<br />

4.1<br />

107<br />

12<br />

92<br />

7.8<br />

Table C.3 Sensitivity analysis experimental matrix<br />

Parameter<br />

UFC test<br />

value<br />

UFC test<br />

window<br />

ORW<br />

Sensitivity analysis<br />

range examined<br />

SRPW<br />

MLW<br />

Incubation time (hours)<br />

Incubation temperature (°C)<br />

24<br />

20.0<br />

±1<br />

±1.0<br />

4-120<br />

6-36<br />

24-120<br />

9-30<br />

12-120<br />

9-33<br />

PH<br />

24-hour chlorine residual (mg/L)<br />

8.0<br />

1.0<br />

±0.2<br />

±0.3<br />

6.5-8.7<br />

0.3-3.1<br />

7.0-8.7<br />

0.5-1.5<br />

7.2-8.8<br />

0.5-1.6<br />

DBF formation was characterized in terms <strong>of</strong> total trihalomethane, the sum<br />

<strong>of</strong> six haloacetic acids, and total organic halide. The HAA species included in HAA6<br />

were monochloroacetic acid, dichloroacetic acid, trichloroacetic acid, monobromoacetic<br />

acid, dibromoacetic acid, and bromochloroacetic acid. All error bars in the<br />

remaining figures represent relative percent differences (RPD) based on analyses <strong>of</strong><br />

duplicate chlorination tests. Pairs <strong>of</strong> vertical dashed lines shown in the figures<br />

indicate the limits <strong>of</strong> the UFC test window for each parameter.<br />

Effect <strong>of</strong> Incubation Time<br />

Figures C.6, C.7, and C.8 show the effect <strong>of</strong> incubation time on the<br />

formation <strong>of</strong> THM, HAA, and TOX, respectively, for ORW. Figures C.9, C. 10, and<br />

C.ll show the equivalent data obtained for SRPW. Figures C.I2, C.I3, and C.I4<br />

show the data obtained for MLW. Incubation time was varied from 4 hours to 5 days<br />

for ORW, from 1 to 5 days for SRPW, and from 12 hours to 5 days for MLW. For<br />

all three waters, most <strong>of</strong> the 3-day DBF formation was already present after the first<br />

24 hours <strong>of</strong> reaction: 24-hour formation averaged 84 percent <strong>of</strong> 3-day formation<br />

(range: 71 -93 percent) for TTHM, HAA6, and TOX. After 24 hours, the increased<br />

formation <strong>of</strong> both TTHM and HAA6 was largely driven <strong>by</strong> the formation <strong>of</strong> the more<br />

chloro-substituted species, while the more bromo-substituted species showed fast<br />

initial rates <strong>of</strong> formation within the first 24 hours and smaller increases after 24


252 <strong>Removal</strong> <strong>of</strong><strong>DBP</strong> <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

hours. This phenomenon occurred with all three waters examined and has been<br />

reported <strong>by</strong> others (Symons et al. 1993).<br />

Effect <strong>of</strong> Incubation Temperature<br />

The effect <strong>of</strong> incubation temperature on DBF formation was examined over<br />

a temperature range <strong>of</strong> 6 to 36°C for ORW (Figures C. 15, C. 16, and C. 17), 9 to 30°C<br />

forSRPW (Figures C.I8, C.I9, and C.20), and 9 to 33°C for MLW (Figures C.21,<br />

C.22, and C.23). While all DBFs analyzed showed an increase in formation with<br />

increasing temperature, the effect was more pronounced for TTHM and HAA6.<br />

This increase in formation after 24 hours at the higher temperatures can be attributed<br />

to an increase in reaction rates at the higher temperatures. The more chlorosubstituted<br />

THMs and HAAs showed larger increases in formation with increasing<br />

temperature than did the more bromo-substituted species. For instance, while<br />

DCAA and MCAA formation in SRPW increased <strong>by</strong> more than 95 percent each<br />

between incubation temperatures <strong>of</strong> 9 and 30°C, DBAA and MBAA formation<br />

showed 40 and 30 percent increases, respectively, over the same temperature range<br />

(Figure C. 19). This impact <strong>of</strong> temperature on speciation was likely due to the slower<br />

formation kinetics <strong>of</strong> the chloro-substituted DBFs (Krasner et al. 1992; Symons et<br />

al. 1993). As shown in Figures C.9 and C. 10, most <strong>of</strong> the bromo-substituted species<br />

were already formed after 24 hours, while the chloro-substituted species were still<br />

forming and thus were more sensitive to temperature changes. Similar effects were<br />

observed in the ORW and MLW studies.<br />

Effect <strong>of</strong> pH<br />

Figures C.24, C.25, and C.26 show the effect <strong>of</strong> pH on THM, HAA, and<br />

TOX formation, respectively, for ORW. The effect <strong>of</strong> pH on THM, HAA, and TOX<br />

formation for SRPW is shown in Figures C.27, C.28, and C.29, respectively, and that<br />

for MLW is shown in Figures C.30, C.31, and C.32. DBF formation was examined<br />

for ORW over a pH range <strong>of</strong> 6.7 to 8.7, for SRPW over a pH range <strong>of</strong> 7.0 and 8.7,<br />

and for MLW over a pH range <strong>of</strong> 7.2 and 8.8. A base catalyzed reaction, TTHM<br />

formation increased with increasing pH, driven <strong>by</strong> an increase in CHC13 formation,<br />

for all waters. For all waters, changes in HAA6 formation with increasing pH were<br />

similar. An increase in pH up to 8.2 did not affect HAA6 formation, and between<br />

8.2 and 8.7 there was a decrease in HAA6 formation for all waters, which is<br />

consistent with results reported <strong>by</strong> others (Stevens et al. 1989; Reckhow and Singer<br />

1985). For all waters, the decrease in HAA6 formation at high pH values was driven<br />

<strong>by</strong> a decrease in TCAA formation. TOX formation was not affected <strong>by</strong> pH within<br />

the range examined, which could represent a balance between the formation <strong>of</strong> base<br />

catalyzed and acid catalyzed DBFs.<br />

Effect <strong>of</strong> 24-Hour Chlorine Residual<br />

The effect <strong>of</strong> 24-hour chlorine residual was investigated from 0.3 to 3.1 mg<br />

C1 2/L for ORW (Figures C.33, C.34, and C.35), from 0.5 to 1.5 mg CL/L for SRPW<br />

(Figures C.36, C.37, and C.38), and from 0.5 to 1.6 mg C1,/L for MLW (Figures<br />

C.39, C.40, and C.41) . For ORW, TTHM formation sho'wed an increase with<br />

increasing residual only above 1.3 mg C12/L, while HAA6 formation was more<br />

strongly affected, showing a linear increase over the entire residual range. The<br />

increase in formation was largely driven <strong>by</strong> increases in TCAA and DCAA. There<br />

was a slight increase in TOX formation with increasing chlorine residual. For


Appendix C 253<br />

SRPW, both TTHM and HAA6 showed slight increases in formation with increas<br />

ing chlorine residual, driven <strong>by</strong> increases in the more chloro-substituted species. For<br />

