Flood-Runoff Analysis - Publications, US Army Corps of Engineers

DEPARTMENT OF THE ARMY EM 1110-2-1417U.S. Army Corps of EngineersCECW-EH Washington, DC 20314-1000ManualNo. 1110-2-1417 31 August 1994Engineering and DesignFLOOD-RUNOFF ANALYSIS1. Purpose. This manual describes methods for evaluating flood-runoff characteristics of watersheds.Guidance is provided in selecting and applying such methods to support the various investigationsrequired for U.S. Army Corps of Engineers (USACE) civil works activities. The manual referencespublications that contain the theoretical basis of the methods and detailed information on their use.2. Applicability. The manual applies to all HQUSACE elements, major subordinate commands,districts, laboratories, and field operating activities having civil works responsibilities for the design ofcivil works projects.FOR THE COMMANDER:WILLIAM D. BROWNColonel, Corps of EngineersChief of Staff

DEPARTMENT OF THE ARMY EM 1110-2-1417U.S. Army Corps of EngineersCECW-EH Washington, DC 20314-1000ManualNo. 1110-2-1417 31 August 1994Engineering and DesignFLOOD-RUNOFF ANALYSISTable of ContentsSubject Paragraph PageChapter 1IntroductionPurpose ....................... 1-1 1-1Applicability .................... 1-2 1-1References ..................... 1-3 1-1Scope and Organization ............ 1-4 1-1Relationship to OtherGuidance ...................... 1-5 1-1Part I Problem Definition andSelection of MethodologyChapter 2Introduction to Flood-Runoff AnalysisGeneral ....................... 2-1 2-1Applications of Flood-RunoffAnalysis ...................... 2-2 2-1Nature of Flood Hydrology .......... 2-3 2-2Data Considerations ............... 2-4 2-3Approaches to Flood-RunoffAnalysis ...................... 2-5 2-3Chapter 3Study Formulation and ReportingGeneral ....................... 3-1 3-1Overview of Corps Flood DamageReduction Studies ............... 3-2 3-1Planning and Managing the HydrologicInvestigation ................... 3-3 3-1Hydrologic Engineering AnalysisStrategy ...................... 3-4 3-2Hydrologic Requirements for PlanningStudies ....................... 3-5 3-3Preconstruction Engineering andDesign (PED) Phase ............. 3-6 3-6Subject Paragraph PageConstruction and Operation ........... 3-7 3-7Reporting Requirements ............. 3-8 3-7Summary ....................... 3-9 3-8Part II Hydrologic AnalysisChapter 4Rainfall AnalysisGeneral ......................... 4-1 4-1Point Rainfall Data ................ 4-2 4-1Rainfall Data From Remote Sensors ..... 4-3 4-1Areal and Temporal Distributionof Rainfall Data ................. 4-4 4-5Chapter 5Snow AnalysisGeneral ......................... 5-1 5-1Physical Processes ................. 5-2 5-1Data Requirements, Collection,and Processing .................. 5-3 5-2Simulating Snow Accumulation ........ 5-4 5-3Simulating Snowmelt ............... 5-5 5-6Chapter 6Infiltration/Loss AnalysisGeneral ......................... 6-1 6-1Gauged versus Ungauged ParameterEstimation ..................... 6-2 6-5Antecedent Moisture Conditions ....... 6-3 6-5Surface Loss Estimation ............. 6-4 6-6Infiltration Methods ................ 6-5 6-6Impervious Areas .................. 6-6 6-20Method Selection .................. 6-7 6-21i

EM 1110-2-141731 Aug 94Subject Paragraph PageChapter 7Precipitation Excess - Runoff TransformationGeneral ........................ 7-1 7-1Runoff Subdivision ............... 7-2 7-1Unit Hydrograph Approach .......... 7-3 7-1Kinematic Wave Approach .......... 7-4 7-12Chapter 8Subsurface Runoff AnalysisGeneral ........................ 8-1 8-1Event-Oriented Methods ............ 8-2 8-1Evapotranspiration ................ 8-3 8-5Continuous Simulation Approachto Subsurface Modeling ........... 8-4 8-11Existing Continuous SimulationModels ....................... 8-5 8-16Parameter Estimation for ContinuousSimulation Models ............... 8-6 8-23Chapter 9Streamflow and Reservoir RoutingGeneral ........................ 9-1 9-1Hydraulic Routing Techniques ........ 9-2 9-2Hydrologic Routing Techniques ....... 9-3 9-5Applicability of Routing Techniques . . . 9-4 9-21Chapter 10Multisubbasin ModelingGeneral ........................ 10-1 10-1General Considerations for SelectingBasin Components ............... 10-2 10-1Selection of Hydrograph ComputationLocations ..................... 10-3 10-2Calibration of Individual Components . . 10-4 10-4Calibration of Multisubbasin Model .... 10-5 10-4Verification of the MultisubbasinModel ....................... 10-6 10-5Part III Methods for Flood-RunoffAnalysisChapter 11Simplified TechniquesIntroduction .................... 11-1 11-1Rational Method ................. 11-2 11-1Regional Frequency Analysis ........ 11-3 11-1Envelope Curves ................. 11-4 11-5Rainfall Data Sources .............. 11-5 11-6Subject Paragraph PageChapter 12Frequency Analysis of Streamflow DataGeneral ......................... 12-1 12-1Frequency Analysis Concepts ......... 12-2 12-1Graphical Techniques ............... 12-3 12-3Numerical Techniques .............. 12-4 12-5Special Considerations .............. 12-5 12-10Chapter 13Analysis of Storm EventsIntroduction ..................... 13-1 13-1Model Development ................ 13-2 13-1Model Calibration ................. 13-3 13-2Simulation of Frequency-BasedDesign Floods ................... 13-4 13-3Simulation of Standard Project andProbable Maximum Floods .......... 13-5 13-5Chapter 14Period-of-Record AnalysisGeneral ......................... 14-1 14-1Simulation Requirements ............ 14-2 14-1Model Calibration ................. 14-3 14-1Applications ..................... 14-4 14-4Part IV Engineering ApplicationsChapter 15Data Collection and ManagementGeneral ......................... 15-1 15-1Data Management Concepts .......... 15-2 15-1Geographic Information Systems ....... 15-3 15-1Data Acquisition and Use ............ 15-4 15-2Chapter 16Ungauged Basin AnalysisGeneral ......................... 16-1 16-1Loss-Model Parameter Estimates ....... 16-2 16-2Runoff-Model Parameter Estimates ..... 16-3 16-3Routing-Model Parameter Estimates ..... 16-4 16-4Statistical-Model Parameter Estimates . . . 16-5 16-5Reliability of Estimates .............. 16-6 16-6Chapter 17Development of Frequency-Based EstimatesIntroduction ..................... 17-1 17-1Choice of Methodology ............. 17-2 17-1ii

EM 1110-2-141731 Aug 94Subject Paragraph PageHypothetical Storm Frequency ........ 17-3 17-2Transfer of Frequency Information withHypothetical Events .............. 17-4 17-3Development of Future-ConditionFrequency Estimates ............. 17-5 17-3Adjustment of Peak Dischargesto Represent Stationary Conditions . . . 17-6 17-4Chapter 18Evaluating ChangeGeneral ........................ 18-1 18-1Evaluating Catchment andConveyance-System Change ........ 18-2 18-1Procedure for Evaluating Damage-Reduction Plans ................ 18-3 18-3Subject Paragraph PageEvaluating Reservoir and DetentionBasins ........................ 18-4 18-4Evaluating Channel Alterationsand Levees ..................... 18-5 18-8Evaluating Other Alternatives ......... 18-6 18-10Appendix AReferencesAppendix BHydrologic Engineering Management Plan forFlood Damage Reduction Feasibility-PhaseStudiesiii

EM 1110-2-141731 Aug 94Chapter 1Introduction1-1. PurposeThis manual describes methods for evaluating flood-runoffcharacteristics of watersheds. Guidance is provided inselecting and applying such methods to support the variousinvestigations required for U.S. Army Corps of Engineers(USACE) civil works activities. The manualreferences publications that contain the theoretical basis ofthe methods and detailed information on their use.1-2. ApplicabilityThis manual applies to HQUSACE elements, major subordinatecommands, districts, laboratories, and field operatingactivities having civil works responsibilities.1-3. ReferencesReferences are listed in Appendix A.1-4. Scope and Organizationa. The manual is organized into four parts. Thefirst, Problem Definition and Selection of Methodology,describes the products of flood-runoff analysis and thetypes of investigation for which these products arerequired. Aspects of flood hydrology are discussed,including physical processes, data availability, and broadapproaches to analysis. Guidance in formulating studyprocedures is provided, which includes criteria for methodselection and recommended content for a hydrologic engineeringmanagement plan (HEMP). The reporting ofstudy results is the focus of the last chapter in Part I.b. Part II, Hydrologic Analysis, provides informationon techniques for simulating various components of thehydrologic cycle, including rainfall, snow, infiltration(loss), surface and subsurface runoff, and flow in channelsand reservoirs. Multisubbasin modeling and design stormdefinition are discussed.c. Part III, Methods for Flood-Runoff Analysis,addresses the application of simplified techniques, frequencyanalysis of streamflow data, precipitation-runoffsimulation of storm events, and period-of-record precipitation-runoffsimulation. Data requirements and calibration/verificationof simulation models are considered.d. Part IV, Engineering Applications, deals withseveral issues associated with the application of methodsfrom Part III. The processing of data can be time-consumingand costly; techniques for efficient data handlingare addressed. The lack of historical streamflow data isthe source of much difficulty and uncertainty in floodrunoffanalysis. Aspects of dealing with “ungauged”basins are discussed. Issues associated with the developmentof frequency-based estimates are covered, includingthe concept of calibration to “known” frequency information.Various aspects of modeling land use change, aswell as the effects of reservoir and other projects, arediscussed. Finally, three examples illustrate some of theprinciples presented in this manual.e. Following Part IV, Appendices A and B providereferences, a generic HEMP, and a set of exampleapplications.1-5. Relationship to Other GuidanceThis engineer manual (EM) relies on references and/ortechnical information in several other guidance documents.Some of those documents are part of this currentguidance effort and others are older documents. The mostrelevant documents are EM 1110-2-1416, River Hydraulics,EM 1110-2-1415, Hydrologic Frequency Analysis,and EM 1110-2-1413, Hydrologic Analysis of InteriorAreas. These documents provide the basic technicalbackground for study procedures closely related to floodrunoffanalysis or information for how the results of floodstudies are used in project analyses. Specific referencesto these and other EM’s are made throughout thisdocument.1-1

PART 1PROBLEM DEFINITION ANDSELECTION OF METHODOLOGY

EM 1110-2-141731 Aug 94Chapter 2Introduction to Flood-Runoff Analysis2-1. GeneralThis chapter describes products of flood-runoff analysisand relates them to the various types of investigationsassociated with the Corps of Engineers Civil Works activities.Flood-runoff analysis as described in this manualcan be regarded as an engineering application of thescience of flood hydrology. Aspects of flood hydrologyare briefly described as a precursor to detailed treatmentin Part II, Hydrologic Analysis. The type, amount, andquality of hydrologic and meteorologic data available fora flood-runoff analysis affect the choice of methodologyand reliability of results. Consequences of data availabilityare discussed. Finally, broad approaches to floodrunoffanalysis are presented. The approaches are aframework for a detailed discussion of methods inPart III, Methods for Flood-Runoff Analysis.2-2. Applications of Flood-Runoff Analysisa. Products of flood-runoff analysis. Products canbe categorized with respect to the type of variable (e.g.,stage, discharge, volume) and the measure of the variable.(1) Measure might be simply the magnitude associatedwith a particular point in time (as in flow forecasting),magnitude associated with a nonfrequency baseddesign flood (e.g., standard project or probable maximum),magnitude associated with duration (e.g., value thatis exceeded, or not exceeded, X-% of the time), or magnitudeassociated with a particular exceedance or nonexceedancefrequency. Exceedance frequency measuresare particularly common for flood prediction and are thebasis for flood risk evaluations (e.g., delineation of the“1-% chance” floodplain for flood insurance purposes), aswell as flood damage analysis for project design. In otherwords, the end product of many flood-runoff analyses is aset of discharge or stage exceedance frequency relations,perhaps for both existing and alternative future conditions,for locations of interest in a watershed. The developmentof probabilistic estimates of flood runoff is dealt with inChapter 12, “Frequency Analysis of Streamflow Data,”and Chapter 17, “Development of Frequency-BasedEstimates.”(2) Generally, water elevation at a location in a riveror on a floodplain is of more direct interest for floodanalysis than magnitude of discharge. Water elevation isdetermined with a hydraulic analysis, which is oftentimesperformed subsequent to a hydrologic analysis. However,the hydraulic characteristics of floodwave movement arean important aspect of hydrologic analysis, and there aresituations where it is best to incorporate detailed hydraulicanalysis directly in the determination of discharge. Chapter9, “Streamflow and Reservoir Routing,” deals withhydraulic aspects of hydrologic analysis, including techniqueswith which water elevations can be determined.b. Types of investigation requiring flood-runoff analysis.Types of investigation include flood risk evaluationof floodplains, flood damage evaluation for project planning,design of hydraulic structures for flood control, andflood-runoff forecasting for project operations.(1) The evaluation of flood risk for floodplains, suchas is required for flood insurance studies, requires discharge-exceedancefrequency estimates for locations alonga stream. Discharges for selected exceedance frequenciesare then used in the hydraulic determination of watersurface profiles from which maps of inundated areas canbe prepared. Hence, the primary product of flood-runoffanalysis for these investigations is a set of dischargeexceedancefrequency relations for current land useconditions.(2) Flood damage evaluations for project planninggenerally require the development of both dischargeexceedancefrequency relations and stage-discharge relationsfor index locations associated with “damage”reaches of a stream. These relations must be developedfor existing conditions as well as future conditions withand without proposed projects. The development of suchrelations is among the most challenging of applications inflood-runoff analysis. Chapter 18, “Evaluating Change,”is particularly pertinent to such studies.(3) Design of hydraulic structures for floods such asthe standard project or probable maximum generallyrequires estimation of the peak stage, discharge, or runoffvolume associated with such events. In the case of alarge dam, the spillway capacity and height of dam aregenerally based on routing the spillway design flood (i.e.,the probable maximum flood) through the reservoir.Because such events are beyond experience, judgment isrequired in establishing parameters for the analysis.Chapter 13, “Analysis of Storm Events,” deals withaspects of such analyses.(4) Real-time estimates of flood runoff are used inmaking operational decisions for reservoirs, reservoirsystems, and other hydraulic structures. Precipitation,stage, and other data are transmitted by telemetry systems2-1

EM 1110-2-141731 Aug 94to water control centers, where the data are processed andforecasts are made. Although flow forecasting is notdealt with explicitly in this manual, pertinent sections arePart II on hydrologic analysis and Part III sections dealingwith precipitation-runoff modeling. Other types of investigationfor which flood-runoff analysis may be requiredinclude those involving the evaluation of applications forpermits to encroach on water bodies and studies involvingthe design of flood warning systems. In both cases, simplifiedtechniques may be appropriate, some of which aredescribed in Chapter 11, “Simplified Techniques.”2-3. Nature of Flood Hydrologya. The hydrologic system.(1) A significant aspect of flood hydrology is the estimationof the magnitude of streamflow at various locationsin a watershed resulting from a given precipitationinput, as illustrated schematically in Figure 2-1.(2) The hydrologic system embodies all of the physicalprocesses that are involved in the conversion of precipitationto streamflow, as well as physical characteristicsof the watershed and atmosphere that influence runoffgeneration. The use of computer models to simulate thehydrologic system is of major significance in the performanceof many flood-runoff analyses. A fundamentalproblem in simulating hydrologic systems is to employthe appropriate level of detail to represent those componentsof the system that have a significant influence onthe phenomena being modeled. An associated problem isto acquire and interpret information on watershed characteristics,etc. to enable appropriate representation of thesystem. Part II, Hydrologic Analysis, is largely devotedto techniques for representing various components of thehydrologic system.b. Physical processes. The hydrologic cycle comprisesall of the physical processes that affect the movementof water in its various forms, from its occurrence asprecipitation near the earth’s surface to it’s discharge tothe ocean. Such processes include interception, waterstorage in depressions, water storage in lakes andFigure 2-1. Hydrologic systemreservoirs, snow accumulation and melt, infiltrationthrough the earth’s surface, percolation to various depthsin the subsurface, the storage of water in the subsurface,the lateral movement of water in both unsaturated andsaturated portions of the subsurface, evaporation fromwater bodies and moist soil, transpiration from vegetation,overland flow, and streamflow. The processes are complexand can be defined with varying degrees of sophistication.Some processes are more significant than othersfor particular types of analysis. For example, if an analysisof runoff from a historical storm with an event-typesimulation model were being performed, it would beappropriate to exclude evapotranspiration during the stormevent from the analysis. On the other hand, if a continuous(moisture accounting) simulation model were beingused for a period-of-record analysis, appropriate representationof evapotranspiration would be very significant.c. Storm characteristics.(1) In Figure 2-1, precipitation is viewed as an inputto a hydrologic system. The precipitation might be associatedwith a historical storm, a design storm, or mayresult from a stochastic generation procedure. Generally,precipitation is averaged spatially (i.e., “lumped”) over asubbasin, or perhaps over a geometric “element,” if a“distributed” model is being used. Likewise, precipitationintensity is averaged over a time interval. Thus, the precipitationinput to the hydrologic system is commonlyrepresented by hyetographs of spatially and temporallyaveraged precipitation. The development of such hyetographsis addressed in Chapter 4, “Rainfall Analysis.”(2) Each storm type (e.g., convective, frontal, orographic)has predominant characteristics regarding thespatial extent and variability, intensity, and duration ofprecipitation. Precipitation fields associated with storms,especially the convective type, exhibit substantial spatialand temporal variability. The sampling of such fieldswith gauge networks of typical density results in precipitationestimates that may be highly uncertain. Indeed, thegauge measurements themselves may exhibit significantuncertainty, primarily due to wind effects. As indicatedin Chapter 4, advances in use of radar-based rainfall datamay offer a significant improvement in capabilities fordefining the spatial and temporal variations of rainfall.d. Watershed characteristics. A key aspect ofsimulating a hydrologic system is representation of thephysical properties of the system. Watersheds are heterogenouswith respect to topography, geology, soils, landuse, vegetation, drainage density, river characteristics, etc.In most applications, the properties are lumped on a2-2

EM 1110-2-141731 Aug 94subbasin basis and represented by simple indices. Therepresentation of physical properties is dealt with in chaptersin Part II that treat components of the hydrologicsystem.e. Scale considerations. The techniques that aremost appropriate for a simulation model are a function ofthe scale of the phenomena being modeled.(1) For example, for small upland basins, a physicallybased model should recognize a variety of storm-runoffproduction mechanisms, including overland flow causedby rainfall exceeding infiltration capacity over the entirebasin, overland flow caused by rainfall exceeding infiltrationcapacity over a portion of the basin (partial areaoverland flow), overland flow caused by a high watertable near the stream system, and subsurface stormflow.Even with capabilities to simulate these processes, suchmodels may not perform satisfactorily because of the lackof information regarding spatial variability of rainfall andof subsurface hydraulic properties.(2) At a larger scale (i.e., larger basins), theprocesses that are dominant at a smaller scale tend toaverage out such that different approaches to modeling areappropriate. Emphasis is given to use of the unit hydrographand (macro scale) kinematic wave methods in thismanual. However, application of these methods requiresthe determination of rainfall excess and the estimation ofsubsurface contributions to runoff, both of which are thesource of substantial uncertainty. Also, at the largerscale, flood wave movement through the stream networkbecomes a dominant factor affecting the magnitude andtiming of flood runoff. Hence, significant attention mustbe given to streamflow routing. The primary focus in thismanual is on basins that are from one to thousands ofsquare miles in size, and for which it is generally necessaryto divide the basin into multiple subbasins and performstreamflow routing to obtain total flow at the outletsof downstream subbasins.2-4. Data Considerationsa. Types and sources of data for flood-runoff analysis.Data may be categorized as that related to physicalattributes of a basin, and data pertaining to the historicalmovement of water (in its various states) through thehydrologic cycle.(1) Physical attributes include area, surficial geometriccharacteristics (area, shape, slope, etc.), soil type, landuse, vegetative cover, subsurface characteristics (location,size and geometry of subsurface features, hydraulicconductivities, etc.), and stream channel characteristics(shape, slope, roughness, etc.). Some of these attributesare static, while others may change seasonally or overlonger time periods. Generally for flood studies,resources are not expended in acquiring subsurface information,as such information can be very costly to acquire,and use of such information is limited.(2) Data related to water movement include precipitation,snow depth and other snow-related information,storage of water in surface water bodies, infiltration, soilmoisture, movement of water in both unsaturated andsaturated portions of the subsurface, evaporation, transpiration,and streamflow (or flow in conduits or other drainagedevices). In addition, meteorologic data such as airtemperature, solar radiation and wind may be used withenergy relations to define water movement. Although anumber of these data types might be used in a particularanalysis, many flood-runoff studies rely primarily onhistorical precipitation and streamflow data.b. Significance of data availability. Because of thecomplex nature of hydrologic processes, storm characteristicsand basin characteristics, the type and amount ofdata available can have a major influence on the choice ofmethodology for performing an analysis and on the reliabilityof results. Part III, Methods for Flood-RunoffAnalysis, describes the data requirements for variousmethods. Streamflow data, in particular, is extremelyvaluable. A relatively long record of streamflow data canbe used to make estimates of flood-runoff probabilitiesthat are far more reliable than could be made by anymethod without such data. Even a short record of streamflowdata is valuable because it can be used in the calibrationof precipitation-runoff simulation models.2-5. Approaches to Flood-Runoff AnalysisIn this section, general approaches to flood-runoff analysisare described. For each approach, there may be severalmethods of analysis. These are described in detail inPart III. Selection of methods is discussed in Chapter 3,“Study Formulation and Reporting.”a. Approaches. Methods of flood-runoff analysis arecategorized under four approaches, as follows:(1) Simplified methods.(2) Frequency analysis of streamflow data.(3) Precipitation-runoff analysis of storm events.2-3

EM 1110-2-141731 Aug 94(4) Period-of-record precipitation-runoff analysis.(a) Simplified methods may involve use of formulas,previously derived regression equations, envelope curves,etc. as a basis for making hydrologic estimates. Themethods may be especially useful for preliminary estimatesof the expected magnitude of a variable, or forproviding an independent check on estimates developedby other means.(b) Where adequate streamflow data are available,frequency analysis of such data can be performed todevelop exceedance frequency relationships. Generalaspects of such analyses are described in Chapter 12;details are provided in EM 1110-2-1415, Hydrologic FrequencyAnalysis.(c) For situations where historical streamflow data isnonexistent or inadequate for required estimates, a precipitation-runoffsimulation model is commonly used forflood-runoff analysis. Generally, such a model must beused if it is intended to evaluate flood runoff effects ofstructural projects or historic or future land use changes.The third approach listed above involves use of a simulationmodel that is designed for analyzing single stormevents. Such models do not perform a continuous waterbalance and, therefore, must be provided input thatdescribes the state of a basin (in terms of base flow andsome measure of wetness) at the beginning of the simulation.Design storms are used with such models todevelop exceedance frequency estimates, or design-floodestimates, of hydrologic variables of interest. Care mustbe exercised in assigning exceedance frequencies to simulatedvalues because the runoff from a storm of specificexceedance frequency does not necessarily have the sameexceedance frequency. Chapter 17, “Development ofFrequency-Based Estimates,” deals with this issue. It isalso possible to use an “event” type model to both analyzeeach of the largest precipitation events of record anddevelop exceedance frequency estimates by statisticalanalysis of the results. This type of discrete event periodof-recordanalysis requires screening of precipitation datafor the largest events and the establishment of initialconditions at the beginning of each event, as discussed inChapter 13, “Analysis of Storm Events.”(d) The fourth approach is to use a precipitationrunoffsimulation model with period-of-record precipitationas an input and to simulate period-of-recordsequences of the variables of interest. If exceedancefrequency relations are desired, they can be developed byconventional statistical analysis of the period-of-recordoutputs. Such a model maintains a continuous moisturebalance; therefore, the state of the basin at the beginningof each storm event is implicitly determined. The use ofsuch models is conceptually attractive. However, themodel requirements in terms of data and the number ofparameters that must be calibrated are substantial.Aspects of continuous moisture accounting are describedin Chapter 8, “Subsurface Runoff Analysis,” andChapter 14, “Period-of-Record Analysis.”b. Factors affecting choice of approach. The choiceof approach for a flood-runoff analysis should take intoaccount required “products” of the analysis, data availability,reliability of results, and resource requirements.With regard to data availability, a key factor is the availabilityof streamflow data adequate for frequency analysis,if frequency estimates are required. Though not alwaysthe case, improved reliability is generally achieved withthe use of more sophisticated and comprehensive methodsof analysis. There is significant uncertainty associatedwith virtually all hydrologic estimates. It is often advisableto produce estimates by two or more independentmethods and to perform a sensitivity analysis to gaininformation regarding reliability of results. Finally, financialand human resources available for a study can be acontrolling factor in choice of methodology. These issuesare discussed in Chapter 3, “Study Formulation andReporting.”2-4

EM 1110-2-141731 Aug 94Chapter 3Study Formulation and Reporting3-1. GeneralThis chapter describes hydrologic engineering analysisstrategies, applications, and reporting for flood damagereduction studies. Hydrologic engineering analysis areperformed for planning investigations, refinements ofprevious study findings due to changed conditions in thedesign phases, and studies that provide information of apotential or impending flood hazard. The primary referencesfor the information of this chapter are: ER 1105-2-100, Guidance for Conducting Civil Works PlanningStudies, and ER 1110-2-1150, Engineering After FeasibilityStudies.3-2. Overview of Corps Flood Damage ReductionStudiesa. General. The Corps undertakes studies of waterand related land resources problems in response to directivesor authorizations from Congress. Congressionalauthorities are contained in public laws or in resolutions.Study authorizations are either unique specific studies orstanding program authorities usually called continuingauthorities. The focus of the studies are to determinewhether a Federal project responding to the problems andopportunities of concern should be recommended withinthe general bounds of Congressional interest. The Corpsstudies for planning, engineering and designing flooddamage reduction projects are predicated upon these legislativerequirements and institutional polices.b. Planning studies. Planning studies are termedfeasibility studies. Most studies are conducted in twophases.(1) The first, or reconnaissance-phase study, is fullyfunded by the Federal Government, normally takes12 months, and determines if there is a Federal interestand non-Federal support.(2) The second, or feasibility-phase study, takes up to3 years to complete, is cost-shared equally between theFederal Government and non-Federal sponsor, and resultsin recommendations to Congress for or against Federalparticipation in solutions to the problems identified in thestudy. The recommendation for Federal participation isgenerally for construction authorization.c. Preconstruction engineering and design (PED)studies. PED is a continuation of planning efforts followingthe feasibility study. This phase of the projectdevelopment encompasses all planning and engineeringnecessary for construction. These studies review previousstudy data, obtain current data, evaluate any changedconditions, and establish the plan for accomplishing theproject and design of the primary features. The preparationof general design memorandums, design memorandums,and plans and specifications are cost-shared asrequired for project construction.d. Engineering and design. Once the preconstructionengineering and design is completed, remaining engineeringand design will continue when the project isfunded for construction or land acquisition. This phaseincludes all remaining feature design memorandums,plans, and specifications needed to construct the project.e. Continuing authorities studies. These studies arestanding study and construction authorities conducted inthe same two-phase process as feasibility studies authorizedby Congress. Section 205 for small flood controlprojects and Section 208 for snagging and clearing forflood control (USACE 1989) with limits of $5,000,000and $500,000, respectively, are continuing authoritiesspecific for flood damage reduction.f. Federal role in flood damage reduction. TheCorps represents the Federal perspective in flood damagereduction actions. Studies are performed in response tocongressional directives. Problems are identified, solutionsproposed and evaluated, and recommendations madeto Congress. The principal Federal interest for flooddamage reduction studies is in furthering the economicdevelopment of the nation. Provided the solution is economicallyfeasible, protection of damageable propertyfrom floods is in the Federal interest (USACE 1989).3-3. Planning and Managing the HydrologicInvestigationa. General. The hydrologic engineering study mustbe planned and detailed to allow the effective and efficientmanagement of the technical work. Before anyhydrologic modeling or analytical calculations are undertaken,considerable planning effort should be performed.b. Scope of study. The scope of the study should beresolved early through meetings with the entire interdisciplinarystudy team and the local sponsor. The time andcost required are a direct function of the study scope and3-1

EM 1110-2-141731 Aug 94amount of detail required to fully evaluate the range ofproblems and potential solutions for the water resourcesproblem(s). The hydrologic engineer should formalizethese scoping meetings and any ideas on addressing theproblems through preparation of hydrologic engineeringwork plans which are presented and upgraded through thevarious phases of the study process. The work plansshould be reviewed by the technical supervisor and shouldbe furnished to the study manager. Unusual problems orsolutions would make it wise to receive division reviewalso. Work plans are especially important to developafter the reconnaissance report has identified the problemsfor further analysis in (and prior to initiating) the feasibilityreport.c. Study team coordination. Every cost-sharedfeasibility study has an interdisciplinary planning team(IPT) assigned, headed by a study manager. The teamconsists of working-level members from economics,hydraulics, geotechnical, design, real estate, environmental,cost estimating, etc. The local sponsor is also a member,although the sponsor may not wish to attend all IPTmeetings. Depending on the level of study activity andcomplexity, frequent meetings of the IPT should be heldranging from once a week to once a month. The advantageof frequent meetings lies in frequent communicationand the exchange of ideas between team members. Themost successful studies are those having free and easycommunication among team members.d. Quality control and review. The assurance ofquality work and an adequate review come from both thetechnical supervisor and the IPT. The development of aHEMP and the supervisor’s concurrence in the methodsand procedures for study analysis give the hydrologicengineer a “road map” for the entire study. Frequentupdates and consultations between the engineer and thetechnical supervisor are important. With these steps followed,technical quality should be acceptable for the finalreport. Similarly, scoping of the problems and necessaryhydrologic information supplied to other IPT memberswill be accomplished through IPT meetings and discussions.Unusual technical problems or policy issues mayrequire the review of higher level authority.e. Relationship with cost-share partner. The costsharepartner is a full member of the IPT and often providesvaluable technical assistance in many areas of thestudy. The partner also has valuable insights on the studyarea and its problems which may not be apparent to thestudy team. The cost-share partner should have as much(or as little) input and access to the planning and technicalanalysis as he/she wants. All hydrologic engineeringnegotiations with the cost-share partner must involve thehydrologic engineer. Sponsor participation in the studyprocess should be continuous. Study layout and scoping,IPT meetings and decisions, alternative evaluation andproject selection, and report recommendations and reviewshould all involve the local cost-share partner.3-4. Hydrologic Engineering Analysis Strategya. Overview. Three interrelated activities proposedas a study strategy are establishing a field presence in thestudy area, performing preliminary analyses, and conductingfull-scoped technical analyses using traditional toolsand methods tailored to the detail defined by the studytype and conditions.b. Field presence. The hydrologic engineer mustspend time in the field throughout all phases of the analysis,from the reconnaissance-phase study through theactual construction. A field presence is required to gatherdata needed for the study and to maintain continuouscontact with local interests involved with the proposedproject. Credibility is quickly lost when the engineersinvolved in the project recommendations have spent littleor no time in the study area. The hydrologic engineer’sfield presence is needed to establish and maintain contactsof local counterparts and determine survey needs, historicevent data, channel and floodplain conveyance characteristics,and operation procedures of existing facilities. Fieldvisits should often include other members of the studyteam and the local sponsor.c. Preliminary analysis techniques. These techniquesrepresent a suitable strategy to scope the complexityof the overall study, identify problems and tentativesolutions, and roughly determine the extent of Federalinterest in continuing the project. A preliminary analysiscould involve all of the following techniques:(1) Simplified techniques--often the application ofan equation for a peak discharge for a specific frequency,like the USGS regional regression equations. A roughestimate for a design discharge could be used to estimatethe required dimensions of a channel modification forcosting purposes. Simplified Techniques are discussed inChapter 11.(2) Field evaluations--experienced hydrologic engineerscan often lay out typical flood reduction measuresduring a field visit, such as, estimating alignment andheight of a levee for protection of a cluster of flood-prone3-2

EM 1110-2-141731 Aug 94structures. Problems associated with certain flood-reductionalternatives can often be ascertained in a fieldinspection.(3) Results of previous studies--most urban areashave flood insurance studies identifying flood profiles forthe 10-, 2-, 1-, and 0.2-percent chance exceedance frequencyfloods. Although not in sufficient detail to rely onfor design studies, this information is often used to estimateexisting flooding and potential damage reductionvalues. Hydrologic studies by other Federal agencies, aswell as State, local, and private agencies are also of value.(4) Application of existing computer models--manystudy areas have been previously analyzed by the Corpsof Engineers or other agencies. An existing computermodel of some or all of the study area is often useful toidentify flood hazard levels and potential flood reductionmeasures.d. Detailed analysis techniques. Detailed studies area suitable approach for the feasibility-phase and designstudies of a project. Detailed analyses are also appropriateduring the reconnaissance-phase investigation,although the analyses may be more abbreviated andapproximated than for subsequent studies. Essentially allfeasibility-phase flood damage reduction studies requiredetailed analysis of precipitation-runoff, floodflow byfrequency and/or modeling, river hydraulics, and storagerouting. Each of these component studies may represent asignificant effort. Therefore, it is not unusual for a hydrologicengineer assigned to a feasibility study to require12 to 24 months of intensive, full-time effort to performthe analyses (USACE 1988).3-5. Hydrologic Requirements for PlanningStudiesa. Overview. The analysis scope and detail requiredto conduct a hydrologic study depends on the type ofstudy, complexity of the study area, problems identified,potential solutions, and availability of needed data andinformation. This is particularly true in the reconnaissance-phaseinvestigation, after which the scope and detailbecomes more focused. A description of the studyrequirements and associated hydrologic analyses methodstypically needed for reconnaissance and feasibility studiesfollows. The methods are variable and should be scopedto specific study needs.b. Reconnaissance-phase study. The reconnaissancephasestudy develops and documents the information for adecision to proceed with feasibility-phase investigations.It also forms the basis for negotiating the feasibility studycost-sharing agreement (FSCA). Reconnaissance-phasestudies are conducted over 12 months or for special cases18 months. Table 3-1 lists the technical elements forconducting the hydrologic engineering analysis of a reconnaissance-phaseflood damage reduction study. Theobjectives are to define the flood problem, determinewhether further study will likely result in a feasible solutionto the flood problem, determine if there is Federalinterest, identify a local cost-sharing sponsor; and, if thefindings are positive, determine the scope and define thetasks for completing the feasibility investigation. Thehydrologic engineer is a key participant in objectives1 and 2 and must formulate in detail the HEMP as part ofthe Initial Project Management Plan (IPMP) for the feasibility-phasestudy (objective 5). Appendix B provides ageneric example of the HEMP for a typical flood damagereduction study. The HEMP should be modified in scopeto meet specific study requirements.(1) Ideally, it is desirable in the reconnaissancephaseto develop the complete hydrologic engineeringanalysis for the existing without-project conditions in thedetail needed for the feasibility-phase study. The reasonfor this detail is that the project feasibility is highly sensitiveto the hydrologic engineering and economic analyses.This concept is possible in some situations. However, inother situations the lack of available data, the complexityof the study area, and limited time may dictate that a lessdetailed analysis be performed.(2) A range of alternatives are formulated thatwould be reasonable to implement and that representdifferent kinds of solutions to the specified problems.The alternatives are analyzed in sufficient detail forapproximate benefit/cost analyses, to eliminate obviouslyinferior alternatives from future consideration, and toprovide for accurately developing the strategy, resourcesand cost of the feasibility study. The benefit and thushydrologic engineering analysis is normally based only onexisting, without-project conditions previously described.The existing with-project conditions are evaluated to thedetail required to determine whether a feasible plan withFederal interest will likely result from further study.Future conditions analyses are normally not required forthe reconnaissance-phase study.c. Feasibility-phase study.(1) The objective of flood damage reduction feasibility-phasestudies is to investigate and recommend solutionsto flood related problems. The feasibility-phase is3-3

EM 1110-2-141731 Aug 94Table 3-1Reconnaissance-Phase Study Technical Elements of Work Plan for Hydrologic Engineering Analysis (USACE 1988)I. Hydrologic engineering study objectivesII.Definition of study area for hydrologic engineering analysisIII. Description of available informationA. Maps, correspondence, documents, and reportsB. Observed flood informationC. Previous study data and analysis resultsIV. Definition of existing conditions flood hazardA. Historic floods documentationB. Hypothetical floods developmentC. Existing without-project conditions flow frequency, water surface profiles, etc.D. Appraisal of special technical issues: such as erosion/sedimentation, unsteady flow, water quality, future development etc.V. Existing with-project conditionsA. Appraisal of broad range of flood loss reduction measures.B. Existing with-project conditions flow frequency, water surface profiles.C. Documentation of flood hazard reduction performance for selected measures.VI. Initial project management plan for feasibility-phase study (HEMP, time, cost, schedule)cost shared 50\50 with a non-Federal sponsor. Typicalstudies are completed in 18 to 36 months. The majorityof hydrologic engineering work is performed in thisphase. The hydrologic engineering analysis must becomplete so that the project recommended in the feasibilityreport is essentially what is constructed after detailedengineering and design are completed.(2) Once the without-project conditions are detailed,the formulation process is iterative, increasing in detailand specificity as the viable measures and plans becomemore defined. The later stages of the feasibility studytherefore show an increase in the engineering and designeffort. Sufficient engineering and design are performed toenable further refinement of the project features, baselinecost estimates, and design and construction schedules.The engineering and design also allow design of theselected plan to begin immediately following receipt ofthe PED funds and the project to proceed through PEDwithout the need for reformulation, General Design Manuals,or postauthorization changes.(3) Working closely with the study manager, economist,cost engineer, and other members of the IPT, thehydrologic engineer completes the with- and withoutprojectevaluations so that an economically feasible planis recommended at the completion of the feasibility phase.This end result requires a continuous exchange of technicalinformation among the various disciplines. The planningprocess within which the hydrologic engineerfunctions consists of six major tasks: specification ofproblems and opportunities, inventory and forecast, alternativeplans, evaluation of effects, comparison of alternativeplans, and plan selection.(a) Specification of problems and opportunities.This initial step establishes the base conditions for theplanning process, defines the potential type and range ofsolutions, and provides the essential insight necessary toperform the remaining steps. The major components aredefinition of flood problem and specification of opportunities.The definition of flood problem component definesthe problems and opportunities for solutions to thoseproblems. The information provides the basis for subsequentproject development. The nature of flooding, locationof threatened properties, and existing project physicaland operational characteristics are determined. Informationis assembled from the reconnaissance-phase study,field reconnaissances, and other information. Hydrologicengineering investigations develop the specific characteristicsof flooding potential in the study area (flood flowsand frequency, flood elevations, and floodplain boundaries),character and variability of flooding (shallow ordeep, swift, debris-laden, etc.). The specification of3-4

EM 1110-2-141731 Aug 94opportunities component defines the general nature ofsolutions that might be appropriate. The general geographyof the watershed, location and density of development,and nature of the flood hazard all interact to revealpossible solutions. Solutions involving reservoirs, levees,and bypasses must be physically possible, reasonable, andnot in obvious conflict with critical community values andenvironmental resources. The community is also a valuablesource of ideas early on and throughout the investigation.It is important at this stage to be comprehensivein the exploration of possible solutions, yet equally importantfor practicality is best use of study time andresources. The hydrologic engineer’s practical experienceon what does and does not work is most helpful in thisphase.(b) Inventory and forecast. This step developsdetailed information about the existing and most-likelyfuture conditions within the watershed and study area.Existing conditions for the study area consist of measuresand conditions presently in place. Base condition refersto the first year that the proposed project is operational.Hydrologic engineering analyses are performed for existingand future without-project conditions. Existing measures,implemented prior to the base year, and measuresauthorized and funded for construction completion prior tothe base year are assumed to be in place and included forboth the with and without conditions. Future withoutconditionanalyses are conducted for the most likelyfuture development condition projected to occur withoutthe project. This includes changes in land use and conveyances.The assessments are performed for specifictime periods. Determination of without-plan conditions isan important aspect of the study process. It is the basisfrom which the alternatives are formulated and evaluated.Assessments of the without-project conditions should beof sufficient detail to establish viable economic (cost andflood damage), social, and environmental impact assessmentsof the with-project conditions without future refinementsthroughout the remaining planning and design studyprocess. Hydrologic analyses include the assembly ofdata for estimating the flood characteristics, developingdischarge-frequency relationships at desired locations, anddefining the performance of the without-project conditions.Specific tasks include the following.• Final data assembly. Most or all of these tasksmay have been conducted previously. These data shouldrepresent the final information used for feasibility anddesign studies.- Obtaining survey and mapping information. Mapsshowing land use, soil types, vegetation, stormsewer layouts, bridge plans, and other informationfrom local agencies.- Precipitation data from the National Weather Serviceor other agencies.- Stream gauge stage, discharge, and sedimentinformation from the U.S. Geological Survey orother agencies. Document historic event highwatermarks and flood characteristics.• Hydrologic analysis. This study aspect developsinformation used in the modeling of the study area andperforms the technical analysis.- Final delineation of watershed and subbasin boundariesbased on stream topology, gauge locations,high-water marks, damage reach flood damageanalysis requirements, and location of existing andpotential flood damage reduction measures.- Develop basic information for hydrologic model(i.e., subbasin areas, rainfall-runoff variables, baseflow, recession, and routing criteria).- Optimize runoff and loss rate variables usinghistoric event data.- Calibrate model to historic event high-watermarks and gauged discharge-frequencyrelationships.- Estimate existing without-conditions dischargefrequencyrelationships at desired ungauged locationsusing hydrologically and meteorologicallysimilar gaged basins data, regression analysis, andinitial hydrologic model results.- Determine best estimate discharge-frequencyrelationships at ungauged locations and, if necessary,adjust initial model variables to calibrate tofrequency relationships.- Adjust the model runoff and routing variables formost likely future without-project conditions forspecific time periods and determine dischargefrequencyrelationships at desired locations.- Provide discharge (or storage)-frequency relationshipsand other information (risk, performance ofthe system for a range of events, warning times,etc.) to economists, cost estimators, environmentalist,study manager, and project manager. The3-5

EM 1110-2-141731 Aug 94information should also be reviewed by the localsponsor counterparts.(c) Alternative plans. Alternative plans are formulatedto address the flood problems and accomplish otherplanning objectives. The alternatives are formulated toachieve the national goal of economic development consistentwith preservation and enhancement of cultural andenvironmental values. One or more measures and one ormore plans should be formulated to enable the full rangeof reasonable solutions to emerge from the investigation.In general, the array of alternatives developed should becomprehensive and not simply a range of sizes of a particularmeasure. The plan formulation exercise is a teamprocess. The hydrologic engineer’s knowledge and experienceis invaluable to this task and critical to the ultimateformulation of meaningful projects. There are numerousfactors to consider when formulating measures and plans.The study authorization should be reviewed as it mayrequire or limit certain actions. The without-conditionsanalysis defines primary damage centers and flood hazardsituations that may tend towards specific types of measures.Real estate and obviously high costs may prohibitcertain measures. Environmental and cultural featuresmay require or negate certain actions. The local sponsormay bring specific insights as to problems and potentialsolutions. In summary, the measures and plansformulated should emphasize comprehensive solutions andalso address specific, clearly localized problems.(d) Evaluation of effects. This step develops theinformation needed to determine and display the accomplishmentsand negative effects of measures and plans ascompared to the without-project condition. The evaluationprocess is conducted across the full perspective ofconcerns - hydrologic engineering, economic, environmental,and others. Hydrologic analysis of flood damagereduction measures and actions are performed for severalcombinations of measures and plans, operation plans, andperformance targets. The initial evaluation should assessthe potential for improved operation of the existing systemif such components are in place. If improved operationprocedures are found viable, they should be detailedand incorporated as part of the existing without-projectconditions. The hydrologic analysis procedures for existingand future with-project conditions are similar to thewithout conditions. The measure effects are incorporatedor determined by the modeling process. Frequency andproject performance information at all important locationsare defined by the without-project condition analysis. Theanalysis includes the full range of hydrologic eventsincluding those that exceed the design levels.(e) Comparison of alternative plans. This step isidentified separately to ensure that the measures are comparedon a consistent basis. Direct application of hydrologicanalysis criteria may include project performanceand safety information (design flows, risk, warning times,consequences of design exceedance, etc.), safety, andoperation considerations. Indirectly, hydrologic analysisinformation is used to assist in determination of flooddamage, stream profiles, fluvial hydraulics, environmentaleffects, and cost aspects. Therefore, the hydrologic engineeris an active participant in the comparison of alternativeplans for flood damage reduction.(f) Plan selection. Plan selection takes place in adiffused decision process. The study manager, technicalstaff, including the hydrologic engineer, and the localsponsor may strongly influence the recommended plan.The selecting officer at the field level is the district engineer.The division and Board of Engineers for Rivers andHarbors perform subsequent independent review and mayrecommend a different plan, but in most circumstancesthe district’s plan is ultimately implemented. Plan selectionat the district field office level must consider existinglaws and regulations applicable to the Corps and otheragencies. The recommended plan must be the plan thatmeets all the statutory tests and maximizes the economiccontribution to the nation. It is at this stage that thehydrologic engineer must demonstrate that the recommendedplan can perform its intended flood damagereduction function safely and reliably over the full rangeof hydrologic events.3-6. Preconstruction Engineering and Design(PED) Phasea. The PED phase begins after the division engineerissues the public notice for the feasibility report and PEDfunds are allocated to the district. Emphasis in this phaseis typically on the hydraulic design aspects, since thehydrologic analyses should have been completed in thefeasibility-phase study. If, however, it is determinedduring the PED phase that a general design memorandum(GDM) will be necessary because the project has changedsubstantially or for other reasons, part or all of the hydrologicanalyses may need redoing. The hydrologic engineeringanalysis would be conducted as a feasibility-phasestudy and reported and documented as such in a GDM.b. The hydrologic engineer is more involved in thedetailed design of the project components, with the overallcomponent capacities, general design, etc., held relativelyconstant from the feasibility report. For instance, the3-6

EM 1110-2-141731 Aug 94feasibility report may have recommended 5 miles ofchannel modifications having specified channel dimensions.The design memorandum would refine thesedimensions to fit the channel through existing buildingand bridge constraints; to perform detailed hydraulicdesign of tributary junctions, bridge transitions, dropstructures, and channel protection; and conduct detailedsediment transport studies to identify operation and maintenancerequirements and other hydraulic design aspects.If necessary, physical model testing is also performedduring the design memorandum phase. No additional planformulation, economics, etc., should be required. Structuraldesign, geotechnical analysis, cost engineering, andother disciplines work to finalize their analyses with theadditional topographic site surveys and subsurface informationnormally obtained in this phase. The hydraulicdesign is often being continuously modified to reflectthese ongoing design problems prior to completion ofdetailed design.3-7. Construction and OperationUnforeseen problems during construction frequentlyinvolve further modification and adaptation of the hydraulicdesign for on-site conditions. Similarly, most projectsrequire detailed operation and maintenance manuals, andhydrologic engineering information can be a critical partof these manuals. The operation of reservoirs, pumpingstations, and other flood mitigation components canrequire considerable hydrologic operation studies to determinethe most appropriate operating procedures. Postconstructionstudies are necessary for most projects. Most ofthese studies monitor sediment deposition and scourcaused by the project to ensure that adequate hydrologicdesign capacity is maintained to monitor the correctnessof the data used in analyzing the project and to estimatethe remaining useful life of the project.3-8. Reporting Requirementsa. General. Reporting requirements for the varioustypes of studies are described in applicable ER’s. Inaddition, hydrologic and hydraulic Engineer TechnicalLetters (ETL’s) summarize the array of hydrologic datathat must be presented for planning reports and suggestdisplay formats. The goal of reporting (investigationfindings) should be to describe in basic terms the natureof the flood problem, status and configuration of theexisting system, the proposed system and alternatives,performance characteristics of the proposed system, andimportant operation plans. This section presents a generalstructure for reporting results of the hydrologic studiescommensurate with the basic concepts of feasibility-phasestudies. Note that it is sometimes suggested that economicand other data be included so that the consequencesof the hydrologic evaluations may be betterjudged. Hydrologic reporting requirements should includea description of the without conditions, an analysis ofalternative flood loss reduction plans, analytical proceduresand assumptions used, and system implementationand operation factors influencing the hydrologic aspects ofthe study.b. Existing system. The existing system should bedefined and displayed schematically and by the use ofmaps, tables, and plates. The layout of the location ofexisting flood damage reduction measures should be indicatedon aerial photographs or other suitable cartographicmaterials. Important environmental aspects, damagelocations, and cultural features should also be indicated.c. Without-project conditions.(1) Physical characteristics and features of existingcondition flood-loss mitigation measures will be describedand shown in tables and plates. Dimensions of gravityoutlets, channels, and other measures shall be specified.Area capacity (storage-area-elevation) data of detentionstorage areas will be presented. Watershed and subbasinboundaries will be shown on a plate or map.(2) The hydrologic analysis approach adopted, criticalassumptions, and other analysis items for existingconditions will be described and illustrated as necessary.Historic and/or hypothetical storms, loss-rate parameters,runoff-transform parameters, routing criteria, and seepagewill be described and depicted via tables and plates.Hydrologic flow characteristics, peak discharge, duration,frequency, and velocity information will be presented forimportant locations (damage centers, high hazard areas,locations of potential physical works). Schematic flowdiagrams indicating peak discharges for a range of eventswill be included for urban areas. Presentation of severalhydrographs of major hydrologic events, including precipitationand loss rates and runoff transforms, can greatlyassist in explaining the nature of flooding.(3) Future without-project conditions will bedescribed as they impact on hydrologic conditions,assumptions, and procedures. Changes in runoff andoperation resulting from future conditions will bedescribed in terms similar to the existing conditionsdescription. Procedures adopted for parameter estimationfor future conditions should be described.3-7

EM 1110-2-141731 Aug 94d. Hydrologic analysis of alternatives.(1) The location, dimensions, and operation criteria ofcomponents of the alternative plans will be described anddepicted on tables and plates. Locations of the alternativemeasures or plans will be displayed on aerial photographsand/or other cartographic materials so that comparisonswith existing conditions may be readily made. Impacts ofmeasures and plans on flood hydrographs (peaks, durations,velocities) for a range of events will be provided atsimilar locations, as for without conditions. Display ofthe effects on hydrographs should be included. Display ofresidual flooding from large (1-percent chance and standardproject flood) events is required.(2) The hydrologic description of the various alternativeplans will include a description of the required localagreements and maintenance requirements. The hydrologicconsequences of failure to adequately fulfill theserequirements will also be presented.(3) Also presented are the basis and results of hydrologicand hydraulic studies required to determine thefunctional design and real estate requirements of all watercontrol projects.(4) The residual flood condition with the selectedplan in place will be described. As a minimum, the informationwill include the following: warning time ofimpending inundation; rate-of-rise, duration, depth andvelocity of inundation; delineation of the best availablemapping of the flood inundation boundaries; identificationof potential loss of public service; access problems; andpotential damages. This information will be developedfor each area of residual flooding for historic, standardproject flood, 1-percent chance flood and the flood eventrepresenting the selected level of protection. Thisinformation will be incorporated into the operation andmaintenance manual for the project and disseminated tothe public (ER 1110-2-1150, EM 1110-2-1413,ER 1105-2-100).3-9. Summarya. The Corps of Engineers utilizes feasibility planning,requiring the local partner to participate financiallyin the study process. These Corps of Engineers fiscalrequirements of the partner must also allow more partnerparticipation in the study selection process. Further localsponsor understanding of the hydrologic engineering analysisrequirements, from the feasibility study through thedetailed design, should allow for a better final product.b. The hydrologic engineering study must be plannedin enough detail to enable effective and efficient managementof the technical analysis. Detailed scoping of thestudy will enable the study manager to identify andaddress any potential problems early. The cost-sharedpartner should be considered a full member of the team.All hydrologic engineering negotiations with the costsharedpartner must involve the hydrologic engineer.3-8