MLW, there was little change in TTHM and HAA6 formation over this chlorine<br />

residual range. TOX formation showed no change in formation for all three waters<br />

within the ranges examined.<br />

DBF Formation Sensitivity Within UFC Test Window<br />

The goal in providing an acceptable window for each UFC test parameter<br />

was to establish limits on the acceptance <strong>of</strong> data for the test; this goal was intended<br />

to allow for some deviation from the required test conditions in a manner that would<br />

result in little change in DBF formation. From the data collected for the sensitivity<br />

analysis, the percent change in DBF formation within the UFC test window was<br />

calculated. Table C.4 summarizes the DBF formation sensitivity to UFC. Within the<br />

acceptable window, DBF formation was least sensitive to incubation time (average<br />

2.3 percent change) and most sensitive to pH and 24-hour chlorine residual (average<br />

5.4 and 5.0 percent change, respectively). For ORW, the average change in DBF<br />

formation within all parameter windows was 3.0 percent, with a minimum <strong>of</strong> 1<br />

percent and a maximum <strong>of</strong> 9 percent. The average change in DBF formation within<br />

all parameter windows for SRPW was 3.8 percent, with a minimum and maximum<br />

<strong>of</strong> 1 and 8 percent, respectively. For MLW, the average change in DBF formation<br />

was 4.3 percent, with a minimum and maximum <strong>of</strong>


254 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> GA C <strong>Adsorption</strong><br />

Chlorine demand tests showed that UFC should yield a 3-5 day detectable<br />

chlorine residual in most treated waters. Future work should include chlorine<br />

demand tests on finished waters with higher TOC concentrations than those<br />

presented here (to determine the presence <strong>of</strong> a residual after 3 and 5 days) and on<br />

finished waters with low alkalinity (to determine the robustness <strong>of</strong> the buffer<br />

system).<br />

DBF formation under UFC was not drastically affected <strong>by</strong> variations within<br />

the acceptable windows given for the waters used in this study: Ohio River water,<br />

Salt River Project water, and Manatee Lake water. The average change in DBF<br />

formation for all three DBFs, all three waters, and at the extremes <strong>of</strong> all UFC<br />

parameter windows was 3.7 percent. For any given DBF, formation at the limits <strong>of</strong><br />

the UFC parameter windows did not exceed 13 percent. Within the UFC test<br />

window, TTHM formation was most sensitive to changes in pH, HAA6 formation<br />

was most sensitive to changes in 24-hour chlorine residual, and TOX was most<br />

sensitive to changes in pH for SRPW and 24-hour chlorine residual for ORW.<br />

Overall, for the waters examined, DBF formation was not overly sensitive to the<br />

conditions chosen for the UFC test or to slight changes that are still acceptable for<br />

the test.


Appendix C 255<br />

o Hillsborough River 16.4<br />

a HarshaLake 6.1<br />

O Winton Lake 5.2<br />

Little Miami River 3.3<br />

DOC (mg/L) Chlorine dose (mg/L)<br />

26<br />

7.7<br />

6.1<br />

5.5<br />

0 -<br />

0123456<br />

Incubation time (days)<br />

Source: Data from Dugan et al. (1995)<br />

Figure C.1 Chlorine decay kinetics for raw waters with moderate to high DOC<br />

concentrations<br />

3.0<br />

—i———•———i———'———i———•-<br />

TOC (mg/L) Chlorine dose (mg/L)<br />

2.5 -<br />

O)<br />

e 2.0 -<br />

"3<br />

O Lake Gaillard<br />

o Ohio River<br />

a Salt River Project<br />

1.4<br />

1.9<br />

2.2<br />

2.2<br />

2.6<br />

2.5<br />

1 1.5-1<br />

£<br />

o<br />

I 1'°"1<br />

O<br />

0.5 -<br />

0.0<br />

—r~<br />

2<br />

T~<br />

4<br />

T-<br />

5<br />

Incubation time (days)<br />

Figure C.2 Chlorine decay kinetics for three finished waters


256 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> GA C <strong>Adsorption</strong><br />

Salt River Project water 24-hour chlorine<br />

TOC = 2.2 mg/L residual (mg/L)<br />

0.0<br />

234<br />

Incubation time (days)<br />

Figure C.3 Effect <strong>of</strong> chlorine dose on chlorine decay kinetics for Salt River Project water<br />

Ohio River water<br />

TOC = 1.3 mg/L<br />

24-hour chlorine<br />

residual (mg/L)<br />

a 1.3<br />

O 0,8<br />

A 0.6<br />

o.o<br />

234<br />

Incubation time (days)<br />

Figure C.4 Effect <strong>of</strong> chlorine dose on chlorine decay kinetics for Ohio River water


Appendix C 257<br />

Manatee Lake water<br />

TOC = 4.1 mg/L<br />

24-hour chlorine<br />

residual (mg/L)<br />

234<br />

Incubation time (days)<br />

Figure C.5 Effect <strong>of</strong> chlorine dose on chlorine decay kinetics for Manatee Lake water<br />

12345<br />

Incubation time (days)<br />

Note: Incubation temperature, 22°C; pH, 7.95±0.02; 24-hour chlorine residual, 1.32±0.03 mg/L<br />

Figure C.6 Effect <strong>of</strong> incubation time on THM formation for ORW


255 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> GA C <strong>Adsorption</strong><br />

012345<br />

Incubation time (days)<br />

Note: Incubation temperature, 22°C; pH, 7.95±0.02; 24-hour chlorine residual, 1.32±0.03 mg/L<br />

Figure C.7 Effect <strong>of</strong> incubation time on HAA formation for ORW<br />

300<br />

o -4<br />

012345<br />

Incubation time (days)<br />

Note: Incubation temperature, 22°C; pH, 7.95±0.02; 24-hour chlorine residual, 1.32±0.03 mg/L<br />

Figure C.8 Effect <strong>of</strong> incubation time on TOX formation for ORW


Appendix C 259<br />

120<br />

100-<br />

80 -<br />

TO<br />

60 -<br />

40 -<br />

20 -<br />

234<br />

Incubation time (days)<br />

Note: Incubation temperature, 20°C; pH, 8.03±0.03; 24-hour chlorine residual, 1.00+0.05 mg/L<br />

Figure C.9 Effect <strong>of</strong> incubation time on THM formation for SRPW<br />

50<br />

I §<br />

40-<br />

30-<br />

20 -<br />

10 -<br />

§ —<br />

6<br />

—————— 5<br />

- —— *r<br />

——————-a——————<br />

———————————— Q——— •—————————<br />

———:——?<br />

§<br />

HAA6<br />

ORWHAA6<br />

5 DCAA<br />

TCAA<br />

BCAA<br />

OBAA<br />

MCAA<br />

MBAA<br />

Incubation time (days)<br />

Note: Incubation temperature, 20°C; pH, 8.03±0.03; 24-hour chlorine residual, 1.00±0.05 mg/L<br />

Figure C.10 Effect <strong>of</strong> incubation time on HAA formation for SRPW


txj<br />

Ox<br />

so<br />

I 1<br />

I<br />

i<br />

C *<br />

3 g<br />

P &<br />

_L 0><br />

ro 5<br />

rn =<br />

=* §<br />

(D 3<br />

a 1<br />

o a<br />

-t. C<br />

5' P<br />

O ro<br />

& 6<br />

O I<br />

3 a,<br />

=•• b<br />

3-. 0 1+<br />


Appendix C 261<br />

120<br />

234<br />

Incubation time (days)<br />

Note: Incubation temperature, 20°C; pH, 8.00±0.02; 24-hour chlorine residual, 1.03±0.07 mg/L<br />

Figure C.13 Effect <strong>of</strong> incubation time on HAA formation for MLW<br />

800<br />

234<br />

Incubation time (days)<br />

Note: Incubation temperature, 20°C; pH, 8.00+0.02; 24-hour chlorine residual, 1.03±0.07 mg/L<br />

Figure C.14 Effect <strong>of</strong> incubation time on TOX formation for MLW


262 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> GA C <strong>Adsorption</strong><br />

20 30<br />

Incubation temperature (°C)<br />

40<br />

Note: Incubation time, 24 hours; pH, 8.04±0.02; 24-hour chlorine residual, 0.99±0.09 mg/L<br />

Figure C.15 Effect <strong>of</strong> incubation temperature on THM formation for ORW<br />

20 30<br />

Incubation temperature (°C)<br />

40<br />

Note: Incubation time, 24 hours; pH, 8.04±0.02; 24-hour chlorine residual, 0.99±0.09 mg/L<br />

Figure C.16 Effect <strong>of</strong> incubation temperature on HAA formation for ORW


Appendix C 263<br />

250<br />

10 20<br />

30<br />

Incubation temperature (°C)<br />

Note: Incubation time, 24 hours; pH, 8.04±0.02; 24-hour chlorine residual, 0.99±0.09 mg/L<br />

Figure C.17 Effect <strong>of</strong> incubation temperature on TOX formation for ORW<br />

0 10 20 30<br />

Incubation temperature (°C)<br />

Note: Incubation time, 24 hours; pH, 8.01 ±0.02; 24-hour chlorine residual, 1.08±0.11 mg/L<br />

Figure C.18 Effect <strong>of</strong> incubation temperature on THM formation for SRPW


264 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

50<br />

HAA6<br />

3- 40-<br />

1<br />

I 30-<br />

(D<br />

ORW<br />

HAA6<br />

20 -<br />

DCAA<br />

10 -<br />

TCAA<br />

BCAA<br />

DBAA<br />

~ MCAA<br />

O MBAA<br />

10 . 20 30<br />

40<br />

Incubation temperature (°C)<br />

Note: Incubation time, 24 hours; pH, 8.01±0.02; 24-hour chlorine residual, 1.08+0.11 mg/L<br />

Figure C.19 Effect <strong>of</strong> incubation temperature on HAA formation for SRPW<br />

200 -<br />

..-• ORW<br />

150 -<br />

..H<br />

100 -<br />

50-<br />

10 20 30<br />

40<br />

Incubation temperature (°C)<br />

Note: Incubation time, 24 hours; pH, 8.01 ±0.02; 24-hour chlorine residual, 1.08±0.11 mg/L<br />

Figure C.20 Effect <strong>of</strong> incubation temperature on TOX formation for SRPW


Appendix C 265<br />

250<br />

20 30<br />

Incubation temperature (°C)<br />

40<br />

Note: Incubation time, 24 hours; pH, 8.01 ±0.02; 24-hour chlorine residual, 0.97±0.05 mg/L<br />

Figure C.21 Effect <strong>of</strong> incubation temperature on THM formation for MLW<br />

120<br />

20 30<br />

Incubation temperature (°C)<br />

40<br />

Note: Incubation time, 24 hours; pH, 8.01 ±0.02; 24-hour chlorine residual, 0.97+0.05 mg/L<br />

Figure C.22 Effect <strong>of</strong> incubation temperature on HAA formation for MLW


266 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

800<br />

700 -<br />

600 -<br />

O<br />

S» 500<br />

*-' /<br />

f 20° -<br />

—!•"—'<br />

.__..— SRPWTOX<br />

8<br />

x 100-<br />

O<br />

ORW TOX<br />

10 20<br />

Incubation temperature (°C)<br />

30<br />

40<br />

Note: Incubation time, 24 hours; pH, 8.01 ±0.02; 24-hour chlorine residual, 0.97±0.05 mg/L<br />

Figure C.23 Effect <strong>of</strong> incubation temperature on TOX formation for MLW<br />

80<br />

60 -<br />

TTHM<br />

O<br />

t 40 -<br />

CHCIj<br />

O<br />

20 -<br />

CHBrCI<br />

O CHBr2CI<br />

6.5 7.0 7.5 8.0<br />

PH<br />

8.5 9.0<br />

9.5<br />

Note: Incubation time, 24 hours; incubation temperature, 20°C; 24-hour chlorine residual, 1.03±0.03 mg/L<br />

Figure C.24 Effect <strong>of</strong> pH on THM formation for ORW


Appendix C 267<br />

40<br />

0 MCAA<br />

D DBAA<br />

O MBAA<br />

'13 q<br />

20 -<br />

HAA6<br />

DCAA<br />

BCAA<br />

TCAA<br />

6.5 7.0 7.5 8.0<br />

PH<br />

8.5 9.0 9.5<br />

Note: Incubation time, 24 hours; incubation temperature, 20°C; 24-hour chlorine residual, 1.03±0.03 mg/L<br />

Figure C.25 Effect <strong>of</strong> pH on HAA formation for ORW<br />

£*JW<br />

d 200 -<br />

O<br />

-<br />

0 150 -<br />

1<br />

c 100 -<br />

8<br />

X<br />

0<br />

K 50 -<br />

T__r<br />

. -<br />

"<br />

0-<br />

————— i ————— | ————— i ————— | ————— r-1<br />

6. 5 7.0 7.5<br />

8.0<br />

PH<br />

-i ————— | ———— , ————— ! ————— , —————<br />

8.5 9.0 9.<br />

Note: Incubation time, 24 hours; incubation temperature, 20°C; 24-hour chlorine residual,1.03±0.03 mg/L<br />

Figure C.26 Effect <strong>of</strong> pH on TOX formation for ORW


00<br />

O<br />

Oo<br />

P &<br />

IO B*<br />

« 5'<br />

fl> CD<br />

a "ro<br />

O t<br />

§<br />

I &<br />

> a<br />

O 5<br />

3<br />

HAA concentration (pg/L)<br />

ro<br />

o<br />

to<br />

o<br />

p &<br />

ro £<br />

-4 o<br />

ro<br />

-a |<br />

X 55<br />

§ if<br />

H &<br />

I £<br />

5 !<br />

O CD<br />

3 1<br />

0) n><br />

r+ Q)<br />

5' c<br />

cn<br />

In<br />

THM concentration (M9/L)<br />

O<br />

I<br />

CT><br />

O<br />

O 3<br />

ro<br />

o V) O<br />

JJ<br />

TJ<br />

o<br />

c<br />

= CD<br />

O g<br />

(/) O<br />

TJ<br />

ro<br />

o<br />

o<br />

o"<br />

CD<br />

o<br />

CD<br />

5<br />

CD<br />

cn<br />

CL<br />

c<br />

0)<br />

o<br />

03<br />

H-<br />

o<br />

CL<br />

SO<br />

O<br />

CO<br />

H-<br />

O<br />

O<br />

3<br />

CQ<br />

CO<br />

in<br />

§<br />

CO<br />

"0<br />

I<br />

2<br />

I<br />

'