PART IIHYDROLOGIC ANALYSIS

EM 1110-2-141731 Aug 94Chapter 4Rainfall Analysis4-1. Generala. The use of rainfall data is essential and fundamentalto the rainfall-runoff process. The rainfall data arethe driving force in the relationship. The accuracy of therainfall data at a point (i.e., at the rain gauge) isextremely significant to all the remaining use of the data.b. This chapter describes the significance of rainfalldata to the rainfall-runoff process. The relationshipbetween point rainfall at a rain gauge and the temporaland spatial distribution of rainfall over the watershed ofinterest is discussed. Limitations and inaccuracies inherentin these processes are also defined.4-2. Point Rainfall Dataa. Rainfall measured at a rain gauge is called pointrainfall. The rain is captured in a container. The standardrain gauge, shown in Figure 4-1, is an 8-in.-diammetal can. A smaller metal tube may be located in thislarger overflow can. An 8-in.-diam receiver cap may beon top of the overflow can and is used to funnel the raininto the smaller tube until it overflows. The receiver caphas a knife edge to catch rain falling precisely in thesurface area of an 8-in.-diam opening.b. Measurements are made using a special measuringstick with graduations devised to account for the 8-in.receiver cap opening, funneling water into the smallertube. When the volume of the smaller tube is exceeded,the volume from the smaller tube is dumped into thelarger overflow can.c. Other types of rain gauges are also available. Incontrast to the nonrecording gauge which requires anobserver to manually measure the rain at regular intervals(i.e. every 24 hours), Figure 4-2 shows a weighing-typerecording gauge which does not require constant observation.The rain is caught in a standard 8-in. opening butstored in a large bucket that sits on a scale. The weightof the water caught during a short time interval isrecorded on a chart graduated to units of linear distance(inches or millimeters) versus time.d. Other variations of these two gauges exist andperform similarly. Although essentially all United Statesgauges have exactly an 8-in. opening and have beencarefully calibrated for exact measurement with an appropriatelygraduated stick or chart, several other conditionsaffect the exact amount of rain caught in the gauge.e. The gauges are affected by wind, exposure, andheight of gauge. Researchers have tried to establish correctioncharts for windspeed effect on the catch, but sinceexposure (including gauge height) has such significantimpacts on the catch, these charts must be viewed withsuspicion. The effect of height has been standardized inthe United States at 31 in. Windshields, Figure 4-2, havebeen used at some locations to minimize the inaccuracy ofmeasurement due to windspeed.f. Other errors are associated with the volume ofwater displaced by the measuring stick (a constant of2 percent) or the inherent errors associated with themechanical aspects of some other types of gauges (i.e.,tipping bucket), which are variable as a function of rainintensity. Variable error associated with mechanicalgauges should be evaluated by comparing recorder dataagainst standard gauge data and correction relationshipsdetermined for future use.4-3. Rainfall Data From Remote Sensorsa. Rain gauges measure the amount of rain that hasfallen at a specific point. However, hydrologists andhydrologic models typically need the amount of rain thathas fallen over an area, which may be different than whatwas measured at a few points. A better estimate of rainfallmay be achieved by installing more rain gauges (adense gauge network), but such a network is very expensive.Alternatively, weather radar, when adjusted withrain gauge data, may provide a relatively accurate measurementof the spatial distribution of rainfall. If the areais in a remote region, where there are few or no raingauges and weather radar is not available, environmentalsatellite data may provide rough estimates of rainfallamounts.b. Radar (Radio Detecting And Ranging) operates onthe principle that an electromagnetic wave will be partiallyreflected by objects or particles encountered by thewave. Generally, a radar system consists of a transmitter,which generates electromagnetic pulses; a movable dishshapedantenna, which serves both to transmit the electromagneticpulses and receive reflected signals; a receiverthat detects and amplifies the reflected signals; and adevice to process and display these signals. The radarantenna transmits electromagnetic pulses into the atmosphereslightly above horizontal. These pulses travel at the4-1

EM 1110-2-141731 Aug 94Figure 4-1. Nonrecording gauge, 8-in. opening (U.S. Weather Bureau standard rain gauge)4-2

EM 1110-2-141731 Aug 94Figure 4-2. Weighing type recording rain gauge (from U.S. Weather Bureau source)4-3

EM 1110-2-141731 Aug 94speed of light. As the pulses encounter raindrops (orother objects), the signal is partially reflected towards theantenna. The power and timing of the received signal (orecho), relative to the transmitted signal, are related to theintensity and location of rainfall.c. Weather radars generally employ electromagneticpulses with a fixed wavelength of between 3 and 20 cm.A radar with a shorter wavelength is capable of detectingfine rain particles, but the signals will be absorbed orattenuated when they encounter larger storms. A longerwavelength radar will have little signal attenuation, but itcannot detect low-intensity rain.d. Doppler radars can detect a “phase shift” (aslightly different frequency of the pulse than when transmitted)of a returned pulse. The velocity of theatmospheric particles which reflected the pulse can becalculated from this phase shift. This information is veryimportant in detecting and predicting severe storm phenomenasuch as tornados but is not generally useful incomputing rainfall intensity.e. The rainfall rate “R,” can usually be computedfrom the reflectivity “Z,” which is related to the amountof power in the returned pulse, using the formula:whereZ = 200 * R 1.6Z = reflectivity, measured in units of mm 6 /m 3R = rainfall rate, given in mm/hrThe constant (200) and the exponent (1.6) vary dependingon the size and type of precipitation encountered. If hailor snow are encountered by the pulse, the reflectivity willbe much higher than that for rain.f. There are several factors which can cause erroneousrainfall rates to be computed from radar data. Themore prevalent problems are:(1) Anomalous propagation, where atmospheric conditionscause the radar beam to bend toward the earth.The beam may be reflected by the ground or objects nearthe ground, producing false echoes and indicating rainfall(usually heavy) where there are none. Anomalous propagationcan be screened by using cloud cover informationfrom satellites or from a knowledge of the atmosphericconditions in the area.(2) Incorrect parameters in the reflectivity-rainfallrate formula (or “Z-R relation”). The parameters givenhave been determined for “typical” rainfall drop sizedistributions, and may vary considerably, depending onthe storm. Also, if the beam encounters other types ofprecipitation, such as snow or hail, these parameterswould greatly overestimate the rainfall amount if notmodified to match the precipitation type.(3) Attenuation is the reduction in power of theradar pulse as it travels from the antenna to the target andback and is caused by the absorption and the scattering ofpower from the beam. Attenuation from precipitationusually appears as a “V” shaped indentation on the farside of a heavy cell and causes the rainfall to beunderestimated in this region.(4) Evaporation and air currents that cause the rainfallrate in the atmosphere, measured by the radar aredifferent than the rate at ground level. Evaporation is themost prominent at the leading edge of a storm, when theair mass near the surface is relatively dry.(5) Hills and buildings near the radar site can reflectthe beam and cause ground clutter. This clutter may alsoreduce the effectiveness of the radar for areas beyondthese objects. Typically, a weather radar is ineffectivewithin a 15- to 20-mile radius.g. The effect of these factors is that rainfall amountscomputed for an area with radar data will typically beinaccurate. However, rain gauge data can be combinedwith the radar data to estimate rainfall amounts that aresuperior to either radar or rain gauge data alone. Itshould be noted that a correct method must be appliedwhen combining the two data sets, or the combined setmay be more erroneous than either set alone.h. In a joint effort of the Department of Commerce,the Department of Defense, and the Department of Transportation,NEXRAD (Next Generation Weather Radar)was developed. The NEXRAD system will incorporateapproximately 175 10-cm Doppler radars across theUnited States. NEXRAD will provide many meteorologicalproducts, including several precipitation products.One of the main graphical products is a 1- or 3-houraccumulation of rainfall, displayed on a 2- by 2-km gridto a range of 230 km from the radar site. An importanthydrological product is the digital array of hourly accumulations.This product gives rain gauge adjusted rainfallamounts for a 4- by 4-km grid for the area covered by asingle NEXRAD radar. Another product “mosaics” the4-4

EM 1110-2-141731 Aug 94digital products from different NEXRAD sites together, toproduce a single-digital rainfall array over a watershed.These digital products can be used as input to rainfall-runoffmodels for improved results in forecasting or in traditionalhydrologic studies.i. Environmental satellites, such as the GOES system,can provide rough estimates of precipitation over aregion. Such satellites cannot measure precipitationdirectly, but can measure spatial cloud cover and cloudtemperature. The approximate height of the top of cloudscan be calculated from the temperatures measured by thesatellite. The colder a cloud is, the higher the top of thecloud is. In general, clouds with higher tops will yieldmore precipitation than those with lower tops. If thecloud temperature satellite image is correlated with a raingauge on the ground, an approximate spatial distributionof the rainfall amounts in that area can be estimated.However, rain gauge data alone provide a more accuratemeasurement of rainfall over an area than that which isestimated with satellite and gauge data.j. Satellites can be useful in estimating rainfallamounts in regions where little or no rain gauge data areavailable, such as areas in Africa. In these regions, estimatesof rainfall may be calculated for hydrologic studies,such as sizing a dam, using satellite data (which mayhave many years of data recorded) when there are no raingauge data available.4-4. Areal and Temporal Distribution of RainfallDataa. Network density and accuracy. For the applicationof point rainfall data to a rainfall-runoff calculation, abasin average rainfall must first be determined.(1) This need raises the question about a proper densityof rain gauges (recording and/or nonrecording gaugesper square mile of drainage area.) No definite answerexists for this question. Adequate coverage is related tothe normal variation in rainfall for a specific region. Ifthunderstorms account for a major source of rainfall in thespecific area, an even denser network of rain gauges isneeded.(2) Average density in the United States is about onegauge for every 250 to 300 square miles. Studies haveshown that with this density, a standard error of about20 percent for a 1,000-square-mile basin is expected ifthunderstorms are the major source of precipitation. Asshown in Figure 4-3, four times the average density ofgauges is required to reduce the error of measurement by10 percent. These results are derived from data in theMuskingum River basin in Ohio. Mountainous terrainrequires a denser network for the same level of error, andplains require a less dense network. If the major sourceof rainfall is the frontal-type storm pattern, rainfall variationsare less than from thunderstorms and less densegauge networks will suffice.b. Areal distribution. Several methods are availableand routinely used to calculate basin average rainfall froman assumption of areal (i.e., spatial) distribution usingpoint rainfall from a gauge network. The most common,useful method is the Thiessen Polygon.(1) The Thiessen method weighs each gauge indirect proportion to the area it represents of the total basinwithout consideration of topography or other basin physicalcharacteristics. The area represented by each gauge isassumed to be that which is closer to it than to any othergauge. The area of influence of each gauge is obtainedby constructing polygons determined by drawing perpendicularbisectors to lines connecting the gauges as shownin Figure 4-4a.(2) The bisectors are the boundaries of the effectivearea for each gauge. The enclosed area is measured andconverted to percent of total basin area. The polygonweighted rainfall is the product of gauge rainfall and theassociated polygon area in percent. The sum of theseproducts is the basin average rainfall.(3) The Thiessen method is usually the best choicefor prairie states during thunderstorms, since elevationdifferences (topographic) are insignificant and gaugedensity is inadequate to use other methods to define theareal pattern of the thunderstorm cells. When analyzingseveral storm events having different gauges reporting foreach event, the Thiessen method becomes more timeconsumingthan other techniques to be discussed.(4) Another popular method is the Isohyetal method,which provides for consideration of topographic effectsand other subjective information about the meteorologicalpatterns in the region. A rainfall-depth contour map isdetermined by tabulating gauge rainfall on a map of theregion and constructing lines of equal rainfall calledisohyets as shown in Figure 4-4b. Average depths areobtained by measuring the areas between adjacentisohyets (zones). Each increment of area in percent oftotal basin area is multiplied by the estimated rainfalldepth for that area. This product for each zone issummed to obtain the basin average rainfall.4-5

EM 1110-2-141731 Aug 94Figure 4-3. Number of rain gauges required for 10 and 15 percent error (U.S. Department of Commerce 1947)4-6

EM 1110-2-141731 Aug 94Figure 4-4. Basin average rainfall analysis techniques(a) The Isohyetal method allows the use of judgmentand experience in drawing the contour map. Theaccuracy is largely dependent on the skill of the personperforming the analysis and the number of gauges. Ifsimple linear interpolation between stations is used fordrawing the contours, the results will be essentially thesame as those obtained by the Thiessen method.(b) The advantages of both the Thiessen andIsohyetal methods can be combined where the area closesto the gauge is defined by the polygons but the rainfallover that area is defined by the contours from the Isohyetalmethod. This combination also eliminates thedisadvantage of having to draw different polygon patternswhen analyzing several different storm events with avariety of reporting gauges. Regardless of the techniqueselected for analysis of basin average rainfall, a regionalmap of areal distribution for the total storm event is alsoproduced.4-7

EM 1110-2-141731 Aug 94c. Temporal distribution. Having already determinedbasin average rainfall, one or more recordinggauges in or near the watershed of interest must belocated and used as a pattern to estimate the temporal(i.e., time) distribution of the basin average rainfall.(1) If only one recording gauge is available, it mustbe assumed that the temporal distribution of the totalstorm rainfall at the recording gauge is proportional to thebasin average rainfall distribution. The calculations necessaryto perform this evaluation are shown in Figure 4-5.(2) If more than one recording gauge is available, aweighted average combination distribution can betabulated and used in the same manner as the distributionat a single gauge. Caution should be used when utilizingmore than one recording gauge to develop the temporaldistribution of a storm event. If the event is a short-duration,high-intensity storm and the timing of the center ofmass of the rainfall is different between the gauges, traditionalaveraging can often result in a storm of longerduration and much lower intensities than what wasrecorded at each of the gauges. If this is the case, it isoften better to use the recording gauge that is closest tothe center of mass of the subbasin as the temporal distribution,and only utilize the other gauges in estimating theaverage depth of rainfall over the subbasin.4-8

EM 1110-2-141731 Aug 94Figure 4-5. Time distribution of basin average rainfall4-9

EM 1110-2-141731 Aug 94Chapter 5Snow Analysis5-1. GeneralThe simulation of flood runoff may involve a key factorwhich affects the determination of precipitation excess;that is, precipitation may or may not fall in its liquid formand thus may not be immediately available for runoff.Furthermore, if snow has accumulated in the basin fromprevious storm events, then water input from this sourcemay be available for a given flood event if hydrometeorologicalconditions permit snowmelt to occur. This chapterwill describe the factors involved in the snow accumulationand ablation process and the techniques used to simulatethese factors for flood runoff analysis. Two distincttypes of floods are usually involved: rain-on-snow events,typical of the winter floods in the Cascade and SierraNevada mountains of the Western United States and theAppalachians in the East; and spring/summer floods -usually involving relatively little rain on the large riversof the interior states, such as the Columbia, Missouri, andColorado.5-2. Physical Processesa. Overview. Chapter 4 described the analysis ofrainfall, leading to the estimation of basin-wide waterexcess that is potentially available for runoff. A specialcase of this hydrometeorological process occurs when airtemperatures are cold enough to cause the precipitation tooccur in its solid form and remain temporarily stored onthe ground as snow. Once in place, a metamorphosis ofthe accumulated snow will eventually occur when heatenergy is supplied from various sources. With enoughheat energy, the snow will be transformed from a solid toliquid state and water will be available for runoff.b. Precipitation, snowfall, and snow accumulation.In the middle latitudes, precipitation usually occurs as aresult of the colloidal instability of a mixed water-icecloud at temperatures below 32 °F. The formation ofsnow and, subsequently, rain in the atmosphere is adynamic process. It has been observed that winter precipitationoccurs initially in the form of snow crystals insubfreezing portions of clouds. As the snowflakes fallthrough the atmosphere, they later melt into raindropswhen they fall through warmer, above-freezing air atlower elevations. The corresponding melting level airtemperature of snowflakes falling through the atmospherevaries from 32 to 39 °F, but it is usually about 34 to35 °F. Accordingly, on the earth’s surface, snowfalloccurs at elevations higher than the melting level, whilerainfall occurs at elevations lower than the melting level.The most significant determinant of the occurrence of rainor snow is the elevation of the melting level. This isparticularly important in mountainous regions. Factorswhich influence the amount and distribution of precipitationin the form of snow and the snowpack water equivalentmay be classified as being meteorologic andtopographic. Meteorologic factors include air temperature,wind, precipitable water, atmospheric circulationpatterns, frontal activity, lapse rate (vertical temperatureprofile), and stability of the air mass. Topographic factorsinclude elevation, slope, aspect, exposure, forest, andvertical curvature. The crystalline form of newly fallensnow is most commonly hexagonal.c. Snow metamorphosis. Freshly fallen snow existsin a clearly defined crystalline state, with sharply definededges and abrupt points in each snow crystal. Metamorphosisof the snow occurs over time as the individualcrystals lose their original distinct form and becomerounded and bound together, ultimately into uniform,coarse, large ice crystals. This process is commonlycalled “ripening.” This transformation may take place inas short a time period as several hours, but commonlyinvolves a period of days or weeks in intercontinentalareas with a large, deep snowpack.(1) The specific gravity of snow (a dimensionlessratio) is commonly called the snow density (which properlywould be mass per unit volume). The density (percentwater equivalent) of the newly fallen snow istypically on the order of 10 percent, with variations of6 to 30 percent dependent upon the meteorological conditionsinvolved, primarily air temperature and wind. Asmetamorphosis occurs, density increases, reaching valuesof 45 to 50 percent for a fully ripe snowpack. A snowpackripe for melt also contains a small amount of freewater, on the order of 3 to 5 percent. A ripe snowpack issaid to be “primed” to produce runoff; that is, when itcontains all the water it can hold against gravity.(2) The temperature of the snowpack varies as afactor in the metamorphosis process. In its early stages,the variation throughout the depth may be marked, fromapproximately 32 °F near the ground to subfreezing temperaturesat shallower depths. As the snow ripens, a moreisothermal pattern develops, and in its “ripe” condition thesnowpack is completely isothermal and near 32 °F. Theamount of heat required per unit area to raise the temperatureof the snowpack to 32 °F is termed the “cold content”of the snow. This is expressed in terms of liquidwater (produced at the surface by rain or melt) which,5-1

EM 1110-2-141731 Aug 94upon freezing within the snowpack, will warm the pack to32 °F.d. Snowmelt. The process of melting snow involvesthe transformation of snow/ice from its solid form toliquid water through the application of heat energy fromoutside sources. While the latent heat of ice is establishedat 80 cal/g, this factor usually must be adjusted to actualsnow conditions since the snowpack is not in the form ofpure ice at 0 °C. The ratio of heat necessary to producewater from snow (and associated free water) to theamount required to melt the same quantity of ice at 32 °Fis termed the “thermal quality” of the snowpack. For afully ripe snowpack, the thermal quality can be on theorder of 0.95 to 0.97.(1) The rate of snowmelt is dependent upon the manydifferent processes of heat transfer to and from thesnow-pack, but it is also somewhat dependent upon thesnow-pack condition. The relative importance of theseprocesses varies widely seasonally, as well as with theday-to-day variation of meteorological factors. The heattransfer processes also vary significantly under variousconditions of forest environment, exposure, elevation, andother environmental factors.(2) The four major natural sources of heat in meltingsnow are absorbed solar radiation, net long-wave (terrestrial)radiation, convective heat transfer from the air, andlatent heat of vaporization by condensation from the air.Two additional minor sources of heat are conduction ofheat from the ground and heat content of rainwater.(3) Solar radiation is the prime source of all energy atthe earth’s surface. The amount of heat transferred to thesnowpack by solar radiation varies with latitude, aspect,season, time of day, atmospheric conditions, forest cover,and reflectivity of the snow surface (termed the “albedo”).The albedo ranges from 40 to 80 percent. Long-waveradiation is also an important process of energy exchangeto the snowpack. Snow is very nearly a perfect blackbody, with respect to long-wave radiation. Long-waveradiation exchange between the snow surface and theatmosphere is highly variable, depending upon conditionsof cloud cover, atmospheric water vapor, nighttime cooling,and forest cover. Heat exchange by convection andcondensation of heat and water vapor from or to the snowsurface and the atmosphere is dependent upon the atmosphericair temperature and vapor pressure gradients,together with the wind gradient in the atmosphere immediatelyabove the snow surface. These processes areparticularly important under storm conditions with warmair advection and high relative humidity. In summary,there is no one process of heat exchange to the snowpackthat may be universally applied, but the relative importanceof each of the processes is dependent on atmospheric,environmental, and geographic conditions for aparticular location and a particular time or season.5-3. Data Requirements, Collection, andProcessinga. Data requirements. Data required for snowanalysis and simulation include those required for rainonlysituations plus additional data necessary for the snowaccumulation/snow melt processes involved. Theseinclude air temperature data and snow measurements as aminimum but could include windspeed, dewpoint, andsolar radiation if energy budget computations are beingperformed.(1) Air temperature data are quite critical in anymodeling or analysis effort, since freezing level must beknown during the snow accumulation process to distinguishbetween precipitation type in the basin. Temperatureis also frequently (almost exclusively) used as anindex to determine snowmelt. An additional parameterneeded in modeling is the lapse rate, which must either bea fixed value or estimated from observed temperaturereadings. If calculated, temperature stations at differentelevations are necessary.(2) Snow data are collected in the form of snowwater equivalent, frequently on a daily basis in the case ofautomated stations using snow pillows, or monthly in thecase of manually read snow courses. Snow water equivalentdata as applied to flood-runoff analysis would beneeded as an independent variable for simplified analysesand seasonal runoff forecasting, and as data to assist incalibrating and verifying simulation models. Since snowstations may be the only source of high-elevation precipitation,they also can be used to help estimate basin-wideprecipitation input to simulation or statistical models.b. Data collection. The collection of precipitationdata in areas subject to snow accumulation presents additionalproblems in gauging, due to considerations of gaugefreezing, “capping” of the gauge by snow, and shieldingof the gauge. Equipment and field procedures for suchconditions are well documented (USACE 1956). Theselection of appropriate precipitation, snow, and temperaturegauges for analysis of a mountainous environmentsubject to snow conditions warrants careful considerationof vertical factors in addition to areal considerations usedin rain-only situations, since the vertical distribution ofprecipitation and the vertical temperature profile must be5-2

EM 1110-2-141731 Aug 94considered. Bearing on this consideration is the applicationinvolved; for simple indexing applications, forinstance, a high-elevation snow gauge may be very important.For detailed simulation, a gauge placed in midelevationsmay be more important for defining thedistribution of precipitation in the vertical direction andgiving a field reference of snow conditions during criticaltimes of snowmelt.c. Data processing. There are no significant additionalrequirements for processing snow-related data ascompared to nonsnow situations. Special treatment ofmonthly snow course data may be required if daily incrementsare to be estimated; this can be accomplishedthrough correlation with a nearby station. Temperature isusually expressed in terms of daily maximum/minimum,or hourly data may be used. In the case of the former,the maximum/minimum data can be expressed as twoseparate stations, and model preprocessors apply weightsto each as desired.5-4. Simulating Snow Accumulationa. Applications. Hydrologic engineering analysesinvolving snow typically require an estimate of snowwater equivalent for the basin being studied as input intothe runoff derivation. This estimate must directly orindirectly consider the process of snow accumulation anddistribution, which includes factors such as the effects ofgeography and elevation in the distribution of snow andthe accounting of the rain/snow threshold. The complexityof this determination can vary depending upon dataavailability and application, from simple estimates of asingle basin value, to detailed simulation using adistributed formulation of the basin. Table 5-1summarizes three possible approaches of varyingcomplexity.b. Watershed definition. Because temperature, andtherefore elevation, play such an important role in definingthe conditions of the basin during a precipitationevent, the watershed being simulated needs to be definedwith independent subunits. The most common approachis to divide the basin into zones or bands of equal elevation.On each band, precipitation, snow, soil moisture,etc. can be independently accounted for as illustrated inFigure 5-1. In a spatially distributed model, the configurationof computational nodes would likewise have toconsider these elevation effects. Available models suchas HEC-1 (USACE 1990a) and SSARR (USACE 1987)provide for the watershed definition to be establishedrelatively easily. Simplifying assumptions, such as definingzone characteristics through generalized functions forthe basin, are often employed. Such assumptions are notunreasonable since detailed information on subbasin definitionis not likely available.c. Simulation elements. Figure 5-2 illustrates theprocess that must be considered in simulating snow accumulation.For a given elevation zone or subbasin elementand a given time period, these steps include: (1) find basetemperature; (2) calculate lapse rate (fixed or variable);(3) calculate temperature at elevation of zone or subelement;(4) calculate zone precipitation; (5) get rain-freezetemperature; (6) calculate breakdown of rain versus snow;and (7) accumulate snow; recalculate snowline. There areno complex equations involved in this process, which islargely a detailed accounting process. The lapse rate isusually taken as a fixed input parameter (often 3.3 degper 1,000 ft of elevation), but may be a specified orTable 5-1Alternatives For Estimating Snow Water Equivalent (SWE)Approach Possible Application CommentSimple estimate of SWEDetailed estimate of SWE, consideringelevation distributionSimulation of snow accumulation throughthe accumulation season1. Single event rain-on-snowcomputation2. Forecasting in raindominatedareasDesign flood derivation, snow-dominatedbasin1. Detailed design floodderivation2. Forecasting water supplySimple estimate based upon historicalrecords. May be adequate where raindominatesMore detailed analysis of historical recordsRequires a continuous-simulation model5-3

EM 1110-2-141731 Aug 94Figure 5-1. Illustration of distributed formulation of a watershed model using elevation bands5-4

EM 1110-2-141731 Aug 94Figure 5-2. Illustration of snow accumulation simulation5-5

EM 1110-2-141731 Aug 94calculated variable. The rain-freeze temperature is likewiseusually a fixed value, usually around 34 °F.d. Alternatives to simulation of snow accumulation.Various alternatives exist to a detailed accounting of snowaccumulation, depending on the hydrologic regimeinvolved and the application desired. For analyzing discreterain-on-snow storm events, such as in design floodanalysis, a simple estimate of snow water quantity at the(beginning) of the storm may be sufficient, particularly ifthe snowmelt contribution is relatively small compared torain runoff. This may be based upon historical records ofsnow. In the Columbia basin, operational forecasting ofspring snowmelt runoff employs simplifying assumptionsof snow accumulation for most basins. In this case, theseasonal accumulation of snow is estimated through theuse of multiple regression models using winter precipitationand snow as independent variables. Errors in thisestimate are accounted for during the simulation of snowmeltby adjusting the model’s estimate of snow basedupon model performance and observed areal distributionof snow.5-5. Simulating Snowmelta. Overview of applications and approaches.Numerous alternatives present themselves in determiningthe best approach for simulating snowmelt in flood-runoffanalysis. These approaches range from simplifiedassumptions on discrete storm events to detailed simulationusing energy budget principles and distributeddefinition of the watershed. The choice of methods isdependent upon the application involved, resources available,and data availability. Table 5-2 summarizes theoptions that are possible and how they tend to relate togiven types of applications. A typical situation that mightbe encountered is that of calculating a hypothetical floodfrom specified rainfall, either of specified frequency orfrom a National Weather Station (NWS) hydrometeorologicalreport. If the meteorological conditions are suchthat rainfall dominates and the duration of the storm isrelatively short, it may be quite satisfactory to use asimple approach to estimating snowmelt (e.g., by establishingan antecedent water content by historical analysisthen using an assumed rate of melt or a temperature indexapplied with a melt rate factor). The simulation of snowconditioning would not be required, since the assumptionof a “ripe” snowpack prior to the storm could beassumed. On the other hand, the derivation of a designflood or the forecasting of flood runoff in a basin that ispredominately snow would likely require a more detailedsimulation of snow conditioning and snowmelt, perhapsthrough the use of theoretical or empirical equations asdescribed below.b. Simulation of energy input. As discussed in paragraph5-2d, the sources of heat energy that cause snowmeltinvolve several factors that can be difficult, if notimpossible, to quantify and measure. In actual practicethen, the theoretical relationships involved are reduced toempirically derived equations that have worked satisfactorilyin simulation models. Two basic approaches arecommonly used: the “energy budget” solution whichemploys simplified equations that represent key causalfactors such as solar radiation, wind, heat from condensationof water vapor, etc.; and the “temperature index”solution which uses air temperature as the primary independentvariable through the use of a fixed or variable“melt-rate factor.” The latter solution is almost exclusivelyused in practical applications of forecasting andanalysis.(1) Energy budget solution. Although variationsexist in the equations that have been developed to simulatesnowmelt, those developed in the 1950’s by theU.S. Army Corps of Engineers remain sound and serve toeasily illustrate the basic principles involved. These werebased on extensive field experiments coupled with theoreticalprinciples, as discussed in the summary report,“Snow Hydrology” (USACE 1956). The several equationsthat were derived are also presented in EM 1110-2-1406, Runoff From Snowmelt, and have been used inseveral applications. The equations presented with abbreviatedexplanation below are described in detail in both ofthese documents.(a) For snowmelt during rain, in which shortwavesolar radiation is relatively unimportant and condensationmelt is relatively high, the following equation (Eq 20,EM 1110-2-1406) applies:whereM (0.029 0.0084kv0.007P r)(T a32) 0.09M = total daily snowmelt, in inches(5-1)k = factor representing the relative exposure of thebasin to wind (for unforested areas, k =1)5-6

EM 1110-2-141731 Aug 94Table 5-2Snowmelt Options 1 Basin Configuration Melt CalculationApplication Example Lumped DistributedSnowTemperatureConditioning Simplified 2 IndexEnergyBudgetSingle-eventanalysis-Rain-on-snowDesign floods incoastal mountainsYes Possibly Assumed“ripe”Possibly Possibly PossiblySingle-eventanalysis-Snow(plus rain)Design floods ininteriorbasinsYes Yes Assumed“ripe”No Yes Yes 3Single-eventforecasting-Rain-on-snowSingle-eventforecasting-Snow (+ rain)Continuoussimulation, anyenvironmentMacro simulationinsmall watershedsShort-term floodforecastingShort-term floodforecastingLong-termflood/droughtforecasting;Detailed analysisfordesignR&D applications;analysis for detaileddesign; specialapplicationsYes Yes Optional Possibly 4 Yes NoYes Yes Optional No Yes NoNo Required Required No Yes PossiblyNo Required Required No No Yes1Qualitative indicator shown for type of option that might typically be used for application. This is a guideline only. “Yes” or “No” indicatessuggested option.2 Simplified approach might be to assume a constant or variable moisture input due to snowmelt.3 Has been used for probable maximum flood (PMF) calculations in the Columbia basin.4 Would be appropriate only in situations where snowmelt is small compared with rain.v = wind velocity at the 50-ft height, in miles perhourP r = daily rainfall, in inchesT a = mean temperature of the saturated air, in degreesFahrenheitThe constants in the equation are based on field investigations.The factor 0.029 relates snowmelt due to longwaveradiation to temperature, and the term 0.0084kv representsthe effects of convection-condensation melt. The factor0.09 accounts for melt from ground heat. If, for example,on a given day the average air temperature is 50 °F, rainfallis 3 in., and wind velocity is 20 mph in an unforestedenvironment, then the melt components would be:Solar radiation (long wave) - 0.5 in.Convection-condensation - 3.0 in.Rain- 0.4 in.Ground heat- 0.1 in.-----------------------------------------------------------------------Total- 4.0 in.5-7

EM 1110-2-141731 Aug 94This example illustrates the importance of the convectioncondensationmelt component, and the correspondingimportance of wind, in a rain-on-snow situation. Theimportance of rain itself in producing melt is relativelysmall.(b) For the case of snowmelt during rain-free periods,direct (short-wave) solar radiation must be accounted for.Several equations are developed in Snow Hydrology(USACE 1956) depending upon the degree of forest canopyinvolved. One, for partly forested areas (Eq 24,EM 1110-2-1406) is as follows:The energy budget equation requires considerably moredata than those previous so that its usage becomes limitedin practical applications. One possibility, however, is inPMF derivations where variables such as insolation,albedo, etc. can be maximized through analysis of historicaldata (USACE 1956) and EM 1110-2-1406. Both ofthe equations presented are available in the HEC-1(USACE 1990a) and SSARR (USACE 1987) computerprograms. The generalized snowmelt equations also providea useful method of estimating relative magnitudes ofmelt components. Table 5-3 presents melt quantitiescalculated from these equations for six hypothetical situations--threewith rain, three without.M k (1 F)(0.0040I i)(1 a) k(0.0084v)(0.22T a0.78T d) F(0.029T a)(5-2)(2) Temperature index solution. Because of thepractical difficulties of obtaining data needed for the energybudget equations, common practice is to simulatesnowmelt by the “temperature index” solution, utilizingthe basic equationwhereM = snowmelt, in inches per periodk′ = basin shortwave radiation melt factor. Itdepends on the average exposure of the openareas to shortwave radiation melt factor. Itdepends on the average exposure of the openareas to shortwave radiation in comparison withan unshielded horizontal surfacewhereM C (T aT b)M = snowmelt, in inches per periodC= melt rate coefficient that is oftenvariable (discussion follows)(5-3)I iak= observed or estimated insolation (solar radiationon horizontal surface), in langleys= observed or estimated average snow surfacealbedo= basin convection-condensation melt factor, asdefined above. It depends on the relative exposureof the area to windT′ a = difference between the air temperature measuredat 10 ft and the snow surface temperature, indegrees Fahrenheit. (Snow surface temperaturecan be assumed to be 32 °F)T′ d = difference between the dewpoint temperaturemeasured at 10 ft and the snow surface temperature,in degrees FahrenheitF = estimated average basin forest canopy cover,effective in shading the area from solar radiation,expressed as a decimal fractionT a= air temperature, in degrees FahrenheitT b = fixed base temperature, near 32 °FGiven the numerous variables contained in the energybudget equations above, it can be seen that the employmentof temperature only as an index to snowmelt resultsin further approximation and inaccuracy; yet, consideringthe other uncertainties involved - particularly in forecastingapplications - this does not usually preclude its use.(a) The melt-rate factor, C, is of course an importantkey in the successful application of the temperature indexequation. Assuming daily melt computation interval, thisfactor would be on the order of 0.02 to 0.04 in./degreeper day when used with maximum air temperature and0.04 to 0.10 in./degree per day when used with averageair temperature. In clear-weather melt situations, thisfactor would typically increase as the snowmelt seasonprogressed because of factors such as the decrease inalbedo, increased short-wave radiation, etc. Because of5-8

EM 1110-2-141731 Aug 94Table 5-3Relative Magnitude of Snowmelt Factorsa. Assumed ConditionsCaseDescriptionAssumed Meteorological ConditionsT aT dI R V1.Clear, hot, summer day. No forest cover. Albedo = 40%70457000.032.Same as Case 1, 50% cloud cover65505000.033.Same as Case 1, fresh snow. Albedo = 70%70457000.03--4.5.6.---------------------------Heavy wind and rain, warm. No forest coverSame as Case 4, but light rain, windySame as Case 5, but light wind--505050--505050--000--3.0"0.5"0.5"--15153T a= Air Temperature, °FT d= Dewpoint Temperature, °FI = Solar Insulation, langleysR = Daily rainfall, in.V = Mean wind velocity, mphb. Daily Melt QuantitiesCaseSnowmelt Components, in.M swM lwM ceM rM gTotalMeltin.Rain +Meltin.1.1.680.000.460.000.022.162.162.1.200.000.460.000.021.691.693.0.840.000.460.000.021.321.32---------------------------------------4.0.070.522.260.380.023.256.255.0.070.522.260.060.022.933.446.0.070.520.450.060.021.121.62M s = Short-wave radiation meltM lw= Long-wave radiation meltM ce= Convection/condensation meltM r= Rain meltM g= Ground heat melt5-9

EM 1110-2-141731 Aug 94this, provision is usually made in simulation models tocalculate this as a variable, perhaps as a function of accumulatedrunoff or accumulated degree-days of airtemperature.(b) The choice of base temperature depends upon thecomputation interval involved and the form of the temperaturedata. If maximum daily temperature is the inputvariable, then this factor would be higher than 32 °F, perhaps40 °F. For a more frequent time interval, the factorwould be at or near 32 °F.(c) The possible range of the melt-rate factor can beillustrated by referring to the hypothetical cases presentedin Table 5-3. Using the daily melt quantity calculated bythe empirical energy budget equations and the temperaturesassumed, the melt-rate coefficients calculatedthrough Equations 5-1 and 5-2 would be as shown onTable 5-4. Table 5-4 generally confirms field experienceregarding the range in variation of the temperature indexmelt-rate factor. For clear-melt conditions, the factor variesbetween 0.03 and 0.06 in./°F and increases as thesnowmelt season progresses. For rain-melt conditions, thefactor can exhibit wide ranging variations from 0.06to 0.20, depending upon wind velocity and, to a lesserextent, the precipitation quantity. These factors would behigher if the temperature index used is the maximumdaily temperature. In forecasting practice, the melt-ratefactors are estimated through the process of calibrating ahydrologic model. Once established for known historicconditions, the factor can be modified by judgment to beapplied to the design condition or forecast situation underconsideration. Use of Equations 5-1 and 5-2 can be usefulguides in this process. Additional discussion of themagnitude of the temperature index melt-rate factor canbe found in the summary report (USACE 1956) and“Handbook of Snow” (Gray and Male 1981).c. Snow conditioning. As discussed in paragraph5-2c, snow conditioning or metamorphosis involvesthe warming of the snow pack to 32 °F, along withchanges in density and character of the snow and thesatisfying of liquid water deficiency. The first step insimulating this process is maintaining an accounting ofthe relative temperature of the snowpack below freezingas a function of time. This can be done through an indexrelation such as proposed by Anderson (1975):whereT s(2) T s(1) F p(T aT s(1))T s = index of the temperature of the snow packT a = temperature of the air(5-4)F p = factor, varying from 0 to 1, representing therelative penetration of the air temperature intothe snowpackIf F p is close to 1.0, the snow temperature will remainclose to that of the air; thus, high values would be appropriatefor a shallow snowpack. For a deep snowpack, alow value of F p will result in a slow cooling or warmingof the snow. The factor T s is limited to a value of 32 °F.(1) Once a snow temperature index is established fora computation period, the cold content (inches of waterrequired to raise the snowpack to 32 °F) can be calculatedthrough an equation such as:CC(2) CC(1) C r(T aT s(2))(5-5)Table 5-4Relative Magnitude of Melt-Rate Factors(Refer to Table 5-3 and Equations 5-1 and 5-2)Case T a T b Melt C in./°F Comment1237065703232322.161.691.320.0570.0510.035Low albedo, high SWECase 1, cloud coverCase 1, fresh snow4565050503232323.252.931.160.1810.1630.064Heavy rain, windyLight rain, windyLight rain, light wind5-10

EM 1110-2-141731 Aug 94whereCC = cold content (inches of water required to raisethe snowpack to 32 °F).C r= factor which converts the increment of temperaturedifferential T a - T s to an increment of coldcontent differentialThe value of C r might typically range from 0.01 to 0.05with higher values associated with late winter or earlyspring season. This factor is typically made a variable insimulation models by relating it to calendar periods or toa cumulative temperature index function.(2) The other factor important in simulating snowmeltis the liquid water deficiency of the snow. This is usuallytaken as a constant percentage of the water equivalent ofthe snowpack on the order of 3 percent. When meltoccurs, or rain falls upon the snowpack, the water generatedmust first be applied to satisfying the cold contentand liquid water deficiency before water is available toenter the ground.d. Snow accounting. As snowmelt progresses, theelevation of the snowline moves upward and the arealsnowcover of the basin decreases. An accounting of thisis necessary to be able to differentiate between snow-freeand snow-covered areas which have different hydrologiccharacteristics; and determine the elevation, of the snowpack,for calculating air temperature for indexing melt. Asecond computation, either associated with snow cover orindependent, is the accounting of the remaining snowwater equivalent of the snowpack.(1) If the basin has been configured into zones ofequal elevation as described in paragraph 5-4b, theaccounting of snow cover and quantity can be done on azone-by-zone basis. One assumption that can be made isto make the zone homogeneous with respect to elevationand either 100 percent snow-covered or snow-free. Thisassumption may require a large number of zones foradequate basin representation. Even with a large numberof zones, the abrupt changes in the snowline can occur asa zone changes from snow-covered to snow-free.Because of this, some models provide an ability to simulatea gradual transition within the zone.(2) An alternative to the distributed approach inaccounting for snow during melt is to employ a snowcoverdepletion curve in conjunction with a “lumped”watershed configuration. A snow-cover depletion curvedescribes the basin’s snow-covered area as a function ofaccumulated snow runoff as a percent of seasonal total.Studies have shown this relationship to be of relativelyuniform shape for a basin. Using historic field and satelliteinformation, a pattern curve can be developed for abasin. This does not have to be followed precisely inactual application if flexibility exists in the program tomake adjustments, for instance based upon real-time satelliteobservations of snow cover. While the snow-coverdepletion curve yields an accounting of snow cover, thismethod still needs to employ an independently derivedestimate of expected total basin snow water equivalent(SWE). The typical approach is to use multiple regressionprocedures as noted in paragraph 5-4d. The accountingof current remaining SWE during the melting of thesnowpack is simply a process of subtraction. Adjustmentsto the estimates of SWE will likely be required, basedupon model performance in simulating runoff.e. Simulation elements. Figure 5-3 illustrates theprocess of simulating snowmelt in a simulation model.For a given time period and subbasin element, theseinclude: (1) rain: is this a dry or wet melt calculation?(2) temperature, lapse rate, elevation of zone, etc.; (3) elevationof snow; (4) calculate temperature at zone pertinentto indexing; (5) melt; (6) type of melt computation;(7) other melt factors as necessary; (8) updated snowcondition status; (9) water available for melt; and(10) updated snowline and SWE.5-11

EM 1110-2-141731 Aug 94Figure 5-3. Illustration of snowmelt simulation5-12

EM 1110-2-141731 Aug 94Chapter 6Infiltration/Loss Analysis6-1. Generala. Role of infiltration/loss computations in floodrunoffanalysis. This chapter describes the methods typicallyavailable for computing the time history of directrunoff volume due to a single rainfall event. This isdetermined by subtracting from the rainfall hyetograph thelosses due to interception, surface storage, and soil infiltration(Figure 6-1). The rainfall excess is routed to thesubbasin outlet, usually by unit hydrograph or kinematicwave techniques, and combined with base flow to obtainthe subbasin hydrograph.b. Physical process. Soil infiltration and surface lossof rainfall involve many different processes at differentscales of observation. The most basic of the processes isthe infiltration of water into an “ideal” soil, a soil ofuniform properties and infinite depth as shown in Figure6-2. Initially, the soil is assumed to have a uniformwater content. The initial water content or an initialcondition related to the water content must be specifiedfor any of the methods which are used for single rainfallevent analysis. At the commencement of rainfall, water isinfiltrated until the rainfall exceeds the capacity of waterto be absorbed by the soil. At this point, the surfacebecomes saturated and rainfall in excess of the soil infiltrationcapacity is assumed to be runoff. As the volumeof infiltrated water increases, the infiltration capacity ofthe soil decreases to a minimum rate equal to the soil’ssaturated hydraulic conductivity. The saturated hydraulicconductivity is a proportionality constant between hydraulicgradient and flow in Darcy’s law for saturated flow inporous (soil) media and is assumed to be a characteristicof the soil.(1) Theoretically, the transport of infiltrated rainfallthrough the soil profile and the infiltration capacity of thesoil is governed by Richards’ equation (Richards 1931and Eagleson 1970). Richards’ equation is derived bycombining an unsaturated flow form of Darcy’s law withthe requirements of mass conservation. Solutions toRichards’ equation show an exponential decrease of infiltrationcapacity with cumulative infiltration. Conceptualor empirical loss-rate equations attempt to duplicate thisin computing rainfall excess.(2) The predictions of infiltration by Richards’ equationmay at best be an approximation to actual field lossesbecause the ideal soil model does not correspond particularlywell to field conditions. The deviations occur forseveral reasons: (a) the soil is heterogenous, usuallylayered and of finite depth; (b) the soil matrix is not aninert structure but is continually being affected by chemicaland biologic processes; (c) surface losses and coverhave a major impact on the available excess; and (d) theideal soil model is a gross approximation to the dynamicsof direct runoff production. Consider the impact of theseadditional processes on rainfall loss rates. Soil heterogeneitymakes both the formulation of a physical modeland the estimation of model parameters much more complicated.Formulating the equations of fluid motion in aheterogenous, layered soil is a difficult problem. Theequations could be formulated, but estimating the parametersof the model, such as soil hydraulic conductivity, istotally impractical given the information typically availableto the engineer. Furthermore, the detail needed tocapture the small scale changes of soil properties isimpractical. At best, some average estimate of soil propertiesfor a relatively large area, a lumped approach tomodeling, must be employed to model infiltration.(3) Far from being inert materials developed strictlyfrom the weathering of bedrock, soils owe their propertiesto the chemistry of rainwater, the chemical properties ofthe parent material, organic matter content and the presenceof roots and burrowing animals. The chemistry ofwater is important because it can affect the shrink/swellpotential of clays and the osmotic pressures within thesoil. Clay soils may shrink and crack resulting in a desiccatedsurface which results in infiltration capacities far inexceedance of anything that would be expected from amaterial with a clay’s saturated hydraulic conductivity.The hydraulic conductivity of the soil, being inverselyproportional to water viscosity, is sensitive to the watertemperature. The soil porosity, the ability to hold water,increases with the organic matter content. Burrowinganimals and decaying tree roots create what has beentermed “macropores” that are very effective in conveyingwater.(4) Surface losses are categorized as being due tointerception, depression, and detention storage. Interceptionstorage results from the absorption of rainfall bysurface cover such as plants and trees. Depression storageresults from micro- and macrorelief depressions in thesurface topography that store water which eventuallyinfiltrates or evaporates. Also a function of topography,detention storage acts as minireservoirs, increasing theretention time of overland flow and providing moreopportunity for infiltration.6-1

EM 1110-2-141731 Aug 94Figure 6-1. Loss rate, rainfall excess hyetographFigure 6-2. Wetting front in ideal soil6-2

EM 1110-2-141731 Aug 94(5) Surface cover also increases loss rates by delayingoverland flow. In addition, surface cover impacts onrainfall losses by protecting the soil surface from theimpact of rainfall, preventing the formation of surfacecrusts that decrease the hydraulic conductivity of the soilsurface.(6) The extent to which surface conditions affectrainfall excess is a function of land use. Forested areasexhibit the greatest surface losses because of their welldevelopedcanopies and significant surface storage providedby surface litter. Range land is less effective instoring water because of sparser cover. The presence ofgrazing further reduces cover and increases runoff potential.Bare surface conditions in agricultural areas canpotentially result in relatively high runoff rates due tocrusted surfaces formed from rainfall impact. Managementpractices, such as contour plowing or mulching,have been employed to protect the soil or store overlandflow. Urban area runoff increases in proportion to theamount of impervious area and how this area is connectedto outflow points by the drainage system.(7) Even if the ideal soil model could account for allthe processes mentioned so far, there would still be theproblem of accounting for the dynamics of direct runoffproduction. Direct runoff can be simulated by either theHorton or Hillslope process (Ward 1967). The Hortonprocess, named for the famous hydrologist, correspondsmore closely to the ideal soil model (Figure 6-3). In thisprocess, overland flow results when all surface storagesare filled and the rainfall rate exceeds the infiltration rate.Overland flow that does not infiltrate along the flow pathto a channel results in direct runoff. The potential forinfiltration along the flow path is not accounted for whenan average soil property is used to calculate runoff in anideal soil model.(8) The Horton process is most likely to be importantin urban and agricultural areas where the infiltrationcapacity of soils is relatively small due to cultural activities.However, overland flow, the cornerstone of theHorton process, rarely occurs in forested soils. Forestsoils generally have extremely high infiltration capacitiesin the upper horizon due to a well-developed surfacecover and extensive tree root structure. In these soils,direct runoff is due to the Hillslope process. In this process,direct runoff results due to the mixture of surfaceand subsurface flow. Prior to direct runoff, the initialwatershed moisture conditions are characterized by drierconditions at the top of the hillslope and wetter conditionsat lower elevations near the channel (Figure 6-4). At thecommencement of rainfall, water infiltrates at the top ofthe hillslope and moves vertically through the soil until itreaches a low conductivity soil zone. Lateral movementof the infiltrated water occurs along the lower conductivitylayer as either saturated or unsaturated flow until itseeps out to the surface nearer the bottom of the hillslope.At this point, the infiltrated water combines with overlandflow generated by rainfall on the initially wetter areasnear the stream channel. These areas are termed variablesource areas because as the rainfall continues they growin size, comprising more of the watershed area. Observationshave shown that the subsurface movement of waterdown the hillslope combined with overland flow from thesource areas is the flood mechanism in forested areas. Insome respects, the apparent rainfall excess in a floodhydrograph in a forested area is a combination of interflow,subsurface flow, and overland flow.(9) In summary, the rainfall infiltration/loss processis complex and affected by many factors. Soil propertiesare important, but chemistry of the water, biologic activity,soil heterogeneity, and surface cover modify the soil’sinfiltration capacity. Surface cover and topography alsoare involved in losses by intercepting, storing, and detainingrainfall. Finally, the dynamics of the rainfall-runoffprocess are important in determining the volume of rainfallavailable for direct runoff. Even though excess maybe generated at some point in an agricultural or urbanarea, some of this excess may infiltrate as overland flowtraveling to a channel. In forested areas, flow that hasinfiltrated is a major contributor to direct runoff.c. Approaches to infiltration/loss analysis. Watershedmodeling for flood prediction is an exercise in findingadequate estimates of watershed properties overwatershed size areas. The methods used to model infiltration/lossrates reflect this approach.(1) The methods can be categorized as physicallybased, conceptual, or empirical. The physically basedmodels, such as Green and Ampt, are based on simplifiedsolutions to the Richards equation. This approach wasdeveloped for three reasons: (a) the solution of the Richardsequation is difficult and not justified given that thisequation is, at best, only a rough approximation of theactual field infiltration; (b) a simplified solution still producesthe exponentially decreasing relationship betweeninfiltration capacity and cumulative infiltration; and(c) the parameters of the methods can be related to soilproperties that can be measured in the laboratory, such asporosity and hydraulic conductivity.6-3

EM 1110-2-141731 Aug 94Figure 6-3. Horton runoffFigure 6-4. Hillslope process6-4

EM 1110-2-141731 Aug 94(2) Methods such as the Holtan loss rate conceptuallymix parameters which have a physical basis such as adeep percolation rate with empirical ones such as anexponent which controls the infiltration capacity as afunction of storage. Empirical methods, such as the SoilConservation Service (SCS) curve number (CN), arebased on correlating parameters estimated from rainfallrunoffrecords to factors affecting loss rates such as soiltype and surface cover. The initial and constant loss ratemethod can be considered either an empirical method or agross representation of an infiltration curve. Each ofthese methods have been applied in watershed models andwill be discussed in the following sections.6-2. Gauged versus Ungauged ParameterEstimationa. Parameter estimation techniques generally arecategorized by application to gauged or ungauged analysis.In the description of loss rate methods, parameterestimation is discussed only with regard to estimatingparameters from the physical characteristics of the watershed.These estimates can be useful in an ungauged orgauged situation. In an ungauged situation, physicalcharacteristics may be the only information available forestimating parameters.b. Gauged estimation procedures are used to estimatemodel parameters, including loss rate parameters,from rainfall-runoff records. The basic element of agauged estimation is to utilize an “optimization” algorithmto choose model parameters so that some measure of thedifference between observed and predicted hydrographs isminimized. This approach to parameter estimation isessentially a regression analysis, as pointed out byDawdy, Lichty, and Bergman (1972). An important principleof regression analysis is parsimony, i.e., inclusion ofthe minimum number of parameters in the model that areneeded to explain the data. In this respect, a simple twoparameterloss rate method, such as the initial and constantloss rate method, is probably adequate because it isparsimonious. Experience has shown that simple empiricalmethods with a minimum number of parameters do asatisfactory job when simulating observed hydrographparameters.c. Although not as parsimonious as simple empiricalmethods, methods with physically based or measurableparameters, such as the Green and Ampt method, can beadvantageous in gauged analysis. The advantage stemsfrom the ability to place bounds on the values of theseparameters. The bounds can be applied using two differentapproaches when applying an optimization procedure.One approach would be to evaluate whether or not thederived parameter estimates are within a reasonable rangebased on the physical characteristics of the watershed. Asecond approach is to constrain the parameter values to areasonable range within the optimization. The secondapproach may prove difficult because of errors in rainfallrunoffdata which dictate that parameters assume unrealisticvalues. Constraining the parameters may prevent areasonable prediction of observed runoff.d. A reasonable procedure to follow when applying aphysically based loss rate method in a gauged analysis isto only perform parameter estimation with a maximum oftwo parameters. Additional parameters in the methodshould be estimated based on the physical characteristicsof the watershed. Certainly, optimized parameters willhave estimated values which are not reasonable due toobservation errors. However, over a number of events,the errors should balance resulting in an acceptable estimateof loss rate parameters. Acceptance can be based onwhat seems reasonable from watershed characteristics.6-3. Antecedent Moisture Conditionsa. The application of the methods discussed requiresan estimation of the antecedent moisture condition (AMC)of the watershed surface cover and soils. Unfortunately,there is no simple answer as to how the AMC might beestablished. The approaches to use are a function of theintended application. Different approaches may be useddepending upon whether individual or design events arebeing simulated or a gauged or ungauged analysis is beingperformed. Consider the simulation of individual events.The gauged analysis is straightforward, with the AMCused as another parameter that is adjusted to improvecorrespondence between the observed and predictedhydrograph. Ungauged analysis is much more difficult inthat some methodology must be developed to determineAMC. The usual technique is to rely on an antecedentprecipitation index (API) which is presumably based onregional information. API is a poor indicator of AMCdue to various factors, most notably the impact of weatherconditions on evapotranspiration. However, it’s the onlyindicator usually available.b. Estimation of AMC for design events depends onthe type of event. AMC for probability-based designstorms might be based on calibration to a gauged orregional discharge or volume frequency curve. In contrast,AMC (and in general loss rates) determination fordeterministic design events such as the probable maximumprecipitation is set by policy.6-5