Appendix C 269<br />

250<br />

O<br />

o><br />

1<br />

200-<br />

150-<br />

100-<br />

X<br />

e 50.<br />

ORW<br />

6.5 7.0<br />

7.5 8.0<br />

PH<br />

8.5 9.0 9.5<br />

Note: Incubation time, 24 hours; incubation temperature, 20°C; 24-hour chlorine residual, 1.03±0.10 mg/L<br />

Figure C.29 Effect <strong>of</strong> pH on TOX formation for SRPW<br />

200<br />

160 -<br />

TTHM<br />

•I 120 -<br />

CHCI,<br />

80 -<br />

SRPW TTHM<br />

ORW TTHM<br />

40 -<br />

A CHBrCI2<br />

O CHBfjCI<br />

6.5<br />

7.0 7.5 8.0<br />

PH<br />

8.5 9.0 9.5<br />

Note: Incubation time, 24 hours; incubation temperature, 20°C; 24-hour chlorine residual, 0.98±0.03 mg/L<br />

Figure C.30 Effect <strong>of</strong> pH on THM formation for MLW


.<br />

^<br />

&<br />

2<br />

8-<br />

•53<br />

5'<br />

•n<br />


Appendix C 271<br />

80<br />

TTHM<br />

s § «o<br />

CHCt,<br />

20H<br />

60-<br />

-£-45-<br />

-5-42-<br />

ACHBfCI,<br />

-O CHBr,<br />

0-0 0.5 1.0 1.5 2.0 2.5<br />

24-hour chlorine residual (mg/L)<br />

3.0 3.5<br />

Note: Incubation time, 24 hours; incubation temperature, 20°C; pH, 7.97±0.02 mg/L<br />

Figure C.33 Effect <strong>of</strong> 24-hour chlorine residual on THM formation for ORW<br />

0.5 1.0 1.5 2.0 2.5<br />

24-hour chlorine residual (mg/L)<br />

3.0 3.5<br />

Note: Incubation time, 24 hours; incubation temperature, 20°C; pH, 7.97±0.02 mg/L<br />

Figure C.34 Effect <strong>of</strong> 24-hour chlorine residual on HAA formation for ORW


272 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> GA C <strong>Adsorption</strong><br />

250<br />

,-» 200-<br />

C<br />

O<br />

8 100-<br />

8<br />

0.0 0.5 1.0 1.5 2.0 2.5<br />

24-hour chlorine residual (mg/L)<br />

3.0<br />

3.5<br />

Note: Incubation time, 24 hours; incubation temperature, 20°C; pH, 7.97±0.02 mg/L<br />

Figure C.35 Effect <strong>of</strong> 24-hour chlorine residual on TOX formation for ORW<br />

0.5 1.0 1.5<br />

24-hour chlorine residual (mg/L)<br />

2.0<br />

Note: Incubation time, 24 hours; incubation temperature, 20°C; pH, 8.02±0.03 mg/L<br />

Figure C.36 Effect <strong>of</strong> 24-hour chlorine residual on THM formation for SRPW


i A<br />

.<br />

i.'<br />

1<br />

ra<br />

s<br />

SI S'<br />

O<br />

NJ<br />

•~* T" r^ < Tl ' > ^-<br />

1 i t i<br />

(D 03 2 5=-<br />

P| 0 |<br />

t» | TOX concentration (MS CI7L) !ij |. HAA concentration (ug/L)<br />

HI ^-f -* -* K> Rj ni .-. — h K5 tJ<br />

=63 0888 §8-3 i 0088<br />

o -° P<br />

2. IV) 0<br />

a £<br />

K i<br />

• y>.<br />

I 3<br />

C c<br />

~« IT<br />

o Si<br />

3" o p<br />

o -' i"<br />

§. | K<br />

33 ^<br />

"> ^ ?<br />

J5 a ^<br />

S2. S o<br />

a ® Z<br />

C M 0<br />

2. °* §• _<br />

o P * b -<br />

3 ^ 3<br />

Hi 52.<br />

g « &<br />

?s 0 o<br />

•* ro —<br />

W rt IT 14- *-~^ ._<br />

29 3<br />

3 ° *S.<br />

1 3° C<br />

%% in'<br />

«^»<br />

O^<br />

(/)<br />

3<br />

i<br />

1 , 1 . 1 . 1<br />

. h H<br />

1 H<br />

(<br />

gj —.—, 1 —. ,— i ,—.,— , I — .—, I<br />

(O<br />

—,—, 0 -<br />

"<br />

£ * § 1<br />

-* ro<br />

a ?<br />

£ s<br />

f* 3<br />

0 I'<br />

C g<br />

-« cr<br />

o a o<br />

Z §' in '<br />

O W<br />

2. 0 D _.<br />

""• o ffl .<br />

0 P * o<br />

3 (0<br />

-a w<br />

ii s<br />

S e i<br />

-* 1^3 ^<br />

S £ I<br />

1 s c<br />

tt 3 r* -<br />

0 (0 en<br />

3 r-<br />

••it<br />

0 -t<br />

w 71<br />

•o<br />

i . i 1 i i,<br />

P f 9? f<br />

T\ : \<br />

h \ \<br />

l\\ '• \<br />

tli il '<br />

JuJUlJ '• 1 _ 1<br />

r~l ' 1 1<br />

II \<br />

^ll<br />

1 1<br />

IR 1 r*HM r-<br />

1 0<br />

no S°Q o o <<br />

^^^^^ |<br />

1 . 1 . 1<br />

0,<br />

1— 1<br />

*~f<br />

I<br />

a<br />

6 S<br />

1


274 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> GA C <strong>Adsorption</strong><br />

200 -<br />

———————— 1 ————————————— | ————————— -T——<br />

,..., _____ , ,.,.. —— .<br />

d 150-<br />

I<br />

1<br />

A concenti 8 i<br />

I 50-<br />

*——<br />

D —— -<br />

ORW TTHM __<br />

-— — $ ————<br />

_- —— 5 ————<br />

^-<br />

—— J TTHM<br />

ṛ.._<br />

7T f^ur^i<br />

.^ SRPW TTHM<br />

I ——————**——————————————————<br />

A ———<br />

| ——-* CHBrCI2<br />

———————— 0 —————————— __.^ CHBr2CI<br />

ft -<br />

0.0 0.5 1.0 1.5<br />

24-hour chlorine residual (mg/L)<br />

2.0<br />

Note: Incubation time, 24 hours; incubation temperature, 20°C; pH, 8.01*0.02 mg/L<br />

Figure C.39 Effect <strong>of</strong> 24-hour chlorine residual on THM formation for MLW<br />