EM 1110-2-141731 Aug 94c. Certainly, the techniques for establishing AMCare varied and subject to some argument. When gaugedinformation is not available, reliance on regional informationis essential in establishing an AMC. Otherwise, theengineer may be forced to assume a conservative estimatefor this parameter.6-4. Surface Loss Estimationa. Rainfall losses are due to both surface storage andsoil infiltration. In the field, the surface storage andinfiltration of rainwater are dynamically interconnected.The interconnection occurs primarily via surface depressionand detention storage. Detention storage increasesinfiltration rate by adding a small (less than an inch)pressure head to the wetting front. This additional head isinsignificant when compared to the suction head whichdrives soil infiltration. Detention storage increases apparentinfiltration by delaying surface flow and providingmore catchment retention time for water to infiltrate. Ingeneral, these effects are minor when compared to theproblem of estimating the magnitude of surface loss andthe in-situ capacity of soils to infiltrate water. Consequently,the typical approach is to separate these twocontributions to rainfall loss unless surface losses areempirically included in the loss rate method. For example,the SCS curve number method includes surface lossesdirectly into the method.b. Surface loss is a function of land use and differsgreatly between forested, agricultural, and urban areas.According to Viessman et al. (1977), interception of rainfallby surface cover is greatest for a forest and decreasesfor agricultural and urban land uses. Schomaker’s (1966)measured values of interception for a spruce forest were30 percent of the annual rainfall and for a birch forestwere 9.5 percent of annual rainfall. Horton (1919)reported that the interception for rainfall events greaterthat 0.25 in. is approximately 25 percent of the total rainfall.The Viessman et al. (1977) conclusion from thisinformation is that interception for forested regions isapproximately 10 to 20 percent of the total precipitation,at least for rainfall events less than 2.0 in. In general,one should not expect interception losses to exceed 0.5 in.for a particular rainfall event.c. Agricultural watershed surface losses are a functionof crop development and management practice.Interception of rainfall by crops was computed byLinsley, Kohler, and Paulhus (1975) using equationsdeveloped by Horton (1919). They found that for a stormdepth of 1.0 in., the interception ranged from 3 to 16 percentfor small grain crops such as wheat and milo. Thiscompares well to the study by Schomaker (1966), sinceinterception by these crops should be less than that of aforest due to the smaller leaves and sparser cover providedby these crops.d. Detention storage in agricultural areas is stronglyaffected by the time since tillage occurred and the overallmanagement practice. Linden (1979) used random roughnessand land surface slope in microrelief models topredict depression storage due to tillage (note randomroughness is essentially a measure of the variation of soilheights from the surface plane). He predicted that depressionstorage could be as high as 0.5 in. immediately aftertillage. The depression storage will decrease with timeafter tillage due to the impact of rainfall. Linden’s resultsdo not account for increased storage capabilities due tomanagement practice such as contour plowing. Horton(1935) estimated that detention storage for agriculturallands, natural grass lands, and forests range from 0.5 to1.5 in.e. Surface losses in urban areas differ for open andimpervious areas. Interception losses for open areas(lawns, parks etc.) can probably be considered of thesame magnitudes as forest or pasture land. However, thedepression storage in the open areas is probably not asgreat as in natural areas because grading has taken placeand there is probably less surface litter. The surface lossfor impervious areas is small and usually taken as 0.1 to0.2 in. Table 6-1 summarizes the surface losses that canbe used for each land use type. The values listed inTable 6-1 are a suggested range based on previousresearch work and experience. If these values are not inline with local experience of a particular watershed, themodeler should by all means use any local information.6-5. Infiltration Methodsa. Green and Ampt. The Green and Ampt method isexplained and illustrated in detail below.(1) Method development. The Green and Ampt(GA) method (Mein and Larson 1973) assumes the samesimple soil model and initial conditions as that of theRichards equation, a uniform soil profile of infinite extent,and constant initial water content. As the water content atthe soil surface increases, the method models the movementof the infiltrated water by approximating the wettingfront with a piston type displacement (Figure 6-5).(a) The piston displacement model, as originallydeveloped, must be modified to account for surface lossesand variable rainfall rates (time varying surface moisture6-6

EM 1110-2-141731 Aug 94Table 6-1Surface LossesInterception LossesAgricultural AreasCropHeightft.Interceptionin.Corn .............................................. 6 0.03Cotton ............................................. 4 0.33Tobacco ............................................ 4 0.07Small grains ......................................... 3 0.16Meadow grass ....................................... 1 0.08Alfalfa ............................................. 1 0.11(from Linsley, Kohler, and Paulhus 1975)Forest Areas (from Viessman et al. 1977)10-20% total rainfall, maximum 0.5 in.Detention Storage (from Horton 1935)Agricultural Areas(Depending on time sense tillage)Forests/Grasslands0.5 - 1.5 in.0.5 - 1.5 in.Total Surface LossUrban AreasOpen AreasImpervious Areas0.1 - 0.5 in.0.1- 0.2 in.conditions). The surface loss is modeled for an initialloss as follows:The cumulative infiltration loss is calculated by the GAmethod:wherer(t) 0 for P(t) ≤ I at ≥ 0r(t) r o(t) for P(t) >I at≥0(6-1)(6-2)P(t) = cumulative precipitation over thewatershedwhereIS f[(i/K) 1]dl/dt=i(t) = infiltration rateKS f[(dl/dt)K = soil’s hydraulic conductivityK] i > K(6-3)r(t)tr o (t) and I a= rainfall intensity adjusted for surface losses= time since the start of rainfall= depth of surface loss assumed to be uniformover the watershedS f= product of the wetting front suction, h f , andthe soil volumetric deficit at the beginning ofthe storm∆θ and I = cumulative infiltration6-7

EM 1110-2-141731 Aug 94Figure 6-5. Green and Ampt piston wetting frontThe GA equation as originally developed, is only strictlyapplicable to a uniform moisture condition at the soilsurface or, in the case of rainfall infiltration, a pondedsurface condition. Modifications were made as suggestedby Mein and Larson (1973) and Morel-Seytoux (1980) touse the GA equation for unponded surface conditions andvariable rainfall rates. In the absence of ponding, infiltrationis estimated for any period by (Figure 6-6):∆I I jI j 1S fr j/K 1j 1Σ r i∆t ii 1r j≥ K(6-4)whereI j and I j-1r j∆I= cumulative depth of infiltration at the end oftime period j and j-1= average rainfall rate over the period ∆t j= potential depth of water infiltrated during theperiod6-8

EM 1110-2-141731 Aug 94Figure 6-6. Green and Ampt application of variable rainfall rate6-9

EM 1110-2-141731 Aug 94If the rainfall rate is less than K or if:∆ t 1j∆Ir j> ∆ t jthen ponding does not occur.time is equal to:t pt j 1∆t 1j(6-5)Otherwise, the ponding(6-6)(c) There may be instances when the rainfall rateduring a storm drops below the infiltration rate after aninitial ponding time has been calculated. In this case, anew ponding time is calculated by keeping track of theaccumulated infiltration and reapplying Equation 6-4; thenEquation 6-7 is applied as before to calculate the excessrate.(d) The infiltrated volume computed by this methodshould always be compared with the total storage volumeavailable in the soil profile. The storage volume in thesoil profile may be computed as:Once ponding has occurred, the cumulative infiltration iscomputed by integrating Equation 6-3 to obtain:S a(∆θ)d(6-10)(I/S f) In(1 (I/s f)) (K/S f)(t t et p)(6-7)whereS a = available initial soil storagewith the initial condition that at t = t p , I = I p where I p isthe cumulative infiltration at ponding and:t e(S f/K) ((I p/S f) In(1 (I p/S f)))(6-8)(b) Prior to ponding, the rainfall excess rate is zero.The rainfall excess rate after ponding is determined bysubtracting the incremental infiltration from the rainfallduring a period:wheree je j(r j∆ t j∆ I)/∆ t j= excess rate during any period(6-9)∆I = incremental infiltration, which is equal to thedifference between applying Equation 6-7 fortimes t j and t j-1d = depth of the soil profileThe GA method is not constrained by storage considerationsbecause of the assumption of an infinite profile.(2) Parameter estimation. Readily available informationfrom soil surveys, texture class, and particle sizedistribution has been used to estimate the GA parameters.Texture class differentiates between types of soils (sand,sandy loam) as shown in Figure 6-7 based on ranges inparticle size distribution, the percent sand, silt, and claycontained in the soil. The general procedure involved hasbeen to relate this information to the GA parameters viathe water retention characteristics of the soil. The moistureretention characteristics are defined by the relationshipof water content to the soil suction (Figure 6-8).Soil suction is essentially a capilary effect, the drier andfiner textured the soil (a clay is a finer textured soil thana sand), the greater the suction. Brooks and Corey (1964)suggested that the water retention versus suction relationshipcould be represented by:S e(θ θ r)/(θ sθ r) (h c/h cb)Notice that Equation 6-7 does not give an explicit expressionfor I. An approximate technique described by Li,Stevens, and Simons (1976) is one approach that can beused to solve for I at any t.whereS e= effective saturation6-10

EM 1110-2-141731 Aug 94Figure 6-7. USDA texture triangle6-11

EM 1110-2-141731 Aug 94Figure 6-8. Water content versus suctionΘθ rθ sh cbλ= volumetric water content at suction, h c= residual saturition= water content at saturation= bubbling pressure= pore size distribution(6-14)Assuming that the initial water content is equal to theresidual saturation, the formula finally derived by Brakensiek(1977) and applied by Rawls and Brakensiek (1982a)is obtained as:The Brooks and Corey parameters are then used to calculatethe wetting front sucfion, h f , by:(6-15)whereh ceh ci= water entry pressure(6-12)(6-13)= water content corresponding to the initial soilwater content of the soil prior to ponded infiltrationResearch performed by Rawls and Brakensiek (1983) andRawls, Brakensiek, and Soni (1983) related the GA parametertotal porosity and the Brooks and Corey parametersto soil texture class as shown in Table 6-2. Theinformation shown in the table can be used together withan estimate of the initial water content via Equation 6-12to estimate h f . Estimates of h f for initial water contentequal to the residual saturation are shown in Table 6-2 forinformational purposes.(a) Attempts made by these researchers to find a relationshipbetween texture class and saturated hydraulic6-12

EM 1110-2-141731 Aug 94conductivity, K, were unsuccessful because the varianceof K within the texture class is too large. However,Rawis and Brakensiek (1983) and Rawls, Brakensiek, andSoni (1983) did provide average estimates of K for thesoils sampled in their survey as shown in Table 6-2.Note that variances about the mean value for each of theparameters are shown in this table except for K becausetexture class was not found to be a discriminator of thisvariable.(b) Additional work has been performed by Ahuja etal. (1988) and Rawis and Brakensiek (1989) to improvepredictions of GA parameters using particle sizedistribution and/or soil porosity. Further modifications tothe estimates for surface cover characteristics, stones, andsurface crusts have been developed by Rawls andBrakensiek (1983); Rawls, Brakensiek, and Soni (1983);and Rawls, Brakensiek, and Savabi (1988).(c) An initial water content θ i must be selected priorto determining ∆θ and h f . A means for estimafing θ i maybe to relate watershed moisture conditions to an antecedentprecipitation index.b. Holtan loss rate method. The Holtan loss ratemethod is expuned and illustrated in detail below.6-13

EM 1110-2-141731 Aug 94(1) Method development. Holtan et al. (1975) used aconceptual soil storage element to compute infiltrationrates based on the formula:(6-18)(6-16)where(6-19)wherei= potenfial infiltration rate in inches per hour(6-20)GI = "growth index" representing the relative maturityof the ground coverAS a= inches per hour per inch of available storageand is an empirical factor discussed in moredetail in the next section= soil storage capacity in inches of equivalentdepth of pore space in the surface layer of thesoil, f c is the constant rate of percolation ofwater through the soil profile below the surfacelayerβ = empirical exponent, typically taken equal to 1.4The available storage, S a , is decreased by the amount ofinfiltrated water and increased at the percolation rate, f c .Note that by calculating S a in this manner, soil moisturerecovery occurs at the deep percolation rate. The methodis applied to a variable rainfall rate by continuously accountingfor storage using the following relationship,given the initial soil deficit S a0 :where(6-17)= storage deficit at the beginning andending of period ∆ti = average infiltration rate during thisperiod(f c ∆ t ) = drainage volume out of storageThe volume draining from storage is limited by the maximumallowable deficit S a . The average infiltration overthe period is the minimum of the available rainfall or thepotential infiltration rate. The potential infiltration rate iscalculated as:The potential infiltration rate (essentially the averageinfiltration rate) must be calculated implicitly or iterativelysince it is a function of the storage deficit at theend of the period. The excess rate is the differencebetween the rainfall rate and average infiltrafion rate.(2) Parameter estimation. The factor "A" is interpretedas an index of the pore volume which is directlyconnected to the soil surface. The number of surfaceconnectedpores is related to the root structure of thevegetation, so the factor "A" is related to the cover cropas well as the soil texture. Since the surface-connectedporosity is related to root structure, the growth index (GI)is used to indicate the development of the root system. Inagricultural basins, GI will vary from near zero when thecrop is planted to 1.0 when the crop is full-grown.(a) Holtan et al. (1975) have made estimates of thevalue of "A" for several vegetation types. Their estimateswere evaluated as the percent of the ground surface occupiedby plant stems or root crowns at plant maturity.Skaggs and Kahleel (1982) provide estimates as shown inTable 6-3.(b) Estimates of f c can be based on either the valuesgiven in Table 6-3 (Skaggs and Kahleel 1982) or thehydrologic soil group given in the SCS Handbook (1972).Musgrave (1955) has given the following values of f c ininches per hour for the four hydrologic soil groups: A,0.45 to 0.30; B, 0.30 to 0.15; C, 0.15 to 0.05; D, 0.05 orless.(c) The total soil storage capacity can be computedusing information in Table 6-2 as:(6-21)6-14

EM 1110-2-141731 Aug 946-15

EM 1110-2-141731 Aug 94where d = depth of the soil horizon. The inifial deficit isgiven by Equation 6-10. The initial water content wouldhave to be determined from an assessment of past conditions.c. Soil Conservation Service curve number method.The SCS curve number method is explained in detailbelow.(1) Method development. The curve number (CN)method depends on the following basic relationship:(6-22)in terms of the precipitation and the parameters of themethods I a and S.(b) The CN method does not incorporate timeexplicitly into the formulation. Consequently, the applicationof the method to a rainfall hyetograph requires thattime be incorporated rather simply into Equation 6-24 as:where(6-25)whereFSQ= watershed retention of water= maximum available retention capacity= direct runoffQ(t) = cumulative runoff at time tP(t) =cumulative rainfall minus I a at time tThe incremental runoff depth over a period ∆t =t 2 -t 1 :(6-26)P= total storm precipitation (in consistent units ofvolume; for example, basin-inches)The retention parameter, S, is related to the CN by arelationship that will be discussed in the next section onparameter estimation. The supposition that F=Sas theamount of precipitation becomes large seems reasonable,since most of the precipitation will directly runoff as thewatershed soils become saturated. Q=Pis a fair approximationfor the same reason.(a) A parametric relationship for calculating directrunoff can be developed by setting F=(P-Q-I a ) andthen solving for Q, assuming that Equation 6-22 applies:(6-23)Note, the computation of cumulative excess by Equation6-25 is entirely dependent on the cumulative precipitationat any time. The total infiltration, therefore (like therunoff) is independent of the storm pattem.(2) Parameter estimation. The parameters of the CNmethod were estimated by examining a great deal of datafrom small (less than 10 acres) agricultural watersheds inthe midwestern United States. The goal was to relate 1.and S to physical characteristics of the watershed. Tosimplify this problem, Equation 6-24 is transformed to useonly a single parameter by developing the following relationshipfrom test watershed data:(6-27)A further simplification was made by relating S to CN as:where I a = basin volume is equal to the initial abstractionof rainfall (i.e., the observed rainfall depth prior to theobservafion of runoff). Solving Equation 6-23 for Qgives the desired direct runoff:(6-24)(6-28)This transformation was performed according to VictorMockus (1964) so that the rainfall-runoff curves fromEquation 6-26 would plot at nearly equal intervals acrossa graph sheet. The CN was assumed to be related to thesoil and cover conditions of a watershed. A search wasmade by Mockus for test watersheds with a single cover6-16

EM 1110-2-141731 Aug 94characteristic and soil type. Total rainfall versus runoffvolumes were analyzed graphically to determine the appropriateCN for the soil type and cover for each watershed.As might be expected, there was a great deal ofscatter in the observed data when plotted in this manner.The CN that resulted in a curve that divided the plotteddata in half was deemed appropriate.(a) A relationship between CN and watershed potentialrunoff was developed by determining enveloping CNfor the scattered data. This results in three sets of curvesthat divide and bound test data for an individual watershed.In the past (SCS 1972), the upper and lower envelopingcurves were assumed to be related to relatively wet(AMC III) and dry (AMC I) watershed soil moisture conditionsand the dividing curve by average soil moistureconditions (AMC II). The CN associated with these differentsoil moisture conditions was then related to the5-day antecedent rainfall. However, the relationshipbetween antecedent rainfall and AMC has been poor andthe SCS no longer relates the potential runoff to an AMC.Rather, the potential runoff defined by the curves envelopingthe scattered data is now related to three antecedentrunoff conditions, ARC(III) for relatively high runoffpotential, ARC(I) for relatively low runoff potential, andARC(II) for average runoff potential.(b) The average CN value for a parficular watershedand the effect of ARC on CN should be determined basedon observed rainfall versus runoff. The SCS now recommendsthat the calibration method used by Mockus or astatistical analysis of rainfall versus runoff data be used todetermine the CN for each ARC value. Table 64 displaysthe effect of ARC condition on curve number basedon the past work by Mockus in developing envelopecurves of CN for observed rainfall versus runoff.McCuen (1989, pg. 299) cautions that this table is onlyapplicable for the region where the CN was calibrated andshould be adjusted based on regional information. Hisrecommended caution refers to the use of the now obsoleteAMC designations but is equally relevant to the ARCdesignations in the table. If data are not available formaking adjustments to the curve number, then theARC(II) curve numbers of Table 64 should be used.(c) The CN corresponding to a large number of soiltypes and cover characteristics are reported by the SCS.Consequently, application of the method requires that soilsurvey information be available for the watershed of interest.A soil survey provides the information needed tochoose CN based on soil type, cover, management practice,and hydrologic condition. Hydrologic group indicatesin-situ infiltration capacity by classifying the soils astype A, B, C, or D, with A having the highest and D thelowest capacities. The CN associated with each group(Table 6-5) is determined based on the cover (agriculturalversus forest), management practice (tillage practice andmulching), and hydrologic condition (degree of grazing orpercentage of area with good cover characteristics). Amore detailed table of curve numbers can be found inSCS TR-55 (SCS 1986) or the National EngineeringHandbook, Chapter 4 (SCS 1972).(d) Although the CN method is easily the most popularmethod for performing ungauged analysis, there hasbeen extensive criticism of the method because it does notlead to accurate reproduction of runoff hydrographs, thepredicted infiltration rates are not in accordance withclassical unsaturated flow theory, the method is applied towatersheds for which it was not calibrated, and the originalcalibration results are not available. As pointed outby Rallison and Miller (1981), p 361:The CN procedure continues to be most satisfactorywhen used for the type of hydrologic problemthat it was developed to solve--evaluating effectsof land use changes and conservation practices ondirect runoff. Since it was not developed to reproduceindividual historical events, only limitedsuccess has been achieved by those using it forthat purpose.Despite this well recognized deficiency, the method remainspopular for simulating rainfall hydrographs.(e) The method has received crificism because it is atvariance with the results of classical unsaturated flowtheory, as can be seen by examining the infiltration rateimplied by Equation 6-25 (Smith 1976, Aron, Miller, andLakatos 1977, and Morel-Seytoux and Verdin 1981):wherei = infiltrafion rater = rainfall intensity(6-29)Morel-Seytoux (1981) points out that i and P areinversely related. As one would expect, the proportionalityof i and r is "in direct disagreement with field experience,laboratory evidence and physical theory," which6-17

EM 1110-2-141731 Aug 94shows that i is independent of r for a ponded surfacecondition.(f) Perhaps the most disturbing aspect of the CNmethod is that the original calibration results obtained byVictor Mockus (1964) have not been preserved. Consequently,the only means of evaluating the observed performanceof the method is to examine current results fromthe literature or from personal experience.(g) However, despite the missing calibration results, itis clear that the method is being used for watershedswhere data did not exist to calibrate the method. Rallisonand Miller (1981) p 361 point out:Data for developing reliable curve numbers are notequally available throughout the United States. Informationon rainfall, runoff and soil is deficient formany range and forest areas, particularly in the WesternStates and, as a consequence, there are many soilcomplexes that are either unclassified or lack data forverification. The sparseness of rainfall-runoff data inurban or urbanizing areas has forced reliance on interpretivevalues with litlle "hard" data available forverification....Despite these caveats about the CN method, engineerscontinue to use the method because it has been the only6-18

EM 1110-2-141731 Aug 94one available that relates readily available watershed characteristicsto a loss rate method.(h) Caution should be used in applications to areaswhere the CN method has not been calibrated. Information on regional rainfall-runoff characteristics should beobtained, if possible, to judge whether or not the CN methodpredictions are useful.(i) Rallison and Miller’s comments with regard toapplications in urban areas are particularly noteworthy.The CN usually chosen for open land uses in urban areas6-19

EM 1110-2-141731 Aug 94are generally based on CN values determined for pastureland use. However, runoff tends to be greater from theopen urban areas than that from a pasture land use. Acommon approach for adjusting for this affect is to reducethe value of I a , thus relaxing the constraint that I a = 0.2S.This approach is not appropriate since the relationshipbetween the initial abstraction and watershed retention iscritical to the reported CN calibration (1986). Eitherattempts should be made to find regional or localinformation for recalibrating CN, or the CN should beadjusted based on some judgment for open land use inurban areas.(j) Researchers have suggested means for utilizing theempirical data present in the curve number method inmore physically based infiltration equations. Hjelmfelt(1980) suggested a procedure for incorporating CN informationinto the Holtan equation. Morel-Seytoux andVerdin (1981) suggested a procedure for doing the samewith the Green and Ampt equation. However, one mightwonder about the efficacy of this approach since there isno information available which details the accuracy of theoriginal CN calibration to observed data or whether or notit is useful for rainfall-runoff simulations.d. Initial and constant loss rate method. The initialand constant loss rate method is described in detail below.(1) Method development. This is a very simple methodand does not need much explanation. An initial loss(units of depth) and a constant loss rate (units of depth/hour) are specified for this method. All rainfall is lostuntil the volume of initial loss is safisfied. After theinitial loss is satisfied, rainfall is lost at the constant rate.As in the case of the GA method, infiltrated volumescomputed by the initial and constant loss rate method arenot constrained by the storage capacity of the soil profile.Consequently, a comparison should be made of the infiltratedvolume and soil storage capacity to be sure that theparameters chosen for the method are appropriate.(2) Parameter estimation. The initial and constantloss rate method, having only two parameters, is valuablein the application of automatic parameter estimation procedures.However, the method could also be used inungauged analysis by assuming a physical interpretationof the parameters. The constant loss might be interpretedas the ultimate infiltration capacity of the soils. Theinitial loss might reflect both antecedent moisture conditionsand losses prior to reaching the ultimate infiltrationcapacity.6-6. Impervious Areasa. Estimation of losses from an urban area is complicatedby the presence of impervious surfaces which arenot hydraulically connected to drainage systems. Typically,these areas are roof tops with downspouts that drainto flower beds or lawns. The critical part of the analysisis to determine if the pervious area can infiltrate the flowreceived from the unconnected impervious area. Amethod applied by SCS (1986) considered this problem indetemining corrections for the curve number based on thepercent of total and unconnected impervious areas asshown in Figure 6-9. The correcfions are only applicablefor areas with up to 30 percent total impervious area. Ifthe percent impervious area exceeded this amount, thenthe assumption was that the unconnected impervious arearunoff would not infiltrate because of the small retentiontime on pervious areas.b. Figure 6-9 was established by calculating theamount of runoff from the unconnected impervious watershedarea due to a given rainfall depth and uniformlydistribufing this volume over the pervious area (McCuen1989). The runoff from the pervious area was then calculatedbased on the pervious area curve number and thecombined volume from rainfall and unconnected imperviousarea runoff. The apparent curve number for the entirewatershed is then back calculated from knowing the totalrainfall and the combined runoff from the pervious areaand connected impervious area. This procedure could beduplicated for methods other than the curve number.c. Caution should be used when applying Figure 6-9because of the assumptions used in its development. Inmany instances, conveyance of flow from unconnectedimpervious areas may not exist or may be very direct. Forexample, portions of a rooftop may directly drain to abackyard which does not drain easily into the street gutter.However, the drainage path from the downspouts drainingthe front portion of the rooftop may be rather short, providinglittle opportunity for infiltration. Certainly, localknowledge of drainage design is needed to judge to whatdegree unconnected impervious area acts as if it werehydraulically connected.d. Caution should also be used when compositeimpervious/pervious values for loss rate parameters areprovided for a particular land use. For example,SCS (1986) provides Table 6-6 for applications in urbanhydrology. Notice that in this table composite curve numberare given for urban land uses as a function of6-20

EM 1110-2-141731 Aug 94Figure 6-9. Correction for unconnected impervious area (SCS 1986)zoning and hydrologic soil group. The assumption madein deriving these values is that the impervious areas havea CN of 98, and the open areas correspond to pastures ingood condition. Weighting these values with percentimpervious area CN’s when computing the CN for aparticular watershed area would lead to double accountingof the impervious area.6-7. Method Seloolona. The selection of the loss rate method is a functionof the data availability, land use, and the purpose of theloss rate calculation. If a reasonably long gauge record isavailable, then any of the methods discussed will be adequatewhen determining parameter estimates with automaticcalibration techniques. A possible exception is theCN method. The loss rate function implied by themethod is very unappealing and should relegate themethod to a last resort application when using an automaticcalibration technique. However, if the record isinadequate due to record length or data errors, thenmethod selection depends on the preferred parameterestimation approach for ungauged analysis.b. The ungauged analysis parameter estimationapproaches are used alternatively: utilize texture class orparticle size distribution in the Green and Ampt method,utilize USDA classifications for the Holtan method,determine the CN from soil hydrologic group and coverclassification, and calibrate any method, the initial andconstant loss rate method being simplest, to a regionalfrequency curve. Each method has its benefits dependingon the purpose of the calculation and the experience thathas been gained with the method.c. A caution at this point concerning the applicationof the Green and Ampt and Holtan methods to forestedareas is warranted. These methods assume an overlandflow-type mechanism which is not entirely appropriate forforested areas where a subsurface mechanism tends tocontrol direct runoff. Applications to forested areas probablyshould rely on empirical methods calibrated toregional information such as regional frequency curves orcorrelation between observed rainfall-runoff characteristicsand watershed characteristics as is done by the CNmethod.6-21

EM 1110-2-141731 Aug 946-22

EM 1110-2-141731 Aug 94Chapter 7Precipitation Excess - RunoffTransformation7-1. GeneralThe transformation of precipitation excess to runoff is akey factor in flood-runoff analysis. Two approaches aredescribed. The first employs the unit hydrograph and isbased on the assumption that a watershed, in convertingprecipitation excess to runoff, acts as a linear, timeinvariantsystem. The second approach is based on mathematicalsimulation of surface runoff using the kinematicwave approximation of the unsteady flow equations forone-dimensional open channel flow. In this chapter, thebasis for the approaches, data requirements, calibrationprocedures, and limitations are described.7-2. Runoff SubdivisionThe methods in this chapter treat total runoff (i.e., streamflow)as being composed of two components, directrunoff and base flow. Direct runoff results from precipitationexcess, which is regarded herein as that portion ofstorm precipitation that appears as streamflow during orshortly after a storm. Base flow results from subsurfacerunoff from prior precipitation events and delayed subsurfacerunoff from the current storm. The differencebetween total storm precipitation and precipitation excessis termed losses (or abstractions). This chapter deals withthe calculation of direct runoff, given precipitation excess.Methods for estimating losses are described in Chapter 6.Base flow must be added to direct runoff to obtain totalrunoff. Base flow estimation is treated in Chapter 8.Precipitation includes both rain and snow. The methodsdescribed in this chapter are generally applied to rainfallexcess. However, some applications involve the combiningof rainfall excess with snowmelt excess as thebasis for direct runoff determination. Chapter 5 dealswith snowmelt estimation.7-3. Unit Hydrograph Approacha. Concepts.(1) The unit hydrograph represents direct runoff at theoutlet of a basin resulting from one unit (e.g., 1 in.) ofprecipitation excess over the basin. The excess occurs ata constant intensity over a specified duration. Assumptionsassociated with application of a unit hydrograph arethe following:(a) Precipitation excess and losses can be treated asbasin-average (lumped) quantities.(b) The ordinates of a direct runoff hydrographcorresponding to precipitation excess of a given durationare directly proportional to the volume of excess (assumptionof linearity).(c) The direct runoff hydrograph resulting from agiven increment of precipitation excess is independent ofthe time of occurrence of the excess (assumption of timeinvariance).(2) Difficulties associated with the first assumptioncan be alleviated by dividing a basin into subbasins sothat the use of lumped quantities is reasonable. Becauserunoff response characteristics of watersheds are notstrictly linear, the unit hydrograph used with a particularstorm hyetograph should be appropriate for a storm ofthat magnitude. Hence, unit hydrographs to be used withlarge hypothetical storms should, if possible, be derivedfrom data for large historical events. In some cases, it isappropriate to adjust a unit hydrograph to account foranticipated shorter travel times for large events. Theduration of precipitation excess associated with a unithydrograph should be selected to provide adequate definitionof the direct runoff hydrograph. Generally, aduration equal to about one-fifth to one-third of the timeto-peakof the unit hydrograph is appropriate.b. Unit hydrograph application and derivation.Application of a unit hydrograph may be performed withthe following equation:whereQ(t)ni 1u[ t o,t (i 1)]I it(7-1)Q(t) = ordinate of direct runoff hydrograph attime tu( t o ,t) = ordinate at time t of unit hydrograph ofduration t oI i = intensity of precipitation excess during blocki of stormn = total number of blocks of precipitationexcess7-1

EM 1110-2-141731 Aug 94Such application is represented graphically in Figure 7-1.The individual direct runoff responses to each block ofprecipitation excess are superimposed to produce the totaldirect runoff.(1) The development of a unit hydrograph for a basinproceeds differently depending on whether a basin isgauged or ungauged. Gauged, in this case, means thatthere is a stream gauge at the basin outlet for which measurementsduring historical storms are available, and datafrom precipitation gauges are available with which hyetographsof basin-average precipitation can be developed forthe storms. Unit hydrographs can be developed and verifiedwith such data, as discussed later in this chapter.(2) For ungauged basins, direct development of a unithydrograph is not possible and techniques for estimating aunit hydrograph from measurable basin characteristics areemployed. Generally, a unit hydrograph is representedmathematically as a function of one or two parameters,and these parameters are related to basin characteristics byregression analysis or other means. Several methods forrepresenting unit hydrographs are described in the nextsection. Chapter 16, “Ungauged Basin Analysis,” discussesthe use of regional analysis for estimating unithydrograph parameters for ungauged basins.c. Synthetic unit hydrographs. Many methods havebeen devised for representing a unit hydrograph as afunction of one or two parameters. The methods can becategorized as those that are strictly empirical and thosethat are based on a conceptual representation of basinrunoff. The five methods described subsequently are theSingle-linear Reservoir, Nash, Clark, Snyder, and SCS.The first three employ conceptual models of runoff; thelatter two are empirical.(1) Single-linear reservoir method. Conceptual modelscommonly employ one or more linear reservoirs aselements. A linear reservoir is a reservoir for whichthere is a linear relationship between storage and outflow:SwhereKO(7-2)K has units of time and is constant for a linear system.(a) A very simple conceptual model would representthe direct runoff from a basin with a single-linear reservoir(SLR). If such a reservoir is filled instantaneouslywith one unit of volume (i.e., representing one unit ofdepth over the basin area) and the reservoir is permittedto drain, it can be shown that the equation for the outflowis:O(t)1K etK(7-3)Figure 7-2 illustrates a single-linear reservoir and theoutflow hydrograph.(b) The above equation represents an instantaneousunit hydrograph (IUH) for the basin because the duration( t o ) of precipitation excess is zero. The IUH can beconverted to a unit hydrograph of finite duration by superposingseveral IUH’s initiated at equal subintervals of aninterval equal to the duration t o and dividing the aggregatedirect runoff by the number of IUH’s. If t o is sufficientlysmall (as is normally the case to provide adequatedefinition to a direct runoff hydrograph), the finite-durationunit hydrograph can be developed by simply averagingthe ordinates of two IUH’s that are separated in timeby t o .(c) A unit hydrograph developed with the SLRmodel involves a single parameter, K. That is, once avalue for K is specified, the unit hydrograph can be determined.This simple model is useful for small basins withshort response times.(2) Nash model. The Nash conceptual model (Nash1957) represents the direct runoff response of a basin bypassing a unit volume of water through a series of identicallinear reservoirs, as depicted in Figure 7-3. As withthe SLR, the unit volume enters the upstream-most reservoirinstantaneously. The outflow from the downstreammostreservoir is the IUH for the basin. The equation forthe IUH is:O(t)1K(n 1)!⎛ t ⎞⎜ ⎟⎝ K ⎠n 1etK(7-4)SKO= volume of water in storage in the reservoir= storage coefficient= rate of outflow from the reservoirA unit hydrograph based on the Nash model has twoparameters: the number of reservoirs, n, and the storagecoefficient, K, which are identical for each reservoir. Themodel is widely used both for unit hydrograph developmentand for streamflow routing.7-2

EM 1110-2-141731 Aug 94Figure 7-1. Superposition of direct runoff hydrographsrouted through a single linear reservoir, and the resultingoutflow represents the IUH for the basin. Figure 7-4illustrates the components of the Clark method.Figure 7-2. Single-linear reservoir model(3) Clark model. The Clark conceptual model (Clark1945) differs from the SLR and Nash models in thateffects of basin shape (and other factors) on time of travelcan be taken into account. As with the previous models,a unit of precipitation excess occurs instantaneously overthe basin. A translation hydrograph at the basin outlet isdeveloped by translating (lagging) the excess based ontravel time to the outlet. The translation hydrograph is(a) The translation hydrograph can be convenientlyderived from a time-area relation, for which area is theaccumulated area from the basin outlet, and time is thetravel time as defined by isochrones (contours of constanttime-of-travel). Such a relationship can be expressed indimensionless form with area as a percent of total basinarea and time as a percent of time of concentration (t c ).The translation hydrograph can be obtained by determiningfrom a time-area relation the portion of the basin thatcontributes runoff at the outlet during each time intervalafter the occurrence of the instantaneous burst of unitexcess. The contributing area associated with a timeinterval (times the unit depth and divided by the timeinterval) yields an average discharge. This is the ordinateof the translation hydrograph for that interval.(b) Isochrones for use in defining the translationhydrograph may be developed by estimating, for a numberof points in the basin, overland flow and channeltravel times to the basin outlet. A simpler approach is toassume a constant travel velocity and base the position ofisochrones on travel distance from the basin outlet, in7-3

EM 1110-2-141731 Aug 94Figure 7-3. Cascade of linear reservoirs (Nash model)which case the translation hydrograph reflects only basinshape.C aR∆t.5∆t(c) An even simpler approach is to use a translationhydrograph that is based on a standard basin shape, suchas an ellipse. For many basins, storage effects representedby the linear reservoir cause substantial attenuationof the translation hydrograph such that the IUH is notvery sensitive to the shape of the translation hydrograph.However, for a basin without a substantial amount ofnatural storage, such as a steep urban basin, the IUH willbe much more sensitive to the shape of the translationhydrograph. For such a basin, the use of a standard shapemay not be appropriate.(d) The routing of the translation hydrograph througha linear reservoir is based on simple storage routing bysolving the continuity equation. An equation for therouting is:andwhereC b1 C aO(t) = ordinate of IUH at time ttI = ordinate of translation hydrographfor interval t -1totR = storage coefficient for linear reservoir= time interval with which IUH is definedO(t) C aI C bO(t 1)The coefficients C a and C b are defined by:(7-5)The two parameters for the Clark method are T c , the timeof concentration (and time base for the translation hydrograph),and R, the storage coefficient for the linear7-4

EM 1110-2-141731 Aug 94Figure 7-4. Clark modelreservoir. Values for these, along with a time-area relation,enable the determination of a unit hydrograph.(e) To calculate direct runoff, the IUH can be convertedto a unit hydrograph (UH) of finite duration. Derivationof a UH of specified duration from the IUH isaccomplished using techniques similar to those employedto change the duration of a UH. For example, if a 2-hourUH is required, a satisfactory approximation may beobtained by first summing the ordinates of two instantaneousunit hydrographs, one of which is lagged 2 hr.This sum represents the runoff from 2 in. of excess precipitation;to obtain the required UH, the ordinates mustbe divided by 2. This procedure is illustrated inFigure 7-5.requires, as a final step, the fitting of a curve (i.e., theunit hydrograph) that has an underlying area consistentwith a unit depth over the basin area.(a) The principal equations of the Snyder methodfrom which the peak of the unit hydrograph can bedefined are:andt lC tLL ca0.3Q p640C pAt l(7-6)(7-7)(4) Snyder method. The Snyder method (Snyder1938) provides equations that define characteristics of theunit hydrograph directly without the use of a conceptualmodel. Equations have been developed to define thecoordinates of the peak and the time base of the unithydrograph. Empirical procedures for defining the unithydrograph width at 50 and 75 percent of the peak dischargehave also been developed. Use of this methodwheret lC t= lag of the “standard” unit hydrograph,in hours= empirical coefficient7-5

EM 1110-2-141731 Aug 94Figure 7-5. Conversion of IUH to UH with specific duration7-6

L = length of main watercourse frombasin outlet to upstream boundary ofbasin, in milest lt REM 1110-2-141731 Aug 94= original unit hydrograph lag time, in hours= desired unit hydrograph duration, in hoursL caQ pC pA= length of main watercourse from basin outlet topoint opposite centroid of basin area, in miles= peak discharge of “standard” unit hydrograph,in cubic feet per second= empirical coefficient= basin area, in square milesThe “standard” unit hydrograph is one for which thefollowing relation holds:t rwheret rt lt l5.5= duration of unit excess(7-8)= time from the center of mass of the unit excessto the time of the peak of the unit hydrographThe time at which the peak of the unit hydrograph occursis therefore t l + t r /2. Thus, the above equations can beused to determine the coordinates of the peak of the“standard” unit hydrograph in terms of two empiricalcoefficients, C t and C p . The following equations can beused to develop the coordinates of the peak of a unithydrograph of any other duration, t R :Q pwheret lRt lRt l0.25(t Rt r)640C pAt lR(7-9)(7-10)= adjusted lag time for unit hydrograph of durationt R , in hourst rQ p= original unit hydrograph duration, in hours= peak discharge of unit hydrograph of durationt REquations for the time base and widths of the unit hydrographare available in several publications (Snyder 1938and Chow, Maidment, and Mays 1988).(b) The original development of this method andvalues for the coefficients C t and C p were made with datafrom the Appalachian Mountain region. Subsequentapplications in other regions produced values for thecoefficients that were substantially different. The coefficientsshould be calibrated with data from the region inwhich they will be applied. Indeed it is not necessary toadopt the form of the original equation for t l ; regressionanalysis can be used to develop expressions for t l and C pthat take into account measurable basin characteristicsother than L and L ca . For example, the variable (LL ca /S ½ ),where S is the slope of the main watercourse, has beenfound useful as an independent variable in relations for t l .According to a number of studies, C p tends to be fairlyconstant in a region.(5) SCS dimensionless unit hydrograph. The SCSdimensionless unit hydrograph (Mockus 1957), which isshown in Figure 7-6, was derived from a large number ofunit hydrographs developed with data from small ruralbasins. The ordinates are expressed as a ratio of the peakdischarge, and the time scale is expressed as a ratio of thetime-to-peak. The time base of the unit hydrograph isfive times the time-to-peak.(a) A characteristic of the dimensionless unit hydrographis that 37.5 percent of the area under the hydrographoccurs from the origin to the peak. The rising limbof the hydrograph is well represented by a straight line.The following equation is based on an expression for thearea of a triangle defined by a linear representation of therising limb and a vertical line from the peak to the x-axis:Q p484 A(7-11)t p7-7

EM 1110-2-141731 Aug 94Figure 7-6. SCS dimensionless unit hydrographwhereQ pt pA= peak discharge of unit hydrograph,in cubic feet per second= basin area, in square miles= time-to-peak of unit hydrograph, in hours(b) A change in volume under the rising limb of theunit hydrograph would be reflected in a change in the“constant” represented by 484 in the above equation.Studies have indicated that the constant varies from about600 for basins with steep slopes to 300 for flat swampybasins. Figure 7-6 is based on the constant 484. Toutilize a constant of 300 or 600, a completely new dimensionlesshydrograph must be developed.7-8

The time-to-peak may be expressed as:Dt pt2 l(7-12)EM 1110-2-141731 Aug 94parameters, there is substantial flexibility for fitting awide variety of runoff responses. Similar plots could bedeveloped with the Nash and Snyder methods.whereDt l= duration of unit excess for unit hydrograph= lag, defined as the time from the centroid ofprecipitation excess to the time of the peak ofthe unit hydrographThe SCS developed the following empirical relationbetween t l and time of concentration:t l0.6t C(7-13)where t C = time of concentration. Thus, if the time ofconcentration for a basin can be estimated, the aboveequation can be used to estimate lag, and the precedingtwo equations can be used to determine the time-to-peakand peak discharge. The coordinates of the dimensionlessunit hydrograph can then be used to completely determinethe unit hydrograph.d. Choice of synthetic unit hydrograph method. Thepreceding section describes five methods for defining aunit hydrograph in terms of parameters. The SLR methodemploys one parameter, a storage coefficient for a linearreservoir. The SCS method is a one-parameter method ifthe value of 484 is adopted for the constant in the equationfor peak discharge. The Nash, Clark, and Snydermethods each employ two parameters, and the Clarkmethod, in addition, requires a time-area relation.(2) If the SCS dimensionless unit hydrograph isapplied as a one-parameter method (by adopting a constantof 484 in the equation for peak discharge), the resultis as shown in Figure 7-7 for the given basin area andtime of concentration. In this case, the unit hydrograph isapproximately equivalent to a Clark unit hydrograph correspondingto a value for R/(t C + R) of about 0.25. Useof a one-parameter unit hydrograph can be very limitingwith respect to ability to fit the runoff response characteristicsof a basin.(3) A number of attempts have been made to relateparameters of a synthetic unit hydrograph to measurablecharacteristics of an observed hydrograph. For example,the time of concentration (t c ) can be estimated as the timefrom the end of a burst of precipitation excess to the pointof inflection on the falling limb of the direct runoffhydrograph. The storage coefficient in the Clark methodcan be estimated by dividing the discharge at the point ofinflection by the slope of the direct runoff hydrograph atthat point. The basis for these estimation procedures isthat, at the point of inflection, inflow to storage hasceased, and from that time on, storage is being evacuated.At the point of inflection, the continuity equation can bestated as:⎛ ⎞ ⎜⎜⎝ dS ⎟⎟⎠O poipoidt(7-14)where the subscript poi indicates “point of inflection.”Since from the storage equation, S = RO, then:(1) Figure 7-7 shows a set of unit hydrographs developedby the Clark method and also a unit hydrographdeveloped with the SCS dimensionless unit hydrograph(based on the 484 constant). The unit hydrographs are fora drainage area of 50 sq mi and a time of concentration of13.3 hr. The parameter that varies for the Clark unithydrographs is the storage coefficient, R. Each of theClark unit hydrographs is labeled with a value for theratio R/(t C + R). This dimensionless ratio has been foundin a number of studies to be fairly constant on a regionalbasis. For a value of this ratio of 0.1, the unit hydrographrises steeply and might be representative of the runoffresponse of an urban basin. For a value of 0.7, the unithydrograph is much attenuated and might be representativeof a flat swampy basin. The point is that with twoO poiSolving for R:R⎛R ⎜⎝dO poidtO poidO poi/dt⎞⎟⎠(7-15)(7-16)7-9

EM 1110-2-141731 Aug 94Figure 7-7. Unit hydrographs by Clark and SCS methods7-10

EM 1110-2-141731 Aug 94Estimates obtained by such methods should not be reliedon too rigorously because the conceptual models are onlycrude approximations at best of real-world phenomena.(4) Any one of the two-parameter methods is adequatefor describing the runoff response of most basins.The choice of method therefore can be based on otherfactors, such as availability of regional relations forparameters, familiarity with a method, or ease of use.Other aspects of the methods should be considered. Forexample, the Snyder method requires explicit curve fitting,and the Clark method permits incorporation of basinshape and timing factors through use of a time-arearelation.e. Unit hydrograph for a gauged basin. A numberof methods have been developed to enable derivation of aunit hydrograph from precipitation and streamflow data.The simplest method involves the analysis of individualstorms for which there are isolated blocks of significantamounts of precipitation excess. After base flow separation,the volume of direct runoff is determined and usedto adjust losses to produce an equivalent volume of precipitationexcess. The duration of precipitation excess isassociated with a unit hydrograph that is obtained bydividing the ordinates of direct runoff by the volume ofdirect runoff expressed as an average depth over thebasin. S-graph methods can be used to convert the unithydrograph of a given duration to one of another duration.The “isolated storm” and S-graph methods are describedin basic hydrology textbooks.(1) Matrix methods. A unit hydrograph can bederived from a complex storm (for which there are severalblocks of precipitation excess) by matrix methods.The first step is to perform a base flow separation on theobserved hydrograph and to develop a precipitation excesshyetograph. Equations are written to define the ordinatesof the direct runoff hydrograph as a function of hyetographordinates and (unknown) unit hydrograph ordinates,and these equations are solved with matrix algebra. Linearregression or optimization methods can be used tofacilitate the search for a unit hydrograph that minimizesthe error in the fitted direct runoff hydrograph (Chow,Maidment, and Mays 1988). A problem with such techniquesis that the derived unit hydrograph may have anoscillatory shape or reflect other irregularities, and asmoothing process is commonly required.(2) Optimization of values for unit hydrographparameters. The methods described thus far produce aunit hydrograph defined by its ordinates. Anotherapproach is to use a synthetic hydrograph technique andassociated parameters to represent the unit hydrograph.The problem then becomes one of finding values for theparameters, generally using trial and error procedures withdata from complex storms. The objective of such proceduresis to obtain parameter values that enable a “best fit”of the observed hydrograph.(a) Optimization methods have been developed forautomated estimation of values for parameters. Suchmethods can optimize values for loss rate parameterssimultaneously with values for unit hydrograph parameters(Ford, Merris, and Feldman 1980). A general scheme isshown in Figure 7-8. A quantitative measure of “bestfit,” termed an objective function, is calculated with eachtrial set of parameters. The optimization scheme isdesigned to adjust parameter values in such a way thatminimization of the objective function is achieved. Onesuch objective function is:FwhereQ obsQ compand⎡⎢⎣ni 1⎤2 Q obsiQ WTicompi ⎥n ⎦F = objective functioni1/2= ordinate of observed hydrograph= ordinate of computed hydrograph= ordinate number(7-17)n = total number of ordinates over which objectivefunction is evaluatedWT iQ obsiQ avg2 Q avg(7-18)where Q avg = average of the observed-hydrograph ordinates.The purpose of the weighting function, WT i ,istoweight deviations between observed and computed ordinatesmore heavily for higher observed discharges. Thiswill tend to produce a relatively good fit for high7-11

EM 1110-2-141731 Aug 94discharges compared with low discharges, which is generallydesired in flood-runoff analysis.(b) Optimization procedures also require initial valuesfor parameters and constraints that define the acceptablerange of magnitude of each parameter. The results ofoptimization should be reviewed carefully, both withrespect to the success of the optimization and the reasonablenessof optimized values.(3) Procedure for unit hydrograph development. Aprocedure for developing a unit hydrograph for a gaugedbasin is given below. It is presumed that the analysis willbe performed with the aid of a computer program that hascapabilities for optimizing values of runoff parameters.(a) Obtain precipitation and discharge data for historicalstorms. It is desirable for the storms to be ofcomparable magnitude to those to which the unit hydrographwill be applied. In the case of application withvery large hypothetical storms, data for the largest stormsof record will be the most useful. Ideally, it would bedesirable to calibrate values for unit hydrographparameters for about five storms and to verify the adoptedvalues with data from about three additional storms.(b) Determine initial streamflow conditions for eachhistorical event and appropriate values for parameters withwhich to define base flow. Select an appropriate methodfor representing losses and a synthetic unit hydrographmethod. Choice of methods will be dependent on thecapabilities of computer software to be used for theanalysis.(c) Perform an optimization of values for loss andunit hydrograph parameters for each storm that has beenselected for calibration. Carefully review optimizationresults and verify that optimized values are reasonable.Extend the analysis (for example with different initialvalues) as appropriate.(d) Based on a review of values for unit hydrographparameters that have been optimized for each calibrationevent, adopt a single set of values. Factors to consider inadopting values include the quality of fit of an observedhydrograph and the magnitude of the event. Events forwhich only a poor fit was possible would be given lessweight in the adoption process. If some events are substantiallylarger than others, these might be given moreweight, if the adopted values are intended for use withlarge events. The adopted values should then be used tocalculate hydrographs for all of the calibration events, andthe results evaluated. Additional adjustment of the valuesmight be warranted to achieve the most satisfactory fittingof the events.(e) If additional events for verification are available,the adopted values should be used to calculate hydrographsfor these events. The quality of results will be ameasure of the reliability of the adopted values. Additionaladjustment of the values may be appropriate.7-4. Kinematic Wave Approacha. Concepts. The application of the kinematic wavemethod differs from the unit hydrograph technique in thefollowing manner. First, the method takes a distributedview of the subbasin rather than a lumped view, like theunit hydrograph approach. The distributed view pointallows the model to capture the different responses fromboth pervious and impervious areas in a single urbansubbasin. Second, the kinematic wave technique producesa nonlinear response to rainfall excess as opposed to thelinear response of the unit hydrograph.(1) When applying the kinematic wave approach tomodeling subbasin runoff, the various physical processesof water movement over the basin surface, infiltration,flow into stream channels, and flow through channelnetworks are considered. Parameters, such as roughness,slope, area, overland lengths, and channel dimensions areused to define the process. The various features of theirregular surface geometry of the basin are generallyapproximated by either of two types of basic flow elements:an overland flow element, or a stream- or channel-flowelement. In the modeling process, overland flowelements are combined with channel-flow elements torepresent a subbasin. The entire basin is modeled bylinking the various subbasins together.(2) In a typical urban system, as shown in Figure7-9, rain falls on two types of surfaces: those that areessentially impervious, such as roofs, driveways, sidewalks,roads, and parking lots; and pervious surfaces,most of which are covered with vegetation and havenumerous small depressions which produce local storageof rainfall. The contribution to the flood hydrograph ofopen areas (pervious surfaces) is characteristically differentthan that from impervious areas. An obvious differenceis that the open areas can infiltrate rainfall whereasthe impervious areas do not infiltrate significant amounts.A less obvious difference is that the open areas are notsewered as heavily as impervious areas, making for longeroverland flow paths to major conveyances such as openchannels and storm sewers. Furthermore, the open areas7-12

EM 1110-2-141731 Aug 94Figure 7-8. Procedure for parameter optimizationhave hydraulically rougher surfaces which impedes theflow to a greater extent than the relatively smoother surfaceof the paved areas. The overall impact of thesedifferences is to cause the runoff from the imperviousareas to have significantly shorter times of concentration,larger peak discharges, and volumes per unit area thanfrom open (pervious) areas.(3) The lumped approach to modeling this type ofbasin (Figure 7-9) would average the runoff characteristicsof both the open and impervious areas into one unithydrograph. In performing this averaging operation, thepeak runoff response of the basin will normally be underestimatedwhen the impervious area is the dominant contributorto the runoff hydrograph. The main advantage ofthe kinematic wave method is that the response of boththe open and impervious areas can be accounted for in asingle subbasin.b. The kinematic wave equations of motion. Thekinematic wave equations are based on the conservationof mass and the conservation of momentum. The conservationprinciples for one-dimensional open channel flow(St. Venant equations) can be written in the followingform:Conservation of massInflow - outflow = the rate in change of channel storageA ∂V∂xVB ∂y∂xB ∂y∂tConservation of momentumSum of forces = gravity + pressure + friction= mass x fluid accelerationq(7-19)7-13

EM 1110-2-141731 Aug 94Figure 7-9. Typical urban basin flow pathswhereAVxBS fS o∂y∂xVg∂V∂x1g= cross-sectional flow area= average velocity of water= distance along the channel= water surface width∂V∂t(7-20)The kinematic wave equations are derived from theSt. Venant equations by preserving conservation of massand approximately satisfying conservation of momentum.In approximating the conservation of momentum, theacceleration of the fluid and the pressure forces are presumedto be negligible in comparison to the bed slope andthe friction slope. This reduces the momentum equationdown to a balance between friction and gravity:S fKinematic wave conservation of momentumFrictional forces = gravitational forcesS o(7-21)ytqS fS og= depth of water= time= lateral inflow per unit length of channel= friction slope= channel bed slope= acceleration due to gravityThis equation states that the momentum of the flow canbe approximated with a uniform flow assumption asdescribed by Manning’s and Chezy’s equations.Manning’s equation can be written in the following form:QαA m(7-22)7-14

EM 1110-2-141731 Aug 94where α and m are related to surface roughness and flowgeometry. Since the momentum equation has beenreduced to a simple functional relationship between areaand discharge, the movement of a floodwave is describedsolely by the continuity equation. Therefore, the kinematicwave equations do not allow for hydrograph diffusion(attenuation). Hydrographs routed with the kinematicwave method will be translated in time but will not beattenuated. The kinematic wave equations are usuallysolved by explicit or implicit finite difference techniques.Any attenuation of the peak flow that is computed usingthe kinematic wave equations is due to errors inherent inthe finite difference solution scheme. In spite of thislimitation, the kinematic wave approximation is very goodfor modeling overland flow at shallow depths or channelflow in moderately steep channels. Application of thekinematic wave equations to a combination of overlandflow and channel flow elements is often used in urbanwatershed modeling.c. Basin representation with kinematic wave elements.The contribution to the flood hydrograph fromopen and impervious areas within a single subbasin ismodeled in the kinematic wave method by using differenttypes of elements as shown in Figure 7-10.(1) The kinematic wave elements shown are an overlandflow plane, collector, and main channel. In general,subbasin runoff is modeled with kinematic wave elementsby taking an idealized view of the basin. Rather thantrying to represent every overland flow plane and everypossible collector channel, subbasins are depicted withoverland flow planes and channels that represent the averageconditions of the basin. Normally two overland flowplanes are used, one to represent the pervious areas andone to represent the impervious areas. The lengths,slopes, and roughnesses of the overland flow planes arebased on the average of several measurements madewithin the subbasin. Likewise, collector channels arenormally based on the average parameters of severalcollector channels in the subbasin.(2) Various levels of complexity can be obtained bycombining different elements to represent a subbasin.The simplest combination of elements that could be usedto represent an urban subbasin are two overland flowplanes and a main channel (Figure 7-11). The overlandflow planes are used to separately model the overlandflow from pervious and impervious surfaces to the mainchannel. Flow from the overland flow planes is input tothe main channel as a uniform lateral inflow. Urbanwatersheds typically have various levels of storm sewers,man-made channels, and natural streams. To modelcomplex urban systems in a manageable fashion, theconcept of typical collector channels must be employed.As shown in Figure 7-12, the complexity of an urban subbasincan be modeled by combining various levels ofchannel elements. An idealized overland flow, subcollector,and collector system are formulated from averageparameters in the subbasin. The runoff contributingto the idealized collector system is assumed to be typicalof the subbasin. The total runoff is obtained by multiplyingthe runoff from the idealized collector system by theratio of the total subbasin area to the contributing area tothe collector system. The total runoff is then distributeduniformly along the main channel and routed to the outlet.d. Estimating kinematic wave parameters. Althoughthe kinematic wave equations are used to route flowthrough both the overland flow planes and channels,different types of data are needed for each elementbecause of differences in characteristic depths of flow andgeometry. The depth of flow over an overland flow planeis much shallower than in the case of a channel. Thisresults in a much greater frictional loss for overland flowthan for channel flow. Frictional losses are accounted forin the kinematic wave equations through Manning’sequation. Typical roughness coefficients for overlandflow are about an order of magnitude greater than forchannel flow. The overland flow roughness coefficients(Table 7-1) will range between 0.1 and 0.5 depending onthe surface cover; whereas the roughness coefficients forchannel flow are normally in the range of 0.012 to 0.10.(1) The estimation of kinematic wave parameters foreach element is an exercise in averaging slopes, lengths,roughness coefficients, and even geometry. The data forthe various kinematic wave parameters can be obtainedfrom readily available topographic, soil, sewer, and zoningmaps, as well as tables of roughness coefficients. Thefollowing data are needed for each overland flow plane:(a)(b)Average overland flow length.Representative slope.(c) Average roughness coefficient (Table 7-1).(d)(e)The percentage of the subbasin area which theoverland flow plane represents.Infiltration and loss rate parameters.Overland flow lengths for impervious surfaces are typicallyshorter than those for pervious surfaces. Imperviousoverland flow lengths range from 20 to 100 ft, while7-15

EM 1110-2-141731 Aug 94Figure 7-10. Kinematic wave elements7-16

EM 1110-2-141731 Aug 94Figure 7-11. Simple kinematic wave subbasin modelpervious overland flow lengths can range from 20 toseveral hundred feet. Overland and channel slopes can beobtained from topographic maps. Overland slopes, aswell as collector channels, should be taken as the averagefrom several measurements made within a subbasin. Themain channel slope can be measured directly. Loss rateparameters must be specified for each overland flowplane. Loss rates for impervious areas are generallyrestricted to a small initial loss to account for wetting thesurface and depression storage. Loss rates for perviousareas are based on the soil types and surface cover. Estimatingthe percent of the subbasin that is actually imperviousarea can be quite difficult. For example, in someareas roof top downspouts are hydraulically connected tothe sewers or drain directly to the driveway; whereas inother areas the downspouts drain directly into flower bedsor lawns. In the former situation, the roof top acts as animpervious area, and in the latter, as a pervious area.(2) The following data are needed to describe collectorand sub-collector channels as well as the mainchannel:(a)Representative channel length.(b) Manning’s n.(c)(d)(e)Average slope.Channel shape.Channel dimensions.(f) Amount of area serviced by the channel element.For collector and subcollector channels, the representativelength and slope is based on averaging the lengths ofseveral collectors and subcollectors within the basin. Themain channel length and slope should be measureddirectly from topographic maps. Manning’s n values canbe estimated from photos or field inspection of thechannels. Channel shapes and dimensions are usuallyapproximated by using simple prismatic geometry asshown in Table 7-2. Collector and subcollector channelsshould be based on the average of what is typical withinthe subbasin. The main channel shape and dimensionsshould be approximated as best as possible with one ofthe prismatic elements shown in Table 7-2.e. Basin modeling. The assumptions made using thekinematic wave approach to model a river basin areessentially the same as those made when applying the unit7-17

EM 1110-2-141731 Aug 94Figure 7-12. Kinematic wave representation of an urban subbasin7-18

EM 1110-2-141731 Aug 94Table 7-1Roughness Coefficients for Overland Flow (from Hjelmfelt 1986)Surface N Value SourceAsphalt & concrete 0.05 - 0.15 aBare packed soil free of stone 0.10 cFallow - no residue 0.008 - 0.012 bConventional tillage - no residue 0.06 - 0.12 bConventional tillage - with residue 0.16 - 0.22 bChisel plow - no residue 0.06 - 0.12 bChisel plow - with residue 0.10 - 0.16 bFall disking - with residue 0.30 - 0.50 bNo till - no residue 0.04 - 0.10 bNo till (20 - 40 percent residue cover) 0.07 - 0.17 bNo Till (60 - 100 percent residue cover) 0.17 - 0.47 bSparse rangeland with debris:0 percent cover 0.09 - 0.34 b20 percent cover 0.05 - 0.25 bSparse vegetation 0.053 - 0.13 fShort grass prairie 0.10 - 0.20 fPoor grass cover on moderately roughbare surface 0.30 cLight turf 0.20 aAverage grass cover 0.40 cDense turf 0.17 - 0.80 a,c,e,fDense grass 0.17 - 0.30 dBermuda grass 0.30 - 0.48 dDense shrubbery and forest litter 0.40 aa) Crawford and Linsley (1966).b) Engman (1986).c) Hathaway (1945).d) Palmer (1946).e) Ragan and Dura (1972).f) Woolhiser (1975).hydrograph technique. Rainfall is assumed to be uniformover any subbasin and there are no backwater effects inchannel routing. The assumption that there are no backwatereffects has some important ramifications for interpretingthe kinematic wave results. Although the channelelements can be used to represent pipe elements, the pipesnever pressurize. The kinematic wave equations are foropen channel flow and cannot represent the effects ofpressure flow.(1) This is not a severe limitation when applying thekinematic wave method for design purposes. Generallyspeaking, sewer systems are designed to convey flow asan open channel. However, in situations where the sewersystem will pressurize, flow will back up into the streetgutters and flow to the nearest low point where it mayenter the sewer system again. In the case where a culvertor a storm sewer pressurizes and creates a largebackwater, the backwater area should be modeled separatelywith a technique that can handle pressure flow.(2) The use of the kinematic wave method for mainchannels and large collector’s should be limited to urbanareas or moderately sloping channels in headwater areas.The limitation results because a hydrograph’s peak dischargedoes not attenuate when it is routed with the kinematicwave technique. This is an adequate approximationin urban areas, or any small, quick responding basin.However, flood waves generally attenuate in most naturalchannels. Consequently, the kinematic wave method willtend to overestimate peak discharges in this type ofstream. Therefore, in natural streams, where it is likelythat hydrograph attenuation will occur, the kinematicwave method should not be used for routing. Alternativerouting methods that can account for attenuation, such asthe Muskingum-Cunge method, should be applied instead.7-19

EM 1110-2-141731 Aug 94Table 7-2Prismatic Elements for Kinematic Wave Channels7-20

EM 1110-2-141731 Aug 94Chapter 8Subsurface Runoff Analysis8-1. Generala. Subsurface runoff analysis considers the movementof water throughout the entire hydrologic cycle(Figure 8-1). The prediction of subsurface runoff is performedwith models of varying complexity depending onthe application requirements and constraints. The modelsused may be categorized as event-oriented or continuoussimulation. Event-oriented models, which have been thefocus of the previous chapters, utilize relatively simpletechniques for estimating subsurface contributions to aflood hydrograph.b. Continuous simulation models continuouslyaccount for the movement of water throughout the hydrologiccycle. Continuous accounting of water movementinvolves the consideration of precipitation, snow melt,surface loss, infiltration, and surface transport processesthat have been discussed previously. Other processes thatneed to be considered are evapotranspiration, soil moistureredistribution, and groundwater transport. The integrationof all these processes in a watershed model is usuallytermed as continuous soil moisture accounting (SMA).The complexity of the SMA model varies greatly dependingon the degree of conceptualization employed in integratingthe subsurface processes.c. Historically, the representation of soil moistureredistribution and subsurface flows has been highly conceptualizedin SMA algorithms by an interconnectedsystem of storages. More recently, a more distributed orsmaller scale representation has been attempted (e.g., theSysteme Hydrologique European SHE model, Abbott,et al. 1986). These models represent overland and subsurfaceflow with finite difference approximations to theSt. Venant and Darcy equations. However, these techniqueshave not yet been widely applied and will not becovered in this manual. Instead, the focus will be on themore highly conceptualized representations of soil moistureredistribution and subsurface flow.d. The purpose of this section is to discuss separatelythe continuous simulation and event orientedapproaches to calculating subsurface flow. A topicimportant to both approaches is hydrograph recessionanalysis. The methods used for event-oriented modelingwill be discussed initially in paragraph 8-2 because recessionanalysis is key to this approach.e. The continuous simulation approach involvesalgorithms that consider a number of processes besideshydrograph recession. Evapotranspiration (ET) is a keyelement in performing continuous simulation. In paragraph8-3, a separate discussion is provided on ETbecause the estimation methods vary greatly. In paragraph8-4, a general discussion is provided of theapproaches used in performing continuous simulation. Inparagraph 8-5, the continuous simulation algorithms usedin public domain models PRMS (U.S. Geological Survey(USGS) 1983) and SSARR (USACE 1987) are presentedfor example purposes. A general discussion of the techniquesthat might be used to estimate parameters in continuoussimulation models is provided in paragraph 8-6.8-2. Event-Oriented Methodsa. Basic model for hydrograph recession modeling.Event-oriented models do not have the capability toaccount for the subsurface water balance. Since the waterbalance is not known, these models use an empiricalapproach to relate model parameters to the recessioncharacteristics of an observed hydrograph. Presumably,the recession of the hydrograph is dominated by subsurfaceresponse at the point where direct runoff from thesurface and near surface ceases. The problem is identifyingthe point at which the direct runoff ceases.(1) The separation of the hydrograph into direct runoffand subsurface response is termed base-flow separation.Base-flow separation methods assume a very simplemodel for the watershed geometry (Figure 8-2). Thewatershed response is assumed to be a sum of directrunoff and base flow due to aquifer discharge. The keyassumption is that the aquifer is homogenous with a singlecharacteristic response. This characteristic responseshould be identifiable from the hydrograph recession.(2) The assumed characteristics of the base-flowrecession are based on simplified equations for flow in aphreatic aquifer. The equations are obtained (Bear 1979)by applying the Dupuit-Forcheimer assumptions to acombination of Darcy’s Law and conservation of masswhich is known as the Boussinesq equation. Theseassumptions require the approximation that flow in theaquifer is essentially horizontal.(3) The Boussinesq equation relates the spatialchange in the square of the phreatic water surface elevationin space to its change in time. Interestingly, theBoussinesq equation results in no approximation in calculatedaquifer discharge to a stream, despite the assumption8-1

EM 1110-2-141731 Aug 94Figure 8-1. The hydrologic cycleof horizontal flow. However, the equation does not preservethe description of the phreatic surface of the aquifer.(4) Linearization of the Boussinesq equation for onedimensional(1-D) flow results in the following differentialequation for aquifer discharge (Figure 8-2):whereT ∂2 h∂x 2S ∂h∂t(8-1)h = phreatic surface or water table height froman arbitrary datum in the aquifer as a functionof position xS = aquifer storativityt = timeSolution of this equation for the recession portion of thebase flow or hydrograph or equivalently for a fallingphreatic surface in an aquifer is of the form:T = average aquifer transmissivity8-2

EM 1110-2-141731 Aug 94Figure 8-2. Simple groundwater model for recession analysiswhereh(0,t) =h(0,t)∼ Ce αtC =α =(8-2)height of the aquifer phreatic surface at thestream interfaceconstant depending on x, aquifer geometry andinitial position of the phreatic surface(T/S)Since the groundwater discharge is proportional to theslope of the phreatic surface given the Dupuit-Forcheimerassumptions, the aquifer discharge or base-flow recessionwill also decay exponentially. Note that the decrease inflow with time or the recession is proportional to anexponential decay.(5) The expected exponential decay in discharge isused to identify the point at which the base-flow recessionbegins. The standard technique is to plot log Q versustime and to determine the point at which the recessionbecomes a straight line.recession limb of the base-flow hydrograph (Figure 8-3).Techniques for determining the rising limb of the baseflowhydrograph vary widely. Viessman et al. (1977)describes various approaches to this problem. As anexample, the approach used in the HEC-1 watershedmodel (USACE 1990a) will be discussed in this section.(1) The HEC-1 model provides means to include theeffects of base flow on the streamflow hydrograph as afunction of three input parameters, STRTQ, QRCSN, andRTIOR. The variable STRTQ represents the initial flowin the river. It is affected by the long-term contributionof groundwater releases in the absence of precipitationand is a function of antecedent conditions (e.g., the timebetween the storm being modeled and the last occurrenceof precipitation). The variable QRCSN indicates the flowat which an exponential recession begins on the recedinglimb of the computed hydrograph. Recession of the startingflow and “falling limb” follow a user specified exponentialdecay rate, RTIOR, which is assumed to be acharacteristic of the basin. RTIOR is equal to the ratio ofrecession limb flow to the recession limb flow occurring1 hr later. The program computes the recession flow Qas:(8-3)Q Q O(RTIOR) n∆t 8-3b. Application of base-flow separation techniques.Base-flow recession analysis characterizes only the

EM 1110-2-141731 Aug 94Figure 8-3. Base-flow separation diagramwhereQ o = STRTQ or QRCSNn∆t = time in hours since recession was initiatedQRCSN and RTIOR can be obtained by plotting the logof observed flows versus time. The point at which therecession limb fits a straight line defines QRCSN and theslope of the straight line is used to define RTIOR. Alternatively,QRCSN can be specified as a ratio of the peakflow. For example, the user can specify that the exponentialrecession is to begin when the “falling limb” dischargedrops to 0.1 of the calculated peak discharge.(2) The rising limb of the streamflow hydrograph isadjusted for base flow by adding the recessed startingflow to the computed direct runoff flows. The fallinglimb is determined in the same manner until the computedflow is determined to be less than QRCSN. At this point,the time at which the value of QRCSN is reached is estimatedfrom the computed hydrograph. From this time on,the streamflow hydrograph is computed using the recessionequation unless the computed flow rises above thebase-flow recession. This is the case of a double-peakedstreamflow hydrograph where a rising limb of the secondpeak is computed by combining the starting flow recessedfrom the beginning of the simulation and the directrunoff.(3) The values for these parameters can be establishedby regionalizing results from gauged basins. As anexample, consider the attempts to determine base-flow8-4

EM 1110-2-141731 Aug 94parameters for the Upper Hudson and Mohawk Rivers inNew York.(4) The starting flow, STRTQ, can be determined byplotting the initial streamflow observed prior to eventsversus drainage area (Figure 8-4). The recession-flowparameters were determined for each event by means ofplotting the recession discharge versus time on semilogpaper (Figure 8-5). QRCSN is the value of the dischargewhere the recession begins to plot as a straight line andRTIOR is related to the slope of this straight line. Figure8-5 is not representative of all efforts to determine therecession parameters. In a significant number ofinstances, a straight line was not easily detectable on thesemilog plot. Note that this study was performed with anolder version of HEC-1 where RTIOR is defined as theratio of the flow to that observed 10 time periods laterrather than 1 time period later as defined in the currentmodel.(5) The results of the analysis indicated that RTIORvaried between 1.1 and 1.7 for the gauges and eventsexamined. Since this range of values does not have alarge affect on the recession limb, an average value of 1.3was assumed for all subbasins. As in the case of STRTQ,QRCSN was graphically related to drainage area as shownin Figure 8-6.8-3. Evapotranspirationa. Introduction. The fundamental water balancerelationship that a continuous simulation model mustsatisfy to accurately represent the hydrologic cycle is:runoff = precipitation - evapotranspirationConsequently, estimating ET is of major importance.This section is dedicated to describing the theory andapplication equations used to estimate ET in continuoussimulation models.b. Basis for computation of evapotranspiration. Asin the case of infiltration, a well developed evapotranspiration(ET) theory exists for ideal conditions, i.e., conditionswhere the properties of the soil and the vegetativecover are well defined. However, the theory, as in thecase of infiltration, is rarely implemented in a watershedmodel because the actual field situation deviates significantlyfrom the ideal conditions assumed in the theory.Instead, the theory is used as a basis to develop manyparametric methods that attempt to capture the essence ofthe evapotranspiration process.(1) The following development is for calculatingpotential evapotranspiration (PET). PET is an estimate ofthe maximum amount of ET that may occur given availablewater. For an open water body, PET and ET areequivalent since the water supply is not limiting. Watersupply is limiting in applications to bare ground or vegetativecover because available soil moisture, conductivityof the soil profile, and/or plant resistance may be limiting.Consequently, PET and ET are not equivalent in soilmoisture accounting algorithms.(2) The various parametric equations used to calculatePET have similarities that can be recognized from arudimentary understanding of evapotranspiration theory.Consequently, the purpose of this section is to describeevapotranspiration theory so that the relationship betweenthe parametric methods used can be related via an overallknowledge of the factors that affect ET.(3) Evaporation theory is most easily developed byconsidering evaporation from a water surface and thenextending these concepts to plant transpiration and evaporationfrom bare surfaces. Diffusion and energy budgetmethods have both been used to compute evaporationfrom a water surface. The diffusion method examines thetransfer of water between water and gaseous states.Water, in a closed system, will evaporate from the watersurface until the water vapor pressure above the surfacereaches the saturation value. At this point, an equilibriumexists between liquid and gaseous phases of water.(4) Practically speaking, equilibrium is not attainedin the field because the atmosphere is unbounded andwind plays a major role in convecting moist air awayfrom the water surface. The diffusion approach modelsthis situation by assuming that a thin film of saturated airabove the water surface is evaporated by convection fromthe wind. The rate at which wind convects water vaporfrom the water surface (the evaporation rate) is determinedbased on thermodynamic and aerodynamic principlesto be proportional to:whereE bu (e se)Eb= evaporation rate= proportionality constant(8-4)8-5

EM 1110-2-141731 Aug 94Figure 8-4. Initial flow versus drainage area Mohawk and Upper Hudson River8-6

EM 1110-2-141731 Aug 94Figure 8-5. Determination of QRCSN and RTIOR for Basin 55, Batten Kill at Battenville, NY, December, 1948 Event8-7

EM 1110-2-141731 Aug 94Figure 8-6. QRCSN versus drainage area for gauged basins, Upper Hudson and Mohawk Basin8-8

EM 1110-2-141731 Aug 94e s = water saturation vaporizationpressuree= vapor pressure at the elevation atwhich u, the wind speed is measuredA s= surface area of the water bodySubstituting Equations 8-7 and 8-8 into Equation 8-6, theevaporation rate is computed as:The diffusion approach is not general because evaporationoccurs in the absence of wind. Consequently, the methodis modified to account for this possibility by adding aconstant so that the evaporation rate is determined by:E(Q iQ r) Q b(Q aQ s)ρ eL eA s(l R)(8-9)E (a bu)(e se)(8-5)where a and b are determined empirically from field data.(5) An alternative approach to computing evaporationis the energy budget approach which computes the rate ofincrease of energy storage within the body, Q s as:Q sQ iQ aQ rQ bQ eQ h(8-6)where the sources and sinks of heat are due to Q i theincoming shortwave radiation from the sun, Q a is the sumof all other sources of heat (due to seepage, rainfall, orother water inflows), Q r reflected shortwave solar radiation,Q b outgoing long wave radiation due to the “blackbody affects,” and Q e is the energy utilized in evaporation(latent heat), and Q h is the conducted and convected heat.This expression can be used to calculate evaporation rateby utilizing the Bowen ratio:RQ hQ e(8-7)and relating the energy used in evaporation to the evaporationrate as:Q eρ eL eEA s(8-8)Application of this equation requires that some measurementof incoming solar radiation is available to estimateQ i and Q r ; and the temperature of the water body and allother inflows of water be known so that the other heatterms can be computed.(6) Penman (1948) combined the best features ofboth the diffusion and energy budget methods to obtain anexpression similar to Equation 8-5, except that the coefficientsa and b are calculable if data are available ontemperature of the water body and net incoming solarradiation.(7) Modification of methods for calculating evaporationfrom water surfaces to vegetative surface requires theconcept of potential evapotranspiration. Unlike waterbodies, water contents available in the soil via plants orbare surfaces may not be sufficient to support the capacityof the atmosphere to retain water. In this case, methodshave been developed to compute the potential evapotranspiration,i.e., the evaporation that would occur if therewere sufficient moisture.(8) The Penman method was modified by Monteith(1965) to compute potential evapotranspiration. Thisrequired that a concept known as diffusion resistance (aresistance to evaporation) be incorporated into the Penmanequation. The resistance to evaporation is divided intocomponents due to atmospheric effects and plant effects.The atmospheric effects are, at least theoretically, calculablefrom thermodynamic and aerodynamic principles.However, the plant effects due to the resistance to moistureflux through plant leaves and the soil must be determinedempirically.whereρ eL e= density of evaporated water= the latent heat of vaporization(9) In summary, the calculation of potential evapotranspirationis based on the theory of evaporation fromwater surfaces. A significant amount of data on windspeed, net influx of solar radiation, temperature, andempirical information is needed for this calculation.8-9

EM 1110-2-141731 Aug 94c. Empirical approaches to calculation potentialevapotranspiration. Numerous empirical approaches forcalculating PET exist. Most basic texts on hydrologysummarize available methods (e.g., Viessman et al. 1977).The difficulty with most of these methods (and with calculationsof ET in general) is that their basis is for openwater bodies rather than land surfaces with vegetativecover.(1) In this section, the empirical methods used byseveral continuous simulation models (PRMS, USGS1983 and SSARR, USACE 1987) are described. PRMSallows the option of using pan evaporation, temperature,or energy-budget methods. The pan evaporation method,probably the most common and popular method for calculatingPET, is estimated as:PETEPAN (EVC (MO))(8-10)whereVDSAT 216.7VPSAT(TAVC 273.3)(8-12)TAVC = mean daily temperature, in degrees CelsiusVPSAT = saturated vapor pressure in millibars atTAVCVPSAT is calculated as:VPSAT 6.108⎡⎤⎢exp ⎛ TAVC ⎞⎜17.26939⎟⎠⎦ ⎥⎣ ⎝ (TAVC 273.3)(8-13)whereEPAN = daily evaporation lossEVC = empirical pan coefficient, less than 1.0, thatvaries monthlyThe energy budget approach by Jensen and Haise (1963)calculates PET by:PETCTS(MO)(TAVF CTX)(RIN)(8-14)The pan coefficient is intended to account for the differencesbetween the thermodynamics of the pan and theprototype (e.g., a reservoir or catchment).whereCTS= coefficient that varies monthlyThe temperature method by Hamon (1961) calculates PETas:TAVF = mean daily temperature, in degreesFahrenheitPETCTS (MO) (DYL 2 )(VDSAT)(8-11)RIN = daily solar radiation, in inches ofevaporationwhereCTS= empirical coefficient that varies monthlyPETCTX= inches per day= coefficient that is a function of humidityand watershed elevationDYL = possible hours of sunshine in units of12 hoursVDSAT= saturated water vapor density at the dailymean temperature in grams per cubic meterCTS is calculated as:CTS [C1 13.0(CH)] 1(8-15)PET= inches per dayVDSAT is computed as (Federer and Lash 1978):8-10

EM 1110-2-141731 Aug 94whereI is the sum of the monthly heat indices:C1= elevation correction factorI (T/5) 1.514(8-20)CH= humidity indexC1 is calculated as:C1 68.0⎡ ⎤⎢3.6 ⎛ E1 ⎞⎜ ⎟⎠⎦ ⎥⎣ ⎝ 1000(8-16)SSARR converts the PET to a daily value and then providesthe capability to adjust this value for snow coveredground, month of the year, elevation of a particular snowband, and for rainfall intensity (i.e., PET is reduced whenit is raining).where E1 = median elevation of the watershed, in feetmsl. CH is calculated as:CH50(e 2e 1)(8-17)where e 2 and e 1 = saturation vapor pressure (mb) forrespectively the mean maximum and minimum air temperaturesfor the warmest month of the year. CTX in Equation8-14 is computed as:CTX 27.5 0.25(e 2e 1)⎛ E2 ⎞⎜ ⎟⎝ 1000 ⎠where E2 = mean elevation for a particular subbasin.(8-18)(2) The SSARR model provides the capability for theuser to supply PET values or calculate a basic PET viathe Thornthwaite (1954) method:PET1.6b (10T/I) a(8-19)(3) In summary, empirical PET methods may bebased on pan evaporation, mean monthly temperature, orenergy budget equations. The pan evaporation approachis probably most popular and is certainly simplest. Afurther discussion of the importance of ET estimation andthe corresponding choice of method will be given in paragraph8-6 on parameter estimation.8-4. Continuous Simulation Approach to SubsurfaceModelinga. Fundamental processes. Continuous simulationmodels attempt to conceptually represent the subsurfacedynamics of water flow. The subsurface flow dynamicscan be separated into wetting and drying phases. In thewetting phase, a wetting front of infiltrated water headsdownward toward the groundwater aquifer as rainfall orsnowmelt falls on the watershed surface. The aquifers ofinterest in this case are termed phreatic in that the aquifersurface is defined by water at atmospheric pressure. Inresponse to this influx of infiltrated water, the groundwaterlevels may rise, if the influx is great enough, andthe rate of water discharging from the aquifer to thestream increases. Streamflow due to aquifer discharge isusually termed base flow. The aquifer may also dischargeto deep percolation depending on the permeability of soilsor bedrock underlying the aquifer.whereTbIaPET= mean monthly temperature= factor to correct for the difference in daysbetween months= annual heat index= cubic function of I= monthly value(1) For the infiltration phase of this process, inChapter 6, the Richards equation describes an infinitelydeep soil profile on infiltration. The consideration ofinfiltration in this instance is complicated because of thetransition between unsaturated flow in the finite thicknesssoil profile and the saturated aquifer flow.(2) The dynamics of the drying phase are not symmetricalwith that of the wetting phase because of theaffects of evapotranspiration and soil hysteresis. Soilhysteresis occurs because the unsaturated hydraulic conductivityis not a unique function of water content. Theusual explanation for this curious behavior is that soil8-11

EM 1110-2-141731 Aug 94pores do not fill and drain in the same sequence. Evaporationalso affects the drying front depending on the vegetativecover and depth of the root zone.(3) At some point during the drying phase, the aquiferlevels must decrease, and the base-flow dischargemust also decrease. This decrease in flow, at least theoretically,can be identified by an exponential decay.(4) The generally accepted method for calculating theflow in this system is to simultaneously solve Richards’equation and Darcy’s Law for a phreatic aquifer. However,this is a rather numerically intense exercise and israrely performed as part of a watershed analysis.(5) As described in Chapter 6, the overall dynamicsof the direct runoff process is rather complicated by anumber of factors. An additional complicating factor thathad not been mentioned previously is the heterogeneity ofthe groundwater aquifer. These heterogeneities make itdifficult to identify the characteristics of the aquiferresponse, particularly the identification of the exponentialdecay of the base flow.(6) In summary, the dynamics of the subsurface processare complex even for an ideal soil profile andaquifer. The dynamics may be modeled using a combinationof Richards’ equation and Darcy’s Law. Practicallyspeaking, this is rarely done in watershed modeling. Theuse of these methods becomes more difficult and impracticalwhen subsurface heterogeneities are considered.b. Conceptual models of subsurface flow. There area multitude of conceptual models that are available toperform continuous moisture accounting. All of thesemodels try to capture the dynamics of subsurface flowwith simple storage elements. As a precursor to discussingany of these models, a useful introduction is to constructa generic model that demonstrates the conceptualnature of the soil-moisture accounting model. Consider amodel that has only rainfall as an input (Figure 8-7). Tobegin with, the storages represent surface effects, unsaturatedzone, and saturated zone or aquifer storages (allstorage shown considers volume in terms of basin-depth,e.g., basin-inches). Consider each zone separately:(1) Surface storage. The surface storage stores waterup to a maximum value of SMAX. Water leaves eitherby evaporation at the potential rate ES, infiltration at arate equal to FS or via an overflow once SMAX isexceeded. The overflow volume might be routed to thestream via the unit hydrograph method.(2) Upper zone storage. The upper zone storeswater up to a maximum value UMAX. Evaporation fromthe zone at the rate EU models the uptake due to vegetation.Water enters the storage at the rate FS and leaveseither by evaporation, infiltration to the lower zone at rateLS, or to the stream via a low-level outlet. If theassumption is made that the upper zone is a linearstorage, then the outflow rate is linearly proportional tothe storage.(3) Lower zone storage. The lower zone storeswater up to a maximum value LMAX. Water enters thestorage from the upper zone at the rate LS and leaves viaa low-level outlet as in the upper zone case or out of thesystem at a deep percolation rate, FD. The computationof the outflow rates is based on the following functions:• Potential evaporation: Compute as a coefficienttimes the pan evaporation amount.• Potential infiltration: The infiltration from onezone to another is based on linearly varying functionof the storage receiving flow:FP FMAX ⎛ V ⎞⎜1⎟⎠⎝ VMAXVS ≤ VMAX(8-21)where FMAX is the maximum infiltration rate into astorage with capacity VMAX and current storage V.• Low-level outlet: the subsurface storages will beconsidered linear reservoirs where the outlet dischargeis computed as:OwhereVKO = outflowK = linear reservoir storage coefficient(8-22)Application of this model to soil moisture accounting andrunoff prediction might be done based on the followingoutflow rule: evaporation takes precedence over infiltrationwhich in turn takes precedence over outflow from alow-level outlet.8-12

EM 1110-2-141731 Aug 94Figure 8-7. Simple example continuous simulation model8-13

EM 1110-2-141731 Aug 94(4) Explicit solution algorithm. An explicit solutionalgorithm would proceed as follows given this rule for theperiod of duration ∆t, or equivalently, between timest i and t i+1 :(a) Surface zone.supply VS as:VS SZ 1RCompute the available surface(8-23)FUP can be calculated simply from the beginning of periodstorage in the upper zone, UZ i . The storage at theend of the period, SZ i+1 , is computed as:or:SZ i 1VSF VSF < SMAXSZ i 1SMAX VSF ≥ SMAX(8-30)(8-31)whereSZ i = storage at the beginning of the periodE VSF SMAX(8-32)R = rainfall volume during the periodThe volume left in storage after evaporation, VSE, iscomputed as:or:VSE VS ESP VS ≥ ESPVSE 0 VS < ESP(8-24)(8-25)where the evaporated volume ES is lost up to the potentialamount ESP if the surface storage is available. The computationof storage, VSF, after infiltration from the surfacezone to the upper zone is computed in a similar manner tothat of evaporation:where E is the excess available if the end of periodstorage exceeds the maximum amount SZM.(b) Upper zone. The soil moisture accounting forthe upper zone proceeds similarly to that of the surfacezone except that outflow is routed based on the linearreservoir outflow relationship. The volume available foroutflow, VU, is:VU UZ iFU(8-33)where UZ i is the beginning of period storage. The volumeleft after evaporation, VUE, is computed as:VUE VU EUP VU ≥ EUP(8-34)VSF VSE FUP VSE ≥ FUP(8-26)EUVUE(8-35)FUFVP(8-27)or:VUE 0 VU < EUP(8-36)or:EUVU(8-37)FU VSE VSE < FUP(8-28)VSF 0(8-29)where FU is the volume infiltrated to the upper zone upto the potential amount FUP if VSE is large enough.where EU is the volume evaporated up to the potentialamount EUP if the storage is available. The volumeremaining, VUF, after infiltration from the upper zone tothe lower zone is computed as:8-14

EM 1110-2-141731 Aug 94VUF VUE FLP VUE ≥ FLP(8-38)or:FD LZ iFL (FL LZ i)

EM 1110-2-141731 Aug 94estimation problem for soil moisture accounting models.Furthermore, the generic model formulation ignored theproblems of surface interception (water that would bestored but not free for outflow or infiltration), snowmeltand snow excess infiltration, partial area or hillslopeeffects, and the routing of base flow through more than alinear reservoir. If these processes were included in themodel, then there would be a significant increase in thenumber of parameters that need to be estimated.(7) Explicit simulation scheme. A second noteworthyaspect of the generic model is the explicit simulationscheme. The explicit simulation scheme can result in apoor simulation if the selected simulation interval, ∆t, isnot appropriately small. For example, computation of theinfiltration loss from one zone to another is dependent onthe beginning of period storage. If the storage changesgreatly over the computation period, then the infiltrationrate computed base on beginning of period storages willbe a poor estimate of the average rate that would occurover the period. Consequently, a computation intervalthat is sufficiently small is needed for accurate numericalsimulation with the model.c. Summary. In summary, the purpose of this sectionwas to introduce the concept of soil moisture accountingvia a description of a simple model. Even though themodel is simple, the number of parameters that must beestimated easily exceeds the number needed for eventoriented estimation. The number of parameters that mustbe estimated poses some very significant parameter estimationproblems.8-5. Existing Continuous Simulation Modelsvegetation, STOR, varies depending on the type of precipitation:winter snow, winter rain, or summer rain.(2) Impervious area. This area represents the fractionof the basin that is impervious. Interception does notoccur, but a surface loss, RETIP, can be specified.(3) Snow pack. The snow pack is assumed touniformly cover the entire basin. The assumption is madethat it is a two-layer system, the surface layer being 3 to5 in. thick. Melt water from the pack is proportionedbetween the pervious and impervious area based on thefraction of the area.(4) Soil zone reservoir. This reservoir represents theactive portion of the soil profile in that soil moistureredistribution is modeled. The capacity of this zone,SMAX, is defined as the difference between the fieldcapacity and wilting point (field capacity is a looselydefined concept being generally defined as the watercontent of the soil after gravity drainage for someextended period from near saturation; the wilting pointdefines the water content at which plants can no longerextract moisture from the soil). The zone is divided intoa recharge zone, capacity REMAX, and lower zone withcapacity LZMX (necessarily the difference betweenSMAX and REMAX). The recharge zone must be fullbefore water can move to a lower zone.(5) Subsurface zone. This zone represents the flowfrom the soil’s unsaturated zone to the stream and groundwaterreservoir. The outflow to stream is based on therelationship:a. Introduction. There are many different continuoussimulation models available which employ differentsoil moisture accounting algorithms. As examples of soilmoisture accounting techniques, two models in the publicdomain, PRMS (USGS 1983), and SSARR (USACE1987) will be described.b. PRMS. The Precipitation-Runoff Modeling System(USGS 1983), PRMS, soil moisture accountingalgorithm is summarized in Figure 8-8. The modelcomponents represent the following watershedcharacteristics:andwhered(RES)dt(INFLOW)0 sO sRCF(RES) RCP(RES) 2(8-49)(8-50)(1) Interception. Interception by vegetation ismodeled as a seasonally varying process for a fraction ofthe basin. The fraction of the basin that has interceptionloss can be specified for winter and summer via parameterCOVDN. The volume of water that can be stored by theRESO sRCF and RCP= storage in the reservoir= outflow= routing parameters8-16

EM 1110-2-141731 Aug 94The outflow to the groundwater zone is determined by:whereO g(RSEP) ⎛ REXPRES ⎞⎜ ⎟⎠ ⎝ RESMXO g= flow to the groundwater zoneRESMX, RSEP,and REXP = parameters to be specified(8-51)(6) Groundwater zone. This zone represents thestorage in a phreatic aquifer and outflow to the streamand deep percolation. Outflow to the stream is based ona linear reservoir assumption, requiring the estimate of astorage coefficient, RCB. Outflow to deep percolation iscomputed by the product of a coefficient GSNK time thecurrent storage in the zone. Model simulation occurs at adaily computation interval if any snowpack exists or atthe minimum of 5 min or a user-specified value if asnow-free ground event is occurring. The procedure forrouting precipitation through the system is performed asfollows:(a) Precipitation. The form of the precipitation isdetermined by either of two methods: a temperature BSTis specified that together with maximum and minimumdaily air temperatures is used to determine if rain, snow,or a mixture of both is the form of the precipitation; or,alternatively, a temperature PAT is specified that is thethreshold for rain to snow formation.(b) Surface interception. The daily potential evapotranspiration,EPT, is computed based on one of threemethods: a pan coefficient method, a method that usesdaily mean temperature and daily hours of sunshine, or amethod that uses daily mean air temperature and solarradiation (see paragraph 8-3). Interception is computedfor the open fraction of the subbasin. The EPT demandfraction for the open portion of the basin is satisfied, ifpossible, from the interception storage either as evapotranspirationor snow sublimation.(d) Runoff available from impervious surface. Runofffrom the impervious fraction is computed by considerationof the available excess, surface storage, and EPT.The surface storage is increased by the amount of thesnowmelt/rainfall excess and depleted by evapotranspirationup to the maximum amount EPT. The remainingamount in excess of surface storage RETIP becomesrunoff excess.(e) Surface runoff - daily mode. A water balance isperformed on the soil zone to determine the fraction ofwater that contributes to subsurface storages and openarearunoff. Inflow to the soil zone is treated differentlyfor snowpack or bare ground. Snowpack infiltration isunlimited until field capacity is reached in the rechargezone. At field capacity, the infiltration rate is limited to aconstant value SRX. Snowmelt excess, including rain onthe snowpack, in excess of SRX contributes to surfacerunoff. Surface runoff due to rain on snow is computedusing a contributing area principle as:SROCAP(PTN)(8-52)where CAP is used to factor the available rain on snowmeltinto surface runoff and infiltrating volumes and PTNis the daily precipitation. CAP may be determined via alinear or nonlinear function of antecedent moisture. Thelinear function is:whereCAPSCN⎡⎤⎢(SCX SCN) ⎛ RECHR ⎞⎜ ⎟⎠⎦ ⎥⎣⎝ REMX(8-53)SCN and SCX = minimum and maximum contributingwatershed area, respectivelyRECHR and REMX = storage parameters defined previouslyfor the soil moisture zoneThe nonlinear function is:(c) Snowpack growth/melt. Snowpack simulation isperformed at a daily time step. The snowpack growth/meltdynamics are based on a complex energy-balanceapproach. A detailed discussion of this algorithm isbeyond the scope of this discussion. However, asdescribed in the previous section on snowmelt, energybudget approaches are rather data intensive.CAP SCN(10 (SC1(SMIDX)) )(8-54)8-17

EM 1110-2-141731 Aug 94Figure 8-8. PRMS, schematic diagram of the conceptual watershed system and its inputs8-18

EM 1110-2-141731 Aug 94whereSCN and SC1 = coefficients to be determinedSMIDX = sum of the current available water in thesoil zone (SMAV) plus one-half PTNThe coefficients of this method might be determined fromsoil moisture data, if available. If data are not available,then the user’s manual suggests determining the coefficientsfrom preliminary model runs. An example of thedetermining the coefficients for the nonlinear method as afunction of an antecedent precipitation index is given inFigure 8-9. (The description in the users manual (USGS1983) of how to establish this relationship from a preliminarymodel is not detailed and would seem to be verydifficult).(f) Surface runoff - event mode. Rainfall infiltrationon snow-free ground is calculated from a potential infiltrationrate adjusted for spatial differences in infiltrationpotential. The potential infiltration rate is based on amodified version of the Green and Ampt equation (Chapter6). The modification involves multiplying the soilmoisture deficit at field capacity by the product of thefraction of the storage available in the recharge zone anda user defined coefficient. The infiltration rate necessarilybecomes zero when the recharge zone reaches maximumcapacity. The spatial variation in infiltration properties isthen accounted for as shown in Figure 8-10. Rainfall notinfiltrated is then routed overland to the stream by thekinematic wave method. Infiltrated rainfall moves to thesoil profile zone. Stored water is first lost to EPT that isnot satisfied by surface interception from the rechargezone and then from the lower zone. In addition, water islost from the lower zone to the groundwater zone up to amaximum rate SEP; and volume available in excess ofthis rate moves to the subsurface zone. Inflow from thesoil zone to the groundwater and subsurface zones isrouted to the stream by the equations describedpreviously.c. SSARR. The Streamflow Synthesis and ReservoirRegulation model (SSARR) performs continuous simulationof watershed runoff and reservoir operations. Watershedrunoff simulation may be performed with either the“depletion curve” or the more general “snow bandmodel.” The more general snow band model will bediscussed.(1) Model simulation. Model simulations are performedat a user specified computation interval. Basintemperature and precipitation are input to the model asconceptualized in Figure 8-11. The model accumulatessnow in different user defined elevation bands (thus theterm snow-band model). The amount of snowaccumulated depends on the elevation band temperaturewhich is a function of the input temperature and elevation-temperaturelapse rate. The soil moisture accountingaspect of the runoff algorithm is performed for each band.The accumulated runoff from the bands is then routedthrough conceptual storages to the outlet of the watershed.(2) Differences. The model differs from PRMS, andmost other conceptual continuous simulation models, inthat the soil moisture accounting is not envisioned as aninterconnected group of conceptual storages. Rather, theprecipitation is routed through the system based on a setof empirical relationships, until the final routing to thebasin outlet. The individual relationships are as follows:(a) Interception. Interception is specified as totalbasin volume. Precipitation in excess of this amountreaches the ground surface. The intercepted volume isdecreased to the potential evapotranspiration.(b) Snowpack. The snowpack is assumed to bedistributed uniformly over the watershed fraction representedby a particular elevation band.(c) Soil moisture input zone. The soil moistureinput zone accounts for the water balance in the waterprofile. This zone receives moisture input either fromsnowmelt or rainfall on bare ground. The amount ofdirect runoff, evapotranspiration, and percolation to thelower zone depends on an empirical index of the watercontent of this zone. The index ranges from a smallpercent representing the wilting point, to a valueapproaching 100 percent representing field capacity. Atthe wilting point there will be very little direct runoff,conversely, at field capacity, the direct runoff wouldapproach 100 percent of available moisture. The soilmoisture index varies based on the following relationship:SMI 2SMI 1(MI RGP)whereSMI 1 and SMI 2PH(ETI)24(8-55)= the soil moisture indexes at the beginningand end of a compute period,respectivelyPH = compute period length, in hours8-19

EM 1110-2-141731 Aug 94Figure 8-9. Sample PRMS partial area corrections. The relation between contributing area (CAP) and soil-moistureindex (SMIDX) for Blue Creek, AL8-20

EM 1110-2-141731 Aug 94Figure 8-10. PRMS function which determines fraction of area contribution runoff due to variation in infiltrationcapacityMI= available excess from snowmelt and rainfallwhere ROP = percent runoff.ETI = evapotranspiration index, in inches per dayPH = computation interval, in fractions of a dayRGP = computed surface runoffA user estimated empirical relationship is used to calculatesurface runoff from the soil moisture index. Thisempirical relationship may consider the intensity of theavailable moisture input to the zone (e.g., Figures 8-12and 8-13). The rate of supply available for outflow iscomputed as:RGP ROP(MI)(8-56)(d) Base-flow separation. An empirical relationbetween a base-flow infiltration index and percent ofrunoff to base-flow is used to divide outflow from the soilmoisture zone into direct runoff and base-flow (e.g., Figure8-14). The base-flow infiltration index is computedas:BII 2BII 124 ⎛ RGP ⎞⎜ ⎟⎠⎝ PH BII 1BII 2≤ BIIMXBIITSPHPH2(8-57)8-21

EM 1110-2-141731 Aug 94Figure 8-11. SSARR “snowbank” watershed model8-22

EM 1110-2-141730 Aug 94TBF BFP ⎛ RGP ⎞⎜ ⎟⎠ ⎝ PH(8-59)where BFP is determined from Figure 8-14 using BII.(e) Lower zone versus base flow. The lower zoneand base-flow components are separated based on a userdefinedfactor PBLZ:LZTBF(PBLZ)(8-60)Figure 8-12. SSARR SMI versus runoff percentwhere LZ is the inflow rate to the lower zone, up to avalue DGWLIM. The difference between LZ and TBF isthe contribution to base flow.(f) Direct runoff. The inflow to direct runoff is thedifference between the outflow from the soil moisturezone and the inflow to the base-flow zone:RGS RG TBF(8-61)Surface and subsurface runoff are distinguished by a userspecifiedempirical relationship (e.g., Figure 8-15). TheSSARR user’s manual provides guidelines for developingthis relationship.Figure 8-13. SSARR SMI versus precipitation intensityand runoff percentor(g) Routing flows to outlet. Surface, subsurface,lower zone, and base flows are routed to the outlet vialinear reservoir routing. The user may separately specifythe number of linear storages for each outflowcomponent.BII 2BIIMX BII 2> BIIMX(8-58)8-6. Parameter Estimation for Continuous SimulationModelswhereBII 1 and BII 2BIITSBIIMX= base-flow indexes at the beginning andending of the computational period= time delay or time of storage= limiting value for the indexThe rate of inflow to the lower and base-flow zone is thencomputed as:a. Parameter estimation. Parameter estimation forcontinuous simulation models is much more difficult thanfor event-oriented models. The reason for this is that acontinuous simulation model must represent the entirehydrologic cycle. This representation requires an increasein model complexity and, correspondingly, an increase inthe number of parameters to be estimated. The parameterestimation process requires an extensive amount of dataand user experience. A totally ungauged parameter estimationprocedure is not practical or advisable.8-23

EM 1110-2-141730 Aug 94Figure 8-14. SSARR base-flow infiltration index (BII) versus base-flow percent (BFP)b. Conceptual model. A conceptual model which isapplicable to all watersheds does not exist. The subsurfacecharacteristics of watersheds, and consequently thebase-flow response, will vary. This variation will requiredifferent model representations to capture the subsurfaceresponse. Consequently, the conceptualization of thebase-flow response by the number of storage zones ortanks in the model is, in some sense, a parameterestimation decision. A single subsurface tank may besufficient for small watersheds with limited base-flowresponse, and multiple zones or tanks might be necessaryfor watersheds that have a complicated base-flowresponse. At the very least, a particular conceptual modelshould allow flexibility in the number of subsurface zones8-24

EM 1110-2-141730 Aug 94Figure 8-15. Surface - subsurface separationthat can be used to model subsurface response. The engineerwould be well advised to find a model that has beensuccessfully calibrated for a watershed that is similar tothe one under investigation and subject to the samemeteorologic conditions. Previous experience will help inselecting the appropriate structure for the model.c. Previous experience. If no previous experienceexists, then the structure of the model required dependson hydrograph recession analysis. The hydrographrecession analysis is an important aspect of an overallparameter estimation procedure which will be discussedsubsequently.(1) A general procedure for estimating parameters isto examine the hydrometeorologic record for errors, performa water balance to determine ET, estimateparameters based on event analysis and watershed physicalcharacteristics, and apply automatic parameter estimationto fine tune parameters. An automatic parameterestimation procedure, if available, can only be used toestimate a handful of parameters, eight at the very most,preferably four or less. The automatic procedure is veryuseful when the number of parameters is limited, as in thecase of event-oriented modeling. However, the largenumber of parameters available for continuous modelsrequires that most of these parameters be estimated priorto application of an automatic procedure.(2) Many of the continuous model parameters have asimilar effect on the predicted hydrograph. An optimizationprocedure cannot distinguish between these parametersfor this reason. The impact of each parameter mustbe examined in context with the physics of the process8-25

EM 1110-2-141730 Aug 94affecting hydrographs. Available automatic parameterestimation algorithms have not been developed which canconsider the physics of the problem as part of the fittingprocedure.d. Experience in applying model. Burnash (1985),who developed and has had extensive experience inapplying the Sacramento Model (a conceptual continuoussimulation model), recommends the first three steps mentionedwhen estimating parameters. Although his recommendationswere directed toward the Sacramento model,they are equally applicable to other continuous models.(1) Examination of hydrometeorologic record forerrors. Burnash is convinced that the major deficiency inhydrometeorologic record is the potential underestimationof rainfall by raingauges due to wind effects. The underestimationis on the order of 10 to 15 percent. The errormay not be consistent and is likely to affect large eventswhere wind speeds are the greatest. Other factors thatcontribute to errors in the record are change in gaugelocation, gauge type, or in the environment surroundingthe gauge which changes local wind patterns.(a) Burnash makes some suggestions to identify andcorrect this problem. For these reasons and others, acareful application of the Sacramento Watershed Modelor, for that matter, the basic water balance equationrequires a continuous comparative analysis of rainfall andrunoff records to describe an unusual pattern which maybe a result of data inconsistencies rather than a true event.Implicit in these comments is the notion that the rainfallinput should be scaled to arrive at a consistent rainfallrunoffrecord.(b) Discharge measurements, particularly for largeflows, may have large errors due to ill-defined ratingcurves. Although not explicitly stated, Burnash seems tobe warning against accepting streamflow measurementsthat are inconsistent with the rest of the record which, inturn, would distort model parameters in the estimationprocess.(2) Water balance preservation. A successful parameterestimation procedure depends on preserving the fundamentalwater balance equation:Runoff = Precipitation - EvapotranspirationEstimation of evapotranspiration is difficult because themost common indicator used is evaporation, most commonlyestimated by evaporation pans. Evaporation is avery different process from ET and a poor indicator aswell. Burnash cautions against using evaporation as thefinal arbitrator of ET; evaporation may be used as an aidin preserving the fundamental water balance equation.(3) Parameters from event analysis. The key to estimatingcontinuous simulation model parameters is toidentify circumstances in the hydrologic record where theindividual parameter has the most effect. This may beaccomplished by examining different events or an aspectof the hydrograph where a particular parameter is of firstorderimportance.(a) The impervious area fraction of the basin may beidentified by examining direct runoff when antecedentprecipitation conditions are extremely dry. The directrunoff in these circumstance would be due to the imperviousfraction.(b) As antecedent precipitation increases, there willbe an increase in direct runoff from a larger portion of thewatershed. The maximum fraction of area that contributesto direct runoff will occur under the wettest conditions.The partial area correction, the relationshipbetween basin contribution to direct runoff and basinmoisture conditions, can be developed from examining thebasin response from wet to dry antecedent conditions.(c) The soil profile zone capacity can be estimatedby examining prediction errors when the soil moisturedeficit should be small. Presumably, an overprediction ofrunoff will indicate that the soil profile capacity has beenunderestimated.(d) The subsurface response characteristics are determinedby performing hydrograph recession analysis asdiscussed in paragraph 8-2 on event-oriented modeling ofbase flow. However, the recession analysis tends to bemore detailed than in the event case. The continuoussimulation analysis endeavors to identify different levelsof aquifer response characteristics by identifying straightline segments on a log-discharge versus time plot.Burnash cautions that deviations from the straight linerecession may occur due to channel losses or riparianvegetation ET. The impact of channel losses may bediscerned by examining the deviations from a straight lineduring periods when ET is low. The recession can thenbe corrected for channel loss and then used to examinethe impact of ET on the recession during high ET periods.(e) Burnash does not discuss the use of automaticparameter estimation or optimization algorithms for estimatingparameters. However, his recommended estimationtechniques should be used to reduce the number of8-26

EM 1110-2-141730 Aug 94parameters that will be used when estimating parametersvia an optimization approach. Optimization techniquesare only useful when the number of parameters are limitedto less than eight and preferably less than four.Consequently, optimization or automatic parameter estimationwill probably be used to fine tune parameter estimatesobtained by event analysis and application of thewater balance equation.8-27

EM 1110-2-141731 Aug 94Chapter 9Streamflow and Reservoir Routing9-1. Generala. Routing is a process used to predict the temporaland spatial variations of a flood hydrograph as it movesthrough a river reach or reservoir. The effects of storageand flow resistance within a river reach are reflected bychanges in hydrograph shape and timing as the floodwavemoves from upstream to downstream. Figure 9-1 showsthe major changes that occur to a discharge hydrograph asa floodwave moves downstream.b. In general, routing techniques may be classifiedinto two categories: hydraulic routing, and hydrologicrouting. Hydraulic routing techniques are based on thesolution of the partial differential equations of unsteadyopen channel flow. These equations are oftenreferred to as the St. Venant equations or the dynamicwave equations. Hydrologic routing employs the continuityequation and an analytical or an empirical relationshipbetween storage within the reach and discharge at theoutlet.c. Flood forecasting, reservoir and channel design,floodplain studies, and watershed simulations generallyutilize some form of routing. Typically, in watershedsimulation studies, hydrologic routing is utilized on areach-by-reach basis from upstream to downstream. Forexample, it is often necessary to obtain a discharge hydrographat a point downstream from a location where ahydrograph has been observed or computed. For suchpurposes, the upstream hydrograph is routed through thereach with a hydrologic routing technique that predictschanges in hydrograph shape and timing. Local flows arethen added at the downstream location to obtain the totalflow hydrograph. This type of approach is adequate aslong as there are no significant backwater effects orFigure 9-1. Discharge hydrograph routing effects9-1