0.5 1.0 1.5<br />

24-hour chlorine residual (mg/L)<br />

2.0<br />

Note: Incubation time, 24 hours; incubation temperature, 20°C; pH, 8.01 ±0.02 mg/L<br />

Figure C.40 Effect <strong>of</strong> 24-hour chlorine residual on HAA formation for MLW


Appendix C 275<br />

/uu -<br />

1<br />

i<br />

i<br />

•<br />

650-<br />

g«:<br />

I ———— i<br />

1<br />

T<br />

f ^<br />

o 250-<br />

/<br />

•<br />

><br />

-<br />

| 2°°-<br />

8 150-<br />

- SRPW TOX<br />

ORWTOX<br />

-<br />

"<br />

%oh- 100-<br />

50-<br />

-<br />

n -<br />

0.0 0.5 1.0 1.5<br />

24-hour chlorine residual (mg/L)<br />

2.0<br />

Note: Incubation time, 24 hours; incubation temperature, 20°C; pH, 8.01 ±0.02 mg/L<br />

Figure C.41 Effect <strong>of</strong> 24-hour chlorine residual on TOX formation for MLW


References<br />

Allgeier, S., and R.S. Summers. 1995. Evaluation <strong>of</strong> NF for DBF Control With<br />

the RBSMT. Jour. AWWA, 87(3):87.<br />

APHA, AWWA, and WEF (American Public Health Association, American<br />

<strong>Water</strong> Works Association, and <strong>Water</strong> Environmental Federation). 1992.<br />

Standard Methods for the Examination <strong>of</strong> <strong>Water</strong> and Waste-water.<br />

Washington D.C.:APHA.<br />

————. 1995. Standard Methods for the Examination <strong>of</strong> <strong>Water</strong> and Wastewater.<br />

19th ed. Washington D.C.: APHA.<br />

Arora, H. 1989. Modelling <strong>of</strong> Fixed Bed Adsorbers for Multi-Component<br />

Mixtures <strong>of</strong> Known and Unknown Components. Ph.D. dissertation,<br />

Michigan Technological University, Houghton, Michigan.<br />

Benz, M. 1989. Untersuchungen Zum <strong>Adsorption</strong>sverhalten von<br />

Vielst<strong>of</strong>fgemischen unbekannterZusammensetzung. PhD. dissertation.<br />

University <strong>of</strong> Karlsruhe, Karlsruhe, Germany.<br />

Berthouex, P.M., andL.C. Brown. 1994. Statistics for Environmental Engineers.<br />

Chelsea, Mich.: Lewis Publishers.<br />

Collins, M.R., G.L. Amy, and C. Steelink. 1986. Molecular Weight Distribution,<br />

Carboxylic Acidity, and Humic Substances Content <strong>of</strong> Aquatic Organic<br />

Matter: Implications for <strong>Removal</strong> During <strong>Water</strong> Treatment. Envir. Sci.<br />

&Technol., 20(10): 1028.<br />

Cowman, G.A., and P.C. Singer. 1994. Effect <strong>of</strong> Pre-Ozonation on Haloacetic<br />

Acid Speciation in Chlorinated <strong>Water</strong>s Containing Bromide. In Proc. <strong>of</strong><br />

the AWWA <strong>Water</strong> Quality Technology Conference. Denver, Colo.:<br />

AWWA.<br />

Crittenden, J.C., J.K. Berrigan, and D.W. Hand. 1986. Design <strong>of</strong> Rapid Small-<br />

Scale <strong>Adsorption</strong> Tests for a Constant Surface Diffusivity. /. <strong>Water</strong><br />

Pollution Control Fed., 58(4):312-319.<br />

Crittenden, J.C., J.K. Berrigan Jr., D.W. Hand, and B.W. Lykins. 1987. Design<br />

<strong>of</strong> Rapid Fixed Bed <strong>Adsorption</strong> Tests for Non-Constant Diffusivities. J.<br />

Env. Engineering, Amer. Society <strong>of</strong> Civil Engineers, 113(2):243-259.<br />

Crittenden, J.C., P.S. Reddy, D.W. Hand, and H. Aurora. 1989. Prediction <strong>of</strong><br />

<strong>GAC</strong> Performance Using Rapid Small-Scale Column Tests. Denver,<br />

Colo.: AWWARF and AWWA.<br />

Crittenden, J.C., P.S. Reddy, H. Aurora, J. Trynoski, D.W. Hand, D.L. Perram,<br />

and R.S. Summers. 1991. Predicting <strong>GAC</strong> Performance With Rapid<br />

Small Scale Column Tests. Jour. AWWA, 83(l):77-87.<br />

Cummings, L. 1992. Activated Carbon <strong>Adsorption</strong> <strong>of</strong> Disinfection By-Product<br />

<strong>Precursors</strong> and Natural and Synthetic Organic Compounds. Master's<br />

Thesis, University <strong>of</strong> Cincinnati, Ohio.<br />

Daniel, W.W, 1987. Biostatistics: A Foundation for Analysis in the Health<br />

Sciences. New York: John Wiley & Sons.<br />

277


275 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

Dressman, R.C. 1980. Total Organic Halide, Method 450.1 —Interim. Cincinnati,<br />

Ohio: United States Environmental Protection Agency, Drinking <strong>Water</strong><br />

<strong>Research</strong> Division, Municipal Environmental <strong>Research</strong> Laboratory.<br />

Dryfuse, M.J., R.J. Miltner, and R.S. Summers. 1995. The <strong>Removal</strong> <strong>of</strong> Molecular<br />

Size and Humic/Nonhumic Fractions <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> Optimized<br />

Coagulation. In Proc. AWWA Annual Conference. Denver, Colo.:<br />

AWWA.<br />

Dugan, N., R.S. Summers, R.J. Miltner, and H.M. Shukairy. 1995. Modeling<br />

Chlorine Demand and Evaluating Chlorine Demand Chemistry. In Proc.<br />

<strong>of</strong> the AWWA <strong>Water</strong> Quality Technology Conference. Denver, Colo.:<br />

AWWA.<br />

Engineering News Record. 1995. July. New York: McGraw-Hill.<br />

Frick, B.R. 1982. Theoretische Betrachtungen zu den Probletnen des Scale-up<br />

von Aktivkohlefestbettadsorbern. Volume 20, in series <strong>of</strong> reports entitled<br />

Veroe-Henlichungen from the <strong>Water</strong> Chemistry Division. University <strong>of</strong><br />

Karlsruhe, Karlsruhe, Germany<br />

Gould, J.P., L.E. Fitchorn, and E. Urheim. 1983. Formation <strong>of</strong> Brominated<br />

Trihalomethanes: Extent and Kinetics. In <strong>Water</strong> Chlorination:<br />

Environmental Impact and Health Effects, Vol. 4. Edited <strong>by</strong> R.L. Jolley<br />

et al. Ann Arbor, Mich: Ann Arbor Sci. Publications.<br />

Hineline, D.W., J.C. Crittenden, and D.W. Hand. 1987. The Use <strong>of</strong> Rapid Small-<br />

Scale Column Tests to Predict Full-Scale Adsorber Capacity and<br />

Performance. In /4WW4 Annual Conference. Denver, Colo.: AWWA.<br />

Hodgeson, J.W., J. Collins, and R.C. Earth. 1990. Method 552: Determination<br />

<strong>of</strong> Haloacetic Acids in Drinking <strong>Water</strong> <strong>by</strong> Liquid-Liquid Extraction,<br />

Derivatization, and Gas Chromatography With Electron Capture<br />

Detection. Cincinnati, Ohio: United States Environmental Protection<br />

Agency, Office <strong>of</strong> <strong>Research</strong> and Development.<br />

Hong, S. 1995. Activated Carbon <strong>Adsorption</strong> <strong>of</strong> Organic Matter: Backwashing,<br />

Desorption, and Attenuation. Ph.D. dissertation. University <strong>of</strong> Cincinnati,<br />

Ohio.<br />

Hong, S., and R.S. Summers. 1994. Impact <strong>of</strong> Backwashing and Desorption on<br />

<strong>GAC</strong> Breakthrough <strong>of</strong> Natural Organic Matter. In Proc. AWWA Annual<br />

Conference. Denver, Colo.: AWWA.<br />

Hooper, S.M. 1996. Granular Activated Carbon <strong>Adsorption</strong> <strong>of</strong> Natural Organic<br />

Matter and Disinfection By-Product Control: Impact <strong>of</strong> Optimized<br />

Coagulation Pretreatment. M.S. thesis, Department <strong>of</strong> Civil and<br />

Environmental Engineering, University <strong>of</strong> Cincinnati, Ohio.<br />

Hooper, S.M., R.S. Summers, H.M. Shukairy, and D.M. Owen. 1994.<br />

Development <strong>of</strong> a New Test for the Assessment <strong>of</strong> Disinfection By-<br />

Product Formation: Uniform Formation Conditions. In Proc. <strong>of</strong> the<br />

AWWA <strong>Water</strong> Quality Technology Conference. Denver, Colo.: AWWA.<br />

Hooper, S.M., S. Hong, G. Solarik, R.S. Summers, and D.M. Owen. 1995. The<br />

Impact <strong>of</strong> Optimized Coagulation on Granular Activated Carbon<br />

<strong>Adsorption</strong> <strong>of</strong> Natural Organic Matter and Disinfection By-Product<br />

Control. In Proc. <strong>of</strong> the AWWA Annual Conference. Denver, Colo.:<br />

AWWA.<br />

Irwin, W. J. 1982. Analytical Pyrolysis. New York: Marcel Dekker.


References 279<br />

Jackson, J., S. Hong, and R.S. Summers. 1993. The Use <strong>of</strong> Ultrafiltration to<br />

Characterize <strong>GAC</strong> Breakthrough <strong>of</strong> Organic Matter Molecular Size<br />

Fractions. In Proc. AWWA <strong>Water</strong> Quality Technology Conference.<br />

Denver, Colo.: AWWA.<br />

Joselyn, B., and R.S. Summers. 1992. Control <strong>of</strong> Disinfection By-Product<br />

<strong>Precursors</strong> <strong>by</strong> Ozonation, Bi<strong>of</strong>iltration and Carbon <strong>Adsorption</strong>. In Proc.<br />

AWWA Annual Conference. Denver, Colo.: AWWA.<br />

Koch, B., R. Chinn, and M.K. Davis. 1987. A Simulated Distribution System<br />

Trihalomethane Formation Potential Method. In Proc. <strong>of</strong> the AWWA<br />

<strong>Water</strong> Quality Technology Conference. Denver, Colo.: AWWA.<br />

Koch, B., S.W. Krasner, M.J. Sclimenti, and W.K. Schimpff. 1991. Predicting<br />

the Formation <strong>of</strong> <strong>DBP</strong>s <strong>by</strong> the Simulated Distribution System. Jour.<br />

AWWA, 83(10):62.<br />

Koechling, M.T., and R.S. Summers. 1995. Evaluation <strong>of</strong> Ultrafiltration and<br />

Hydrophobic Separation for Characterizing Natural Organic Matter. In<br />

Proc. AWWA <strong>Water</strong> Quality Technology Conference. Denver, Colo.:<br />

AWWA.<br />

Koechling, M.T., H.M. Shukairy, and R.S. Summers. 1996. Effect <strong>of</strong> Ozonation<br />

and Biotreatment on Molecular Size and Hydrophilic Fractions <strong>of</strong><br />

Natural Organic Matter. In <strong>Water</strong> Disinfection and Natural Organic<br />

Matter: Characterization and Control. Edited <strong>by</strong> R.A. Minear and G.L.<br />

Amy. Washington, D.C.: American Chemical Society.<br />

Krasner, S.W., M.J. Sclimenti, E.G. Means, J.M. Symons, and L. Simms. 1992.<br />

The Impact <strong>of</strong> Chlorine Dose, Residual Chlorine, Bromide, and Organic<br />

Carbon on Trihalomethane Speciation. In Proc. <strong>of</strong> the AWWA <strong>Water</strong><br />

Quality Technology Conference. Denver, Colo.: AWWA.<br />

Malcolm Pirnie, Inc. 1994. City <strong>of</strong>Tempe Safe Drinking <strong>Water</strong> Act Compliance<br />

Study (Phase II). Unpublished report to the city.<br />

McCreary, J.J., and V.L. Snoeyink. 1980. Characterization and Activated<br />

Carbon <strong>Adsorption</strong> <strong>of</strong> Several Humic Substances. <strong>Water</strong> <strong>Research</strong>,<br />

14:151.<br />

Metz, D.H., R.S. Summers, and J. DeMarco. 1993. The Assessment <strong>of</strong><br />

Preozonation, Biotreatment, and <strong>GAC</strong> <strong>Adsorption</strong> <strong>of</strong> <strong>DBP</strong> <strong>Precursors</strong><br />

and Ozone <strong>DBP</strong>s Using the Rapid Small Scale Column Test. In<br />

Proceedings <strong>of</strong> AWWA <strong>Water</strong> Quality Technology Conference. Denver,<br />

Colo.: AWWA.<br />

Miller, R., D.J. Hartman, J. DeMarco, and B.W. Lykins, Jr. 1982. Feasibility<br />

Study <strong>of</strong> Granular Activated Carbon <strong>Adsorption</strong> and On-Site<br />

Regeneration. EPA-600/S2-82-087. Washington, D.C.: USEPA.<br />

Miltner, R.J. and R.S. Summers. 1992. A Pilot-Scale Study <strong>of</strong> Biological<br />

Treatment. In Proc. AWWA Annual Conf. Denver, Colo.: AWWA.<br />

Mogren, E., P.V. Scorpino, and R.S. Summers. 1990. Measurement <strong>of</strong><br />

Biodegradable Dissolved Organic Carbon in Drinking <strong>Water</strong>. In Proc.<br />

AWWA Annual Conference. Denver, Colo.: AWWA.<br />

Owen, D.M., G. Amy, and Z.K. Chowdhury. 1992. Characterization <strong>of</strong> Natural<br />

Organic Matter and Its Relationship to Treatability. Denver, Colo.:<br />

AWWARF and AWWA.