EM 1110-2-141731 Aug 94discontinuities in the water surface because of jumps orbores. When there are downstream controls that will havean effect on the routing process through an upstreamreach, the channel configuration should be treated as onecontinuous system. This can only be accomplished with ahydraulic routing technique that can incorporate backwatereffects as well as internal boundary conditions, such asthose associated with culverts, bridges, and weirs.d. This chapter describes several different hydraulicand hydrologic routing techniques. Assumptions, limitations,and data requirements are discussed for each. Thebasis for selection of a particular routing technique isreviewed, and general calibration methodologies are presented.This chapter is limited to discussions on 1-D flowrouting techniques in the context of flood-runoff analysis.The focus of this chapter is on discharge (flow) ratherthan stage (water surface elevation). Detailed presentationof routing techniques and applications focused on stagecalculations can be found in EM 1110-2-1416.9-2. Hydraulic Routing Techniquesa. The equations of motion. The equations thatdescribe 1-D unsteady flow in open channels, the SaintVenant equations, consist of the continuity equation,Equation 9-1, and the momentum equation, Equation 9-2.The solution of these equations defines the propagation ofa floodwave with respect to distance along the channeland time.qS fS og= lateral inflow per unit length of channel= friction slope= channel bed slope= gravitational accelerationSolved together with the proper boundary conditions,Equations 9-1 and 9-2 are the complete dynamic waveequations. The meaning of the various terms in thedynamic wave equations are as follows (Henderson 1966):(1) Continuity equation.A ∂V∂xVB ∂y∂xB ∂y∂tqprism storagewedge storagerate of riselateral inflow per unit length(2) Momentum equation.S ffriction slope (frictional forces)A ∂V∂xVB ∂y∂xS fS o∂y∂xB ∂y∂tVg∂V∂xq1g∂V∂t(9-1)(9-2)S o∂y∂xV ∂Vg ∂xbed slope (gravitational effects)pressure differentialconvective accelerationwhereA = cross-sectional flow area1 ∂Vg ∂tlocal accelerationV = average velocity of waterx = distance along channelB = water surface widthy = depth of water(3) Dynamic wave equations. The dynamic waveequations are considered to be the most accurate andcomprehensive solution to 1-D unsteady flow problems inopen channels. Nonetheless, these equations are based onspecific assumptions, and therefore have limitations. Theassumptions used in deriving the dynamic wave equationsare as follows:t= time9-2

EM 1110-2-141731 Aug 94(a) Velocity is constant and the water surface is horizontalacross any channel section.(b) All flows are gradually varied with hydrostaticpressure prevailing at all points in the flow, such thatvertical accelerations can be neglected.(c) No lateral secondary circulation occurs.The use of approximations to the full equations forunsteady flow can be justified when specific terms in themomentum equation are small in comparison to the bedslope. This is best illustrated by an example taken fromHenderson’s book Open Channel Flow (1966).Henderson computed values for each of the terms on theright-hand side of the momentum equation for a steepalluvial stream:(d) Channel boundaries are treated as fixed; therefore,no erosion or deposition occurs.Term: S o∂y∂xV ∂Vg ∂x1 ∂Vg ∂t(e) Water is of uniform density, and resistance toflow can be described by empirical formulas, such asManning’s and Chezy’s equation.(f) The dynamic wave equations can be applied to awide range of 1-D flow problems; such as, dam breakfloodwave routing, forecasting water surface elevationsand velocities in a river system during a flood, evaluatingflow conditions due to tidal fluctuations, and routingflows through irrigation and canal systems. Solution ofthe full equations is normally accomplished with anexplicit or implicit finite difference technique. The equationsare solved for incremental times ( t) and incrementaldistances ( x) along the waterway.b. Approximations of the full equations. Dependingon the relative importance of the various terms of themomentum Equation 9-2, the equation can be simplifiedfor various applications. Approximations to the fulldynamic wave equations are created by combining thecontinuity equation with various simplifications of themomentum equation. The most common approximationsof the momentum equation are:Magnitude (ft/mi): 26 .5 .12-.25 .05These figures relate to a very fast rising hydrograph inwhich the flow increased from 10,000 to 150,000 cfs anddecreased again to 10,000 cfs within 24 hr. Even in thiscase, where changes in depth and velocity with respect todistance and time are relatively large, the last three termsare still small in comparison to the bed slope. For thistype of flow situation (steep stream), an approximation ofthe full equations would be appropriate. For flatterslopes, the last three terms become increasingly moreimportant.(1) Kinematic wave approximation. Kinematic flowoccurs when gravitational and frictional forces achieve abalance. In reality, a true balance between gravitationaland frictional forces never occurs. However, there areflow situations in which gravitational and frictional forcesapproach an equilibrium. For such conditions, changes indepth and velocity with respect to time and distance aresmall in magnitude when compared to the bed slope ofthe channel. Therefore, the terms to the right of the bedslope in Equation 9-3 are assumed to be negligible. Thisassumption reduces the momentum equation to thefollowing:S fS o(9-4)Equation 9-4 essentially states that the momentum of theflow can be approximated with a uniform flow assumptionas described by Manning’s or Chezy’s equation.Manning’s equation can be written in the following form:(9-5)Q αA m 9-3

EM 1110-2-141731 Aug 94where α and m are related to flow geometry and surfaceroughnessS fS o∂y∂x(9-8)Since the momentum equation has been reduced to asimple functional relationship between area and discharge,the movement of a floodwave is described solely by thecontinuity equation, written in the following form:∂A∂t∂Q∂xq(9-6)Then by combining Equations 9-5 and 9-6, the governingkinematic wave equation is obtained as:∂A∂tαmA (m 1) ∂A∂xq(9-7)Because of the steady uniform flow assumptions, thekinematic wave equations do not allow for hydrographdiffusion, just simple translation of the hydrograph intime. The kinematic wave equations are usually solvedby explicit or implicit finite difference techniques. Anyattenuation of the peak flow that is computed using thekinematic wave equations is due to errors inherent in thefinite difference solution scheme.(a) The application of the kinematic wave equation islimited to flow conditions that do not demonstrate appreciablehydrograph attenuation. In general, the kinematicwave approximation works best when applied to steep(10 ft/mile or greater), well defined channels, where thefloodwave is gradually varied.(b) The kinematic wave approach is often applied inurban areas because the routing reaches are generallyshort and well defined (i.e., circular pipes, concrete linedchannels, etc.).(c) The kinematic wave equations cannot handlebackwater effects since, with a kinematic model flow,disturbances can only propagate in the downstream direction.All of the terms in the momentum equation that areused to describe the propagation of the floodwaveupstream (backwater effects) have been excluded.(2) Diffusion wave approximation. Another commonapproximation of the full dynamic wave equations is thediffusion wave analogy. The diffusion wave model utilizesthe continuity Equation 9-1 and the following simplifiedform of the momentum equation:The diffusion wave model is a significant improvementover the kinematic wave model because of the inclusionof the pressure differential term in Equation 9-8. Thisterm allows the diffusion model to describe the attenuation(diffusion effect) of the floodwave. It also allows thespecification of a boundary condition at the downstreamextremity of the routing reach to account for backwatereffects. It does not use the inertial terms (last two terms)from Equation 9-2 and, therefore, is limited to slow tomoderately rising floodwaves (Fread 1982). However,most natural floodwaves can be described with the diffusionform of the equations.(3) Quasi-steady dynamic wave approximation. Thethird simplification of the full dynamic wave equations isthe quasi-steady dynamic wave approximation. Thismodel utilizes the continuity equation, Equation 9-1, andthe following simplification of the momentum equation:S fS o∂y∂xVg∂V∂x(9-9)In general, this simplification of the dynamic wave equationsis not used in flood routing. This form of themomentum equation is more commonly used in steadyflow-water surface profile computations. In the case offlood routing, the last two terms on the momentum equationare often opposite in sign and tend to counteract eachother (Fread 1982). By including the convective accelerationterm and not the local acceleration term, an error isintroduced. This error is of greater magnitude than theerror that results when both terms are excluded, as in thediffusion wave model. For steady flow-water surfaceprofiles, the last term of the momentum equation (changesin velocity with respect to time) is assumed to be zero.However, changes in velocity with respect to distance arestill very important in the calculation of steady flow-watersurface profiles.c. Data requirements. In general, the data requirementsof the various hydraulic routing techniques arevirtually the same. However, the amount of detail that isrequired for each type of data will vary depending uponthe routing technique being used and the situation it isbeing applied to. The basic data requirements for hydraulicrouting techniques are the following:9-4

EM 1110-2-141731 Aug 94(1) Flow data (hydrographs).(2) Channel cross sections and reach lengths.(3) Roughness coefficients.(4) Initial and boundary conditions.(a) Flow data consist of discharge hydrographs fromupstream locations as well as lateral inflow and tributaryflow for all points along the stream.(b) Channel cross sections are typically surveyedsections that are perpendicular to the flow lines. Keyissues in selecting cross sections are the accuracy of thesurveyed data and the spacing of the sections along thestream. If the routing procedure is utilized to predictstages, then the accuracy of the cross-sectional dimensionswill have a direct effect on the prediction of the stage. Ifthe cross sections are used only to route discharge hydrographs,then it is only important to ensure that the crosssection is an adequate representation of the dischargeversus flow area of the section. Simplified cross-sectionalshapes, such as 8-point cross sections or trapezoids andrectangles, are often used to fit the discharge versus flowarea of a more detailed section. Cross-sectional spacingaffects the level of detail of the results as well as theaccuracy of the numerical solution to the routing equations.Detailed discussions on cross-sectional spacing canbe found in the reference by the Hydrologic EngineeringCenter (HEC) (USACE 1986).(c) Roughness coefficients for hydraulic routingmodels are typically in the form of Manning’s n values.Manning’s coefficients have a direct impact on the traveltime and amount of diffusion that will occur when routinga flood hydrograph through a channel reach. Roughnesscoefficients will also have a direct impact on predictedstages.(d) All hydraulic models require that initial and boundaryconditions be established before the routing cancommence. Initial conditions are simply stated as theconditions at all points in the stream at the beginning ofthe simulation. Initial conditions are established by specifyinga base flow within the channel at the start of thesimulation. Channel depths and velocities can be calculatedthrough steady-state backwater computations or anormal depth equation (e.g., Manning’s equation).Boundary conditions are known relationships betweendischarge and time and/or discharge and stage. Hydraulicrouting computations require the specification ofupstream, downstream, and internal boundary conditionsto solve the equations. The upstream boundary conditionis the discharge (or stage) versus time relationship of thehydrograph to be routed through the reach. Downstreamboundary conditions are usually established with a steadystaterating curve (discharge versus depth relationship) orthrough normal depth calculations (Manning’s equation).Internal boundary conditions consist of lateral inflow ortributary flow hydrographs, as well as depth versus dischargerelationships for hydraulic structures within theriver reach.9-3. Hydrologic Routing TechniquesHydrologic routing employs the use of the continuityequation and either an analytical or an empirical relationshipbetween storage within the reach and discharge at theoutlet. In its simplest form, the continuity equation canbe written as inflow minus outflow equals the rate ofchange of storage within the reach:IwhereIO∆S∆t= the average inflow to the reach during ∆tO = the average outflow from the reach during ∆tS = storage within the reacha. Modified puls reservoir routing.(9-10)(1) One of the simplest routing applications is theanalysis of a floodwave that passes through anunregulated reservoir (Figure 9-2a). The inflow hydrographis known, and it is desired to compute the outflowhydrograph from the reservoir. Assuming that all gateand spillway openings are fixed, a unique relationshipbetween storage and outflow can be developed, as shownin Figure 9-2b.(2) The equation defining storage routing, based onthe principle of conservation of mass, can be written inapproximate form for a routing interval t. Assuming thesubscripts “1” and “2” denote the beginning and end ofthe routing interval, the equation is written as follows:O 1O 22I 1I 22S 2S 1∆t(9-11)9-5

EM 1110-2-141731 Aug 94Figure 9-2. Reservoir storage routingThe known values in this equation are the inflow hydrographand the storage and discharge at the beginning ofthe routing interval. The unknown values are the storageand discharge at the end of the routing interval. With twounknowns (O 2 and S 2 ) remaining, another relationship isrequired to obtain a solution. The storage-outflow relationshipis normally used as the second equation. Howthat relationship is derived is what distinguishes variousstorage routing methods.(3) For an uncontrolled reservoir, outflow and waterin storage are both uniquely a function of lake elevation.The two functions can be combined to develop a storageoutflowrelationship, as shown in Figure 9-3. Elevationdischargerelationships can be derived directly fromhydraulic equations. Elevation-storage relationships arederived through the use of topographic maps. Elevationarearelationships are computed first, then either averageend-area or conic methods are used to compute volumes.(4) The storage-outflow relationship provides the outflowfor any storage level. Starting with a nearly emptyreservoir, the outflow capability would be minimal. If theinflow is less than the outflow capability, the water wouldflow through. During a flood, the inflow increases andeventually exceeds the outflow capability. The differencebetween inflow and outflow produces a change in storage.In Figure 9-4, the difference between the inflow and theoutflow (on the rising side of the outflow hydrograph)represents the volume of water entering storage.(5) As water enters storage, the outflow capabilityincreases because the pool level increases. Therefore, theoutflow increases. This increasing outflow with increasingwater in storage continues until the reservoir reaches amaximum level. This will occur the moment that theoutflow equals the inflow, as shown in Figure 9-4. Oncethe outflow becomes greater than the inflow, the storagelevel will begin dropping. The difference between theoutflow and the inflow hydrograph on the recession sidereflects water withdrawn from storage.(6) The modified puls method applied to reservoirsconsists of a repetitive solution of the continuity equation.It is assumed that the reservoir water surface remainshorizontal, and therefore, outflow is a unique function ofreservoir storage. The continuity equation, Equation 9-11,can be manipulated to get both of the unknown variableson the left-hand side of the equation:⎛S 2⎜⎝ ∆t⎞O 2⎟2 ⎠⎛S 1⎜⎝ ∆t⎞O 1⎟2 ⎠O 1I 1I 22(9-12)Since I is known for all time steps, and O 1 and S 1 areknown for the first time step, the right-hand side of theequation can be calculated. The left-hand side of theequation can be solved by trial and error. This is accomplishedby assuming a value for either S 2 or O 2 , obtainingthe corresponding value from the storage-outflow relationship,and then iterating until Equation 9-12 is satisfied.9-6

EM 1110-2-141731 Aug 94Figure 9-3. Reservoir storage-outflow curveFigure 9-4. Reservoir routing exampleRather than resort to this iterative procedure, a value oft is selected and points on the storage-outflow curve arereplotted as the “storage-indication” curve shown inFigure 9-5. This graph allows for a direct determinationof the outflow (O 2 ) once a value of storage indication(S 2 / t + O 2 /2) has been calculated from Equation 9-12(Viessman et al. 1977). The numerical integration ofEquation 9-12 and Figure 9-5 is illustrated as an examplein Table 9-1. The stepwise procedure for applying themodified puls method to reservoirs can be summarized asfollows:9-7

EM 1110-2-141731 Aug 94and discharge at the outlet of a channel is not a uniquerelationship, rather it is a looped relationship. An examplestorage-discharge function for a river is shown inFigure 9-7.(1) Application of the modified puls method torivers. To apply the modified puls method to a channelrouting problem, the storage within the river reach isapproximated with a series of “cascading reservoirs” (Figure9-8). Each reservoir is assumed to have a level pooland, therefore, a unique storage-discharge relationship.The cascading reservoir approach is capable of approximatingthe looped storage-outflow effect when evaluatingthe river reach as a whole. The rising and falling floodwaveis simulated with different storage levels in thecascade of reservoirs, thus producing a looped storageoutflowfunction for the total river reach. This is depictedgraphically in Figure 9-9.(2) Determination of the storage-outflowrelationship.Figure 9-5. Storage-indication curve(a) Determine a composite discharge rating curve forall of the reservoir outlet structures.(b) Determine the reservoir storage that correspondswith each elevation on the rating curve for reservoir outflow.(c) Select a time step and construct a storage-indicationversus outflow curve [(S/ t) +(O/2)] versus O.(d) Route the inflow hydrograph through the reservoirbased on Equation 9-12 and the storage-indication curve.(e) Compare the results with historical events toverify the model.b. Modified puls channel routing. Routing in naturalrivers is complicated by the fact that storage in a riverreach is not a function of outflow alone. During the passingof a floodwave, the water surface in a channel is notuniform. The storage and water surface slope within ariver reach, for a given outflow, is greater during therising stages of a floodwave than during the falling(Figure 9-6). Therefore, the relationship between storage(a) Determining the storage-outflow relationship fora river reach is a critical part of the modified puls procedure.In river reaches, storage-outflow relationships canbe determined from one of the following:• steady-flow profile computations,• observed water surface profiles,• normal-depth calculations,• observed inflow and outflow hydrographs, and• optimization techniques applied to observedinflow and outflow hydrographs.(b) Steady-flow water surface profiles, computedover a range of discharges, can be used to determinestorage-outflow relationships in a river reach(Figure 9-10). In this illustration, a known hydrograph atA is to be routed to location B. The storage-outflowrelationship required for routing is determined by computinga series of water surface profiles, corresponding to arange of discharges. The range of discharges shouldencompass the range of flows that will be routed throughthe river reach. The storage volumes are computed bymultiplying the cross-sectional area, under a specific flowprofile, by the channel reach lengths. Volumes arecalculated for each flow profile and then plotted against9-8

EM 1110-2-141731 Aug 94Table 9-1Storage Routing Calculation(1) (2) (3) (4) (5) (6) (7)Time(hr)Inflow(cfs)AverageInflow(cfs)S∆tO2 Outflow(cfs)(cfs)S∆t(cfs)S(acre-ft)0 3,000 8,600 3,000 7,100 1,7603,1303 3,260 8,730 3,150 7,155 1,7743,4456 3,630 9,025 3,400 7,325 1,8163,8259 4,020 9,450 3,850 7,525 1,8664,25012 4,480 9,850 4,300 7,700 1,909..etc.the corresponding discharge at the outlet. If channel orlevee modifications will have an effect on the routingthrough the reach,modifications can be made to the crosssections, water surface profiles recalculated, and a revisedstorage-outflow relationship can be developed. Theimpacts of the channel or levee modification can beapproximated by routing floods with both pre- and postprojectstorage-outflow relationships.(c) Observed water surface profiles, obtained fromhigh water marks, can be used to compute storage-outflowrelationships. Sufficient stage data over a range of floodsare required for this type of calculation; however, it is notlikely that enough data would be available over the rangeof discharges needed to compute an adequate storagedischarge relationship. If a few observed profiles areavailable, they can be used to calibrate a steady-flowwater surface profile model for the channel reach ofinterest. Then the water surface profile model could beused to calculate the appropriate range of values to calculatethe storage-outflow relationship.(d) Normal depth associated with uniform flow doesnot exist in natural streams; however, the concept can beused to estimate water depth and storage in natural riversif uniform flow conditions can reasonably be assumed.With a typical cross section, Manning’s equation is solvedfor a range of discharges, given appropriate “n” valuesand an estimated slope of the energy grade line. Underthe assumption of uniform flow conditions, the energyslope is considered equal to the average channel bedslope; therefore, this approach should not be applied inbackwater areas.(e) Observed inflow and outflow hydrographs can beused to compute channel storage by an inverse process offlood routing. When both inflow and outflow are known,the change in storage can be computed, and from that astorage versus outflow function can be developed. Tributaryinflow, if any, must also be accounted for in thiscalculation. The total storage is computed from somebase level storage at the beginning or end of the routingsequence.(f) Inflow and outflow hydrographs can also be usedto compute routing criteria through a process of iterationin which an initial set of routing criteria is assumed, theinflow hydrograph is routed, and the results are evaluated.The process is repeated as necessary until a suitable fit ofthe routed and observed hydrograph is obtained.9-9

EM 1110-2-141731 Aug 94Figure 9-6. Rising and falling floodwaveFigure 9-7. Looped storage-outflow relationship for a river reach9-10

EM 1110-2-141731 Aug 94Figure 9-8. Cascade of reservoirs, depicting storage routing in a channelFigure 9-9. Modified puls approximation of the rising and falling floodwaves9-11

EM 1110-2-141731 Aug 94Figure 9-10. Storage-outflow relationships(9-13)(3) Determining the number of routing steps. InKreservoir routing, the modified puls method is appliedNSTPS∆twith one routing step. This is under the assumption thatKLV wthe travel time through the reservoir is smaller than thecomputation interval t. In channel routing, the travel wheretime through the river reach is often greater than thecomputation interval. When this occurs, the channel must K = floodwave travel time through the reachbe broken down into smaller routing steps to simulate thefloodwave movement and changes in hydrograph shape. L = channel reach lengthThe number of steps (or reach lengths) affects the attenuationof the hydrograph and should be obtained by calibration.V w = velocity of the floodwave (not average velocity)The maximum amount of attenuation will occurwhen the channel routing computation is done in one step. NSTPS = number of routing stepsAs the number of routing steps increases, the amount ofattenuation decreases. An initial estimate of the number The time interval t is usually determined by ensuringof routing steps (NSTPS) can be obtained by dividing thetotal travel time (K) for the reach by the computationthat there is a sufficient number of points on the risingside of the inflow hydrograph. A general rule of thumb isinterval t.that the computation interval should be less than 1/5 ofthe time of rise (t r ) of the inflow hydrograph.9-12

∆t ≤ t r5(9-14)S prism storage wedge storageS KO KX(I O)EM 1110-2-141731 Aug 94c. Muskingum method. The Muskingum method wasdeveloped to directly accommodate the looped relationshipbetween storage and outflow that exists in rivers.With the Muskingum method, storage within a reach isvisualized in two parts: prism storage and wedge storage.Prism storage is essentially the storage under the steadyflowwater surface profile. Wedge storage is the additionalstorage under the actual water surface profile. Asshown in Figure 9-11, during the rising stages of thefloodwave the wedge storage is positive and added to theprism storage. During the falling stages of a floodwave,the wedge storage is negative and subtracted from theprism storage.(1) Development of the Muskingum routing equation.(a) Prism storage is computed as the outflow (O)times the travel time through the reach (K). Wedge storageis computed as the difference between inflow andoutflow (I-O) times a weighting coefficient X and thetravel time K. The coefficient K corresponds to the traveltime of the floodwave through the reach. The parameterX is a dimensionless value expressing a weighting of therelative effects of inflow and outflow on the storage (S)within the reach. Thus, the Muskingum method definesthe storage in the reach as a linear function of weightedinflow and outflow:whereS K [XI (1 X)O]S = total storage in the routing reachO = rate of outflow from the routing reachI = rate of inflow to the routing reachK = travel time of the floodwave through the reach(9-15)X = dimensionless weighting factor, ranging from0.0 to 0.5(b) The quantity in the brackets of Equation 9-15 isconsidered an expression of weighted discharge. WhenX = 0.0, the equation reduces to S = KO, indicating thatstorage is only a function of outflow, which is equivalentto level-pool reservoir routing with storage as a linearfunction of outflow. When X = 0.5, equal weight is givento inflow and outflow, and the condition is equivalent to auniformly progressive wave that does not attenuate. Thus,“0.0” and “0.5” are limits on the value of X, and withinthis range the value of X determines the degree of attenuationof the floodwave as it passes through the routingFigure 9-11. Muskingum prism and wedge storage concept9-13

EM 1110-2-141731 Aug 94reach. A value of “0.0” produces maximum attenuation,and “0.5” produces pure translation with no attenuation.(c) The Muskingum routing equation is obtained bycombining Equation 9-15 with the continuity equation,Equation 9-11, and solving for O 2 .(a) Using Seddon’s law, a floodwave velocity can beapproximated from the discharge rating curve at a stationwhose cross section is representative of the routing reach.The slope of the discharge rating curve is equal to dQ/dy.The floodwave velocity, and therefore the travel time K,can be estimated as follows:O 2C 1I 2C 2I 1C 3O 1(9-16)V w1BdQdy(9-20)The subscripts 1 and 2 in this equation indicate the beginningand end, respectively, of a time interval t. Therouting coefficients C 1 , C 2 , and C 3 are defined in terms oft, K, and X.KwhereLV w(9-21)C 1∆t 2KX2K(1 X) ∆t(9-17)V w= floodwave velocity, in feet/secondC 2∆t 2KX2K(1 X) ∆t(9-18)B = top width of the water surfaceL = length of the routing reach, in feetC 32K(1 X) ∆t2K(1 X) ∆t(9-19)Given an inflow hydrograph, a selected computation intervalt, and estimates for the parameters K and X, theoutflow hydrograph can be calculated.(2) Determination of Muskingum K and X. In agauged situation, the Muskingum K and X parameters canbe calculated from observed inflow and outflow hydrographs.The travel time, K, can be estimated as the intervalbetween similar points on the inflow and outflowhydrographs. The travel time of the routing reach can becalculated as the elapsed time between centroid of areasof the two hydrographs, between the hydrograph peaks, orbetween midpoints of the rising limbs. After K has beenestimated, a value for X can be obtained through trial anderror. Assume a value for X, and then route the inflowhydrograph with these parameters. Compare the routedhydrograph with the observed outflow hydrograph. Makeadjustments to X to obtain the desired fit. Adjustments tothe original estimate of K may also be necessary to obtainthe best overall fit between computed and observed hydrographs.In an ungauged situation, a value for K can beestimated as the travel time of the floodwave through therouting reach. The floodwave velocity (V w ) is greaterthan the average velocity at a given cross section for agiven discharge. The floodwave velocity can be estimatedby a number of different techniques:(b) Another means of estimating floodwave velocityis to estimate the average velocity (V) and multiply it by aratio. The average velocity can be calculated fromManning’s equation with a representative discharge andcross section for the routing reach. For various channelshapes, the floodwave velocity has been found to be adirect ratio of the average velocity.Channel shape Ratio V w /VWide rectangular 1.67Wide parabolic 1.44Triangular 1.33For natural channels, an average ratio of 1.5 is suggested.Once the wave speed has been estimated, the travel time(K) can be calculated with Equation 9-21.(c) Estimating the Muskingum X parameter in anungauged situation can be very difficult. X variesbetween 0.0 and 0.5, with 0.0 providing the maximumamount of hydrograph attenuation and 0.5 no attenuation.Experience has shown that for channels with mild slopesand flows that go out of bank, X will be closer to 0.0.For steeper streams, with well defined channels that donot have flows going out of bank, X will be closer to 0.5.Most natural channels lie somewhere in between thesetwo limits, leaving a lot of room for “engineering judgment.”One equation that can be used to estimate theMuskingum X coefficient in ungauged areas has been9-14

EM 1110-2-141731 Aug 94developed by Cunge (1969). This equation is taken fromthe Muskingum-Cunge channel routing method, which isdescribed in paragraph 9-3e. The equation is written asfollows:Xwhere12⎛⎜1⎝Q ⎞oBS ⎟oc∆x⎠Q o = reference flow from the inflowhydrographcS oBx= floodwave speed= friction slope or bed slope= top width of the flow area= length of the routing subreach(9-22)The choice of which flow rate to use in this equation isnot completely clear. Experience has shown that a referenceflow based on average values (midway between thebase flow and the peak flow) is in general the most suitablechoice. Reference flows based on peak flow valuestend to accelerate the wave much more than it would innature, while the converse is true if base flow referencevalues are used (Ponce 1983).(3) Selection of the number of subreaches. TheMuskingum equation has a constraint related to the relationshipbetween the parameter K and the computationinterval t. Ideally, the two should be equal, but tshould not be less than 2KX to avoid negative coefficientsand instabilities in the routing procedure.2KX < ∆t ≤K(9-23)d. Working R&D routing procedure. The WorkingR&D procedure is a storage routing technique that accommodatesthe nonlinear nature of floodwave movement innatural channels. The method is useful in situationswhere the use of a variable K (reach travel time) wouldassist in obtaining accurate answers. A nonlinear storageoutflowrelationship indicates that a variable K is necessary.The method is also useful in situations wherein thehorizontal reservoir surface assumption of the modifiedpuls procedure is not applicable, such as normally occursin natural channels.(1) The working R&D procedure could be termed“Muskingum with a variable K” or “modified puls withwedge storage.” For a straight line storage-discharge(weighted discharge) relation, the procedure is the samesolution as the Muskingum method. For X = 0, the procedureis identical to Modified Puls.(2) The basis for the procedure derives from theconcept of a “working discharge,” which is a hypotheticalsteady flow that would result in the same natural channelstorage that occurs with the passage of a floodwave.Figure 9-12 illustrates this concept.whereI= reach inflowO = reach outflowD = working value discharge or simply workingdischarge(3) The wedge storage (WS) may be computed inthe following two ways: As in the Muskingum techniquewhere X is a weighting factor and K is reach travel time:WS KX (I O)(9-24)A long routing reach should be subdivided intosubreaches so that the travel time through each subreachis approximately equal to the routing interval t. That is:or using the working discharge (D) concept:WS K (D O)(9-25)Number of subreachesK∆tequating and solving for O:K (D O) KX (I O)(9-26)This assumes that factors such as channel geometry androughness have been taken into consideration in determiningthe length of the routing reach and the travel time K.9-15

EM 1110-2-141731 Aug 94Figure 9-12. Illustration of the “working discharge” conceptorLetODX1 X(I D)(9-27)R S (1 X) 0.5D∆t(9-30)The continuity equation may be approximated by:S 2S 1∆t0.5 (I 1I 2) 0.5 (O 1O 2)(9-28)where R is termed the “working value of storage” orsimply working storage and represents an index of thetrue natural storage. Equation 9-29 may therefore bewritten:R 2R 10.5∆t (I 1I 2) D 1∆t(9-31)whereS= storagetransposing t results in the equation used in routingcomputations:t = time incrementSubstituting Equation 9-27 into 9-28 and appending theappropriate subscripts to denote beginning and end ofperiod and performing the appropriate algebra yields:0.5∆t(I 1I 2) [S 1(1 X) 0.5D 1∆t][S 2(1 X) 0.5D 2∆t](9-29)R 2∆tR 1∆t0.5 (I 1I 2) D 1(9-32)The form of the relationship for R (working discharge) isanalogous to storage indication in the modified puls procedure.R 2 / t may be computed from information knownat the beginning of a routing interval. The outflow at theend of the routing interval may then be determined from a9-16

EM 1110-2-141731 Aug 94rating curve of working storage versus working discharge.The cycle is then repeated stepping forward in time.(4) The solution scheme using this concept requiresdevelopment of a rating curve of working storage versusworking discharge as stated above. The following columnheadings are helpful in developing the function whenstorage-outflow data are available.1 2 3SWorkingStorage (S) (1 X)∆tDischarge (D)4 5D S(1 X)2 ∆tD2• Repeat process until finished.e. Muskingum-Cunge channel routing. TheMuskingum-Cunge channel routing technique is a nonlinearcoefficient method that accounts for hydrograph diffusionbased on physical channel properties and theinflowing hydrograph. The advantages of this methodover other hydrologic techniques are the parameters of themodel are more physically based; the method has beenshown to compare well against the full unsteady flowequations over a wide range of flow situations (Ponce1983 and Brunner 1989); and the solution is independentof the user-specified computation interval. The majorlimitations of the Muskingum-Cunge technique are that itcannot account for backwater effects, and the methodbegins to diverge from the full unsteady flow solutionwhen very rapidly rising hydrographs are routed throughflat channel sections.(1) Development of equations.(5) Column 2 of the tabulation is obtained from column1 by using an appropriate conversion factor andappropriate X. The conversion factor of 1 acre-ft/hour =12.1 cfs is useful in this regard. Column 5 is the sum ofcolumns 2 and 4. Column 3 is plotted against column 5on cartesian coordinate paper and a curve drawn throughthe plotted points. This represents the working dischargeworkingoutflow rating curve. An example curve isshown in Figure 9-13.(6) The routing of a hydrograph can be performed asthe one shown in Table 9-2. The procedure, in narrativeform is:(a) The basic formulation of the equations is derivedfrom the continuity Equation 9-33 and the diffusion formof the momentum Equation 9-34:∂A∂t∂Q∂xS fS o∂Y∂x(9-33)q l(9-34)(b) By combining Equations 9-33 and 9-34 andlinearizing, the following convective diffusion equation isformulated (Miller and Cunge 1975):• Conditions known at time 1: I 1 , O 1 , D 1 , and R 1 / t.• At time 2, only I 2 is known, therefore:∂Q∂tc ∂Q∂xµ ∂2 Q∂x 2cq L(9-35)R 2∆tR 1∆t0.5 (I 1I 2) D 1whereQ = discharge, in cubic feet per second• Enter working storage, working discharge function,and read out D 2 .• Calculate O 2 as follows:O 2D 2X1 X (I 2D 2)AtxY= flow area, in square feet= time, in seconds= distance along the channel, in feet= depth of flow, in feet9-17

EM 1110-2-141731 Aug 94Figure 9-13. Rating curve for working R&D routingq L = lateral inflow per unit of channel lengthwhere B is the top width of the water surface. The convectiveµQ(9-37)2BS odiffusion Equation 9-35 is the basis for theS f = friction slopeMuskingum-Cunge method.S o = bed slope(c) In the original Muskingum formulation, withc = the wave celerity in the x direction as definedlateral inflow, the continuity Equation 9-33) is discretizedon the x-t plane (Figure 9-14) to yield:belowQ n 1(9-38)j 1 C 1Q nj C 2Q n 1j C 3Q nj 1 C 4Q LThe wave celerity (c) and the hydraulic diffusivity (µ) areexpressed as follows:It is assumed that the storage in the reach is expressed asdQthe classical Muskingum storage:c(9-36)dAS K [XI (1 X)O](9-39)9-18

EM 1110-2-141731 Aug 94Table 9-2Working R&D Routing ExampleTimehrInflowcfsAverageInflowcfsK∆ tDcfscfsOcfs3,000 7,100 3,000 3,0003,1303 3,260 7,230 3,100 3,0603,4456 3,630 7,575 3,300 3,2203,8259 4,020 8,100 3,800 3,7454,25012 4,480 8,550 4,400 4,420whereS = channel storageK = cell travel time (seconds)X = weighting factorI = inflowO = outflowTherefore, the coefficients can be expressed as follows:C 1C 2∆tK∆tK∆tK∆tK2X2(1 X)2X2(1 X)∆t2(1 X)KC 3∆t2(1 X)KQ Lq L∆XC 4∆tK2( ∆tK )2(1 X)(d) In the Muskingum equation the amount of diffusionis based on the value of X, which varies between 0.0and 0.5. The Muskingum X parameter is not directlyrelated to physical channel properties. The diffusionobtained with the Muskingum technique is a function ofhow the equation is solved and is therefore considerednumerical diffusion rather than physical. Cunge evaluatedthe diffusion that is produced in the Muskingum equationand analytically solved for the following diffusioncoefficient:µ nc ∆x ⎛ 1 ⎞⎜ X ⎟⎠ ⎝ 2(9-40)In the Muskingum-Cunge formulation, the amount ofdiffusion is controlled by forcing the numerical diffusionto match the physical diffusion of the convective diffusionEquation 9-35. This is accomplished by setting Equations9-37 and 9-40 equal to each other. TheMuskingum-Cunge equation is therefore considered anapproximation of the convective diffusion Equation 9-35.As a result, the parameters K and X are expressed asfollows (Cunge 1969 and Ponce and Yevjevich 1978):9-19

EM 1110-2-141731 Aug 94Figure 9-14. Discretization of the continuity equation on x-t planeK∆Xc(9-41)averaging scheme is used to solve for c, B, and Q. Thisprocess has been described in detail by Ponce (1986).X12⎛⎜1⎝⎞QBS ⎟oc∆x⎠(2) Solution of the equations.(9-42)(a) The method is nonlinear in that the flow hydraulics(Q, B, c), and therefore the routing coefficients(C 1 , C 2 , C 3 , and C 4 ) are recalculated for every x distancestep and t time step. An iterative four-point(b) Values for t and x are chosen for accuracyand stability. First, t should be evaluated by looking atthe following three criteria and selecting the smallestvalue: (1) the user-defined computation interval, (2) thetime of rise of the inflow hydrograph divided by 20( Tr / 20 ), and (3) the travel time through the channel reach.Once t is chosen, x is defined as follows:∆xc∆t(9-43)9-20

EM 1110-2-141731 Aug 94but x must also meet the following criteria to preserveconsistency in the method (Ponce 1983):∆x < 1 2⎛⎜c∆t⎝Q ⎞oBS ⎟oc⎠(9-44)where Q o is the reference flow and Q B is the baseflowtaken from the inflow hydrograph as:Q oQ B0.50 (Q peakQ B)(3) Data requirements.(a) Data for the Muskingum-Cunge method consist ofthe following:• Representative channel cross section.• Reach length, L.• Manning roughness coefficients, n (for mainchannel and overbanks).• Friction slope (S f ) or channel bed slope (S o ).(b) The method can be used with a simple cross section(i.e., trapezoid, rectangle, square, triangle, or circularpipe) or a more detailed cross section (i.e., cross sectionswith a left overbank, main channel, and a right overbank).The cross section is assumed to be representative of theentire routing reach. If this assumption is not adequate,the routing reach should be broken up into smaller subreacheswith representative cross sections for each. Reachlengths are measured directly from topographic maps.Roughness coefficients (Manning’s n) must be estimatedfor main channels as well as overbank areas. Ifinformation is available to estimate an approximate energygrade line slope (friction slope, S f ), that slope should beused instead of the bed slope. If no information is availableto estimate the slope of the energy grade line, thechannel bed slope should be used.(4) Advantages and limitations. The Muskingum-Cunge routing technique is considered to be a nonlinearcoefficient method that accounts for hydrograph diffusionbased on physical channel properties and the inflowinghydrograph. The advantages of this method over otherhydrologic techniques are: the parameters of the modelare physically based, and therefore this method will makefor a good ungauged routing technique; several studieshave shown that the method compares very well with thefull unsteady flow equations over a wide range of flowconditions (Ponce 1983 and Brunner 1989); and the solutionis independent of the user-specified computationinterval. The major limitations of the Muskingum-Cungetechnique are that the method can not account for backwatereffects, and the method begins to diverge from thecomplete unsteady flow solution when very rapidly risinghydrographs (i.e., less than 2 hr) are routed through flatchannel sections (i.e., channel slopes less than 1 ft/mile).For hydrographs with longer rise times (T r ), the methodcan be used for channel reaches with slopes less than1 ft/mile.9-4. Applicability of Routing Techniquesa. Selecting the appropriate routing method. Withsuch a wide range of hydraulic and hydrologic routingtechniques, selecting the appropriate routing method foreach specific problem is not clearly defined. However,certain thought processes and some general guidelines canbe used to narrow the choices, and ultimately the selectionof an appropriate method can be made.b. Hydrologic routing method. Typically, in rainfallrunoffanalyses, hydrologic routing procedures are utilizedon a reach-by-reach basis from upstream to downstream.In general, the main goal of the rainfall-runoff study is tocalculate discharge hydrographs at several locations in thewatershed. In the absence of significant backwatereffects, the hydrologic routing models offer theadvantages of simplicity, ease of use, and computationalefficiency. Also, the accuracy of hydrologic methods incalculating discharge hydrographs is normally well withinthe range of acceptable values. It should be remembered,however, that insignificant backwater effects alone do notalways justify the use of a hydrologic method. There aremany other factors that must be considered when decidingif a hydrologic model will be appropriate, or if it is necessaryto use a more detailed hydraulic model.c. Hydraulic routing method. The full unsteadyflow equations have the capability to simulate the widestrange of flow situations and channel characteristics.Hydraulic models, in general, are more physically basedsince they only have one parameter (the roughness coefficient)to estimate or calibrate. Roughness coefficients canbe estimated with some degree of accuracy from inspectionof the waterway, which makes the hydraulic methodsmore applicable to ungauged situations.d. Evaluating the routing method. There are severalfactors that should be considered when evaluating whichrouting method is the most appropriate for a given9-21

EM 1110-2-141731 Aug 94situation. The following is a list of the major factors thatshould be considered in this selection process:(1) Backwater effects. Backwater effects can beproduced by tidal fluctuations, significant tributaryinflows, dams, bridges, culverts, and channel constrictions.A floodwave that is subjected to the influences ofbackwater will be attenuated and delayed in time. Of thehydrologic methods discussed previously, only the modifiedpuls method is capable of incorporating the effects ofbackwater into the solution. This is accomplished bycalculating a storage-discharge relationship that has theeffects of backwater included in the relationship. Storagedischargerelationships can be determined from steadyflow-water surface profile calculations, observed watersurface profiles, normal depth calculations, and observedinflow and outflow hydrographs. All of these techniques,except the normal depth calculations, are capable ofincluding the effects of backwater into the storage-dischargerelationship. Of the hydraulic methods discussedin this chapter, only the kinematic wave technique is notcapable of accounting for the influences of backwater onthe floodwave. This is due to the fact that the kinematicwave equations are based on uniform flow assumptionsand a normal depth downstream boundary condition.(2) Floodplains. When the flood hydrograph reachesa magnitude that is greater than the channels carryingcapacity, water flows out into the overbank areas.Depending on the characteristics of the overbanks, theflow can be slowed greatly, and often ponding of watercan occur. The effects of the floodplains on the floodwavecan be very significant. The factors that are importantin evaluating to what extent the floodplain willimpact the hydrograph are the width of the floodplain, theslope of the floodplain in the lateral direction, and theresistance to flow due to vegetation in the floodplain. Toanalyze the transition from main channel to overbankflows, the modeling technique must account for varyingconveyance between the main channel and the overbankareas. For 1-D flow models, this is normally accomplishedby calculating the hydraulic properties of the mainchannel and the overbank areas separately, then combiningthem to formulate a composite set of hydraulic relationships.This can be accomplished in all of the routingmethods discussed previously except for the Muskingummethod. The Muskingum method is a linear routingtechnique that uses coefficients to account for hydrographtiming and diffusion. These coefficients are usually heldconstant during the routing of a given floodwave. Whilethese coefficients can be calibrated to match the peakflow and timing of a specific flood magnitude, they cannot be used to model a range of floods that may remain inbank or go out of bank. When modeling floods throughextremely flat and wide floodplains, the assumption of1-D flow in itself may be inadequate. For this flow condition,velocities in the lateral direction (across the floodplain)may be just as predominant as those in thelongitudinal direction (down the channel). When thisoccurs, a two-dimensional (2-D) flow model would give amore accurate representation of the physical processes.This subject is beyond the scope of this chapter. Formore information on this topic, the reader is referred toEM 1110-2-1416.(3) Channel slope and hydrograph characteristics.The slope of the channel will not only affect the velocityof the floodwave, but it can also affect the amount ofattenuation that will occur during the routing process.Steep channel slopes accelerate the floodwave, while mildchannel slopes are prone to slower velocities and greateramounts of hydrograph attenuation. Of all the routingmethods presented in this chapter, only the completeunsteady flow equations are capable of routing floodwavesthrough channels that range from steep toextremely flat slopes. As the channel slopes becomeflatter, many of the methods begin to break down. Forthe simplified hydraulic methods, the terms in themomentum equation that were excluded become moreimportant in magnitude as the channel slope is decreased.Because of this, the range of applicable channel slopesdecreases with the number of terms excluded from themomentum equation. As a rule of thumb, the kinematicwave equations should only be applied to relatively steepchannels (10 ft/mile or greater). Since the diffusion waveapproximation includes the pressure differential term inthe momentum equation, it is applicable to a wider rangeof slopes than the kinematic wave equations. The diffusionwave technique can be used to route slow risingfloodwaves through extremely flat slopes. However,rapidly rising floodwaves should be limited to mild tosteep channel slopes (approximately 1 ft/mile or greater).This limitation is due to the fact that the accelerationterms in the momentum equation increase in magnitude asthe time of rise of the inflowing hydrograph is decreased.Since the diffusion wave method does not include theseacceleration terms, routing rapidly rising hydrographsthrough flat channel slopes can result in errors in theamount of diffusion that will occur. While “rules ofthumb” for channel slopes can be established, it should berealized that it is the combination of channel slope andthe time of rise of the inflow hydrograph together thatwill determine if a method is applicable or not.(a) Ponce and Yevjevich (1978) established a numericalcriteria for the applicability of hydraulic routing9-22

EM 1110-2-141731 Aug 94techniques. According to Ponce, the error due to the useof the kinematic wave model (error in hydrograph peakaccumulated after an elapsed time equal to the hydrographduration) is within 5 percent, provided the followinginequality is satisfied:whereTTS ou od o≥ 171S ou od o= hydrograph duration, in seconds= friction slope or bed slope= reference mean velocity= reference flow depth(9-45)When applying Equation 9-45 to check the validity ofusing the kinematic wave model, the reference valuesshould correspond as closely as possible to the averageflow conditions of the hydrograph to be routed.(b) The error due to the use of the diffusion wavemodel is within 5 percent, provided the following inequalityis satisfied:⎛ ⎞TS ⎜⎜⎝ g ⎟⎟⎠od o1/2≥ 30(9-46)where g = acceleration of gravity. For instance, assumeS o = 0.001, u o = 3 ft/s, and d o = 10 ft. The kinematicwave model will apply for hydrographs of duration largerthan 6.59 days. Likewise, the diffusion wave model willapply for hydrographs of duration larger than 0.19 days.(c) Of the hydrologic methods, the Muskingum-Cunge method is applicable to the widest range of channelslopes and inflowing hydrographs. This is due to thefact that the Muskingum-Cunge technique is an approximationof the diffusion wave equations, and therefore canbe applied to channel slopes of a similar range in magnitude.The other hydrologic techniques use an approximaterelationship in place of the momentum equation. Experiencehas shown that these techniques should not beapplied to channels with slopes less than 2 ft/mi.However, if there is gauged data available, some of theparameters of the hydrologic methods can be calibrated toproduce the desired attenuation effects that occur in veryflat streams.(4) Flow networks. In a dendritic stream system, ifthe tributary flows or the main channel flows do not causesignificant backwater at the confluence of the twostreams, any of the hydraulic or hydrologic routing methodscan be applied. If significant backwater does occur atthe confluence of two streams, then the hydraulic methodsthat can account for backwater (full unsteady flow anddiffusion wave) should be applied. For full networks,where the flow divides and possibly changes directionduring the event, only the full unsteady flow equationsand the diffusion wave equations can be applied.(5) Subcritical and supercritical flow. During aflood event, a stream may experience transitions betweensubcritical and supercritical flow regimes. If the supercriticalflow reaches are long, or if it is important to calculatean accurate stage within the supercritical reach, thetransitions between subcritical and supercritical flowshould be treated as internal boundary conditions and thesupercritical flow reach as a separate routing section.This is normally accomplished with hydraulic routingmethods that have specific routines to handle supercriticalflow. In general, none of the hydrologic methods haveknowledge about the flow regime (supercritical or subcritical),since hydrologic methods are only concerned withflows and not stages. If the supercritical flow reaches areshort, they will not have a noticeable impact on the dischargehydrograph. Therefore, when it is only importantto calculate the discharge hydrograph, and not stages,hydrologic routing methods can be used for reaches withsmall sections of supercritical flow.(6) Observed data. In general, if observed data arenot available, the routing methods that are more physicallybased are preferred and will be easier to apply.When gauged data are available, all of the methods shouldbe calibrated to match observed flows and/or stages asbest as possible. The hydraulic methods, as well as theMuskingum-Cunge technique, are considered physicallybased in the sense that they only have one parameter(roughness coefficient) that must be estimated or calibrated.The other hydrologic methods may have morethan one parameter to be estimated or calibrated. Manyof these parameters, such as the Muskingum X and thenumber of subreaches (NSTPS), are not related directly tophysical aspects of the channel and inflowing hydrograph.Because of this, these methods are generally not used inungauged situations. The final choice of a routing modelis also influenced by other factors, such as the requiredaccuracy, the type and availability of data, the type ofinformation desired (flow hydrographs, stages, velocities,etc.), and the familiarity and experience of the user with agiven method. The modeler must take all of these factors9-23

EM 1110-2-141731 Aug 94into consideration when selecting an appropriate routingtechnique for a specific problem. Table 9-3 contains alist of some of the factors discussed previously, alongwith some guidance as to which routing methods areappropriate and which are not. This table should be usedas guidance in selecting an appropriate method for routingdischarge hydrographs. By no means is this table allinclusive.Table 9-3Selecting the Appropriate Channel Routing TechniqueFactors to consider in the selection of arouting technique.1. No observed hydrograph data availablefor calibration.2. Significant backwater that will influencedischarge hydrograph.3. Flood wave will go out of bank into theflood plains.Methods that are appropriate for thisspecific factor.* Full Dynamic Wave* Diffusion Wave* Kinematic Wave* Muskingum-Cunge* Full Dynamic Wave* Diffusion Wave* Modified Puls* Working R&D* All hydraulic and hydrologic methods thatcalculate hydraulic properties of mainchannel separate from overbanks.Methods that are not appropriate for thisfactor.* Modified Puls* Muskingum* Working R&D* Kinematic Wave* Muskingum* Muskingum-Cunge* Muskingum4. Channel slope > 10 ft/mile* All methods presented * NoneandTS ou od o≥ 1715. Channel slopes from 10 to 2 ft/mile andTS ou od o< 1716. Channel slope < 2 ft/mile and1/2⎛ ⎞TS ⎜⎜⎝ g ⎟⎟⎠o≥ 30d o* Full Dynamic Wave* Diffusion Wave* Muskingum-Cunge* Modified Puls* Muskingum* Working R&D* Full Dynamic Wave* Diffusion Wave* Muskingum-Cunge* Kinematic Wave* Kinematic Wave* Modified Puls* Muskingum* Working R&D7. Channel slope < 2 ft/mile and⎛ ⎞TS ⎜⎜⎝ g ⎟⎟⎠od o1/2

EM 1110-2-141731 Aug 94Chapter 10Multisubbasin Modeling10-1. Generala. The foregoing chapters have described the variouscomponents of the watershed-runoff process. This chapterdescribes how these components are combined into a riverbasin model, as shown in Figure 10-1. The componentscan be thought of as building blocks for a comprehensiveriver basin analysis model or detailed pieces of a smallerwatershed model. Those components may include small,component watersheds (subbasins) which are integralpieces of the larger basin; river reaches which connect thesubbasins; confluences of rivers; lakes; and various manmadefeatures. There are many man-made features in ariver basin which affect the rate and volume of runoff.Some of the main features are reservoirs, urbanization,diversions, channel improvements, levees, and pumps.The multisubbasin model refers to the collection of allthese natural and man-made components which describethe runoff process in a river basin.b. This chapter describes the components of a riverbasin and provides criteria for subdividing basins intothese components. Once the river basin model is built, ittoo must be calibrated and verified. Even though each ofthese components may have been individually calibratedas described in the previous chapters, the whole modelmust also be calibrated. The synergistic effect of all thecomponents acting together may produce different resultsthan their individual calibrations. This can be due to thespatial variation in precipitation and runoff, and/or nonlinearitiesin the river routing process. Thus, methods forcalibrating and verifying a river basin model are alsoprovided here.10-2. General Considerations for Selecting BasinComponentsa. General considerations for selecting the size andlocation of basin components are discussed in this paragraph.More detailed hydrologic, hydraulic, and projectengineering and management criteria for sizing and locatingbasin components are given in paragraph 10-3. Thereare three general considerations for selecting and sizingbasin components: where is information needed; whereFigure 10-1. Components of a river basin model10-1

EM 1110-2-141731 Aug 94are data available; and where do hydrologic/hydraulicconditions change?b. The first consideration addresses why the analysisis being made. It includes input from all study teammembers such as economists, engineers, and environmentalists.Typical places where information is needed arelocations subject to flood damage, sites where projects areproposed, wetlands, and archeological sites. The entirestudy team should be involved in identifying these informationneeds.c. The second consideration addresses how muchinformation one has to perform the analysis. Data availabilityis key to successful modeling. Thus, the basinmodel must be structured to take advantage of availabledata. Some of the primary types of information whichinfluence basin subdivision are stream gauges, precipitationgauges, river-geometry surveys, and previous studies.By subdividing a river basin at these data locations, themodel can be calibrated much more easily.d. The third consideration addresses where hydrologicand hydraulic processes change so that they can beadequately represented by the basin components. Hydrologicmodeling assumes that uniform conditions prevailwithin each of the basin components. The term “lumpedparameter” refers to that condition where the same processis assumed to occur equally over the entire component.Some examples are a watershed with a mixture ofurban and rural areas and a river which goes from anarrow canyon into a broad floodplain. One cannotassume that the same hydrologic or hydraulic parameterscan describe the runoff process in areas where the physicalprocesses are quite variable. One of the keys to successfulhydrologic modeling is to select basin componentswhich are “representative” of the heterogeneous processesof nature.10-3. Selection of Hydrograph ComputationLocationsSubbasins, channel reaches, and confluences are the generallocations where hydrographs are computed. Subbasinsare usually part of all the other basin componentdecisions. That is, whenever a computational point isidentified in a basin, the local tributary watershed runoffis also computed at that point. For example, the landsurface runoff characteristics of a basin may be constant,but the river hydraulic characteristics may change in themiddle of that basin. In that case, two subbasins for landsurface runoff may be used so that the upstream runoffcan be added to the hydrograph before routing through thedownstream reach. Too large a subbasin may “averageout” important watershed and river dynamics. Too smalla subbasin increases data handling and processingexpenses.a. Hydrologic criteria. Variation in precipitationand infiltration are two important hydrologic criteria. Adetailed explanation of both criteria follows.(1) Precipitation variation. The subbasin runoffcomputation process involves determination of subbasinaverageprecipitation and infiltration, then transformationof the resulting moisture excess into streamflow at theoutlet of the subbasin. Thus, subbasins should be sized tocapture variations in precipitation and infiltration. Itwould be desirable to have a recording precipitation gaugein every subbasin, but usually there are many more subbasinsneeded than there are precipitation gauges. Themain consideration must be that there are enough subbasinsto capture the spatial variation in precipitation.(2) Infiltration variation. Infiltration characteristicsof the land surface are a major part of subbasin selection.The desire is to have a subbasin with uniform infiltrationcharacteristics. Thus, forested areas should be separatedfrom grasslands, urban from rural, agricultural from natural,etc. Different soil types also have different infiltrationcharacteristics. The general consideration is land use.Where land use changes, infiltration characteristicschange. The problem is that watersheds are not made upof uniform land uses. The objective is to select areaswhich are “representative” of a particular infiltration condition.This average infiltration consideration becomesespecially difficult in urban areas which are inherently amixture of many land uses. The basic concept is thesame: subdivide the watershed into like-watershedresponses. The urban case requires consideration of additionalfeatures such as:(a) presence of storm drains,(b) roof downspouts directly connected to the street,(c) large parks or shopping centers, and(d) local drainage requirements.Oftentimes these conditions will change with differentland developers and city/county ordinances. In the urbancase, it is often desirable to compute runoff from perviousand impervious areas separately within the same subbasin.10-2

EM 1110-2-141731 Aug 94This procedure allows infiltration characteristics to vary,capturing the two main land-use characteristics in anurban subbasin.(3) Runoff variation. The land-use condition alsodetermines the rate of runoff. Land slope is also a majorfactor determining the rate of runoff. These land-use andslope conditions are represented by different unit hydrographor kinematic wave land surface runoff parameters.Because only one set of parameters may be specified for asubbasin, it is important to have them “representative” ofthe average subbasin characteristics. Ideally, a subbasinwould have similar land use, soils, slope, and streamnetworkpatterns. Urbanization obviously can make drasticchanges to the runoff network.b. Hydraulic criteria. Natural channel variation andman-made variations are two important hydraulic criteria.A detailed explanation of both criteria follows.(1) Natural channel variation. Hydraulic criteria referto where the river or stream channel changes in a significantenough manner to affect the routing processes.Examples of these hydraulic controls are constrictions inthe channel, major changes in the slope of the channel,broadening of the channel into a floodplain, and confluencewith tributaries. Because only a single routingmethod can be used in a single river reach, that reachshould have reasonably uniform hydraulic characteristics.Tributaries are important because many of the river routingprocesses are nonlinear and depend upon the magnitudeof the flow. Thus, where tributaries increase theflow significantly, separate routing reaches should beincorporated upstream and downstream of the tributary.(2) Man-made variations. Manmade structures andmodifications of the river channels usually have a majorimpact on the flow routing process. Examples of manmadefeatures are bridges, floodplain encroachments,diversions, pumps, dams, weirs, culverts, levees andfloodwalls, and channel clearing and cleaning. All ofthese changes to the natural river system usually have amajor impact on the routing process; and thus, channelswhere these variations occur should be modeledseparately.c. Information needs criteria. Information needs aredependent on the purpose of the project. If flood damagereduction is a project purpose, then flood damagelocations (cities, towns, industrial sites, etc.) must beincluded as a hydrograph computation point. Such pointsdo not enhance the hydrologic/hydraulic computations, butthey are necessary to provide the flow and stageinformation for computing flood damage. The same istrue for any other project purpose which depends on riverflow/stage for its evaluation.d. Data availability criteria. Data are necessary tocalibrate the river basin model to observed historicalevents. Data are usually insufficient, so every gaugelocation must be carefully reviewed and used to the fullestextent possible. Stage or flow gauging stations are themost important for determining hydrograph computationpoints.(1) A subbasin/river reach break point will usuallybe made at every stream gauge so that the computedhydrograph can be compared with the observed hydrographat that location.(2) Sometimes special river basin subdivisions willbe made differently than the general river basin model tomake use of the data for calibration of subbasin runoffparameters. That is, the total area contributing to a gaugemay be used as a single subbasin for the purposes ofcalibrating watershed runoff parameters to different sizedbasin areas. Those special, large basins are only used forthe subbasin parameter calibration and regionalizationneeds; those large subbasins are disaggregated back intotheir logical components for the generalized river basinmodel.e. General criteria.(1) These considerations must also take into accountthe practical engineering economy of the analysis and thepurpose for which the study is being made. There isalways a tradeoff between a detailed representation of ariver basin and the practical information needs andresources available for the study. For instance, the surchargingof a culvert may be critical to local informationneeds in an urban area, but it may have very little effecton the peak discharge farther downstream in a large riverbasin. Thus, the watershed modeler’s job is to weighthese information needs against the watershed and riverdynamics to obtain a representation of the basin whichprovides the needed information within a reasonable costand time.(2) The river basin modeler should be careful toinclude all major components in the river basin model.That is, do not lump together too large of an area justbecause data are not available for a particular area of theriver basin. A logical physical network of subbasins andriver reaches should still be maintained even though dataare not available. One example is to treat major10-3

EM 1110-2-141731 Aug 94tributaries as separate components rather than lump themin with the local tributary area of the main river. Buildinga logical basin model will help identify where additionaldata are needed, and then it will be ready to includethe data when available.10-4. Calibration of Individual ComponentsThe first step in the calibration of the river basin model isthe calibration of the individual components whereobserved precipitation runoff data are available; such as,calibration of the subbasin infiltration, runoff transformation,base-flow parameters, and calibration of the riverreach flood routing parameters, as described in the precedingchapters. Obviously, not all of the subbasins androuting reaches in the river basin will have gauged eventsfor this calibration. Thus, it is common to take calibrationresults from gauged basins and regionalize that datafor use in ungauged areas. The regionalization processrelates the best estimate of the parameters from thegauged locations to readily measurable basin characteristics.That relationship (usually a regression equation) isthen used in the ungauged area where the same basincharacteristics can be measured and the relation used toestimate the parameters. The ungauged area analysis andregionalization process are described in Chapter 16. Inperforming regional analysis, it is imperative to have theparameters make good physical sense and not to use theregional equations outside the range of the data fromwhich they were derived.10-5. Calibration of Multisubbasin ModelThe river basin model must be calibrated as a wholebecause the individual runoff and routing processes werecalibrated for the gauged subbasins only. Also, the nonlinearitiesof the runoff and routing processes will causedifferences from the individual component calibrations.The river basin model is calibrated using the observedprecipitation data and streamflow measurements at alllocations in the basin. The primary points of comparisonare at gauges on the main stem of the river. Severalsubbasins, routing reaches, and confluences will undoubtedlyhave been used to compute a hydrograph at thegauge location. Several considerations and methods arenecessary to calibrate the basin model. The timing, magnitude,and volume of the hydrograph must be calibrated.Sometimes one or the other is more important dependingon the purpose of the study; e.g., peak flow is importantfor channel design, and volume is important for reservoiranalysis.a. Analysis of components. There are several componentsof the multisubbasin model. A detailed analysisof these components follows.(1) General. The total flow at a gauge location isthe result of several upstream processes. Sometimes theindividual upstream components of the basin are routed tothe gauge location to compare their relative contributionto the total hydrograph. Caution must be exercised to besure that nonlinearities in the upstream routing processesdo not adversely affect the component parts when routedindividually. The component parts should be evaluatedfor timing, magnitude, and volume with respect to theobserved hydrograph. Any inconsistent results should betraced back to their origin. The source of the problemmay be in a precipitation gauge, runoff conversion processes,or the routing. At this point, it is hydrologicdetective work to determine and resolve the sources ofinconsistencies. Do not be too quick to blame poorresults on data; check the river basin model formulationfirst, and be sure that the tributary drainage area iscorrect.(2) Analysis of volume errors. Volume errors arethe result of incorrect precipitation, detention and retentionstorage, and infiltration calculations. Check basinaverage precipitation assumptions with other gauges in thearea. Check the infiltration method assumptions for theungauged portions of the tributary basin.(3) Analysis of timing errors. Timing errors can bethe result of almost every part of the precipitation-runoffprocess. The precipitation and infiltration rates and patternscan cause differences in timing. More commonly,the unit hydrograph or kinematic wave land surface runoffand channel routing parameters are reviewed first. Thechannel routing parameter sensitivity can be easily analyzedby looking at the routed components separately.Always compare physically based estimates of traveltimes with the model results.(4) Analysis of magnitude errors. Errors in peakflow can be caused by inaccurate precipitation intensities,incorrect subbasin runoff parameters, incorrect timing oftributaries, or the wrong amount of attenuation in thechannel routing reaches. Precipitation distributions shouldbe reviewed to ensure that periods of high intensity havenot been averaged out by weighing the measurements ofmore than one recording gauge. Unit hydrograph andkinematic wave parameters should be checked with10-4

EM 1110-2-141731 Aug 94respect to physical characteristics of the basin. Routingreaches should be reviewed for unreasonable amounts ofattenuation.b. Consistency checks. In performing the analysesof components, it is easy to independently change theindividual parameters to obtain the desired result. But,such change should not be made without maintainingconsistency with respect to all like land uses, soils, channels,etc. throughout the river basin. For example, if theinfiltration rate is changed in one subbasin to improve thefit with the observed flows, then a corresponding changemust be made in all subbasins with the same or similarland use and soil types. Consistency in runoff and routingparameters must be maintained throughout the entire riverbasin. Oftentimes a compromise is reached where thechange helps in one location but increases the error inanother location.10-6. Verification of the Multisubbasin ModelVerification is the name of a procedure for independentchecking of the parameters selected for the basin components,that is, checking the performance of the model withdata not used in calibration. Independent checks of theparameters can also be made with simple measures of thephysical process.always hard to do because there never seem to be enoughdata for adequate calibration. Good verification resultsgive high credibility to the model and, thus, should beperformed if at all possible. Floods selected for verificationshould be of the size and type for which the projectis being designed. If the verification results are not good,then further calibration of the model must be made usingthe verification flood event in addition to the previouscalibration data. Any changes in the parameters must bejustified with respect to all storms and the physics of theprocess.b. Physics of the runoff process. The physics of therunoff and routing processes are often used to help withthe calibration of the basin components. They are equallyhelpful in checking the river basin model results.Approximate travel times in subbasin and routing reachescan be calculated by Manning’s equation. Saturated soilinfiltration rates can be checked with modeled losses.Runoff per unit area (e.g., cfs per square mile) can becalculated for various points in the basin and checked forconsistency with respect to precipitation. Volume checkscan be made to verify overall continuity of moisture input,outflows, and water still in the system. Each one of thesemeasures helps build confidence in the model’s representationof the basin and helps gain insight into the hydrologicand hydraulic processes of the basin.a. Other storm runoff data not used in calibration.Reserving some data for verification analysis only is10-5

PART IIIMETHODS FOR FLOOD-RUNOFFANALYSIS

EM 1110-2-141731 Aug 94Chapter 11Simplified Techniques11-1. Introductiona. Simplified techniques include numerousapproaches for determining the approximate magnitude ofthe peak flow expected for events of varying frequency.These approaches are useful for an approximate answerwith a minimum of effort. They are often used inungauged drainage areas.b. This chapter describes the role of simplified techniquesfor flood-runoff analysis. Various methods forestimating the peak flow associated with varying frequencieswill be discussed including the rational method,regression techniques, SCS methods, and maximumexpected envelop curves.11-2. Rational Methoda. The so-called rational method is a popular, easyto-usetechnique for estimating peak flow in any smalldrainage basin having mixed land use. It generally shouldnot be used in basins larger than 1 square mile. The peakflow can be calculated by the following equation:Qwhere:QCIA= peak flow, in cubic feet per second(11-1)for runoff to travel from the most distant point of thewatershed to the watershed outlet. T c influences the shapeand peak of the runoff hydrograph and is a parameterused in many simplified techniques. Numerous methodsexist in the literature for estimating T c . The SCS hasdeveloped a method that takes a physically basedapproach to calculating T c , which can be found in Chapter2 of SCS (1986).d. Use of the rational method for large drainageareas should be discouraged because of the greater complexityof land use and drainage pattern and the unlikelihoodof having uniform rainfall intensity for a durationequal to the time of concentration. The method assumesthat the peak flow occurs from uniform rainfall intensityover the entire area once every portion of the basin iscontributing to runoff at the outlet.11-3. Regional Frequency Analysisa. Regional frequency analysis usually involvesregression analysis of gauged watersheds within the generalregion. Through this very powerful technique, sufficientlyreliable equations can often be derived for peakflow of varying frequency given quantifiable physicalbasin characteristics and rainfall intensity for a specificduration. Once these equations are developed, they canthen be applied to ungauged basins within the sameregion.b. A regional analysis usually consists of the followingsteps:(1) Select components of interest, such as mean andpeak discharge.CIA= runoff coefficient= rainfall intensity, in inches per hour= drainage area, in acres(2) Select definable basin characteristics of gaugedwatershed: drainage area, slope, etc.(3) Derive prediction equations with single- or multiple-linearregression analysis.b. The coefficient is the proportion of rainfall thatcontributes to runoff. Table 11-1 is an example of therelationship between this coefficient and land use. Inbasins having a significant nonhomogeneity of land use,an average coefficient can easily be determined by multiplyingthe percentage of each land use in the basin by itsappropriate coefficient from Table 11-1.c. The rainfall intensity is specifically defined for anevent or the frequency of interest and for a duration equalto or greater than the time of concentration of the watershed.Time of concentration (T c ) is defined as the time(4) Map and explain the residuals (differencesbetween computed and observed values) that constitute“unexplained variances” in the statistical analysis on aregional basis.c. This procedure for development of the regressionequation from gauged basin data is illustrated in Figure11-1. The equation can then be used in ungaugedareas within the same region and for data of similar magnitudeto that used in the development process. Much11-1

EM 1110-2-141731 Aug 94Table 11-1Typical C Coefficients (for 5- to 10-year Frequency Design)DESCRIPTIONOF AREABusinessRUNOFFCOEFFICIENTDowntown areas 0.70 - 0.95Neighborhood area 0.50 - 0.70ResidentialSingle-family areas 0.30 - 0.50Multiunits, detached 0.40 - 0.60Multiunits, attached 0.60 - 0.75Residential (suburban) 0.25 - 0.40Apartment dwelling areas 0.50 - 0.70IndustrialLight areas 0.50 - 0.80Heavy areas 0.60 - 0.90Parks, cemeteries 0.10 - 0.25Playgrounds 0.20 - 0.35Railroad yard areas 0.20 - 0.40Unimproved areas 0.10 - 0.30StreetsAsphaltic 0.70 - 0.95Concrete 0.80 - 0.95Brick 0.70 - 0.85Drives and walks 0.75 - 0.85Roofs 0.75 - 0.95Lawns, Sandy soilFlat, 2% 0.05 - 0.10Average, 2-7% 0.10 - 0.15Steep, 7% 0.15 - 0.20Lawns, Heavy soilFlat, 2% 0.13 - 0.17Average, 2-7% 0.18 - 0.22Steep, 7% 0.25 - 0.35(from Viessman et al. 1977)11-2

EM 1110-2-141731 Aug 94Figure 11-1. Regional analysis11-3

EM 1110-2-141731 Aug 94more detail on regression and regional frequency analysisis available in EM 1110-2-1415, Hydrologic FrequencyAnalysis.d. Regional equations have already been developedby the U.S. Geological Survey (USGS) and published forthe various areas of the United States. An example ofthis type of equation is the following:kSA,L,I&G= log Pearson type II deviates= standard deviation of the logarithms annualseries peak flood events, in cubic feet persecond= various (some are logarithmic) quantifiablephysical basin characteristicsQ 10019.7 A 0.88 P 0.84 H 0.33(11-2)a&e= represent regression constantswhereQ 100 = the 1 percent chance flood peak, in cubic feetper secondA = drainage area, in square milesP = mean annual precipitation, in inchesH = average main channel elevation at 10 and85 percent points along the main channellength, in 1,000 fte. Table 11-2 illustrates various examples ofregional equations for the entire state of California.These equations make no assumptions regarding statisticaldistribution or skew. Both characteristics are inherent inthe data used to develop the regression equations. Thesepredeveloped USGS regional equations may or may notbe as good as ones developed specifically for the regionof interest; but they are already available, and developmentof regional equations is an expensive approach.f. In contrast to the USGS regional equations shownabove, the USACE usually develops regional frequencyequations as documented in EM 1110-2-1415. TheUSACE type equations are of the following form:Q X kS(11-3)(11-4)X aA b L c (1 I) d (11-5)SwhereQXeA f G g L h= flood peak for varying frequency, in cubic feetper second= mean of the logarithms of annual series peakflood events, in cubic feet per secondb,c,d,f,g&h = represent regression coefficientsg. The USACE methods assume a log Pearson typeIII distribution for “k” values and a weighted skew coefficientfor peak flood events. The equation provides a peakflow for various frequency levels associated with thevalue of “k.” Values of “k” are found in various USACEliterature such as the EM 1110-2-1415.h. Other governmental agencies (i.e., city andcounty) have developed regional frequency equations, butthey may be difficult to locate.i. Regardless of the source of the equations, the usermust identify the standard error of estimate (SE) associatedwith the equation. The SE of estimate defines thepossible range of error in the value of flow predicted bythe regression equation. Assuming the error is log normallydistributed, there is a 68 percent chance that the“true value” of flow is within ± 1 SE and a 95 percentchance that it is within ±2SE.j. For the example of the USGS equation for Q 100(the Central Coast region of California), the standard erroris 0.41 log units. The true value of Q 100 is within ± antilogof (0.41 + log Q 100 ). It can then be stated with68 percent confidence that for the example above wherethe equation predicted the Q 100 to be 1,000 cfs, the truevalue is between 2,570 and 389 cfs. Since the calculatedflows (Q 100 ) for this data set vary from 159 cfs to30,682 cfs, the example of Q 100 at 1,000 cfs is not anunlikely case. This large range in confidence limits is notunusual for a regression approach. Often this approach isthe best available technique to estimate the flow frequencyat ungauged locations.k. Again, it bears repeating that when using regressionequations from any source, make sure the equationswere developed within the region of interest, the basincharacteristics for the watershed of interest are within therange of those used to derive the equations, and the11-4

EM 1110-2-141731 Aug 94Table 11-2Regional Flood-Frequency Equations for CaliforniaNORTH COAST REGION 1 NORTH EAST REGION 2Q 2 = 3.52 A .90 P .89 H -.47 (1) Q 2 = 22 A .40 (7)Q 5 = 5.04 A .89 P .91 H -.35 (2) Q 5 = 46 A .45 (8)Q 10 = 6.21 A .88 P .93 H -.27 (3) Q 10 = 61 A .49 (9)Q 25= 7.64 A .87 P .94 H -.17 (4) Q 25= 84 A .54 (10)Q 50= 8.57 A .87 P .96 H -.08 (5) Q 50= 103 A .57 (11)Q 100 = 9.23 A .87 P .97 (6) Q 100 = 125 A .59 (12)SIERRA REGIONCENTRAL COAST REGIONQ 2 = 0.24 A .88 P 1.58 H -.80 (13) Q 2 = 0.0061 A .92 P 2.54 H -1.10 (19)Q 5 = 1.20 A .82 P 1.37 H -.64 (14) Q 5 = 0.118 A .91 P 1.95 H -.79 (20)Q 10 = 2.63 A .80 P 1.25 H -.58 (15) Q 10 = 0.583 A .90 P 1.61 H -.64 (21)Q 25 = 6.55 A .79 P 1.12 H -.52 (16) Q 25 = 2.91 A .89 P 1.26 H -.50 (22)Q 50= 10.4 A .78 P 1.06 H -.48 (17) Q 50= 8.20 A .89 P 1.03 H - .41 (23)Q 100 = 15.7 A .77 P 1.02 H -.43 (18) Q 100 = 19.7 A .88 P 0.84 H -.33 (24)SOUTH COAST REGION SOUTH - COLORADO DESERTREGION 2Q 2 = 0.41 A .72 P 1.62 (25) Q 2 = 7.3 A .30 (31)Q 5= 0.40 A .77 P 1.69 (26) Q 5= 53 A .44 (32)Q 10= 0.63 A .79 P 1.75 (27) Q 10= 150 A .53 (33)Q 25 = 1.10 A .81 P 1.81 (28) Q 25 = 410 A .63 (34)Q 50= 1.50 A .82 P 1.85 (29) Q 50= 700 A .68 (35)Q 100= 1.95 A .83 P 1.87 (30) Q 100= 1080 A .71 (36)where:Q = Peak discharge, in cubic feet per secondA = Drainage area, in square milesP = Mean annual precipitation, in inchesH = Altitude index, in thousands of feetNotes:1 In the north coast region, use a minimum value of 1.0 for altitude index (H).2 These equations are defined only for basins of 25 square miles or less.confidence of the predicted peak flow value is evaluatedby assessing the magnitude of ±1SE.11-4. Envelope Curvesa. The maximum “credible” peak discharge at anysite (usually ungauged) can be estimated by using envelopecurves. Although the result has no frequency associatedwith it, the maximum peak discharge may be usefulfor comparison with a family of peak discharges at variousfrequencies obtained by techniques discussed in previousparagraphs 11-2 and 11-3 of this manual.b. Figure 11-2 is first used to determine the regionnumber for the geographical area of interest. Select theappropriate envelope curve for the region of interest. Anexample regional envelope curve is shown in Figure 11-3.With the known drainage area, determine the maximumpeak discharge.c. More extensive discussion regarding envelopecurves can be found in USGS Water Supply Papers 1887(Crippen and Bue 1977) and 1850-B (Matthai 1969);Water Resources Investigations 77-21 (Waananen and11-5

EM 1110-2-141731 Aug 94Figure 11-2. Map of the conterminous United States showing flood-region boundariesCrippen 1977); and the American Society of Civil Engineers,Hydraulic Journal (Crippen 1982).11-5. Rainfall Data SourcesThis section lists the most current 24-hour rainfall datapublished by the National Weather Service (NWS) forvarious parts of the country. For the area generally westof the 105th meridian, TP-40 has been superseded by the(NOAA) Atlas 2, “Precipitation-Frequency Atlas of theWestern United States,” published by the NOAA.a. East of 105th Meridian (Hershfield 1961).“Rainfall Frequency Atlas of the United States for Durationsfrom 30 Minutes to 24 Hours and Return Periodsfrom 1 to 100 Years,” U.S. Department of Commerce,Weather Bureau, Technical Paper No. 40, Washington,DC. For durations of 1 hour and less, TP40 has beensuperseded by Hydrometeorological Report No. 35,U.S. Department of Commerce, National Weather Service,Silver Springs, MD.b. West of 105th Meridian (Miller, Frederick, andTracey 1973). “Precipitation-Frequency Atlas of theWestern United States, Volume I, Montana; Volume II,Wyoming; Volume III, Colorado; Volume IV, NewMexico; Volume V, Idaho; Volume VI, Utah; VolumeVII, Nevada; Volume VIII, Arizona; Volume IX, Washington;Volume X, Oregon; Volume XI, California,”U.S. Department of Commerce, National Weather Service,NOAA Atlas 2, Silver Springs, MD.c. Alaska (Miller 1963). “Probable Maximum Precipitationand Rainfall-Frequency Data for Alaska forAreas to 400 Square Miles, Durations to 24 Hours andReturn Periods From 1 to 100 Years,” U.S. Department ofCommerce, Weather Bureau, Technical Paper No. 47,Washington, DC.11-6

EM 1110-2-141731 Aug 94Figure 11-3. Peak discharge versus drainage area, and envelope curve for Region 1d. Hawaii (U.S. Department of Commerce 1962).“Rainfall-Frequency Atlas of the Hawaiian Islands forAreas to 200 Square Miles, Duration to 24 Hours andReturn Periods From 1 to 100 Years,” U.S. Department ofCommerce, Weather Bureau, Technical Paper No. 43,Washington, DC.Maximum Precipitation and Rainfall-Frequency Data forPuerto Rico and Virgin Islands for Areas to 400 SquareMiles, Durations to 24 Hours, and Return Periods From1 to 100 years,” U.S. Department of Commerce, WeatherBureau, Technical Paper No. 42, Washington, DC.e. Puerto Rico and Virgin Islands (U.S. Departmentof Commerce 1961). “Generalized Estimates of Probable11-7

EM 1110-2-141731 Aug 94Chapter 12Frequency Analysis of Streamflow Data12-1. GeneralFrequency analysis of recorded streamflow data is animportant flood-runoff analysis tool. This chapterdescribes the role of frequency analysis and summarizesthe technical procedures. EM 1110-2-1415 describes theprocedures in greater detail.a. Role of frequency analysis.(1) The traditional solution to water-resource planning,designing, or operating problems is a deterministicsolution. With a deterministic solution, a critical hydrometeorologicalevent is selected. This event is designatedthe design event. Plans, designs, or operating policies areselected to accommodate that design event. For example,the maximum discharge observed in the last 40 years maybe designated the design event. A channel modificationmay be designed to pass, without damage, this designevent. If this design event is not exceeded in the next1,000 years, the design may not be justified. On the otherhand, if the discharge exceeds the design event 20 timesin the next 30 years, the channel modification may beunderdesigned.(2) A probabilistic solution employs principles ofstatistics to quantify the risk that various hydrometeorologicalevents will be exceeded. Risk is quantified interms of probability. The greater the risk, the greater theprobability. If an event is certain to occur, its probabilityis 1.00. If an event is impossible, its probability is 0.00.For flood-runoff analyses, the probability of exceedance isusually the primary interest. This is a measure of the riskthat discharge will exceed a specified value. Decisionsare taken so that the risk of exceedance is acceptable.For example, the channel modification described abovecould be designed for a discharge magnitude with anannual exceedance probability of 0.01. In that case, therisk is known and is accounted for explicitly in the decisionmaking.b. Definition of frequency analysis.(1) The objective of streamflow frequency analysis isto infer the probability of exceedance of all possible dischargevalues (the parent population) from observed dischargevalues (a sample of the parent population). Thisprocess is accomplished by selecting a statistical modelthat represents the relationship of discharge magnitudeand exceedance probability for the parent population. Theparameters of the models are estimated from the sample.With the calibrated model, the hydrologic engineer canpredict the probability of exceedance for a specified magnitudeor the magnitude with specified exceedance probability.This magnitude is referred to as a quantile.(2) For convenience, a statistical model may bedisplayed as a frequency curve. Figure 12-1 is an exampleof a frequency curve. The magnitude of the event isthe ordinate. Probability of exceedance is the abscissa.For hydrologic engineering studies, the abscissacommonly shows “percent chance exceedance.” This isexceedance probability multiplied by 100.(3) In some sense, frequency analysis is a modelfittingproblem similar to the precipitation-runoff analysisproblem described in Chapter 8. In both cases, a modelmust be selected to describe the desired relationship, andthe model must be calibrated with observed data.c. Summary of streamflow frequency analysis techniques.Techniques for selecting and calibrating streamflowfrequency models may be categorized as graphical ornumerical. With graphical techniques, historical observationsare plotted on specialized graph paper and thecurves are fitted by visual inspection. Numerical techniquesinfer the characteristics of the model from statisticsof the historical observations. The procedures for bothgraphical and numerical analysis are presented in detail inEM 1110-2-1415 and are summarized herein for readyreference.12-2. Frequency Analysis Conceptsa. Data requirements. Statistical models of streamflowfrequency are established by analyzing a sample ofthe variable of interest. For example, to establish a statisticalmodel of annual peak discharge, the sample will be aseries of annual peaks observed throughout time. Theprocedures of statistical analysis require the following ofany time series used in frequency analysis:(1) Data must be homogeneous. That is, the datamust represent measurements of the same aspect of eachevent. For example, daily discharge observations shouldnot be combined with peak discharge observations.Furthermore, all sample points must be drawn from thesame parent population. For example, rain-flood data andsnowmelt-flood data should not be combined if they canbe identified and analyzed separately. Likewise, dischargedata observed after development upstream shouldnot be combined with predevelopment data.12-1

EM 1110-2-141731 Aug 94Figure 12-1.Frequency curve example(2) Data must be spatially consistent. All data shouldbe observed at the same location. Data observed atdifferent locations may be used to develop probabilityestimates. However, these data must be adjusted to representconditions at a common location.(3) Time series must be continuous. Statistical analysisprocedures require an uninterrupted series. If observationsare missing, the missing values must be estimated,or techniques for analysis of broken records must be used.b. Probability estimates from historical data.(1) Streamflow probability is estimated from analysisof past occurrence. The simplest model of the relationshipof streamflow magnitude and probability is a relativefrequency model. This model estimates the probability ofexceeding a specified magnitude as the fraction of timethe magnitude was exceeded historically. For example, ifthe mean daily discharge at a given location exceeds80 cfs in 6,015 of 8,766 days, the relative frequency is0.68. The estimated probability of exceedance of 80 cfsis 0.68.(2) Figure 12-2 is a graphical representation of therelative frequency models of mean daily flow in FishkillCreek at Beacon, NY. Such a plot is commonly referredto as a duration curve. The abscissa of this plot shows“percent of time exceeded.” This equals relativefrequency multiplied by 100, so it is consistent with theterm “percent chance exceedance.”(3) The reliability of a relative frequency modelimproves as the sample size increases; with an infinitesample size, relative frequency exactly equals the probability.Unfortunately, sample sizes available for streamflowfrequency analysis are small by scientific standards.Thus, relative frequency generally is not a reliable estimatorof probability for hydrologic engineering purposes.12-2

EM 1110-2-141731 Aug 94Figure 12-2.Graphical representation of relative frequency model(4) The alternative to the empirical relative frequencymodel is a theoretical frequency model. With a theoreticalmodel, the relationship of magnitude and probabilityfor the parent population is hypothesized. The relationshipis represented by a frequency distribution. Acumulative frequency distribution is an equation thatdefines probability of exceedance as a function of specifiedmagnitude and one or more parameters. An inversedistribution defines magnitude as a function of specifiedprobability and one or more parameters.c. Distribution selection and parameter estimation.In certain scientific applications, one distribution oranother may be indicated by the phenomena of interest.This is not so in hydrologic engineering applications.Instead, a frequency distribution is selected because itmodels well the data that are observed. The parametersfor the model are selected to optimize the fit. A graphicalor numerical technique can be used to identify the appropriatedistribution and to estimate the parameters.12-3. Graphical Techniquesexceedance probability of that data. If a best-fit line isdrawn on the plot, the probability of exceeding variousmagnitudes can be estimated. Also, any desired quantilescan be estimated. Graphical representations also providea useful check of the adequacy of a hypothesizeddistribution.a. Plotting-position estimates of probability.(1) Graphical techniques rely on plotting positions toestimate exceedance probability of observed events. Themedian plotting position estimates the exceedance probabilityas:where:P m(m 0.3)(N 0.4)(12-1)P m = exceedance probability estimate for the mthlargest eventSome of the early and simplest methods of frequencyanalysis were graphical techniques. These techniquespermit inference of the parent population characteristicswith a plot of observed magnitude versus estimatedmN= the order number of the event= the number of events12-3

EM 1110-2-141731 Aug 94For example, to estimate annual exceedance probability ofannual maximum discharge, N = the number of years ofdata. To express the results as percent-chance exceedance,the results of Equation 12-1 are multiplied by 100.(2) Table 12-1 shows plotting positions for annualpeak discharge on Fishkill Creek. Column 4 of the tableshows the discharge values in the sequence of occurrence.Column 7 shows these same discharge values arranged inorder of magnitude. Column 5 is the order number ofeach event. Column 8 shows the plotting position. Theseplotting positions are values computed with Equation 12-1and multiplied by 100. The values in columns 7 and 8thus are an estimate of the peak-discharge frequencydistribution.b. Display and use of estimated frequency curve.(1) The estimated frequency distribution is displayedon a grid with the magnitude of the event as the ordinateand probability of exceedance (or percent-chance exceedance)as the abscissa. The plot thus provides a useful toolfor estimating quantiles or exceedance probabilities.Specialized plotting grids are available for the display.These grids are constructed with the abscissa scaled so aselected frequency distribution plots as a straight line.For example, a specialized grid was developed by Hazenfor the commonly used normal frequency distribution.(2) The specialized normal-probability grid is auseful tool for judging the appropriateness of the normalTable 12-1Annual Peaks, Sequential and Ordered with Plotting Positions (Fishkill Creek at Beacon, NY)Events AnalyzedOrdered EventsMon(1)Day(2)Year(3)Flow, cfs(4)Rank(5)WaterYear(6)Flow, cfs(7)MedianPlot Pos(8)3 5 1945 2,290. 1 1955 8,800. 2.8712 27 1945 1,470. 2 1956 8,280. 6.973 15 1947 2,220. 3 1961 4,340. 11.073 18 1948 2,970. 4 1968 3,630. 15.161 1 1949 3,020. 5 1953 3,220. 19.263 9 1950 1,210. 6 1952 3,170. 23.364 1 1951 2,490. 7 1962 3,060. 27.463 12 1952 3,170. 8 1949 3,020. 31.561 25 1953 3,220. 9 1948 2,970. 35.669 13 1954 1,760. 10 1958 2,500. 39.758 20 1955 8,800. 11 1951 2,490. 43.8510 16 1955 8,280. 12 1945 2,290. 47.954 10 1957 1,310. 13 1947 2,220. 52.0512 21 1957 2,500. 14 1960 2,140. 56.152 11 1959 1,960. 15 1059 1,960. 60.254 6 1960 2,140. 16 1963 1,780. 64.342 26 1961 4,340. 17 1954 1,760. 68.443 13 1962 3,060. 18 1967 1,580. 72.543 28 1963 1,780. 19 1946 1,470. 76.641 26 1964 1,380. 20 1964 1,380. 80.742 9 1965 980. 21 1957 1,310. 84.842 15 1966 1,040. 22 1950 1,210. 88.933 30 1967 1,580. 23 1966 1,040. 93.033 19 1968 3,630. 24 1965 980. 97.1312-4

EM 1110-2-141731 Aug 94distribution as a model of the parent population. If datadrawn from a normally distributed parent population areassigned plotting positions using Equation 12-1 and areplotted on Hazen’s grid, the points will fall approximatelyon a straight line. If the points do not, then either thesample was drawn from a population with a different distributionor sampling variation yielded a nonrepresentativesample.(3) A specialized plotting grid has been developedalso for another commonly used frequency distribution,the log-normal distribution. Figure 12-3 is an example ofsuch a grid. The values from columns 7 and 8 ofTable 12-1 are plotted on this grid, and a frequency curveis fitted. If the data are truly drawn from the distributionof a log-normal parent population, the points will fall on astraight line. The Fishkill Creek data, shown byFigure 12-3, do not fall on a straight line, so the assumptionthat the parent population is a log-normal distributionis suspect.12-4. Numerical TechniquesNumerical techniques define the relationship betweenstreamflow magnitude and probability with analyticaltools, instead of the graphical tools.a. Steps of numerical techniques. With numericaltechniques, the following general steps are used to derivea frequency curve to represent the population (McCuenand Snyder 1986):Figure 12-3.Log-normal probability grid12-5

EM 1110-2-141731 Aug 94(1) Select a candidate frequency model of the parentpopulation. Three distributions are commonly used forfrequency analysis of hydrometeorological data: the normaldistribution, the log-normal distribution, and the logPearson type III distribution.where:Q pQ K pS(12-2)(2) Obtain a sample.(3) Use the sample to estimate the parameters of themodel identified in step 1.Q p = the quantile with specified exceedanceprobability pQ= the sample mean(4) Use the model and the parameters to estimatequantiles to construct the frequency curve that representsthe parent population.SK p= the sample standard deviation= a frequency factorb. Numerical parameter estimation.(1) Parameters of a statistical model are commonlyestimated from a sample with method-of-moments estimators.The method-of-moments parameter estimators aredeveloped from the following assumptions:The sample mean and standard deviation are computedwith the following equations:QNQ i(12-3)(a) The streamflow-probability relationship of theparent population can be represented with a selected distribution.The moments (derivatives) of the distributionequation can be determined with calculus. One momentis determined for each parameter of the distribution. Theresulting expressions are equations in terms of the parametersof the distribution.Swhere:Q i⎛⎜⎝⎞(Q iQ ) 2⎟(N 1) ⎠0.5= observed event i(12-4)(b) Moments of a sample of the parent population canbe computed numerically. The first moment is the meanof the sample; the second moment is the variance; thethird moment is the sample skew. Other moments can befound if the distribution selected has more than threeparameters.(c) The numerical moments of the sample are the bestestimates of the moments of the parent population. Thisassumption permits development of a set of simultaneousequations. The distribution parameters are unknown inthe equations. Solution yields estimates of theparameters.(2) When the parameters of the distribution are estimated,the inverse distribution defines the quantiles of thefrequency curve. Chow (1951) showed that with themethod-of-moments estimates, many inverse distributionscommonly used in hydrologic engineering could be writtenin the following general form:N= number of events in sample(3) The frequency factor in Equation 12-2 dependson the distribution selected. It is a function of the specifiedexceedance probability and, in some cases, otherpopulation parameters. The frequency factor function canbe tabulated or expressed in mathematical terms. Forexample, normal-distribution frequency factors correspondingto the exceedance probability p (0 < p < 0.5)can be approximated with the following equations(Abramowitz and Stegun 1965):K pw⎡⎢⎣2.515517 0.802853 w 0.010328 w 21 1.432788 w 0.189269 w 2 0.001308 w 3⎤⎥⎦(12-5)12-6

EM 1110-2-141731 Aug 94wwhere:w⎡ ⎛ ⎞⎤⎢ln1⎜ ⎥⎣ ⎝⎟ ⎠⎦P 2 0.5(12-6)= an intermediate variable, if p > 0.5, (1 - p) isused in Equation 12-6, and the computed valueof K p is multiplied by -1.(4) For the log-normal distribution, Equation 12-2 iswritten as:where:X pX K pSX pXSK p= the logarithm of Q p , the desired quantile= mean of logarithms of sample(12-7)= standard deviation of logarithms of sample= the frequency factorThis frequency factor is the same as that used for thenormal distribution. X and S are computed with the followingequations:Xlog Q iN(12-8)c. Recommended procedure for annual maximumdischarge.(1) The U.S. Water Resources Council (USWRC)(1967, 1976, 1977) recommended the log Pearson type IIIdistribution for annual maximum streamflow frequencystudies. This recommendation is followed by USACE.Current guidelines are presented in Bulletin 17B (USWRC1981).(2) The log Pearson type III distribution models thefrequency of logarithms of annual maximum discharge.Using Chow’s (1951) format, the inverse log Pearson typeIII distribution iswhere:X pX KSX pXSK(12-10)= the logarithm of Q p , the desired quantile= mean of logarithms of sample= standard deviation of logarithms of sample= the Pearson frequency factorX and S are computed with the Equations 12-8 and 12-9.(3) For this distribution, the frequency factor K is afunction of the specified probability and of the skew ofthe logarithms of the sample. The skew, G, is computedwith the following equation:S⎛⎜⎝⎞(log Q iX ) 2⎟(N 1) ⎠0.5(12-9)GN (X iX) 3(N 1) (N 2) S 3(12-11)where:Q iN= observed peak annual discharge in year i= number of years in sampleFor the annual peak discharge values shown inTable 12-1, these values are as follows: X = 3.3684, andS = 0.2456.For the values of Table 12-1, the skew computed withEquation 12-11 is 0.7300.(4) The log Pearson type III frequency factors forselected values of skew and exceedance probability aretabulated in Bulletin 17B (USWRC 1981) and inEM 1110-2-1415. Alternatively, an approximating12-7

EM 1110-2-141731 Aug 94function can be used. If the skew equals zero, thePearson frequency factors equal the normal distributionfactors. Otherwise, the following approximation suggestedby Kite (1977) can be used:K K pK 2 ⎛ 1 ⎞p k ⎜ ⎟ K 3p 6K⎝ 3pk 2 K 2p 1 k 3⎠K pk 4 ⎛ 1 ⎞ ⎜⎝ ⎟⎠3 k 5 ~ (12 12)where k = G/6.d. Analysis of special cases.(1) In hydrologic engineering applications, frequencyanalysis of annual maximum discharge is complicated byspecial cases. These include broken records, incompleterecords, zero-flow years, outliers, historical data, andsmall samples. Bulletin 17B provides guidance for dealingwith these cases.(5) An outlier is an observation that departs significantlyfrom the trend of the remaining data. Proceduresfor treating outliers require hydrologic and mathematicaljudgment. Bulletin 17B describes one procedure foridentifying high and low outliers and for censoring thedata set. High outliers are treated as historical data ifsufficient information is available. Low outliers aretreated as zero-flow years.(6) Large floods outside the systematically recordedtime series may be used to extend that record. The procedurerecommended for analysis of these historical flows isas follows:(a) Assemble known historic peaks and determinethe historic record length.(b) Censor the systematic record by deleting allpeaks less than the minimum historical peak. Estimatethe model parameters for the remaining record.(c) Compute a weight with the following equation:(2) If 1 or more years of data are missing from atime series of annual maximum discharge due to reasonsnot related to flood magnitude, the record is broken. Foranalysis, the record segments are combined, and the combinedrecord is analyzed as previously described.Wwhere:(H Z)(N L)(12-13)(3) If data are missing because the events were toolarge to record, too small to record, or the gauge wasdestroyed by a large event, the record is incomplete. Anymissing large events should be estimated and the estimatesincluded in the time series. Missing small eventsare treated with the conditional probability adjustmentrecommended for zero-flow years.(4) The log Pearson type III distribution is not suitedto analysis of series which include zero-flow years. If thesample contains zero-flow years, the record is analyzedusing the conditional probability procedure. With thisprocedure, the subseries of nonzero peaks is analyzed asdescribed previously. The resulting frequency curve is aconditional frequency curve. The exceedance frequenciesfrom this curve are scaled by the relative frequency ofnon-zero flow years. The log Pearson type III modelparameters are estimated for the upper portion of thecurve. With these parameters, a synthetic frequency curveis developed. Paragraph 3-6 of EM 1110-2-1415describes the procedure.W = the weightH = number of years in historic recordZNL= number of historic event= number of years in censored systematic record= number of zero-flow years, low outliers,missing years excluded from systematic record(d) Adjust the model parameters with this weight.Equations for the adjustments are presented in Appendix 6of Bulletin 17B (USWRC 1981). Compute the quantileswith these modified parameters and Equation 12-10.(7) Small samples adversely affect the reliability ofestimates of the skew. This parameter is difficult to estimateaccurately from a small sample. A more reliableestimate is obtained by considering skew characteristics ofall available streamflow records in a large region. An12-8

EM 1110-2-141731 Aug 94adopted skew is computed as a weighted sum of thisregional skew and the skew computed withEquation 12-11. The weights chosen are a function of thesample skew of the logs, the sample record length, thegeneralized skew, and the accuracy in developing thegeneralized values. The generalized skew can be determinedfrom a map included in Bulletin 17B, or it can bedetermined from detailed analyses if additional data areavailable.(8) The impact of uncertainty due to small samplesize can be quantified further with the expectedprobability adjustment. This adjustment is based on theargument that the x percent-chance discharge estimatemade with a given sample is approximately the median ofall estimates that would be made with successive samplesof the same size. However, the probability distribution ofthe estimate is skewed, so the average of the samplesexceeds the median. The consequence of this is that if avery large number of estimates of flood magnitude aremade over a region, more x percent-chance floods willoccur than expected on the average (Chow, Maidment,and Mays 1988). For example, more “100-year floods”will occur in the United States annually than expected.Paragraph 3-4 of EM 1110-2-1415 describes how eitherthe probability associated with a specified magnitude orthe magnitude for a specified probability can be adjustedto obtain a frequency curve with the expected number ofexceedances.e. Verification of frequency estimates.(1) The reliability of frequency estimates depends onhow well the proposed model represents the parent population.The fit can be tested indirectly with a simplegraphical comparison of the fitted model and the sampleor with a more rigorous statistical test. The reliability canalso be illustrated with confidence limits.(2) A graphical test provides a quick method forverifying frequency estimates derived with numericalprocedures. The test is performed by plotting observedmagnitude versus plotting-position estimates ofexceedance probability. The postulated frequency curvewith best-estimate parameters is plotted on the same grid.Goodness-of-fit is judged by inspection, as describedpreviously.(3) Because of the complexity of the log Pearson typeIII distribution, no single specialized plotting grid ispractical for this graphical test. Instead, the log-normalgrid is used to display data thought to be drawn from alog Pearson type III distribution. The fit is judged byinspection. Figure 12-4 illustrates this. The observedpeaks and plotting positions from columns 7 and 8 ofTable 12-1 are plotted here. Quantiles computed withEquation 12-10 are plotted on the same grid. Theestimated values of the terms of Equation 12-5 areX = 3.3684; S = 0.2456; and G = 0.700. The skew wasadjusted here with a regional skew. The computed frequencycurve fits well the plotted observations.(4) Rigorous statistical tests permit quantitativejudgement of goodness of fit. These tests compare thetheoretical distribution with sample values of the relativefrequency or cumulative frequency function. For example,the Kolmogorov-Smirnov test provides bounds withinwhich every observation should lie if the sample actuallyis drawn from the assumed distribution. The test is conductedas follows (Haan 1977):(a) For each observation in the sample, determinethe relative exceedance frequency. This is given by m/N,where m = the number of observations in the samplegreater than or equal to the observed magnitude, andN = the number of observations.(b) For each magnitude in the sample, determine thetheoretical exceedance frequency using the hypothesizedmodel and the best estimates of the parameters.(c) For each observation, compute the difference inthe relative exceedance frequency and the theoreticalexceedance frequency. Determine the maximum differencefor the sample.(d) Select an acceptable significance level. This is ameasure of the probability that the sample is not drawnfrom the candidate distribution. Values of 0.05 and 0.01are common. Determine the corresponding Kolmogorov-Smirnov test statistic. This statistic is a function of thesample size and the significance level. Test statistics aretabulated or can be computed with the following equation(Loucks, Stedinger, and Harth 1981):where:C⎡⎢n 0.5 0.12⎣⎛ ⎞⎤0.11⎜ ⎟⎥⎝ N 0.5⎠⎦C = 1.358 for significance level 0.05C = 1.628 for significance level 0.01(12-14)12-9

EM 1110-2-141731 Aug 94Figure 12-4.Plot for verification(e) Compare the maximum difference determined instep c with the test statistic found in step d. If the valuein step c exceeds the test statistic, the hypothesized distributioncannot be accepted with the specified significancelevel.(5) The reliability of a computed frequency curve canbe illustrated conveniently by confidence limits plotted onthe frequency grid. Confidence limits are establishedconsidering the uncertainty in estimating population meanand standard deviation from a small sample. For convenience,Appendix 9 of Bulletin 17B (USWRC 1981)includes a table of frequency factors that permit definitionof 1 percent to 99 percent confidence limits. These frequencyfactors are a function of specified exceedanceprobability and sample size. As the sample size increases,the limits narrow, indicating increased reliability.(6) Figure 12-5 shows the 5 and 95 percent confidencelimits for the Fishkill Creek frequency curve. Theprobability is 0.05 that the true quantile for a selectedexceedance probability will exceed the value shown onthe 5 percent curve. The probability is 0.95 that the truequantile will exceed the 95 percent-curve value and only0.05 that it will be less than the 95 percent curve.12-5. Special Considerationsa. Mixed populations. In certain cases, observedstreamflow is thought to be the result of two or moreindependent hydrometeorological conditions. The sampleis referred to as a mixed-population sample. For example,the spring streamflow in the Sacramento River, CA, is theresult of both rainfall and snowmelt. For these cases, thedata are segregated by cause prior to analysis, if possible.Each set can be analyzed separately to determine theappropriate distribution and parameters. The resultingfrequency curves are then combined using the followingequation to determine probability of union:12-10

EM 1110-2-141731 Aug 94Figure 12-5.Frequency curve with confidence limitswhere:P cP 1P 2P 1P 2P cP 1P 2(12-5)= annual exceedance probability of combinedpopulations for a selected quantile= annual exceedance probability of samemagnitude for sample 1= annual exceedance probability of samemagnitude for sample 2This assumes that the series are independent. Otherwise,coincident frequency analysis must be used.b. Coincident frequency analysis. In some planning,designing, or operating problems, the hydrometeorologicalevent of interest is a function of two or more randomhydrometeorological events.(1) For example, discharge at the confluence of twostreams is a function of the coincident discharge in thetributary streams. The objective of coincident frequencyanalysis is to estimate the frequency distribution of theresult if the frequency distributions of the components areknown. The specific technique used depends on themathematical form of the function relating the variables.Benjamin and Cornell (1970) describe a variety of solutions,including analytical closed-form solutions andMonte Carlo simulation.(2) In hydrologic engineering, the variable of interestoften is the sum of components. In that case, the frequencydistribution of the sum can be found throughconditional probability concepts. For illustration, considerthe total discharge downstream of a confluence, Q T . Thisis computed as the sum of tributary discharge Q 1 andtributary discharge Q 2 . The frequency of Q 1 and Q 2 areestablished using procedures described previously.Roughly speaking, the probability that Q T equals some12-11

EM 1110-2-141731 Aug 94specified value, q T , is proportional to the probability thatQ 1 equals a specified value, q 1 , times a factor proportionalto the probability that Q 2 equals q T -q 1 . This product issummed over all possible values of Q 1 . To develop afrequency curve for the sum, the process is repeated forall possible values of Q T . Chapter 11 of EM 1110-2-1415presents a detailed example of coincident frequency analysis.c. Regional frequency analysis.(1) Methods of frequency analysis described previouslyin this chapter apply to data collected at a singlesite. If a large sample is available at that site, the resultingfrequency analysis may be sufficiently reliable forplanning, designing, or operating civil-works projects.However, samples commonly are small. In fact, it is notunusual that risk information is required at sites for whichno data are available. Regional frequency analysis techniquesmay be used to develop this information.(2) Regional frequency analysis procedures relateparameters of a streamflow-frequency model to catchmentcharacteristics. Briefly, the following general steps arefollowed to derive such a relationship:(a) Select long-record sites within the region, andcollect streamflow data for those sites.(b) Select an appropriate distribution for the data, andestimate the parameters using the procedures describedherein.(c) Select catchment characteristics that should correlatewith the parameters. Measure or observe thesecharacteristics for the long-record sites. Typical characteristicsfor streamflow frequency model parametersinclude the following: contributing drainage area, streamlength, slope of catchment or main channel, surface storage,mean annual rainfall, number of rainy days annually,infiltration characteristics, and impervious area.(d) Perform a regression analysis to establish predictiveequations. The dependent variables in the equationsare the frequency model parameters. The independentvariables are the catchment characteristics.(3) EM 1110-2-1415 provides additional guidance inestablishing regional equations.d. Frequency of other hydrometeorological phenomena.The procedures described for dischargefrequencyanalysis apply to analysis of otherhydrometeorological phenomena. The same general stepspresented in paragraph 12-4 are followed. For example, ifthe variable of interest is streamflow volume, rather thandischarge, the time series will be a sequence of volumesfor a specified duration. The procedures for selecting,calibrating, and verifying a frequency model are the sameas previously described.e. Volume-frequency and precipitation-depth-duration-frequencyanalyses. These analyses present someunique problems. Because of the small samples fromwhich parameters must be estimated, the set of frequencycurves for various durations may be inconsistent. Forexample, the 1-day volume should not exceed the 3-dayvolume for all probabilities. Yet, for a small sample, thecomputed curves may not follow this rule. To overcomethis, the computed curves may be “smoothed,” adjustedby inspection of plots. Alternatively, the statistical modelparameters can be adjusted to maintain consistency. Paragraph3-8c of EM 1110-2-1415 describes a typicalprocedure.12-12

EM 1110-2-141731 Aug 94Chapter 13Analysis of Storm Events13-1. IntroductionThis chapter is concerned with the application of eventtypesimulation models for flood-runoff analysis. Suchmodels are commonly used with frequency-based hypotheticalstorms to develop discharge-frequency estimatesor with standard project or probable maximum storms todevelop associated flood estimates. The chapter beginswith a discussion of initial development of a simulationmodel. This is followed by consideration of methods forcalibration/verification of the model. Applications issuesassociated with design storms are the focus of the remainderof the chapter.13-2. Model DevelopmentSteps in the initial development of a simulation model areas follows: assess data requirements and availability;acquire and process data; develop subbasin configurationof model; and develop initial estimates for modelparameters.a. Assessment of data requirements and availability.(1) It is essential that the model developer be fullyaware of the study objectives and requirements, includingthe intended use of modeling products. Types of datarequired for model development include(a) historical precipitation and streamflow data;(b) runoff-parameter data from past studies;(c) data associated with watershed characteristics suchas drainage areas, soil types, and land use;(d) characteristics of rivers and other drainage-system(natural or artificial) features; and(e) existence and characteristics of storage elementssuch as lakes, detention basins, etc.(2) A field reconnaissance of the study basin shouldbe performed. Information acquired from field observationscan significantly enhance one’s understanding of therunoff-response characteristics of the watershed and perhapsenable recognition of important watershed featuresthat might otherwise be overlooked.b. Acquisition and processing of data. Aspects ofdata management are treated in Chapter 17. Much precipitationand streamflow data are stored on electronicmedia, which can greatly facilitate data acquisition. It isgenerally desirable to place data in a data base and reviewit with graphics software. As the study proceeds, simulationresults can also be stored in the data base, and utilitysoftware can be used to produce graphs, tables, etc. ofkey information. A careful review should be made ofpast studies and of the basis for all of the data beingacquired.c. Development of subbasin configuration.(1) For most studies, it is necessary to divide a basininto subbasins to enable development of information atlocations of interest and to better represent spatially variablerunoff characteristics. A subbasin outlet should belocated:(a) at each stream location where discharge estimatesare required,(b) at each stream gauge, and(c) at dams and other significant hydraulic structures.(2) Nondistributed models use lumped (spatiallyaveraged) values for precipitation and loss (infiltration)parameters. Subbasins should be sufficiently small so thatspatial-averaging of this information is reasonable. Basinsubdivision may also be performed to tailor rainfall-runofftransformations to particular land-use conditions. Forexample, rural and urban portions of a basin might berepresented separately. If flood-damage or other modeldependentanalyses are to be performed, subbasin delineationshould be coordinated with the users of model results.There may be reasons other than hydrologic that affect thechoice of locations of subbasin outlets.d. Development of initial estimates for parametervalues.(1) After defining the subbasin configuration, askeleton input file can be developed which contains allrequired information (such as drainage areas, subbasinlinkages, etc.) except values for runoff parameters. Suchparameters might be required for defining unit hydrograph,kinematic wave, loss-rate, base-flow, or routingrelationships. At this point, initial estimates of values forrunoff parameters can be made and entered into the inputfile. Estimates can be derived from13-1