280 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> GA C <strong>Adsorption</strong><br />

Randtke, S.J., and V.L. Snoeyink. 1983. Evaluating <strong>GAC</strong> Adsorptive Capacity.<br />

Jour. AWWA, 75(8):406.<br />

Reckhow, D. A., and P.C. Singer. 1985. Mechanisms <strong>of</strong> Organic Halide Formation<br />

During Fulvic Acid Chlorination and Implications With Respect to<br />

Preozonation. In <strong>Water</strong> Chlorination: Chemistry, Environmental Impact<br />

and Health Effects, Vol. 5. Edited <strong>by</strong> R.L. Jolley. Chelsea, Mich.: Lewis<br />

Publishers.<br />

Roberts, P. V., and R.S. Summers. 1982. Granular Activated Carbon Performance<br />

for Organic Carbon <strong>Removal</strong>. Jour. AWWA. 74(2):113-118.<br />

Semmens, M.J., and A.B. Staples. 1986. The Nature <strong>of</strong> Organics Removed<br />

During Treatment <strong>of</strong> Mississippi River <strong>Water</strong>. Jour. AWWA, 78(2):76.<br />

Semmens, M.J., G.E. Norgaard, G. Hohenstein, and A.B. Staples. 1986. Influence<br />

<strong>of</strong> pH on the <strong>Removal</strong> <strong>of</strong> Organics <strong>by</strong> Granular Activated Carbon. Jour.<br />

AWWA, 78(5):89.<br />

Shukairy, H.M. 1994. The Control <strong>of</strong> Disinfection By-Product Formation <strong>by</strong><br />

Ozonation and Biotreatment. Ph.D. dissertation. University <strong>of</strong> Cincinnati,<br />

Ohio.<br />

Shukairy, J.M. and R.S. Summers. 1993. Biological Treatment for the Control<br />

<strong>of</strong> Disinfection By-Products: A Review. In Strategies and Technologies<br />

for Meeting SDWA Requirements. Edited <strong>by</strong> R.M.Clark and R.S.<br />

Summers. Lancaster, Pa.: Technomic Publications.<br />

Shukairy, J.M., R.J. Miltner, and R.S. Summers. 1994. Bromide's Effect on<br />

<strong>DBP</strong> Formation, Speciation and Control: Part 1, Ozonation. Jour.<br />

AWWA, 86(6):72.<br />

Sontheimer, H., J.C. Crittenden, and R.S. Summers. 1988. Activated Carbon for<br />

<strong>Water</strong> Treatment. 2d ed. Karlsruhe, West Germany: DVGW-<br />

Forschungsstelle, Universitat Karlsruhe (distributed in the United States<br />

<strong>by</strong> AWWA).<br />

Stevens, A.A., and J.M. Symons. 1977. Measurement <strong>of</strong> Trihalomethane and<br />

Precursor Concentration Changes. Jour. AWWA, 69(10):546.<br />

Stevens, A.A., L.A. Moore, and R.J. Miltner. 1989. Formation and Control <strong>of</strong><br />

Non-Trihalomethane Disinfection By-Products. Jour. A WWA, 81 (8):54.<br />

Summers, R.S. 1986. Activated Carbon <strong>Adsorption</strong> <strong>of</strong>Humic Substances: Effect<br />

<strong>of</strong> Molecular Size and Heterodispersity. Ph.D. dissertation, Department<br />

<strong>of</strong> Civil Engineering, Stanford University, Stanford, Calif.<br />

Summers, R.S., and P.V. Roberts. 1988. Activated Carbon <strong>Adsorption</strong> <strong>of</strong>Humic<br />

Substances, II: Size Exclusion and Electrostatic Interactions. J. Colloid<br />

and Interface Sci., 122(2)382-397.<br />

Summers, R.S., and J.C. Crittenden. 1989. The Use <strong>of</strong> Mini-Columns for the<br />

Prediction <strong>of</strong> Full-Scale <strong>GAC</strong> Behavior. In Design and Use <strong>of</strong> Granular<br />

Activated Carbon: Practical Aspects. Conference proceedings. Denver,<br />

Colo.: AWWARF and AWWA.<br />

Summers, R.S., M. Benz, P. H<strong>of</strong>fman, and H. Sontheimer. 1989. <strong>Adsorption</strong> <strong>of</strong><br />

Dissolved Organic Carbon <strong>by</strong> Activated Carbon Scale-Up, Modeling<br />

and Desorption. In Proc. <strong>of</strong> the AWWA Annual Conference. Denver,<br />

Colo.: AWWA.


References 281<br />

Summers, R.S., L. Cummings, J. DeMarco, D. Hartman, D. Metz, E.W. Howe,<br />

B. MacLeod, and M. Simpson. 1992. Standardized Protocol for the<br />

Evaluation <strong>of</strong><strong>GAC</strong>. In Proc. <strong>of</strong>theAWWA Annual Conference. Denver,<br />

Colo.: AWWA.<br />

Summers, R.S., S. Hong, S.M. Hooper, and G. Solarik. 1994a. <strong>Adsorption</strong> <strong>of</strong><br />

Natural Organic Matter and Disinfection By-Product <strong>Precursors</strong>. In<br />

Proc. <strong>of</strong> the AWWA Annual Conference. Denver, Colo.: AWWA.<br />

Summers, R.S., S.M. Hooper, S. Hong, and G. Solarik. 1994b. The Use <strong>of</strong><br />

RSSCTs to predict <strong>GAC</strong> Field-Scale Performance. In Proc. <strong>of</strong> the<br />

AWWA Annual Conference. Denver, Colo.: AWWA.<br />

Summers, R.S., S.M. Hooper, H.M. Shukairy, and D.M. Owen. 1994c.<br />

Development <strong>of</strong> Uniform Formation Conditions for the Assessment <strong>of</strong><br />

Disinfection By-Product Formation. In Proc. <strong>of</strong> the AWWA Annual<br />

Conference. Denver, Colo.: AWWA.<br />

Summers R.S., S.M. Hooper, G. Solarik, D.M. Owen, and S. Hong. 1995.<br />

Bench-Scale Evaluation <strong>of</strong> Activated Carbon for the Control <strong>of</strong> Natural<br />

Organic Matter and Disinfection By-Products. Jour. AWWA. 87(8):69.<br />

Summers, R.S., S.M. Hooper, H.M. Shukairy, G. Solarik, and D.M. Owen.<br />

1996. Assessing DBF Yield: Uniform Formation Conditions. Jour.<br />

AWWA, 88(6):80.<br />

Symons, J.M., S.W. Krasner, L.A. Simms, and M. Sclimenti. 1993. Measurement<br />

<strong>of</strong> THM and Precursor Concentrations Revisited: The Effect <strong>of</strong> Bromide<br />

Ion. Jour. AWWA, 85(1):93.<br />

Thurman, E.M., and R.L. Malcolm. 1981. Preparative Isolation <strong>of</strong> Aquatic<br />

Humic Substances. Environ. Sci Tech., 15:463.<br />

USEPA (United States Environmental Protection Agency). 1990. Methods for<br />

the Determination <strong>of</strong> Organic Compounds in Drinking <strong>Water</strong> Supplement<br />