EM 1110-2-141731 Aug 94(a) past studies,(b) application of previously developed regional relationships,and(c)physical characteristics of the subbasins.(2) If there were no streamflow data available for thebasin, there may be little basis for improving the initialestimates. However, generally, there are some streamflowdata for locations in or near the basin which can be usedin a calibration process to improve the initial estimates.13-3. Model CalibrationCalibration here refers to the process of using historicalprecipitation and streamflow data to develop values forrunoff parameters. Verification refers to the testing ofcalibrated values, generally with data not used for calibration.Topics in this section pertain to calibration strategy,selection of historical events, calibration techniques, andmodel verification.a. Calibration strategy. Calibration of simulationmodels must be done carefully with due consideration forthe reliability of historical data and for the simplisticnature of model components used to represent complexphysical processes in heterogeneous basins. The insightthat an experienced analyst brings to bear in accommodatingthese factors is, in many cases, the single most importantelement of the calibration process.(1) The calculation of a discharge hydrograph at alocation in a basin may be a function of few or manyrunoff parameters. A headwater subbasin is one forwhich there are no subbasins upstream. The simulation ofrunoff from a headwater subbasin is a function of parametersassociated solely with that subbasin. Calculatedrunoff for the outlet of a downstream subbasin is a functionnot only of the parameters of the subbasin, but alsoof those for all upstream subbasins and routing reaches.For this reason, the calibration of values of parameters forgauged headwater subbasins is often more direct andreliable than calibration associated with downstreamgauges.(2) In a multisubbasin model, subbasins with streamgaugesat the outlet are generally a small proportion ofthe total number of subbasins. Hence, the generalapproach is to first calibrate parameter values for allgauged headwater subbasins and to use the results as anaid in setting or adjusting values for all other subbasins.The next (and generally most difficult) step is to reviewcalculated versus simulated results at all downstreamgauges and to manually adjust key parameter values toprovide basin-wide simulations that are as reasonable andconsistent as possible.b. Selection of historical events. Model componentsthat employ unit hydrographs or other linear entities produceoutputs proportional to inputs. Because watershedsdo not respond in a truly linear manner, the events chosenfor calibration and verification should, if possible, beconsistent in magnitude with the magnitude of hypotheticalevents to which the model will be applied. In manycases, this is not feasible because the hypothetical eventsare more rare than those that have been experienced historically.Nevertheless, the largest historical events forwhich data are available generally provide the best basisfor calibration/verification.(1) In addition to the size of a historical event, thestate of the basin at the time of occurrence is significant.The model must represent land-use and other conditionsconsistent with the time of occurrence of the historicalevent. If existing basin conditions are of primary interestand a historical event occurred when the basin conditionswere markedly different, the event may be of little valuefor calibration.(2) Also important are the amount and quality ofdata associated with historical events. If precipitation dataare lacking or if only daily values are available and amodel with small subbasins is being calibrated, an eventmay be of limited value for calibration.(3) In general, it is desirable to use several events(say, four to six) for calibration. It is also desirable toreserve a couple of events for verification. Sometimes theamount of useful data is limited so that there are fewevents for calibration and no events for verification.c. Calibration techniques for gauged headwaterbasins. Computer software can be used for automatedcalibration of parameter values for gauged headwatersubbasins. Figure 7-7 shows in simple terms the procedurethat may be used. As may be noted, it is necessaryto specify initial values for the parameters to be optimized.The simulation is performed with these valuesand the results compared with the observed dischargehydrograph. The quantitative measure of goodness of fit,the objective function, is often defined in terms of a rootmean square error, where error is the difference betweencomputed and observed discharge ordinates. For floodrunoffanalysis, the errors may be weighted with afunction that gives more weight to higher flows than13-2

EM 1110-2-141731 Aug 94lower flows, as illustrated in Equation 7-18 inparagraph 7-3e.(1) Parameter values are adjusted in automatic calibrationto minimize the magnitude of the objective function.Because of interdependence between parameters andother factors, a global minimum is not always achieved,which results in suboptimal values for parameters.Another aspect of calibration is that constraints on acceptableparameter values are often imposed. For example,negative loss rates would be unreasonable. Parametervalues obtained by calibration should be reviewedcarefully; values that are unreasonable or inconsistentshould be rejected. Generally, the quality of fit betweenthe observed and computed hydrographs is best judged byreviewing plots of the hydrographs and associated rainfalland rainfall-excess hyetographs, rather than simply lookingat statistical measures of the fit.(2) The analyst should thoroughly understand theoptimization procedure being implemented and have sufficientoutput information to enable verification of its performance.Suboptimal results can sometimes be improvedby reoptimization with different initial conditions, restrictingthe optimization region, or other means.(3) The parameter values optimized for each historicalevent will be unique. Criteria are required for choosing asingle set of values to represent the runoff characteristicsof the subbasin. Consideration should be given to factorssuch as(a) the quality of fit between the observed and computedhydrographs,(b) the magnitude of the event, and(c) the quality of the precipitation and streamflowdata for the event.Generally, estimates based on the larger events would begiven more weight if the calibrated model is intended forapplication to rare events. Once a set of parameter valueshas been adopted, the historical events should be rerunwith these values. Further refinement may be needed toachieve the best compromise in matching available data.d. Calibration techniques for downstream gauges.The calibration process for downstream locations involvessimulating runoff at each streamgauge and ascertainingwhat parameter-value adjustments, if any, should be madefor upstream subbasins and/or routing reaches. The calibrationshould be performed starting at the upstreamgauges and working downstream. Adjustments shouldgenerally not be tailored to any one event. Rather, themodel performance should be judged for all calibrationevents. When a consistent bias is noted, for example ifthe timing of runoff is consistently too early or too late,the most likely cause of the bias should be sought and themodel adjusted accordingly. Often, poor results are dueto erroneous definition of precipitation or other data problems.If the problems cannot be reconciled, the datashould be rejected for calibration purposes. Numeroussimulations may be required to determine a final set ofparameter values that are most reasonably consistent withknowledge of the basin and the data associated with thecalibration events.e. Verification. Verification enables assessment ofthe reliability of the calibrated model. It is performed bysimulating historical events not used for calibration. Withan event-type model, there is always uncertainty associatedwith loss rates, and they are critical in their impacton runoff volumes. For purposes of verification, the antecedentrainfall-runoff conditions should be assessed andloss rates chosen that are consistent with similar antecedentconditions associated with calibration events.Adjustment of the loss rates may be required to obtainreasonable agreement with the observed runoff volumes.Once this agreement has been achieved, a critical assessmentof the simulated results can be made. Good agreementbetween simulated and observed hydrographsengenders confidence in the model performance, at leastfor events similar in magnitude to those simulated. Ifresults are poor, reasons for such results should be ascertained,if possible. Parameter-value modificationsrequired to produce reasonable simulations of the verificationevents should be determined. If such modificationscan be made without significant degradation of the resultsobtained for the calibration events, the modifications canbe adopted. If degradation of calibration results wouldoccur, it may be appropriate to redo the calibration withincorporation of the verification events. In either case,the poor results are cause for associating a higher level ofuncertainty with model application.13-4. Simulation of Frequency-Based DesignFloodsEvent-type models are commonly used with frequencybasedhypothetical storms for the development ofdischarge-frequency estimates. Issues discussed in thissection include design-storm definition, depth-area adjustments,and association of runoff frequency with rainfallfrequency. Other issues such as transfer of frequencyinformation from gauged to ungauged locations,13-3

EM 1110-2-141731 Aug 94conversion of nonstationary to stationary peak discharges,and development of future-condition frequency estimatesare discussed in Chapter 17.a. Design-storm definition.(1) The NOAA has published generalized rainfallcriteria for the United States. Appendix A lists a numberof these publications. The criteria consist of maps withisopluvial lines of point precipitation for various frequenciesand durations. Generally, the maps for mountainousregions are substantially more detailed because of orographiceffects.(2) The rainfall depths obtained from NOAA criteriaare point values commonly assumed to apply up to10 square miles. For larger areas, the average precipitationover the area is less than the value for a point, andadjustments are required. Figure 13-1 shows adjustmentcriteria provided in NOAA publications.(3) The rainfall depths from NOAA criteria are basedon a partial duration series. If value of the annual seriesis desired, adjustment factors are applied to recurrenceintervals of 10 years or less. No adjustment is appliedfor larger recurrence intervals larger than 10 years, as thetwo series essentially merge at that recurrence interval.(4) The NOAA criteria do not contain specific guidancefor establishing the temporal distribution of designstormrainfall. A common approach is to arrange therainfall to form a balanced hyetograph; that is, the depthassociated with each duration interval of the storm satisfiesthe relation between depth and duration for a givenfrequency. For example, for a 1 percent-chance(100-year) 24-hr storm, the depths for the peak 30-min,1-hr, 2-hr, ..., 24-hr durations would each equal the1 percent-chance depth for that duration. Although suchstorms do not preserve the random character of naturalstorms, use of a balanced storm ensures an appropriatedepth (in terms of frequency), regardless of the timeresponsecharacteristics of a particular river basin.(5) The SCS has developed four 24-hr synthetic rainfalldistributions (USDA 1986) from available NationalWeather Service duration-frequency data. Types I and IArepresent the Pacific maritime climate with wet wintersand dry summers. Type III represents Gulf of Mexicoand Atlantic coastal areas where tropical storms bringlarge 24-hr rainfall amounts. Type II represents the restof the country. Other approaches for defining the temporaldistribution of design storms are reported in theliterature. If none of the synthetic distributions areapplicable to the area being modeled, the hydrologistshould look at historical information, as well as regionaldata, to develop an adequate temporal distribution.b. Depth-area considerations.(1) The area-adjustment criteria of Figure 13-1 havea nonlinear effect on storm hyetographs. That is, abalanced hyetograph for one storm size is not a simpleproportion of a balanced hyetograph for a different stormsize. Each storm size will have its own unique depth andtemporal distribution. This creates a problem in situationswhere it is desired to develop a consistent set of frequencyestimates for numerous sites in a basin. It wouldbe necessary to develop a unique storm hyetograph forevery location. For a basin with many subbasins andstream junctions, the computational requirements could besubstantial.(2) An approach for dealing with this situation isbased on calculating index discharge hydrographs at eachlocation of interest from a set of index hyetographs forstorm areas that encompass the full range of drainageareas from the area of the smallest subbasin to the totalbasin area. The hydrograph for a given location isobtained by interpolating, based on drainage area, betweentwo index hydrographs for that location. This isillustrated in Figure 13-2.(3) A semi-logarithmic interpolation equation (usedin computer program HEC-1) is as follows:Qwhere⎛Q1⎜⎝log A2Axlog A2A1⎞⎟⎠⎛Q2⎜⎝log AxA1log A2A1⎞⎟⎠(13-1)Q = instantaneous discharge for the interpolatedhydrographAx = drainage area represented by the interpolatedhydrographA1 = index drainage area that is closest to, butsmaller than, AxA2 = index hydrograph closest to, but largerthan, Ax13-4

EM 1110-2-141731 Aug 94Figure 13-1.Area-adjustment of point rainfallQ1 = instantaneous discharge for the indexhydrograph corresponding to A1Q2 = instantaneous discharge for the indexhydrograph corresponding to A2An illustration of this approach is given in the HEC-1User’s Manual.c. Association of runoff frequency with rainfall frequency.Although the NOAA rainfall criteria associatefrequency with depth, it does not follow that the samefrequencies should be associated with the design stormsor the calculated flood-runoff.(1) In addition to rainfall, runoff is a function of lossrates and base flow, the magnitudes of which vary withtime and antecedent moisture conditions. A very dryantecedent condition associated with a 100-year stormmight produce runoff with a significantly smaller recurrenceinterval.(2) Because of the uncertainty of the frequency ofdesign-storm runoff, it is best to utilize statistically basedfrequency information (for locations with at least 10 yearsof streamflow data) wherever possible to ’calibrate’ theexceedance frequency to associate with particular combinationsof design storms and loss rates. This importantconcept is discussed further in Chapter 17.13-5. Simulation of Standard Project and ProbableMaximum FloodsStandard project and probable maximum floods are usedas design events and also as reference events for comparisonwith flood magnitudes developed by other means.They are generally developed by simulation (with anevent-type model) of runoff from design storms. Theevents represent very rare occurrences, generally wellbeyond the range of events for which reliable frequencyestimates (from statistically based frequency curves) couldbe made. This section defines each design flood and discussesissues associated with their derivation.13-5

EM 1110-2-141731 Aug 94Figure 13-2.Index and interpolated hydrographs13-6

EM 1110-2-141731 Aug 94a. Standard project flood. The standard projectflood (SPF) is the flood that can be expected from themost severe combination of meteorologic and hydrologicconditions that are considered reasonably characteristic ofthe region in which the study basin is located. The SPFis generally based on analysis (and transposition) of majorstorms that have occurred in the region and selection of astorm magnitude and temporal distribution that is assevere as any of the transposed storms, with the possibleexception of any storm or storms that are exceptionallylarger than others and are considered to be extremely rare.Studies compiled in the United States indicate that SPFpeak discharges are usually of the order of 40 to 60 percentof probable maximum peak discharges.(1) The SPF is intended as a practicable expression ofthe degree of protection to be considered for situationswhere protection of human life and high-valued propertyis required, such as for an urban levee or floodwall. Italso provides a basis of comparison with the recommendedprotection for a given project. Although aspecific fre quency cannot be assigned to the SPF, areturn period of a few hundred to a few thousand years iscommonly associated with it.(2) Because the standard project storm (SPS) is notwidely used outside the USACE, only a limited number ofpublications describe its derivation and use. EM 1110-2-1411 describes SPS derivation for the United States eastof the 105° longitude. Computer program HEC-1 containsan option for automatically applying this criteria.SPS development for the remainder of the United States isbased on various published and unpublished Corps reportsand procedures. Sometimes the SPF is developed as aproportion (e.g., 50 percent) of the probable maximumflood.(3) Associated with SPF simulation is the choice ofloss rate and base flow parameter values and perhapsantecedent snowpack and related information. Loss ratesand base flow should be commensurate with values consideredreasonably likely to occur during storms of suchmagnitude. They should be estimated on the basis ofrates observed in floods that have occurred in the basin orin similar areas. EM 1110-2-1406 is a source of informationrelating to snowpack and snowmelt assumptions toassociate with an SPF.b. Probable maximum flood. The probable maximumflood (PMF) is the flood that may be expected fromthe most severe combination of critical meteorologic andhydrologic conditions that are reasonably possible in theregion. It is used in the design of projects for whichvirtually complete security from flood-induced failure isdesired. Examples are the design of dam height andspillway size for major dams and protection works fornuclear power plants.(1) The PMF is calculated from a probable maximumstorm (PMS), generally with an event-type model.The PMS is based on probable maximum precipitation(PMP), criteria developed by the HydrometeorologicalBranch of the Office of Hydrology, NWS. Figure 13-3shows regions of the contiguous United States for whichgeneralized PMP criteria have been developed. The hydrometeorologicalreports shown in the figure are listed inthe references. Hydrometeorological Report (HMR)No. 52 (U.S. Department of Commerce 1982) providescriteria for developing a PMS based on PMP criteria fromReports No. 51 and 55 (U.S. Department of Commerce1978 and 1988) (for the United States east oflongitude 105°). A computer program (USACE 1984b)has been developed to apply the criteria in Report No. 52.Hydrometeorological criteria are being updated for variousareas of the country. A check should be made for themost recently available criteria prior to performing astudy. In regions where there are strong orographic influences,it is sometimes desirable for basin-specific criteriato be developed. Such studies require considerable timeand dollar resource commitments, and their need shouldbe well established. The Hydrometeorological Branch ofthe NWS is partially funded by the USACE and is availableto serve in a consulting capacity.(2) The technical basis for PMP estimation isdescribed in the various hydrometeorological reports.NOAA Technical Report NWS 25 contains maps indicatingstorms of record that produced rainfall within 50 percentof PMP. (Other maps show the ratio of point PMPto 100-year values.) Such information shows that PMPvalues are consistent with reasonable extrapolation of themajor storms of record; in some cases, the extrapolation isless than about 10 percent.(3) Ground conditions that affect losses during thePMS should be the most severe that can reasonably existin conjunction with such an event. The lowest loss ratesthat have been developed for historical storms might beused if there is reasonable assurance that such stormsrepresent severe conditions. Where it is possible for theground to be frozen at the start of a rain flood or snowmeltflood, it can be concluded that zero or near-zero lossrates should be used. If there is a seasonal variation inminimum loss rates, the values selected should be representativeof extreme conditions for the season for whichthe PMF is being developed.13-7

EM 1110-2-141731 Aug 94Figure 13-3. Regions covered by generalized PMP studies(4) For situations where snowpack/snowmelt is afactor, it is generally not feasible to compute maximumsnowpack accumulation from winter precipitation, temperature,and snowmelt losses. Rather, a probable maximumsnowpack for floods that are primarily snowmelt floodscan be estimated by extrapolation of historical snowpackdata. In the case of rain floods that have some snowmeltcontribution, snowpack used for probable maximum rainfloodcomputation should be the maximum that can contributeto the peak flow and runoff volume of the floodwithout inhibiting the direct runoff from rainfall. Thecritical snowpack in mountainous regions will ordinarilybe located at elevations where most of the rain-floodrunoff originates. Snowpack is ordinarily greater athigher elevations and less at lower elevations, and hencecritical snowpack will not exist at all elevations. Factorsto be considered in selecting temperature sequences forsnowmelt simulation are discussed in EM 1110-2-1406.(5) Runoff parameter values used for the transformationof rainfall/snowmelt to runoff should be appropriatefor the magnitude of the event being simulated. Traveltimes tend to be significantly shorter during major events.Indices of travel time, such as values for unit hydrographparameters and routing coefficients, are frequentlyadjusted downward from their magnitudes based on historicalevents to reflect the severe conditions. In applicationsfor spillway design, allowance should be made forthe acceleration effect of a reservoir in relation to thestream reaches that are inundated.(6) In spillway design applications, flood conditionsthat precede the PMF may have substantial influence onthe regulatory effect of the reservoir. In such cases, it isappropriate to precede the PMF with a flood of majormagnitude at a time interval that is consistent with thecausative meteorological conditions. While a specialmeteorological study is desirable for this purpose, it isoften assumed that the PMF is preceded by a SPF 4 or5 days earlier.13-8

EM 1110-2-141731 Aug 94Chapter 14Period-of-Record Analysis14-1. GeneralPeriod-of-record analysis is seeing increasing interest andusage due to the continuing decrease in the costs of computerprocessing and the increased availability of hydrologicmodels with continuous simulation capability. Asused in this document for flood-runoff analysis, period-ofrecordanalysis refers to applying a precipitation-runoffmodel to simulate a continuous period of record ofstreamflow, including the detailed simulation of floodevents. This method requires a relatively sophisticatedhydrologic model capable of simulating throughout thehydrologic cycle; it implies a more complete model calibrationeffort; and, it requires extensive data and dataprocessing. Because of these factors, it is not an inexpensiveapproach to flood-runoff analysis and therefore notan economical application in many situations. However,certain engineering applications, e.g., the detailed evaluationof the effects of urbanization in a basin, are readilysuited to this type of analysis.14-2. Simulation RequirementsBecause period-of-record analysis requires the continuousand detailed simulation of stream flow from precipitation,additional modeling requirements are required beyondthose normally associated with the simulation of discretestorm events. Previous chapters in Part II of this manualhave described the processes associated with individualflood analysis, including precipitation/runoff transformationand routing techniques. These techniques are alsoapplicable to continuous simulation. Beyond these, however,are several additional factors that must be treated ina continuous modelling effort, summarized as follows:a. Evapotranspiration.b. Lake and reservoir evaporation.c. Long-term subsurface simulation.d. Distributed watershed formulation.e. Interception.f. Data processing requirements.14-3. Model CalibrationThe process of deriving characteristics, equation constants,weighting factors, and other parameters that serve todefine the model for a particular watershed is termed“calibration.” (Strictly speaking, “calibration” is distinguishedfrom “verification,” as described below.) Incontinuous simulation, the calibration process is generallymore rigorous and complex than is model developmentfor discrete storm analysis, in that more parameters areusually involved in a continuous model; a much greateramount of hydrometeorological data is involved; the fittingof the model requires a greater number of hydrologicfactors (i.e., short- and long-term volumes, year-to-yearcarryover of volume, low-flow streamflow reproduction--aswell as peak flow and flood-runoff timing); andmore rigorous statistical procedures are usually employedto ensure that an unbiased fitting of the model isachieved.a. Calibration process. The following is an outlineof the steps typically followed in calibrating a continuoussimulation model.(1) Data development. The data base developmentfor the model can be a time-consuming process, requiringcareful attention. Although digital sources now exist foreasy downloading of streamflow, precipitation, temperature,and snow data, these sources may not include anadequate frequency of observations. For example, a smallbasin may require hourly observations, for satisfactorysimulation, that are not readily available from commondata sources. Calibration of a continuous simulationmodel typically employs from 5 to 30 or more years ofcontinuous records, so the data processing task is relativelylarge if a frequent timestep is required. The presenceof poor quality data can be a problem. Precheckingthe data by such techniques as graphics display or bydouble-mass curve analysis and other station cross-checkingprocedures is desirable. The use of a data managementsystem such as HECDSS is useful in this regard.(2) Station selection. The choice of which precipitationand temperature stations best represent the basin-widemeteorological input might take several iterations throughthe entire calibration process. However, reasonablyappropriate choices can be made prior to calibrationthrough intuitive inspection of station location andcharacteristics, use of normal annual isohyetal maps,simple correlations of precipitation with runoff, etc.These factors were described in detail in Chapter 8.14-1

EM 1110-2-141731 Aug 94(3) Initial model parameters. The initial choice ofmodel parameters is not a critical concern since adjustmentswill be made during calibration. However, thoseparameters that have physical relevance should be determinedto reduce the possibilities for future adjustment asthe calibration proceeds. Table 14-1 lists the modelparameters that are typically encountered in continuoussimulation models and indicates those factors that can bedetermined by independent analysis. For other parametersthat need to be empirically determined, the initial valuemight be based upon known factors in previous simulationstudies, by examples given in user manuals, or by defaultvalues in the computer program.(4) Water balance. A desirable, if not essential, partof the calibration process is to make an independent estimateof the basin’s water balance. This calculation wouldyield annual, or perhaps monthly, estimates of basin precipitation,evapotranspiration, and soil moisture that canbe helpful in calibrating the model.(5) Parameter adjustment. Trial simulation runs aremade and model output is compared with observedstreamflow and runoff data as described in paragraph13-3. Based upon those comparisons, parameteradjustments are then made to improve the fit of themodel. This process requires an experienced and knowledgeableperson, both in the use of the model and understandingbasic hydrologic principles. Adjustments aremade first to those factors which have the greatest impacton the model fit, then proceeding to variables with lessersensitivity. The process may be expressed as three basicsteps (each having several trials) as follows:(a) Achieve fit of runoff volumes throughout the year(monthly water balance). This process primarily involvesadjustment of precipitation weighting, loss-rate functions,and evapotranspiration factors. Calibration fit is usuallyjudged by comparing monthly and annual runoff volumes.(b) Develop hydrograph shape. This step involvesworking with runoff distribution and routing factors, particularlyin the lower-zone components. If snow is afactor, then temperature and snow accumulation/ablationfactors may need to be adjusted.(c) Refine hydrograph fit. This final step involvesworking with surface runoff factors and other parametersto refine the hydrograph shape.(6) Table 14-1 describes this priority order in moredetail and gives relative sensitivity of the variables. Mostof the parameters in a continuous simulation modelrepresent a physical process. It is imperative that parametervalues remain physically reasonable throughout thecalibration process to keep the fit from being a localoptimization that will not work when extrapolated to newdata. The verification step described below is highlydesirable to ensure that the fit is a general solution, notone unique only to the calibration data used.b. Calibration comparison tools. Continuous simulationmodels use and create an immense amount of data,particularly if a long period of record is involved. Judgingthe fit of the final streamflow output alone is difficult,but reviewing the intermediate output such as soil moisturelevels, snow pack, and runoff component hydrographs,makes the task more difficult. Accordingly, it isalmost mandatory that special techniques be employed tofacilitate comparisons of calibration runs and make modeladjustments. These techniques are preferably built intothe computer program being utilized. The following areexamples of tools that are typically employed:(1) Tabular summaries:(a) Monthly and annual volume summaries in unitsof runoff volume and in inches.(b) Summary tabulations of model internalcomputations.(2) Graphical displays:(a) Hydrographs of observed and computedstreamflow.(b) Hydrographs of model internal component output(e.g., soil moisture, subsurface flow).(c) Flow-duration curves, observed and computedstreamflow.(d) Scatter-plots, monthly runoff volumes.(e) Period residuals (observed - computed flow) oraccumulated errors versus time.(3) Statistical calculations:(a) Statistical summaries of monthly volumes.(b) Root-mean square error of period deviations(computed minus observed).14-2

EM 1110-2-141731 Aug 9414-3

EM 1110-2-141731 Aug 94c. Verification. After calibration of the model iscomplete, it is good practice to then simulate an independentperiod of record and compare the results withobserved data. This procedure will help to ensure that thecalibration is not unique and limited to the data setemployed for calibration.14-4. ApplicationsIt is important in approaching a possible application ofperiod-of-record analysis to be certain that it is a necessaryand appropriate approach to solving the problem,since the commitment of time and resources is relativelyhigh. On the other hand, this type of analysis is availableas a potentially powerful tool in hydrologic analysis andforecasting for types of applications that may not be obvious.To assist in the decision-making on applications, andfor providing references if similar studies are undertaken,the following actual and potential applications aredescribed:a. Extension of streamflow records. In situationswhere weather records in a basin have a longer period ofrecord than streamflow stations, continuous simulationwould be a logical method of extending a record ofstreamflow, particularly if a continuous flow record isdesired (as opposed to, say, just peak flows). The modelused would be calibrated and verified on the observeddata and extended as meteorological data permit.b. Derivation of ungauged streamflow records. Thisapplication is quite feasible and has been utilized in theprofession. Since the effort involved is not small, it islikely that it would not be used in ordinary planninginvestigations but might be appropriate for special situations,e.g., cases with legal or controversial ramifications.The method relies on the fact that most likely adjacentbasins will have similar subterranean characteristics, sothat if a detailed simulation model is developed on a basinwith streamflow data, subsurface and groundwater characteristicscan be transferred to the ungauged basin with arelatively high degree of confidence. Surface characteristicsalso can be based upon the gauged basin but arelikely to be modified as necessary by observable factorssuch as slope, terrain, etc.characteristics of a river basin are quite well suited toperiod-of-record analysis with a continuous simulationmodel. The model can be calibrated by relating observedphysical conditions (past records might likely exist reflectingeither no development or partial development) toobserved hydrometeorological data. Then, the period ofrecord of hydrometeorological data can be simulatedutilizing the observed or forecasted physical conditions tobe evaluated. The resulting flows will be a large sampleof data for statistical representation, reflecting a consistentlevel of basin development. The model used in this typeof analysis would have to be capable of representing thephysical changes involved; i.e., increase in imperviousarea, changes in runoff response, etc.d. Interior runoff analysis. A stochastic analysismay be required in the planning and design of interiordrainage facilities for leveed areas, particularly when therelative timing and magnitude of the main river and theinterior runoff are important in determining the economicsof the project. Although the main river would likely havean adequate record of streamflow data, most interiordrainage areas do not. By using continuous simulation,the rainfall-runoff calculation required for the interior areacan be performed and conveniently joined with the mainchannel streamflow, which would either be derived by therain-runoff model or based upon observed streamflowdata.e. Long-term runoff forecasting. Continuous simulationhas been used to produce long-term forecasts ofstreamflow for operational purposes. In a techniquecalled “extended streamflow prediction” (ESP), the NWSand others have combined period-of-record weatherrecords for a given future period of up to several monthswith current basin hydrologic conditions to produce astatistical representation of future conditions. The procedureis best suited for the Western interior river basinswith large winter snowpacks, where a snowpack inJanuary plays a relatively large role in determining runoffin May and June. The statistical analysis produced by theperiod-of-record simulation reflects the variations in subsequentprecipitation and temperature combined with thecurrent snow conditions. Successive forecasts made asspringtime approaches have less and less variance.c. Analysis of basin modifications. The assessmentof urbanization effects and other changes in the physical14-4

PART IVENGINEERING APPLICATIONS

EM 1110-2-141730 Aug 94Chapter 15Data Collection and Management15-1. Generala. Water resource studies tend to be data intensive.One reason is that the physical systems involved are oftenlarge and complex (e.g., watersheds, precipitation fields,river-reservoir systems, etc.), and substantial quantities ofdata are required for their representation. Another reasonis that the investigations themselves are complex, with avariety of interdependent computational elements (e.g.,precipitation-runoff simulation, and statistical, systems,and economic analyses, etc.). The transfer of data generatedwith one element to another is a significant requirementin such investigations.b. The acquisition, processing, and management ofdata can require a substantial portion of the resourcesallotted for an investigation. Performance of these tasksin an efficient and reliable manner can be of considerablevalue. This chapter describes aspects of datamanagement.15-2. Data Management Conceptsa. Figure 15-1 illustrates components of a datamanagement system for a water resource study. Elementsof the system include a data loading module for enteringdata from various sources into the management system,“application” programs that read information from andwrite information to data storage, and utility programs thatperform functions such as data editing, displaying data ingraphical and tabular form, and mathematical transformationsof data. With such a system, basic data can beloaded into storage, reviewed, and perhaps edited withutility programs. Interdependent application programs canbe used to perform the analysis, using the data storage topass information generated with one program to the next.Utility programs can be used to prepare summaries ofresults in various forms, including report-quality tablesand graphs.b. Common data types for flood-runoff analysisinclude individual element, time series, and paired function.Individual element data include items such as basinproperties (e.g., drainage area, percent imperviousness,soil types), values for runoff parameters (e.g., unit hydrograph,kinematic wave parameters, baseflow, loss rate),structure dimensions, inventories of gauge types and locations,etc. Time series data consist of values of a variablefor sequential points in time such as discharge and stagehydrographs, precipitation hyetographs, air temperaturerecords, etc. Paired function data are sets of interrelatedvariables for which each value of one variable is pairedwith a value of another, such as discharge elevation,exceedance frequency-stage, reservoir storage-elevation,etc.c. There are a number of commercial data bases thatare well suited for the storage of individual element data.Such data bases are relational and permit queries such as“list all gauges within specified latitude and longitudebounds,” or “list all subbasins for which the soils are insoil group A.” Whether or not it is desirable to utilize adata base for individual element data depends on the datatype and the frequency of use intended for the data.d. Hydrologic studies generally make extensive useof time series data. Data bases that are designed specificallyfor this data type gain efficiency by treating suchdata in blocks (i.e., groups or sets) rather than asindividual data items. Storage and retrieval is performeda block at a time. Block size might be, for example, onemonth for hourly data. A system designed for use withtime series data is the data storage system (DSS) developedby the Hydrologic Engineering Center. It is configuredas in Figure 15-1, with a number of water resourceapplication programs having the capability to communicatewith DSS files. A comprehensive set of data loadingand utility programs supports the system.e. Paired function data are also widely used in hydrologicstudies. An advantage of central storage of discharge-elevationrating “curves,” or any paired data, isthat changes made to the data need to be made in oneplace only, and all application programs that use the datawill have access to the revised data. The DSS system isdesigned to accommodate paired function data.15-3. Geographic Information SystemsA powerful data management tool for spatial (i.e., geographicallyoriented) data is the geographic informationsystem (GIS). Such systems enable the storage andretrieval of information that is associated with spatial elementssuch as rectangles, triangles, or irregularly shapedpolygons. Variables such as slope, orientation, elevation,soil type, land use, average annual rainfall, etc. can bestored for each element. The data may then be retrievedand tabulated, displayed graphically, or used directly byapplication programs. Several commercial GIS’s areavailable.15-1

EM 1110-2-141730 Aug 94Figure 15-1. Data management15-4. Data Acquisition and Usea. The use of data typically involves the following:based on the purpose and scope of the study, determinethe types of data that will be required; determine thesources and availability of the data; acquire and processthe data; perform the analysis; and archive the data andstudy results. The first two steps are components of studyformulation. As stated previously, the type, amount, andquality of data available for a study can have a significantimpact on the choice of methodology and reliability ofresults.b. Data for hydrologic investigations are generallyobtained from several sources. For example, streamflowrecords are commonly available from the USGS, anddaily and hourly rainfall values from the NWS. Also,commercial firms obtain such data from collectionagencies and make them available in useful form (e.g., oncompact disk). Various formats are used to encode thedata, and these must be interpreted as part of the processof loading data into a data base. There are a number ofdata loading programs associated with DSS, includingprograms to read data formats used on commerciallyavailable compact disks, NWS data formats andU.S. Geological Survey WATSTORE formats. In addition,software is available to read the Standard HydrometeorologicalExchange Format (SHEF), which isaccepted as a standard for data exchange by a number ofagencies. The function of the loading programs is to readdata in the appropriate format and enter that data into aDSS file. After the data has been loaded, utility programscan be used to graph or tabulate the data and perhaps editor apply transforms to it (such as stage to discharge).15-2

EM 1110-2-141730 Aug 94c. Application programs that have the capability toaccess data storage must be “told” what and how muchdata to retrieve. Such instructions are part of programinput, as are instructions specifying the calculated informationthat is to be written to data storage. Thecapability to review application program results in tabularor graphical form with utility programs can be verypowerful in facilitating the performance of a study. Finalresults can then be produced in report-quality form.d. Upon completion of a study, data and studyresults should be prepared for long-term storage. Becauseformats used in specific data management systems maychange over time, data should be stored in a systemindependentformat. For example, information from aDSS file can be transferred to a text (ASCII) file.15-3

EM 1110-2-141731 Aug 94Chapter 16Ungauged Basin Analysis16-1. Generala. Problem definition. Earlier chapters of this manualdescribed various flood-runoff analysis models. Some ofthe models are causal; they are based on the laws ofthermodynamics and laws of conservation of mass,momentum, and energy. The St. Venant equationsdescribed in Chapter 9 are an example. Other models areempirical; they represent only the numerical relationshipof observed output to observed input data. A linearregressionmodel that relates runoff volume to rainfalldepth is an empirical model.(1) To use either a causal or empirical flood-runoffanalysis model, the analyst must identify model parametersfor the catchment or channel in question. Paragraph7-3e described a method for finding rainfall-runoffparameters for existing conditions in a gauged catchment.Through systematic search, parameter values are found toyield computed runoff hydrographs that best matchobserved hydrographs caused by observed rainfall. Withthese parameter values, runoff from other rainfall eventscan be estimated with the model. A similar search can beconducted for routing model parameters, given channelinflow and outflow hydrographs.(2) Unfortunately, as Loague and Freeze (1985) pointout, “...when it comes to models and data sets, there is asurprisingly small intersecting set.” The rainfall andrunoff data necessary to search for the existing-conditioncalibration parameters often are not available. Streamflowdata may be missing, rainfall data may be sparse, or theavailable data may be unreliable. Furthermore, forUSACE civil-works project evaluation, runoff estimatesare required for the forecasted future and for with-projectconditions. Rainfall and runoff data are never availablefor these conditions. In the absence of data required forparameter estimation for either existing or future conditions,the stream and contributing catchment are declaredungauged. This chapter presents alternatives for parameterestimation for such catchments.b. Summary of solutions. To estimate runoff from anungauged catchment, for existing or forecasted-future conditions,the analyst can use a model that includes onlyparameters that can be observed or inferred from measurements,or extrapolate parameters from parametersfound for gauged catchments within the same region.In practice, some combination of these solutions typicallyis employed, because most models include both physicallybased and calibration parameters.c. Using models with physically based parameters.Model parameters may be classified as physically basedparameters or as calibration parameters.(1) Physically based parameters are those that can beobserved or estimated directly from measurements ofcatchment or channel characteristics.(2) Calibration parameters, on the other hand, arelumped, single-valued parameters that have no directphysical significance. They must be estimated from rainfalland runoff data. If data necessary for estimating thecalibration parameters are not available, one solution is touse a flood-runoff analysis model that has only physicallybased parameters. For example, the parameters of theMuskingum-Cunge routing model described in paragraph9-3a(6) are channel geometry, reach length, roughnesscoefficient, and slope. These parameters may beestimated with topographic maps, field surveys, photographs,and site visits. Therefore, that model may beused for analysis of an ungauged catchment.d. Extrapolating calibration parameters. If thenecessary rainfall or runoff data are not available to estimatecalibration parameters using a search procedure suchas that described in paragraph 7-3e, the parameters maybe estimated indirectly through extrapolation of gaugedcatchmentresults. This extrapolation is accomplished bydeveloping equations that predict the calibration parametersfor the gauged catchments as a function of measurablecatchment characteristics. The assumption is that theresulting predictive equations apply for catchments otherthan those from which data are drawn for development ofthe equations. The steps in developing predictive relationshipsfor calibration parameters for a rainfall-runoff modelare as follows:(1) Collect rainfall and discharge data for gaugedcatchments in the region. The catchments selected shouldhave hydrological characteristics similar to the ungaugedcatchment of interest. For example, the gauged andungauged catchments should have similar geomorphologicaland topographical characteristics. They should havesimilar land use, vegetative cover, and agriculturalpractices. The catchments should be of similar size.Rainfall distribution and magnitude and factors affectingrainfall losses should be similar. If possible, data shouldbe collected for several flood events. These rainfall and16-1

EM 1110-2-141731 Aug 94discharge data should represent, if possible, events consistentwith the intended use of the model of the ungaugedcatchment. If the rainfall-runoff model will be used topredict runoff from large design storms, data from largehistorical storms should be used to estimate the calibrationparameters.(2) For each gauged catchment, use the data to estimatethe calibration parameters for the selected rainfallrunoffmodel. The procedure is described in Chapter 7,and guidelines for application of the procedure are presentedin Chapter 13 of this document.(3) Select and measure or estimate physiographiccharacteristics of the gauged catchments to which therainfall-runoff model parameters may be related.Table 16-1 lists candidate catchment characteristics.Some of these characteristics, such as the catchment area,are directly measured. Others, such as the Horton ratios,are computed from measured characteristics.Table 16-1Catchment Characteristics for Regression ModelsTotal catchment areaArea below lowest detention storageStream lengthSteam length to catchment centroidAverage catchment slopeAverage conveyance slopeConveyance slope measured at 10% and 85% of stream length(from mouth)Height differentialElevation of catchment centroidAverage of elevation of points at 10% and 85% of stream lengthPermeability of soil profileSoil-moisture capacity average over soil profileHydrologic soil groupPopulation densityStreet densityImpervious areaDirectly-connected impervious areaArea drained by storm sewer systemPercent of channels that are concrete linedLand useDetention storageRainfall depth for specified frequency, durationRainfall intensity for specified frequency, durationHorton’s ratios (Horton 1945)Drainage density (Smart 1972)Length of overland flow (Smart 1972)(4) Develop predictive equations that relate the calibrationparameters found in step 2 with characteristicsmeasured or estimated in step 3. In a simple case, theresults of steps 2 and 3 may be plotted with the ordinate arainfall-runoff model parameter and the abscissa acatchment characteristic selected in step 3. Each point ofthe plot will represent the value of the parameter and theselected characteristic for one gauged catchment. Withsuch a plot, a relationship can be “fitted by eye” andsketched on the plot. Regression analysis is an alternativeto the subjective graphical approach to defining a predictiverelationship. Regression procedures numericallydetermine the optimal predictive equation. Details ofregression analysis are presented in EM 1110-2-1415 andin most statistics texts, including those by Haan (1977)and McCuen and Snyder (1986).(a) To apply a parameter-predictive equation for anungauged catchment, the independent variables in theequation are measured or estimated for the ungaugedcatchment.(b) Solution of the equation with these values yieldsthe desired flood-runoff model parameter. This parameteris used with the same model to predict runoff from theungauged catchment.16-2. Loss-Model Parameter Estimatesa. Options. Two of the rainfall loss modelsdescribed in Chapter 6 of this document are particularlyuseful for ungauged catchment analysis: the Green-Amptmodel and the SCS model. The Green-Ampt model is acausal model with quasiphysically based parameters. TheSCS loss model is an empirical model with parametersthat have been related to catchment characteristics. Otherloss models may be used if parameter-predictive equationsare developed from gauged catchment data.b. Physically based parameter estimates for Green-Ampt model. The Green-Ampt model is derived fromDarcy’s law for flow in porous media. The model predictsinfiltration as a function of time with three parameters:volumetric moisture deficit, wetting-front suction,and hydraulic conductivity. In application, an initial lossmay be included to represent interception and depressionstorage. Additional details of the Green-Ampt model arepresented in Chapter 6.(1) Brakensiek, and Onstad (1988), McCuen, Rawls,and Brakensiek (1981), Rawls and Brakensiek (1982a),Rawls, Brakensiek, and Saxton (1982b), Rawls and16-2

EM 1110-2-141731 Aug 94Brakensiek (1983a), Rawls, Brakensiek, and Soni (1983b),and Rawls and Brakensiek (1985) propose relationships ofthe Green-Ampt model parameters to observable catchmentcharacteristics, thus permitting application of themodel to an ungauged catchment. The relationshipsdefine model parameters as a function of soil textureclass.(2) Texture class, in turn, is a function of soil particlesize distribution. This distribution can be estimated froma sample of catchment soil. For example, a soil that is80 percent sand, 5 percent clay, and 10 percent silt isclassified as a loamy sand. For this texture class, Rawlsand Brakensiek (1982a) and Rawls, Brakensiek, andSaxton (1982) suggest that the average saturated hydraulicconductivity is 6.11 cm/hr. The other parameters can beestimated similarly from the soil sample.c. Predictive equations for SCS model parameters.The SCS loss model, described in detail in Chapter 6, isan empirical model with two parameters: initial abstractionand maximum watershed retention (maximum loss).Often both parameters are related to a single parameter,the curve number (CN). Using data from gauged catchmentsin the United States, the SCS developed a tabularrelationship that predicts CN as a function of catchmentsoil type, land use/ground cover, and antecedent moisture.Table 16-2 is an excerpt from this table (U.S. Departmentof Agriculture (USDA) 1986).(1) To apply the SCS loss model to an ungaugedcatchment, the analyst determines soil type from a catchmentsoil survey. For many locations in the UnitedStates, the SCS has conducted such surveys and publishedsoil maps. The analyst determines existing-condition landuse/ground cover from on-site inspection or throughremote sensing. For future conditions, the land use/ground cover may be determined from developmentplans. The analyst selects an appropriate antecedentmoisture condition for catchment conditions to bemodeled (wet, dry, or average). With these three catchmentcharacteristics estimated, the tabular relationshipmay be used to estimate CN. For example, for a residentialcatchment with 2-acre lots on hydrologic soil group C,the CN found in Table 6-6 for average antecedent moistureis 77. With this CN, the initial abstraction and maximumwatershed retention can be estimated, and the lossfrom any storm can be predicted.(2) Publications from the SCS provide additionaldetails for estimating the CN for more complex cases.16-3. Runoff-Model Parameter Estimatesa. Options. Chapter 7 presents a variety of modelsfor estimating runoff due to excess rainfall. For anungauged catchment, the analyst may use the kinematicwavemodel, a UH model with physically-based parameters,or a UH model with predictive equations for thecalibration parameters.b. Physically based parameter estimates for kinematicwave model. The kinematic-wave model describedin Chapter 7 is particularly well suited to analysis of anungauged urban catchment.(1) This causal model, which is described in furtherdetail in HEC documents (USACE 1979, 1982, 1990a),represents the catchment rainfall-runoff process by solvingtheoretical equations for flow over planes. Catchmentrunoff is estimated by accumulating the flow from manysuch planes.(2) Application of the model requires identificationof the following parameters: catchment area, flow length,slope, and overland-flow roughness factor. The area,length, and slope are physically based and are estimatedfor existing catchment conditions from maps, photographs,or inspection. For forecasted-future condition, theseparameters are forecasted from development plans. Theoverland-flow roughness factor is a quasiphysically basedparameter that describes resistance to flow as a functionof surface characteristics. Published relationships, basedon hydraulic experimentation, are used to select this coefficientfor existing or forecasted conditions. Thus allparameters of the kinematic wave model can be estimatedwithout gauged data.c. Physically based parameter estimates for Clark’sIUH and SCS UH. Parameters of Clark’s and the SCSempirical UH models have a strong link to the physicalprocesses and thus can be estimated from observation ormeasurement of catchment characteristics. Clark’s IUHaccounts for translation and attenuation of overland andchannel flow. Translation is described with the time-dischargehistogram. To develop this histogram, the time ofconcentration is estimated and contributing areas are measured.Likewise, the SCS UH hydrograph peak and timeto peak are estimated as a function of the time of concentration.The time of concentration, t c , can be estimatedfor an ungauged catchment with principles of hydraulics.The SCS suggests that t c is the sum of travel times for all16-3

EM 1110-2-141731 Aug 94consecutive components of the drainage conveyance system(USDA 1986).That is,wheret ct 1t 2... t mt i= travel time for component im = number of components(16-1)Each component is categorized by the type of flow. Inthe headwaters of streams, the flow is sheet flow across aplane. Sheet-flow travel time is estimated via solution ofthe kinematic-wave equations. The SCS suggests a simplifiedsolution. When flow from several planes combines,the result is shallow concentrated flow. The traveltime for shallow concentrated flow is estimated with anopen-channel flow model, such as Manning’s equation.Shallow concentrated flow ultimately enters a channel.The travel time for channel flow is estimated also withManning’s equation or an equivalent model.d. Predictive equations for UH calibration parameters.The procedure described in paragraph 16-1d can beused to develop predictive equations for UH modelparameters for ungauged catchments. For example,Snyder (1938) related unit hydrograph lag, t p , to a catchmentshape factor using the following equation:If they are, the analyst must be especially cautious. He orshe should review derivation of the equations. Conditionsunder which the equations were derived should be examinedand compared with conditions of the catchments ofinterest.Table 16-2Example UH Parameter Prediction EquationsEquationC t =7.81/I 0.78C p= 0.89 C t0.46R=cT cReferenceWright-McLaughlinEngineers (1969)Wright-McLaughlinEngineers (1969)Russell, Kenning,and Sunnell (1979)T c/ R = 1.46 - 0.0867 L 2 /A Sabol (1988)T c = 8.29 (1.00 + I) -1.28 (A/S) 0.28 USACE (1982)Note: In the above equations,C t= calibration coefficient for Snyder’s UH (see paragraph 7-3c)C p= calibration coefficient for Snyder’s UH (see paragraph 7-3c)T c= time of concentration, in hoursR = Clark’s IUH storage coefficient, in hoursI = impervious area, in percentL = length of channel/ditch from headwater to outlet, in mileswheret pC t( LL ca) 0.3(16-2)S = average watershed slope, in feet per footc = calibration parameter (for forested catchments = 8 - 12, forrural catchments = 1.5 - 2.8, and for developed catchments= 1.1 - 2.1)t p= basin lag, in hoursA = catchment area, in square milesC tL= predictive-equation parameter= length of main stream, in milesL ca = length from outlet to point on stream nearestcentroid of catchment, in milesThe value of C t is found via linear regression analysiswith data from gauged catchments. A wide variety ofpredictive equations for UH model calibration parametershave been developed by analysts. Table 16-2 showsexample equations for Snyder’s and Clark’s UH parameters.In general, these equations should not be used inregions other than those for which they were developed.16-4. Routing-Model Parameter Estimatesa. Candidate models. The routing models describedin Chapter 9 account for flood flow in channels. Of themodels presented, the Muskingum-Cunge, modified puls,and kinematic-wave are most easily applied in ungaugedcatchments. Parameters of each of these models arequasiphysically based and can be estimated from channelcharacteristics.b. Physically based parameter estimates for modifiedpuls routing model. The modified puls (level-pool)routing model is described in detail in Chapter 9-3. The16-4

EM 1110-2-141731 Aug 94parameters of this model, as it is applied to a river channel,include the channel storage versus outflow relationshipand the number of steps (subreaches). The former isconsidered a physically based parameter, while the latteris a calibration parameter.(1) For an ungauged catchment, the channel storageversus outflow relationship can be developed with normaldepth calculations or steady-flow profile computations. Ineither case, channel cross sections are required. Thesemay be measured in the field, or they may be determinedfrom previous mapping or aerial photography. Both proceduresalso require estimates of the channel roughness.Again, this may be estimated from field inspection orfrom photographs. With principles of hydraulics, watersurfaceelevations are estimated for selected discharges.From the elevations, the storage volume is estimated withsolid geometry. Repetition yields the necessary storageversus outflow relationship. These computations can beaccomplished conveniently with a water-surface profilecomputer program, such as HEC-2 (USACE 1990b).(2) The second parameter, the number of steps, is acalibration parameter. Paragraph 9-3a suggests estimatingthe number of steps as channel reach length/velocity ofthe flood wave/time interval (Eq. 9-13). Strelkoff (1980)suggests that if the flow is controlled heavily from downstream,one step should be used. For locally controlledflow typical of steeper channels, he suggests the moresteps, the better. He reports that in numerical experimentswith such a channel, the best peak reproduction wasobserved with:whereNSTPS2 L S oY oNSTPS = number of stepsL = entire reach length, in milesS 0 = bottom slope, in feet per mileY 0 = baseflow normal depth, in feet(16-3)So, for example, for a 12.4-mile reach with slope2.4 ft/mile and Y 0 = 4 ft, the number of steps would beestimated as 15.c. Physically based parameter estimates for kinematicwave model. The physical basis of the kinematic-wavemodel parameters makes that model useful for someungauged channels. In particular, if the channels aresteep and well-defined with insignificant backwatereffects, the kinematic-wave model works well. Theselimitations are met most frequently in channels in urbancatchments.(1) The parameters of the kinematic-wave channelrouting model include the channel geometry and channelroughness factor. The necessary channel geometryparameters include channel cross section and slope data.Since these are physically based, they may be estimatedfor existing conditions from topographic maps or fieldsurvey.(2) For modified channel conditions, the geometrydata are specified by the proposed design. The roughnessgenerally is expressed in terms of Manning’s n. This is aquasiphysically based parameter that describes resistanceto flow as a function of surface characteristics. Publishedrelationships predict this coefficient for existing or modifiedconditions.d. Physically based parameter estimates forMuskingum-Cunge model. If the channel of interest is notsteep and well-defined as required for application of thekinematic-wave channel routing model, a diffusion modelmay be used instead. In the case of an ungauged channel,the Muskingum-Cunge model is a convenient choice,since the parameters are physically based.(1) Parameters of the Muskingum-Cunge channelrouting model include the channel geometry and channelroughness factor. The necessary channel geometryparameters include channel cross section and slope data,which may be estimated for existing conditions fromtopographic maps or field survey.(2) For modified channel conditions, the geometrydata are specified by the proposed design. The roughnessis expressed in terms of Manning’s n.16-5. Statistical-Model Parameter EstimatesIn some hydrologic-engineering studies, the goal is limitedto definition of discharge-frequency relationships.EM 1110-2-1415 describes procedures for USACEflood-frequency studies. Chapter 12 of this documentsummarizes those procedures and describes the statisticalmodels used. All the models described are empirical.Observed data are necessary for calibration.Consequently, these statistical models cannot be applieddirectly to an ungauged catchment. Options available tothe analyst requiring frequency estimates for an ungauged16-5

EM 1110-2-141731 Aug 94where Q 100 = 100-year (0.01 probability) discharge.stream include development of frequency-distributionQ 1001040 A 0.91 (16-5)parameter predictive equations, and development of distributionquantile predictive equations.a. Parameter predictive equations. The log Pearson16-6. Reliability of EstimatesThe reliability of a runoff estimate made for an ungaugedtype III distribution (model) is used for USACE annual catchment is a function of the reliability of the floodrunoffmaximum discharge-frequency studies. As described inmodel, the form of the predictive equation and itsChapter 12, this model has three parameters. These are coefficients, and the talents and experience of the analyst.estimated from the mean, standard deviation, and skewcoefficient of the logarithms of observed peak discharges. a. Model reliability. Linsley (1986) relates theresults of a 1981 pilot test by the Hydrology Committee(1) In the absence of flow data, regional-frequencyanalysis procedures described in paragraph 12-5c may beof the USWRC that found that all runoff models testedwere subject to very large errors and exhibited a pronouncedapplied to develop distribution parameter predictive equations.bias to overestimate. He shows that errors ofAs with the equations for rainfall-runoff model plus or minus 10 percent in estimating discharge for aparameters, these equations relate model parameters to desired 100-year (0.01 probability) event may, in fact,catchment characteristics. For example, for the Shellpot yield an event as small as a 30-year event or as large as aCreek Catchment, Delaware, the following predictive 190-year event for design. Lettenmaier (1984) categorizesequation was developed (USACE 1982):the sources of error as model error, input error, andparameter error. Model error is the inability of a modelS 0.311 0.05 log A(16-4) to predict runoff accurately, even given the correct parametersand input. Input error is the result of error in specifyingrainfall for predicting runoff or in specifying rainfallwhereand runoff for estimating the model parameters. Thisinput error may be due to measurement errors or timingS = standard deviation of logarithmserrors. Parameter error is the result of inability toproperly measure physically based parameters or toA = catchment drainage area, in square milesproperly estimate calibration parameters. The net impactof these errors is impossible to quantify. They are identifiedWith similar equations, other parameters can be estimated.here only to indicate sources of uncertainty in dis-charge prediction.(2) To apply a distribution parameter-predictive equationfor an ungauged catchment, the independent variables b. Predictive equation reliability. Predictive equationsin the equation are measured or estimated for theare subject to the same errors as runoff models. Theungauged catchment. Solution of the equation with these form and parameters of the equations are not known andvalues yields the desired statistical distribution parameter. must be found by trial and error. The sample size uponThe frequency curve is then computed as described inEM 1110-2-1415 and Chapter 12.which the decision must be based is very small by statisticalstandards because data are available for relatively fewgauged catchments. Overton and Meadows (1976) go sob. Quantile predictive equations. The frequencydistributionfar as to suggest that the reliability of a regionalizedquantiles for an ungauged catchment also may model can always be improved by incorporating a largerbe defined with predictive equations. Such a predictive data base into the analysis. Predictive equations are alsoequation is developed by defining the frequency distributionsfor streams with gauged data, identifying from thesubject to input error. Many of the catchment characteristicsused in predictive equations have considerable uncertaintydistributions specified quantiles, and using regressionin their measured values. For example, theanalysis procedures to derive a predictive equation. For accuracy of stream length and slope estimates are a functionexample, for the Red Lion Creek Catchment, Delaware,the following quantile predictive equation was developed(USACE 1982):of map scale (Pilgrim 1986). Furthermore, many ofthe characteristics are strongly correlated, thus increasingthe risk of invalid and illogical relationships.16-6