I. EPA-600/4-90-020. Cincinnati, Ohio: USEPA.<br />

————. 1991. Methods for the Determination <strong>of</strong> Organic Compounds in<br />

Drinking <strong>Water</strong>. EPA-600/4-88-039. Cincinnati, Ohio: USEPA.<br />

————. 1992. Methods for the Determination <strong>of</strong> Organic Compounds in<br />

Drinking <strong>Water</strong>—Supplement II. EPA-600/R-92-129. Cincinnati, Ohio:<br />

USEPA.<br />

————. 1993. Technologies and Costs for the Control <strong>of</strong> Disinfectants and<br />

Disinfection By-Products. Cincinnati, Ohio: USEPA.<br />

————. 1994. Federal Register, 59(145):38668-38829.<br />

————. 1996. ICR Manual for Bench- and Pilot-Scale Treatment Studies.<br />

Cincinnati, Ohio: USEPA, Technical Support Division, Office <strong>of</strong> Ground<br />

<strong>Water</strong> and Drinking <strong>Water</strong>.<br />

Weber, W.J., Jr., et al. 1983. <strong>Adsorption</strong> <strong>of</strong> Humic Substances: The Effects <strong>of</strong><br />

Heterogeneity and System Characteristics. Jour. AWWA, 75(12):612.<br />

West, G.S., and J.T. Hepworth. 1991. Statistical Issues in the Study <strong>of</strong> Temporal<br />

Data: Daily Experiences, Journal <strong>of</strong> Personality, 59(3):609-661.


Abbreviations<br />

ACF autocorrelation function<br />

amu atomic mass unit<br />

a level <strong>of</strong> significance<br />

aQ, a,, BO, Cj parameters used in regression<br />

analysis modeling <strong>of</strong><br />

pretreatment<br />

Avg. average<br />

AWWA American <strong>Water</strong> Works<br />

Association<br />

AWWARF American <strong>Water</strong> Works<br />

Association <strong>Research</strong> Foundation<br />

BAT best available technology<br />

BCAA bromochloroacetic acid<br />

BDOC biodegradable dissolved organic<br />

carbon<br />

Br:DOC bromide-to-DOC ratio<br />

BrrTOC bromide-to-TOC ratio<br />

BV bed volumes<br />

C concentration<br />

C. concentration <strong>of</strong> immediate<br />

imm<br />

breakthrough<br />

CQ influent concentration<br />

C Q average influent concentration<br />

CD constant diffusivity<br />

CH chloral hydrate<br />

°C degrees Celsius<br />

cm centimeter<br />

I/cm per centimeter<br />

CT conventional treatment<br />

CUR carbon usage rate<br />

CWC Culp-Wesner-Culp<br />

CWW Cincinnati <strong>Water</strong> Works<br />

D dalton<br />

d day<br />

D-<strong>DBP</strong> Rule Disinfectants-Disinfection By-<br />

Products Rule<br />

DBAA dibromochloroacetic acid<br />

DBF disinfection <strong>by</strong>-products<br />

283<br />

<strong>DBP</strong>FP disinfection <strong>by</strong>-product formation<br />

potential<br />

DCAA dichloroacetic acid<br />

DCE cw-l,2-dichloroethene<br />

DOC dissolved organic carbon<br />

EBCT empty bed contact time<br />

eV electron volt<br />

FGW Florida groundwater<br />

FP formation potential<br />

ft foot<br />

g gram<br />

<strong>GAC</strong> granular activated carbon<br />

gal gallons<br />

GC gas chromatography<br />

gpm gallons per minute<br />

h hour<br />

HAA haloacetic acid<br />

HAAS sum <strong>of</strong> five haloacetic acids<br />

HAA6 sum <strong>of</strong> six haloacetic acids<br />

ICR Information Collection Rule<br />

in. Hg inch mercury<br />

JPWD Jefferson Parish <strong>Water</strong> District<br />

k' ratio <strong>of</strong> resin void volume to<br />

volume <strong>of</strong> sample passed<br />

kW'h kilowatt-hour<br />

L liter<br />

LC large column<br />

LCW laboratory clean water<br />

LGW Lake Gaillard water<br />

Ibs pounds<br />

M molar<br />

m meter


284 <strong>Removal</strong> <strong>of</strong> DBF <strong>Precursors</strong> <strong>by</strong> <strong>GAC</strong> <strong>Adsorption</strong><br />

MBAA monobromoacetic acid<br />

MCAA monochloroacetic acid<br />

MCL maximum contaminant level<br />

mg milligrams<br />

mgd million gallons per day<br />

jig microgram<br />

uL microliter<br />

mil gal million gallon<br />

min minutes<br />

mL milliliter<br />

MLW Manatee Lake water<br />

mm millimeter<br />

MRW Mississippi River water<br />

MS molecular size (synonymous to<br />

apparent molecular weight,<br />

AMW); mass spectrometry<br />

N normal<br />

n bromine incorporation factor for<br />

THM speciation<br />

n' bromine incorporation factor for<br />

HAA speciation<br />

nm nanometer<br />

NOM natural organic matter<br />

O&M operation and maintenance<br />

OC optimized coagulation<br />

ORW Ohio River water<br />

PD proportional diffusivity<br />

pH negative logarithm <strong>of</strong> hydrogen<br />

ion concentration<br />

PRW Passaic River water<br />

PTFE polytetrafluoroethylene<br />

PVWC Passaic Valley <strong>Water</strong> Commission<br />

PY-GC-MS pyrolysis gas chromatographymass<br />

spectrometry<br />

rpm<br />

RSSCT<br />

s<br />

SC<br />

SCCRWA<br />

SDS<br />

SF<br />

SOC<br />

SRP<br />

SRPW<br />

TCAA<br />

THM<br />

'in,<br />

TOC<br />

TOX<br />

TTHM<br />

UC<br />

UFC<br />

UM<br />

USEPA<br />

UV<br />

UV 254<br />

WIDB<br />

WTP<br />

solid phase concentration<br />

correlation coefficient<br />

revolutions per minute<br />

rapid small-scale column test<br />

second<br />

small column<br />

South Central Connecticut<br />

Regional <strong>Water</strong> Authority<br />

simulated distribution system<br />

scaling factor<br />

synthetic organic chemical<br />

Salt River Project<br />

Salt River Project water<br />

trichloroacetic acid<br />

trihalomethane<br />

time at which adsorbable fraction<br />

begins to show breakthrough<br />

total organic carbon<br />

total organic halide<br />

total trihalomethane<br />

University <strong>of</strong> Cincinnati<br />

uniform formation conditions<br />

University <strong>of</strong> Massachusetts<br />

United States Environmental<br />

Protection Agency<br />

ultraviolet<br />

ultraviolet light absorbance at 254<br />

nanometers<br />

<strong>Water</strong> Industry Data Base<br />

<strong>Water</strong> treatment plant


American <strong>Water</strong> Works Association<br />

RESEARCH FOUNDATION<br />

6666 W. Quincy Avenue, Denver, CO 80235<br />

(303)347-6100<br />

1 P-4.5C-90744-2/99-CM ISBN 0-89867-941-9

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!