EM 1110-2-141731 Aug 94c. Role of hydrologic engineer. Loague and Freeze(1985) suggest that hydrologic modeling is more an artthan a science. Consequently, the usefulness of theresults depends in large measure on the talents and experienceof the hydrologic engineer and her or his understandingof the mathematical nuances of a particular model andthe hydrologic nuances of a particular catchment. Thisposition is especially true in estimation of runoff from anungauged catchment. The hydrologic engineer must exercisewisdom in selecting data for gauged catchments, inestimating flood-runoff model parameters for these catchments,in establishing predictive relationships, and finally,in applying the relationships.16-7

EM 1110-2-141731 Aug 94Chapter 17Development of Frequency-BasedEstimates17-1. IntroductionFrequency-based estimates of flood discharge are a fundamentalrequirement for flood-risk investigations andflood-damage analysis. The development of such estimatesis a challenging task that requires sound interpretationof regional historical flood-related data andappropriate application of various analytical techniques.This chapter deals with issues such as choice of methodology,use of hypothetical storms in frequency determinations,transfer of frequency-based information fromgauged to ungauged sites, development of futureconditionfrequency estimates, and adjustment of peakdischarges to represent stationary conditions.17-2. Choice of Methodologya. Choice of methodology for frequency curvedevelopment will depend on the purpose of a study andcharacteristics of available data. Possible methodsinclude the following:(1) Statistical analysis of observed streamflow data.(2) Regional frequency analysis.(3) Event-type precipitation-runoff analysis withhypothetical storms.(4) Period-of-record precipitation-runoff analysis.b. Key questions related to study purpose are asfollows:(1) Will effects of future land-use changes or projectalternatives be evaluated?(2) Is period-of-record type information requiredbecause of the nature of the study?(3) What are accuracy and reliability requirements?c. The answer to the first question is a primarydeterminant for choice of methodology. If it is necessaryto model future land-use changes and/or the effects ofprojects, application of a precipitation-runoff simulationmodel is generally essential. The answer to the secondquestion will determine whether the simulation modelshould have capability for period-of-record analysis. Forexample, analysis of pond stage on the interior (landward)side of a levee often requires such analysis toreflect the coincident effects of exterior (main river)stage and interior runoff. If modeling of future land-usechanges and/or projects is not required, choice of methodologywill depend on availability of data and accuracyand reliability requirements. Key questions related toavailable data are as follows:(1) Are long-term historical discharge records availablefor the location(s) of interest?(2) Are long-term historical discharge records availablefor nearby sites?(3) Are short-term discharge records available forthe location(s) of interest?(4) Are (applicable) regional-frequency relationshipsavailable for the location(s) of interest?d. “Long-term” as used here refers to a length ofrecord sufficient to enable development of statisticallybased frequency estimates of reasonable reliability.Long-term data is extremely valuable and generally providesthe most reliable basis for frequency determinations.If land-use conditions have changed during theperiod of collecting the long-term data, or if there arereservoirs (with significant capacity to store flood runoff)upstream of the location(s) of interest, the period-ofrecordpeak discharges must generally be adjusted torepresent a stationary, storage-free condition. The makingof such adjustments can require substantial analysisand application of simulation.e. If long-record data are not available for the location(s)of interest, it may be possible to transfer (andadjust) discharge data or frequency-based informationfrom nearby, similar locations. Such transfer can bedifficult, and reliability of results is affected by the transferprocess. However, use of such data can be of substantialvalue and can result in frequency estimates thatare significantly more reliable than could be producedwithout such data.f. “Short-term” as used here implies that dischargerecords are insufficient to enable development of statisticallybased frequency estimates of reasonable reliability.Short-term data can, however, be adequate to enable thecalibration of a precipitation-runoff simulation model.17-1

EM 1110-2-141731 Aug 94Hence, the availability of such data is very significantwhen a simulation approach is required.g. When land-use changes and/or project conditionsare not a factor, it may be possible to employ regionalfrequencyrelations. Previously determined relationsshould be applied carefully; their applicability should beverified, and independent variables should be evaluatedproperly.h. In many cases, frequency estimates should bedeveloped by several independent techniques. Differentsegments of the adopted frequency curve may be derivedfrom different sources depending on the basis for, andreliability of, the individual estimates. All the means atone’s disposal should be used to verify resulting estimates.For example, it may be reasonable to expect thatthe standard project flood would have a magnitude withina certain range of exceedance frequency. The range canbe used as a rough check for the upper end of a derivedfrequency curve. Historical accounts of flooding shouldbe used, if possible, to verify estimates. Peak dischargeenvelope curves may also be useful.17-3. Hypothetical Storm Frequencya. The magnitude and spatial and temporal characteristicsof every natural storm is unique. Hence, it isonly possible to determine probabilities for average stormdepths over specific areas and for specific durations.Although generalized rainfall criteria such as that providedin NOAA publications associate recurrence intervalswith rainfall depths, the recurrence interval (orexceedance frequency) of a hypothetical storm developedfrom such depths is indeterminate. To label a storm as a“100-year” or “25-year” storm can therefore, bemisleading.b. What is generally of primary interest is theexceedance frequency of streamflow peaks and volumes.Attempts are, therefore, made to devise hypotheticalstorms that can be associated with the generation ofstreamflow peaks and/or volumes of specified exceedancefrequency. However, the runoff generated by a particularstorm will be a function of the state of the watershedwhen the storm occurs. A major storm occurring on avery dry watershed can result in moderate runoff, and amoderate storm on a saturated watershed can result insubstantial runoff. Streamflow peaks or volumes of aspecified frequency can be caused by an infinite numberof combinations of storms and watershed states.c. Paragraph 13-4 addresses the development of abalanced hypothetical storm. With such a storm, theaverage depth of rainfall for a duration equal to the timeof concentration for a watershed will have a ’known’exceedance frequency, as will the average depth for anyother duration. However, the triangular temporal distributionof rainfall will generally not be representative ofnatural storms. For watersheds with substantial naturalstorage, the streamflow at the outlet may be relativelyinsensitive to the temporal distribution, whereas for awatershed with a short response time, the resultingstreamflow may be quite sensitive. Methods have beendeveloped (Huff 1967, Pilgrim and Cordery 1975) thatbase the time distribution of a hypothetical storm ondistributions observed in historical storms.d. Associated with application of a hypotheticalstorm is selection of a storm duration. When a balancedhypothetical storm is used, the duration is generallychosen to equal or exceed the time of concentration for awatershed. The infiltration rate that pertains during theperiod of peak storm intensities will depend on how drythe watershed is initially and on how much infiltrationoccurs during the early part of the storm. If a stormduration substantially longer than the time of concentrationis used, the infiltration rate during the period of peakstorm intensities may be unreasonably low because of thelarge volume of infiltration that occurs initially. Sensitivityanalysis can be useful to determine the effects ofstorm duration.e. Another issue is the spatial distribution of hypothetical-stormrainfall. A common assumption is that thedistribution is uniform. Such an assumption is consistentwith use of a nondistributed model for an elemental basin(i.e., one which is not subdivided). However, for a large,subdivided basin, such an assumption may not be reasonable,especially if orographic or other effects tend toresult in substantial deviations from a uniform distribution.Analysis of storm patterns for historical events canprovide insight as to the variability of the spatial distributionand whether or not there is a tendency for relativelygreater concentrations of rainfall in some subbasins andless in others. It may be appropriate to distribute hypothetical-stormrainfall in accordance with a representativepattern based on such analysis.f. Also, with large basins, it may be unreasonable toassume that the temporal distribution of rainfall is thesame for all subbasins. Such an assumption implies thatstorm movement and other phenomena affecting the17-2

EM 1110-2-141731 Aug 94timing of rainfall are not important. Analysis of historicalprecipitation data can provide a basis for evaluatingtemporal characteristics of large storms over the basin.17-4. Transfer of Frequency Information withHypothetical Eventsa. A situation commonly occurs where there areone or more gauged locations with long-term streamflowrecords in the vicinity of ungauged locations for whichdischarge-frequency estimates are required. When this isthe case, it may be possible to develop dischargefrequencyestimates for an ungauged location by transferringfrequency-based information from a gauged locationusing simulation of runoff from hypothetical storms. Aprerequisite for this approach is that storm-occurrencecharacteristics for the gauged and ungauged basin beessentially the same; that is, there should be about equallikelihood that a storm of a given magnitude could occurover either basin. The procedure is as follows:(1) Let Basin A be a gauged basin for which there isa sufficient length of record to enable development of adischarge frequency relation by statistical procedures.Develop and calibrate an event-type precipitation-runoffmodel for Basin A.(2) Let Basin B be the basin for which a dischargefrequency relation is required. There may be no streamflowdata for the basin, or there may be a limited amountof stream-flow data which may be adequate to enablecalibration of a precipitation-runoff model. In any case,develop (and if possible, calibrate) a precipitation-runoffmodel for the basin.(3) Apply a set of hypothetical storms to Basin A.These may correspond to the various recurrence intervalsassociated with NOAA criteria, or they may simply beproportions of a single hypothetical storm. If storms ofspecific recurrence intervals are used, adjust loss rates, ifpossible, so that an x percent-chance storm produces anx percent-chance peak discharge as defined by the statisticallyderived frequency curve. If this is not possible, orif loss rates so determined are not reasonable, use reasonableloss rates and determine the percent-chanceexceedances of the resulting peak discharges for Basin Afrom the statistically derived frequency curve.(4) Apply the storm and antecedent moisture conditioncombinations used for Basin A in the Basin Bmodel. Associate resulting peak discharges for Basin Bwith the exceedance frequencies of the events as establishedfor Basin A.b. An advantage of using storms with definedexceedance frequencies rather than proportioned stormsand adjusting loss rates as required to produce peak dischargesof the same frequencies is that the loss rates soderived can be checked for consistency. Typically, lossrates decrease with decreasing storm frequency. However,it is often not possible to reconcile storms, lossrates, and the ’known’ frequency curve in a reasonablefashion, in which case the storm and antecedent moisturecondition combinations are treated simply as index eventswithout regard to assigning predetermined exceedancefrequencies to them.17-5. Development of Future-ConditionFrequency Estimatesa. The development of frequency estimates forfuture conditions based on estimates for existing conditionscan be accomplished using an approach similar tothat for transferring frequency information from a gaugedto an ungauged location. A procedure is as follows:(1) Develop an existing-condition frequency curveby whatever means is appropriate considering studyrequirements and data availability.(2) Develop and calibrate (if possible) an event-typerainfall-runoff simulation model to represent existingconditions.(3) Apply a set of hypothetical storms with the existing-conditionmodel and associate exceedance frequenciesof the storm and antecedent moisture condition combinationswith the exceedance frequencies of the resultingpeak discharges from the existing-condition frequencycurve.(4) Adjust the existing-condition simulation model torepresent future conditions. This may involve, for example,changes to values for percent imperviousness, unithydrograph or kinematic wave parameters, and routingparameters. Chapter 18 is concerned with techniques formodeling watershed changes.(5) Apply the same storm and antecedent moisturecombinations used for existing conditions to simulatecorresponding future-condition peak discharges. Assignexceedance frequencies determined for the events forexisting conditions to the future-condition peakdischarges.b. If the future condition is to include new storageelements such as detention reservoirs, such elements must17-3

EM 1110-2-141731 Aug 94be added to the future-condition model. The modeling ofstorage elements involves additional considerations, however,because as antecedent storage conditions can bevery significant. A period-of-record analysis may be themost viable approach in this case. Chapter 18 providesfurther discussion on this topic.17-6. Adjustment of Peak Discharges to RepresentStationary Conditionsa. A common problem in statistical analysis ofannual peak discharges is that watershed changes haveoccurred during the period of record so that the annualvalues reflect nonstationary conditions. If the changesare primarily due to the construction of storage reservoirs,it is possible to adjust hydrographs at a downstreamgauge to natural conditions by routing reservoirholdouts (increments of stored water) to the gauge andadding the routed discharge to the observed discharge. Astatistical analysis of the adjusted peaks could then beperformed to produce a natural-condition frequencycurve.b. If watershed changes are due to effects of urbanizationsuch as land use and channel modifications, it isgenerally much more difficult to make adjustments tostationary conditions. An approach for adjusting peakdischarges to existing conditions is as follows:(1) Develop and calibrate a rainfall-runoff model forexisting basin conditions and for conditions at severalother points in time during the period of record. In theexample illustrated in Figure 17-1, rainfall-runoff modelswere developed to represent existing basin conditions (forthe year 1975, in this example), and conditions in theyears 1960, 1950, and 1940.recurrence interval is arbitrary as it is not assumed in thisapproach that runoff frequency is equal to rainfall frequency.The purpose of adopting a specific magnitude isto establish a base storm to which ratios can be appliedfor subsequent steps in the analysis.(3) Apply several ratios to the hypothetical stormdeveloped in step (2) so that the resulting calculated peakdischarges at the gauge cover the range desired for frequencyanalysis. Input the storms to the rainfall-runoffmodels for each of the basin conditions and determinepeak discharges at the gauged location.(4) Plot curves representing peak discharge versusstorm ratio for each basin condition, as illustrated inFigure 17-1.(5) Use the curves developed in the previous step toadjust the observed annual peak discharges. For example,an annual peak discharge for 1963 would be used toenter the family of storm-ratio curves to interpolate astorm ratio consistent with that peak. This storm ratiocan then be used to intersect the base-condition (forexample, existing condition) curve to determine theadjusted peak discharge. The adjustment method isapplied for each of the annual peaks of record.(6) A statistical analysis of the adjusted peak dischargescan then be performed.c. The above approach can also be extended to applyto a future condition. For example, a basin model couldbe developed to represent year 2020 conditions, and acorresponding storm-ratio curve developed. Theobserved annual peak discharges could then be adjustedas in step (5) above to year 2020 conditions.(2) Develop an x percent-chance hypothetical stormfor the basin using generalized rainfall criteria. The17-4

EM 1110-2-141731 Aug 94Figure 17-1. Conversion of nonstationary to stationary peak discharges17-5

EM 1110-2-141731 Aug 94Chapter 18Evaluating Change18-1. Generala. Sources of change and methods of evaluation.(1) Flood-runoff from a catchment may change as aconsequence of human action. Some human actions aretaken with the expressed goal of altering the runoff.Construction of a reservoir in the catchment is an example.Other human actions alter the catchment and conveyancesystem only as a side effect. Nevertheless, theactions alter the runoff. An example of this is conversionof an agricultural field to a residential neighborhood.(2) Flood-runoff from a catchment may change alsoas a consequence of natural phenomena, if the phenomenachange the catchment or conveyance system. Forexample, a lightning-caused range fire may alter thevegetative cover, and consequently, the rate of runofffrom a catchment.b. Illustration. This chapter illustrates the use of theinfiltration, runoff, routing, and statistical modelsdescribed in previous chapters of this document to evaluatethe impacts of human action and natural phenomena.Here, the evaluation is limited to analysis of changes torunoff hydrographs, discharge-frequency curves, andrating curves.18-2. Evaluating Catchment and Conveyance-System Changea. Effects of change on floods. Catchments andconveyance systems may be modified by human action,such as urbanization, or by natural phenomena, such aslightning-caused range fire. These changes alter runoffhydrographs from single events. Consequently, thesechanges also alter the discharge-frequency relationship.According to Leopold (1968),... the two principal factors governing flow regimenare the percentage of (catchment) area madeimpervious and the rate at which water is transmittedacross the land to stream channels. Theformer is governed by the type of land use; thelatter is governed by the density, size, and characteristicsof tributary channels...Development or urbanization in a catchment typically isaccompanied by an increase in impervious area. As theimpervious area increases, the infiltration decreases. Asinfiltration decreases, the volume of runoff from a stormincreases. As the volume increases, the magnitude of theflood peak increases. An increase in impervious areaalso speeds the flow of water across the land, and thisincreases the flood peak. Likewise, improvements to orexpansion of the catchment conveyance system speedsthe flow and increases the peak.b. Evaluation with a rainfall-runoff model. Theimpact of watershed changes can be estimated convenientlywith a rainfall-runoff model that includes onlyparameters that are measurable or parameters that aredirectly related to catchment characteristics. Given adescription of the proposed changes to the catchment orthe conveyance system, these parameters can beestimated. An example of a (pseudo) physically basedrainfall-runoff model is the kinematic-wave model.Application of this model requires identification of catchmentarea, flow length, slope, and overland-flow roughnessfactor. To evaluate the impact of catchment orconveyance-system changes with this model, theseparameters are estimated from maps, photographs,inspection, or, in the case of future conditions, fromdevelopment plans. With the modified parameters, runoffcan be estimated for any storm.(1) The impact on the discharge-frequency curve canbe evaluated with a rainfall-runoff model via period-ofrecordanalysis. The period-of-record analysis computesrunoff from the entire time series of historical rainfall orfrom a lengthy series of equally likely rainfall (Chapter12 of EM 1110-2-1415). The resulting series ofrunoff is analyzed with the statistical-analysis proceduresdescribed in Chapter 12 to define the modified-conditiondischarge-frequency curve. This analysis is straightforwardbut data-intensive and time-consuming.(2) Simulation of selected historical events is analternative to a complete period-of-record analysis. Thisprocedure uses historical rainfall and runoff data. Theexisting, present-condition discharge-frequency curve isdetermined by statistical analysis of the discharge timeseries. To estimate the modified discharge-frequencycurve, a rainfall event is selected from the historicalrecord. The probability of the historical runoff peakcorresponding to the event is determined from the existingconditions discharge-frequency curve. Runoff due tothe rainfall after catchment and conveyance-system18-1

EM 1110-2-141731 Aug 94changes is estimated by simulation, with model parametersselected to represent the modified condition. Thispeak discharge is assigned the same probability as theexisting-condition peak. This is repeated for a range ofrainfall events to adequately define the modified discharge-frequencycurve.(3) If historical rainfall and runoff data are not available,the modified-condition discharge-frequency curvecan be estimated with hypothetical rainfall. To estimatethe discharge-frequency curve, a design storm of specifiedprobability is developed. Runoff due to the rainfallevent after catchment and conveyance-system changes isestimated by simulation, with model parameters selectedto represent the modified condition. The computedmodified-condition peak is assigned the same probabilityas the design storm. This is repeated for a range ofhypothetical rainfall events to adequately define themodified discharge-frequency curve. This procedure isdescribed in Chapter 17.c. Evaluation with regional rainfall-runoff modelparameters. The impact of watershed changes can beestimated with an rainfall-runoff model with calibrationparameters, using parameter-predictive equations. Withgauged data, these parameters are determined by trial anderror, comparing computed hydrographs with observedhydrographs. As described in Chapter 16, predictiveequations may be developed to permit estimation of theparameters for ungauged catchments. These predictiveequations relate the calibration parameters to catchmentcharacteristics. A simple example is the following equation,proposed by Wright-McLaughlin Engineers (1969)to predict a parameter for Snyder’s synthetic unit hydrograph,in the Denver metropolitan area:whereC t7.81I 0.76(18-1)C t = Snyder’s unit hydrograph parameter(paragraph 7-3c)I= catchment impervious area, in percentage.As a natural catchment is developed, the impervious areatypically increases. With (Equation 18-1) and Snyder’smodel, the resulting change in the unit hydrograph can bepredicted. Application of the unit hydrograph permitsestimation of the runoff from any storm. Similarequations can be developed and applied to estimateparameters for other rainfall-runoff models.(1) The SCS loss and unit hydrograph models areespecially convenient empirical models for estimatingmodifications to runoff due to catchment and conveyance-systemchanges (USDA 1986). The SCS lossmodel parameter is predicted as a function of land use,soil type, and antecedent-moisture condition. The unithydrograph model parameter may be predicted as a functionof land use, soil type, antecedent-moisture condition,slope, and flow length. For existing, current conditions,these can be observed or measured. For modified conditions,these can be forecasted.(2) A GIS is helpful for developing the physicalfeaturedata base required for evaluation of changes. AGIS is a computerized data base management systemwith spatial references for all data. The simplest GIS isa rectangular grid superimposed on a map of the catchment.Pertinent characteristics are determined and storedin a data base for each cell of the grid. For example, forthe SCS models, land-use type, soil type, moisture condition,slope, and length can be stored. Once stored, thecharacteristics can be retrieved and mapped. They alsocan be manipulated for use with parameter predictiveequations, such as those that predict loss rate parametersfor the SCS model. A GIS is convenient for evaluatingrunoff changes due to future catchment or conveyancesystems (DeBarry and Carrington 1990). With proposedland-use types stored in the GIS, the modified-conditionmodel parameters can be determined easily, and therunoff can be computed. Of course, the reliability is afunction of the quality of the data stored and the reliabilityof the parameter-predictive equations.(3) Given rainfall-runoff model parameters determinedwith predictive equations, the impact of watershedand conveyance-system changes on the dischargefrequencycurve can be evaluated using the same proceduresdescribed for the model with physically basedparameters. A period-of-record analysis can be performedto develop a modified condition time series.Alternatively, selected historical or hypothetical eventscan be simulated.d. Evaluation with regional frequency-model parameters.The lumped impact of watershed and conveyancesystemchanges on the discharge-frequency curve can beevaluated with frequency-based model parameter predictiveequations. Paragraph 16-6 of this document18-2

EM 1110-2-141731 Aug 94describes how frequency-based model parameters ordischarge-frequency relationships may be related tocatchment characteristics. If these characteristics canreflect catchment and conveyance-system changes, theequations can directly predict the modified-conditiondischarge-frequency curve.(1) Quantiles for the modified discharge-frequencycurve can be estimated with a predictive equation. Forexample, Sauer et al. (1983) propose the following equationto estimate the 0.01-probability peak discharge for adeveloped urban catchment:parameters. With these parameters and the distributionequation, the discharge-frequency relationship is defined.18-3. Procedure for Evaluating Damage-Reduction Plansa. Damage-reduction measures. Flood damage canbe reduced by decreasing flow rate, decreasing the depthof water, and decreasing directly the damage caused byflooding. Table 18-1 lists measures that reduce flooddamage, classifying each by impact. A mitigation plancomprises one or more of these measures.UQ100 2.50 A 0.29 SL 0.15 (RI2 3) 1.76(ST 8) 0.52 (13 BDF)SIP 0.28IA 0.06RQ100 0.63(18-2)b. Plan evaluation criterion. The effectiveness ofany plan is quantified in terms of inundation-damagereduction benefit. Guidelines for Federal water-resourcesplanning define this as:whereUQ100 = discharge, in cubic feet per secondwhereE B IRE D existE D plan(18-3)A = catchment contributing area, in square milesB IR= inundation-reduction benefitSL = channel slope, in feet per mileRI2 = basin rainfall, in inchesST = basin storage, in percentageBDF = basin development factor (0 to 12)IA = impervious area, in percentageRQ100 = equivalent rural peak discharge, in cubicfeet per secondRQ100 is estimated independently with statistical analysisof the historical time series. For forecasted or proposedchanges, the slope, storage, development factor, andimpervious area can be estimated. With (Equation 18-2),the modified 0.01-probability discharge is estimated.Similar equations can be developed for other quantiles orwith other catchment characteristics.(2) Equations can also be developed to predict thestatistical model parameters as a function of catchmentcharacteristics. For example, the standard deviation in(Equation 12-9) can be correlated with catchment characteristics.The resulting equation could permit estimationof current, future, existing, or proposed conditionD exist = existing-condition flood-damage cost(without a plan)D plan= flood-damage cost with the plan in placeE = the expected value (USWRC 1983).Chapter 7 of EM 1110-2-1415 describes alternativeapproaches to computing the expected value. The mostwidely used approach in USACE is the frequency technique.To compute expected damage with the frequencytechnique, the damage-frequency curve is derived bytransforming the annual-maximum discharge-frequencycurve with the elevation-discharge (rating) function andthe elevation-damage function. This is illustrated byFigure 18-1. The expected damage is the area beneath(the integral of) this damage-frequency relationship. TheHydrologic Engineering Center’s Expected Annual FloodDamage (EAD) computer program derives the damagefrequencycurve following this procedure and integratesthe result numerically (USACE 1984a).(1) For computation of expected damage, the hydrologicengineer must define the discharge-frequency curveand rating functions for existing and proposed conditions,accounting for current and future catchment and conveyance-systemconditions. Table 18-2 shows how the18-3

EM 1110-2-141731 Aug 94Table 18-1Damage-Reduction Measures, Classified by ImpactDecreaseDecreaseDepth ofDamageDecrease Flow Rate Flooding DirectlyReservoir Channel FloodplainalterationmanagementDiversionWatershedmanagementLevee/floodwallFloodproofingFlood warningand preparednessplanningfunctions are modified by each of the damage-reductionmeasures listed in Table 18-1.(2) Mathematical tools described in Part II of thisdocument and in EM 1110-2-1416, EM 1110-2-1413, andEM 1110-2-1415 are used for the analysis.c. Summary of evaluation procedures. The economicimpact of catchment and conveyance system changesand of flood-damage mitigation measures isdetermined via solution of Equation 18-3. This may beaccomplished as follows:(1) Define the existing-condition discharge-frequencycurve, rating, and elevation-damage functions. To definethe discharge-frequency curve, rainfall-runoff and routingmodels or statistical models are used. To define the ratingfunction, routing models or the hydraulics modelsdescribed in EM 1110-2-1416 may be employed.(2) Derive the damage-frequency curve using theprocedure illustrated by Figure 18-1. Integrate to computeexpected inundation damage for the existingcondition.(3) Identify the plan to be evaluated. Perform theanalyses necessary to define modifications to the discharge-frequencycurve, rating, and elevation-damagefunctions due to the plan. These analyses may requirerainfall-runoff and routing models, statistical models, orhydraulics models.(4) Derive the modified-condition damage-frequencycurve, using the modified functions. Integrate the damage-frequencycurve to compute expected damage withthe changes.(5) Solve (Equation 18-3) to compute inundationreductionbenefit.(6) If catchment, channel, and economic conditionsare dynamic, repeat steps 1-5 for each year of analysis.d. The remainder of this chapter describes technicalprocedures for evaluating changes to the discharge-frequencycurve and rating function as a consequence offlood-damage reduction plans.18-4. Evaluating Reservoir and Detention Basinsa. Reservoir performance. A reservoir stores floodrunoff and then releases it downstream to the channelover a longer period of time. This operation reduces thepeak flow rate, resulting in lower water-surface elevationand less damage. The primary impact of the reservoir ismodification of the discharge-frequency curve, as illustratedby Figure 18-2.(1) The effectiveness of the reservoir depends on itscapacity, location, and operation rules.(2) The capacity limits the amount of runoff that canbe collected and held for release at a nondamaging rate.(3) The location governs the amount of runoff thatthe reservoir can control, since a reservoir will store onlyinflow from the area upstream. The reservoir operationrules determine the manner of release.b. Reservoir modeling fundamentals. The performanceof a reservoir or detention basin is evaluated withthe routing procedures described in Chapter 9. The fundamentalrelationship used is the continuity relationship:18-4

EM 1110-2-141731 Aug 94Figure 18-1. Derivation of damage-frequency curve from discharge-frequency curve, rating function, and elevationdamagefunction18-5

EM 1110-2-141731 Aug 94Table 18-2Evaluation Requirements of Damage Mitigation MeasuresFunction(s) modified by measures in categoryCategory of Discharge- Elevation- Elevation-Measure probability discharge damageReservoir X - -Diversion X - -Watershedmanagement X - -Channel X 1 X -alterationLevee/floodwall X 1 X XFloodplain - - XmanagementFloodproofing - - XFlood warningand preparednessplanning - - X 21 If floodplain storage altered significantly.2 Evaluation requires subjective analysis.whereS t 1I tdt O tdt S tS t-1 = storage at the end of time interval t -1(18-4)I t = average reservoir inflow rate during interval tdt = length of time intervalO t = average reservoir outflow rate during interval tS t = reservoir storage at the end of interval tThis equation is solved recursively to determine thereservoir storage and release hydrographs. Solutionrequires specification of the initial volume in storage inthe reservoir (S t for t = 0), specification of the reservoiroperation rules, and specification of the reservoir inflowhydrograph (I t for all t).(1) The initial storage selected for solution of Equation18-4 depends on the reservoir condition to be evaluated.If the proposed reservoir has no permanent pool,the initial storage is zero. If the impact of successivestorms is of interest, the initial storage for each event,after the first, is the final storage of the preceding storm.If the reservoir is a multiple-purpose reservoir, a portionof the reservoir is allocated to flood control, and a portionis allocated to conservation. The reservoir operatorstrives to keep the conservation pool full, as releases orwithdrawals from this pool satisfy water supply andenergy demands. The operator tries to keep the floodcontrolpool empty. For analysis of reservoir operationduring a flood, the initial storage depends on the successor likely success in meeting the goal. If the flood-controlpool is empty, the total flood-control volume is available.Most reservoir flood-control operation studies assumethis to be the case.(2) The reservoir operation rules relate inflow, storage,and outflow. For a simple detention pond, the rulesare fixed by the hydraulic characteristics of the structure.For example, for a simple detention pond with an uncontrolledconduit outlet and an ungated spillway, the operationrules can be determined via the orifice and weirequations. These equations will define the outflow as afunction of reservoir water-surface elevation. With a site18-6

EM 1110-2-141731 Aug 94Figure 18-2. Discharge-frequency curve modification due to reservoirelevation-area description, the elevation can be relatedtostorage. This will permit solution of (Equation 18-4)and simulation of reservoir performance. For a gatedflood-control reservoir, the rules are constrained byhydraulics and defined by economic, environmental,social, and political criteria.(3) The reservoir inflow hydrograph depends on thestudy objective. If the goal is to define the modifieddischarge-frequency curve, one option is to evaluatereservoir performance with a long series of historical orsynthetic inflows. The operation is simulated with theseries to define the reservoir outflow. Statistical analysisprocedures described in Chapter 12 are applied to theoutflow series to estimate the modified dischargefrequencycurve.(4) Alternatively, the discharge-frequency curve canbe estimated by evaluating performance for a limitednumber of historical events. The current, withoutreservoircondition discharge-frequency curve is foundwith methods of Chapter 12. To estimate the modifieddischarge-frequency curve, a runoff event is selectedfrom the historical inflow record. The probability of thehistorical runoff peak corresponding to the event is determinedfrom the discharge-frequency curve. The peakwith the reservoir is estimated by simulation. This controlledpeak discharge is assigned the same probability asthe existing-condition peak. This is repeated for a rangeof runoff events to adequately define the modifieddischarge-frequency curve.(5) The modified-condition discharge-frequencycurve can be estimated also with hypothetical runoffevents. Such a runoff event is developed from rainfallrunoffanalysis with rain depths of known probability orfrom discharge duration-frequency analysis. In the firstcase, a design storm of specified probability is developedwith procedures described in Chapter 13. The correspondingrunoff hydrograph is computed with a rainfallrunoffmodel. This runoff hydrograph is inflow to thereservoir. In the second case, a balanced inflow hydrographis developed. This balanced hydrograph has volumesfor specified durations consistent with established18-7

EM 1110-2-141731 Aug 94volume-duration-frequency relations. For example, a0.01-probability balanced hydrograph is developed so thepeak 1-hr volume equals the volume with probability0.01 found through statistical analysis of runoff volumes.Likewise, the hydrograph’s 24-hr volume equals thevolume with probability 0.01. With either of the hypotheticalinflow events, reservoir operation is simulatedand the outflow peak is assigned the same probability asthe inflow hydrograph. This procedure is repeated for arange of hypothetical rainfall events to adequately definethe modified discharge-frequency curve. Strictly speaking,this is appropriate only if the reservoir has no permanentpool. Otherwise, the outflow depends on theinflow and the initial storage.c. Dam-safety studies. The discharge-reductionbenefit of a reservoir is accompanied by the hazard ofdam failure. The impact of this failure can be estimatedwith hydraulics models described in EM 1110-2-1416 orwith the routing models of Chapter 9 of this document.Three aspects of dam failure must be considered:formation of a breach, an opening in the dam as it fails;flow of water through this breach; and flow in the downstreamchannel. For analysis, the reservoir outflowhydrograph is computed with Equation 18-4 as before.However, the operating rules change with time as thebreach grows. For convenience in analysis, a breach isassumed to be triangular, rectangular, or trapezoidal andto enlarge at a linear rate. At each instant that the breachis known, the flow through the breach can be determinedwith principles of hydraulics. Flow through the downstreamchannel is modeled with one of the routingmodels.18-5. Evaluating Channel Alterations and Leveesa. Channel-alteration performance. Channel alterationsinclude enlarging the channel, smoothing the channel,straightening the channel, and removing or minimizingobstructions in the channel. Enlarging the channelincreases its flow-carrying capacity. The other alterationslessen the energy loss, thus permitting a given dischargeto flow at a lesser depth. The primary impact of increasingthe flow-carrying capacity or lessening the energyloss is modification of the rating function, as illustratedby Figure 18-3.b. Channel-alteration modeling. The performance ofa channel alteration is evaluated with river hydraulicsmodels described in EM 1110-2-1413. These physicallybased models have physically based parameters that aremodified to reflect changes to channel characteristics.(1) The HEC-2 computer program (USACE 1982) isa well-known tool for evaluating channel alterations.This program implements a model of gradually variedsteady flow in a rigid-boundary channel. That modeluses the physical dimensions of the channel and indicesof channel roughness directly in estimating flow depth.To evaluate the impact of proposed channel enlargement,the channel dimensions are modified in the programinput to reflect the changes. Repeated solution of thegradually varied steady-flow equations with HEC-2 yieldsthe rating function for a specified channel configuration.(2) For modeling the impacts of changes in an alluvialchannel, a movable-bed model should be used.Program HEC-6 (USACE 1990c) implements such amodel.c. Levee performance. A levee or floodwall reducesdamage by reducing floodplain flooding depth. It doesso by blocking overflow from the channel onto the floodplainwhen the capacity of the channel is exceeded. Therating function, as modified by a levee, is shown inFigure 18-4. A levee may also modify the dischargefrequencycurve. The levee restricts flow onto the floodplain,eliminating the natural storage provided by thefloodplain. This restriction may increase the dischargedownstream of the levee for a specified probability.Further, as the natural channel is narrowed by the levee,the velocity may increase. This too may increase thedischarge for a given probability.d. Levee modeling.(1) Introduction of a levee alters the effective channelcross section. The impact of this change can bedetermined with the physically based river hydraulicsmodels. As with channel alteration, the impact of alevee can be determined by modifying the parameterswhich describe the channel dimensions. Repeated applicationof the model with various discharge magnitudesyields the rating function for a specified leveeconfiguration.(2) Modifications to the discharge-frequency curvedue to a levee are identified with the river hydraulicsmodels or with routing models described in Chapter 9.Either models the impact of storage on the dischargehydrograph and will reflect the loss of this storage. Forexample, the modified puls routing model determines thechannel outflow hydrograph with a relationship of channeldischarge to channel storage. A levee will reduce thechannel storage for discharge magnitudes that exceed the18-8

EM 1110-2-141731 Aug 94Figure 18-3. Rating function modification due to channel alterationFigure 18-4. Rating function modification due to levee18-9

EM 1110-2-141731 Aug 94channel capacity. Historical or hypothetical runoffhydrographs can be routed with the selected model todetermine discharge peaks with the proposed levee.e. Interior drainage. A levee or floodwall blocks thenatural drainage of local runoff into the channel. Thislocal runoff may cause flooding and must be consideredin levee planning. Rainfall-runoff and routing modelsdescribed in this document can be used to estimate thevolume and time distribution of local runoff. Facilitiesfor managing the water are described in EM 1110-2-1413. Often, a detention pond is used to store the interiordrainage. The water is pumped from the pond intothe channel. The performance of the pond can be simulatedwith routing models similar to those used for analysisof a reservoir or detention pond. Analysis proceduresare described in detail in EM 1110-2-1416.18-6. Evaluating Other Alternativesa. Diversion. A diversion reduces the peak flowdownstream of its location by reducing the volume ofwater flowing in a channel reach. This discharge reductioncauses the discharge-frequency curve to be modifiedas illustrated by Figure 18-5. Figure 18-6 is a plan viewof a diversion. This diversion includes a bypass channeland a control structure. The control structure could be asimple overflow weir, a pipe through an embankment, ora gated, operator-controlled weir. When the flow rate inthe main channel reaches a threshold, the control structurediverts a portion of the flow into the bypass channel.The volume and flow rate in the main channel isreduced, thus eliminating or reducing damage to thedownstream property. Downstream, the bypass and themain channel join. There, the diverted water flows intothe main channel.(1) The performance of a diversion is evaluated withrouting models described in Chapter 9 of this document.At the control structure, a hydraulics model estimates thedistribution of flow into the bypass and flow in the mainchannel. This model may be as complex as the2-D models described in EM 1110-2-1416 or a simple asa rating curve, based on 1-D steady-flow analysis, whichdefines diversion-channel flow as a function of mainchannelflow. Passage of flow in the diversion channeland in the main channel is modeled with a routingmodel, such as the puls model.Figure 18-5. Discharge-frequency curve modified due to diversion18-10

EM 1110-2-141731 Aug 94Figure 18-6.Plan view of diversion(2) The impact of a diversion on the dischargefrequencycurve can be evaluated via period-of-recordanalysis or simulation of selected events. With the period-of-recordanalysis, the historical discharge time seriesis analyzed to estimate channel flow when the proposeddiversion operates. The resulting modified main-channeldischarge time series is analyzed with statistical proceduresto define the discharge-frequency curve. Otherwise,operation of the diversion with selected historicalor hypothetical runoff hydrographs can be simulated. Aswith a reservoir, the resulting peaks are assigned probabilitiesequal the probabilities of the peaks without thediversion. For small events, the diversion has little or noimpact on the discharge-frequency curve, since little orno water is diverted from the main channel. As the dischargemagnitude increases, the diversion functions anddiverts water up to its capacity. For larger events, thedischarge reduction possible is constrained by thecapacity of the diversion.b. Watershed management. Watershed managementincludes vegetation and crop management, terracing andcontour plowing, and drainage control. Whereas urbanizationin a catchment increases the volume and speedsrunoff, these measures decrease the volume and/or slowthe runoff.(1) Vegetation and crop management ensure thatland is covered with vegetation during the rainy season.This increases infiltration by impeding flow and makingthe soil more permeable.(2) Terracing and contour plowing alter the shape ofcatchment surfaces, increasing storage, slowing flow, andincreasing infiltration.(3) Storm drainage control intercepts runoff anddiverts or detains it, much like a reservoir or detentionbasin does. This reduces the runoff peak by spreadingthe runoff volume over a longer time period.(4) The impact of watershed management measuresis evaluated with the same procedures used to evaluatecatchment and conveyance-system changes. A statisticalmodel may be used with predictive equations for themodel parameters. These predictive equations mustinclude terms descriptive of watershed managementmodifications. Otherwise, the impacts of watershed18-11

EM 1110-2-141731 Aug 94management may be predicted with a rainfall-runoffmodel. As described in paragraph 18-2, such a modelpermits evaluation of changes to runoff hydrographs.Through period-of-record analysis or by simulatingselected historical or hypothetical events, the modifiedconditiondischarge frequency curve can be estimated.c. Floodplain management. Floodplain managementdecreases future damage by reducing vulnerability offuture development. This may be accomplished withland-use ordinances, subdivision regulations, zoning laws,building codes, or real estate statutes.(1) A floodplain land-use ordinance could restrictland uses that are dangerous due to water or erosionhazards. This will change the future elevation-damagefunction.(2) Floodplain management may also modify thefuture discharge-frequency curves and future rating functions.For example, if future development in the floodplainis restricted, the impervious area may increase asold structures are razed and land is returned to a naturalstate. The impact of such modification can be evaluatedusing procedures described in paragraph 18-2.18-12

EM 1110-2-141731 Aug 94Appendix AReferencesA-1. Required PublicationsER 1105-2-100Guidance for Conducting Civil Works Planning StudiesER 1110-2-1150Engineering and Design for Civil Works ProjectsEM 1110-2-1406Runoff from SnowmeltEM 1110-2-1411Standard Project Flood DeterminationsEM 1110-2-1413Hydrologic Analysis of Interior AreasEM 1110-2-1415Hydrologic Frequency AnalysisEM 1110-2-1416River HydraulicsAbbot et al. 1986Abbot, M. B., Bathurst, J. C., Cunge, J. A., O’Connell,P. E., and Rasmussen, J. 1986. “Introduction to theEuropean Hydrological System - Systeme HydrologiqueEuropeen, ’SHE’: history and philosophy of a physicallybaseddistributed modelling system,” Journal of Hydrology,Vol 87, pp 45-49.Abramowitz and Stegun 1965Abramowitz, M., and Stegun, I. A. 1965. Handbook ofMathematical Functions, U.S. Government PrintingOffice, Washington, DC.Ahuja et al. 1988Ahuja, L. R., Bruce, R. R., Cassel, D. K., and Rawls,W. J. 1988. “Simpler Field Measurement and Estimationof Soil Hydraulic Properties and Their Spatial Variabilityfor Modeling.” ASAE Proceedings, Modeling AgriculturalForest and Rangeland Hydrology, Pub 07-88, St. Joseph,MO.Model,” National Oceanic and AtmosphericAdministration Tech. Memo, NWS HYDRO-17, NationalWeather Service, Silver Springs, MD.Aron, Miller, and Lakatos 1977Aron, G., Miller, A. C., and Lakatos, D. F. 1977. “InfiltrationFormula Based on SCS Curve Number,” Journal ofthe Irrigation and Drainage Division, American Society ofCivil Engineers, New YorkBear 1979Bear, J. B. 1979. Hydraulics of Groundwater, McGraw-Hill International Book Company, New York.Benjamin and Cornell 1970Benjamin, J. R., and Cornell, C. A. 1970. Probability,Statistics, and Decision for Civil Engineers. McGraw-HillBook Co., New York.Brakensiek 1977Brakensiek, D. L. 1977. Estimating the Effective CapillaryPressure in the Green and Ampt Infiltration Equation,Water Resource Research, Vol 13, No. 3, pp 680-682.Brakensiek and Onstad 1988Brakensiek, D. L., and Onstad, C. A. 1988. “ParameterEstimation of the Green and Ampt Infiltration Equation,”Water Resource. Research., Vol 13, No. 6, pp 1009-1012.Brooks and Corey 1964Brooks, R. H., and Corey, A. T. 1964. “Hydraulic Propertiesof Porous Media,” Colorado State UniversityHydrology Paper No. 3, Fort Collins, CO.Brunner 1989Brunner, G. W. 1989. “Muskingum-Cunge ChannelRouting,” Lecture Notes, Hydrologic Engineering Center,U.S. Army Corps of Engineers, Davis, CA.Chow 1951Chow V. T. 1951. “A General Formula For HydrologicFrequency Analysis,” Transactions, American GeophysicalUnion, Vol 32, No. 2, pp 231-237.Chow, Maidment, and Mays 1988Chow, V. T., Maidment, D. R., and Mays, L. W. 1988.Applied Hydrology. McGraw-Hill, New York.Anderson 1975Anderson, E. A. 1975. “National Weather Service RiverForecast System - Snow Accumulation and AblationA-1

EM 1110-2-141731 Aug 94Clark 1945Clark, C. O. 1945. “Storage and the Unit Hydrograph.”Transactions of the American Society of Civil Engineers,Vol 110, pp 1419-1446.Crawford and Linsley 1966Crawford, Norman H., and Linsley, Ray K. 1966. “DigitalSimulation in Hydrology: Stanford Watershed ModelIV,” Civil Engineering Technical Report No. 39, StanfordUniversity, Palo Alto, CA.Crippen 1982Crippen, J. R. 1982. “Envelope Curves for ExtremeFlood Events,” Hydraulic Journal, American Society ofCivil Engineers, Vol 108, No. HY10, pp 1208-1212.Crippen and Bue 1977Crippen, J. R., and Bue, Conrad D. 1977. “MaximumFloodflows in the Conterminous United States,” WaterSupply Paper 1887, U.S. Geological Survey, Washington,DC.Cunge 1969Cunge, J. A. 1969. “On the Subject of a Flood PropagationComputation Method (Muskingum Method),”Journal of Hydraulic Research, Vol 7, No. 2 pp 205-230.Dawdy, Lichty, and Bergman 1972Dawdy, D. R., Lichty, R. W., and Bergman, J. M. 1972.“A rainfall-runoff Simulation Model for Estimation ofFlood Peaks for Small Drainage Basins,” GeologicalSurvey Professional Paper, 506-B, U.S. GovernmentPrinting office, Washington, DC.DeBarry and Carrington 1990DeBarry, P. A., and Carrington, J. T. 1990. “ComputerWatersheds.” Civil Engineering, Vol 6, No. 7, pp 68-70.Eagleson 1970Eagleson, P. S. 1970. Dynamic Hydrology. McGraw-Hill Book Company, New York.Engman 1986Engman, E. T. 1986. “Roughness Coefficients for RoutingSurface Runoff,” Journal of Irrigation and Drainage,American Society of Civil Engineers, Vol 112, No. 4,pp 39-53.Federer and Lash 1978Federer, Anthony C., and Lash, Douglas. 1978. “Brook:A Hydrologic Simulation Model for Eastern Forests,Durham, New Hampshire,” University of New Hampshire,Water Resource Research Center, Research ReportNo. 19.Ford, Morris, and Feldman 1980Ford, D. T., Morris, E. C., and Feldman, A. D. 1980.“Corps of Engineers Experience With AutomaticCalibration of a Precipitation-Runoff Model,” Water andRelated Land Resource Systems, Y. Haimes and J.Kindler, ed., Pergamon Press, New York.Fread 1982Fread, D. L. 1982. “Flood Routing: A Synopsis of Past,Present, and Future Capabilities,” Rainfall-Runoff Relationships,Proceedings of the International Symposium onRainfall-Runoff Modeling. Mississippi State University,Mississippi State, MS.Gray and Male 1981Gray, D. M., and Male, D. H. 1981. Handbook of Snow,Pergaman Press, New York.Haan 1977Haan, C. T. 1977. Statistical Methods in Hydrology,Iowa State University Press, Ames, IA.Hamon 1961Hamon, W. R. 1961. “Estimating Potential Evapotranspiration:Proceedings of American Society of Civil Engineers,”Journal of the Hydraulic Division, Vol 87, HY3,pp 107-120.Hathaway 1945Hathaway, G. A. 1945. “Symposium on Military Airfields- Design of Drainage Facilities,” Transactions,American Society of Civil Engineers, Vol 110,pp 697-733.Henderson 1966Henderson, F. M., 1966. Open Channel Flow,MacMillan Publishing Co. Inc., New York, New York.Hershfield 1961Hershfield, D. M. 1961. “Rainfall frequency atlas of theUnited States for durations from 30 minutes to 24 hoursand return periods from 1 to 100 years,” Technical PaperNo. 40, superseded by Hydrometeorological ReportNo. 35, U.S. Department of Commerce, National WeatherService, Silver Springs, MD.Hjelmfelt 1980Hjelmfelt, A. T. 1980. “Curve Number Procedure as anA-2

EM 1110-2-141731 Aug 94Infiltration Method,” Journal of the Hydrology Division,American Society Civil Engineers, HY6, pp 1107-1111.Holtan et al. 1975Holtan, H. N., Stitner, G. J., Henson, W. H., and Lopez,N. C. 1975. “USDAHL-74 revised model of watershedhydrology,” Technical Bulletin No. 1518, AgriculturalResearch Service, U.S. Department of Agriculture,Washington, DC.Horton 1919Horton, R. E. 1919. “Rainfall Interception,” MonthlyWeather Review, Vol 47, pp 603-623.Horton 1935Horton, R. E. 1935. “Surface Runoff Phenomena; Part I,Analyses of the Hydrograph,” Hydrology Laboratory,Pub 101, Edwards Feathers, Inc. Ann Arbor, MI.Horton 1945Horton, R. E. 1945. “Erosional Development of Streamsand Their Drainage Basins: Hydrophysical Approach toQuantitative Morphology.” Bull. Geol. Soc. of America,Vol 56, pp 275-370.Huff 1967Huff, F. A. 1967. “Time Distribution of Rainfall inHeavy Storms,” Water Resources Research, Vol 3,No. 4, pp 1007-1019.Jensen and Haise 1963Jensen, M. E., and Haise, H. R. 1963. “EstimatingEvapotranspiration Form Solar Radiation: Proc. of theAmerican Society of Civil Engineers, Journal of Irrigationand Drainage, Vol 89, No. IR4, pp 15-41.Kite 1977Kite, G. W. 1977. “Frequency and Risk Analysis inHydrology,” Water Resources Publications, Ft. Collins,CO.Leopold 1968Leopold, Luna. 1968. “Hydrology for urban land planning- a guidebook on the hydrologic effects of urbanland use,” Geological Survey Circular 554. U.S. GeologicalSurvey, Washington, DC.Lettenmaier 1984Lettenmaier, D. P. 1984. “Limitations on seasonal forecastaccuracy. J. Water Res. Plng. and Mgmt. Div.,American Society Civil Engineers, Vol 110, pp 255-269.Li, Stevens, and Simons 1976Li, R., Stevens, M. A., and Simons, D. B. 1976. “Solutionsto Green-Ampt Infiltration Equation,” Journal ofIrrigation and Drainage Division, American Society CivilEngineers, IR2, pp 239-248.Linden 1979Linden, R. L. 1979. A model to predict soil waterstorage as affected by tillage practices. Thesis presentedto the University of Minnesota, at Minneapolis, MN.Linsley 1986Linsley, R. K. 1986. “Flood estimates: how good arethey?” Water Resources Research, Vol 22, No. 9,pp 159S-164S.Linsley, Kohler, and Paulhus 1975Linsley, R. K., Kohler, M. A., and Paulhus, J. L. H.1975. Hydrology for Engineers, McGraw-Hill, NewYork.Loague and Freeze 1985Loague, K. M., and Freeze, R. A. 1985. “A Comparisonof Rainfall-runoff Modeling Techniques on Small UplandCatchments,” Water Resources Research, AGU, Vol 21,No. 2, pp 229-248.Loucks, Stedinger, and Haith 1981Loucks, D. P., Stedinger, G., and Haith, D. 1981. WaterResources Systems Planning and Analysis, Prentice-Hall,Inglewood Cliffs, NJ.Matthai 1969Matthai, H. F. 1969. “Flood of June 1965 in SouthPlatte River Basin, Colorado,” Water Supply Paper1850-B, U.S. Geological Survey, Washington, DC.McCuen 1989McCuen, R. H. 1989. Hydrologic Analysis and Design,Prentice Hall, Englewood Cliffs, NJ.McCuen and Snyder 1986McCuen, R. H., and Snyder, W. M. 1986. Hydrologicmodeling: Statistical Methods and Applications,Prentice-Hall, Englewood Cliffs, NJ.McCuen, Rawls, and Brakensiek 1981McCuen, R. H., Rawls, W. J., and Brakensiek, D. L.1981. “Statistical Analysis of the Brooks-Corely and theGreen-Ampt Parameters Across Soil Textures.” WaterResources Research, Vol 17, No. 4, pp 1005-1013.A-3

EM 1110-2-141731 Aug 94Mein and Larson 1973Mein, R. G., and Larson, C. L. 1973. “Modeling InfiltrationDuring Steady Rain,” Water Resources Research,Vol 2, pp 384-394.Miller 1963Miller, J. F. 1963. “Probable Maximum Precipitationand Rainfall-Frequency Data for Alaska for Areas to400 Square Miles,” Technical Paper No. 47, U.S. Departmentof Commerce, Weather Bureau, Washington, DC.Miller and Cunge 1975Miller, W. A., and Cunge, J. A. 1975. “Simplified Equationsof Unsteady Flow,” Unsteady Flow in Open Channels,Vol I, K. Mahmood and V. Yevjevich, ed.,pp 183-249.Miller, Frederick, and Tracey 1973Miller, J. F., Frederick, R. H., and Tracey, R. J. 1973.“Precipitation-Frequency Atlas of the Western UnitedStates, Volume I, Montana; Volume II, Wyoming; VolumeIII, Colorado; Volume IV, New Mexico; Volume V,Idaho; Volume VI, Utah; Volume VII, Nevada; VolumeVIII, Arizona; Volume IX, Washington; Volume X,Oregon; Volume XI, California,” U.S. Department ofCommerce, National Weather Service, NOAA Atlas 2,Silver Springs, MD.Mockus 1957Mockus, Victor. 1957. “Use of Storm and WatershedCharacteristics in Synthetic Hydrograph Analysis andApplication,” U.S. Department of Agriculture, Soil ConservationService, Latham, MD.Mockus 1964Mockus, Victor. 1964. Letter from Victor Mockus toOrrin Ferris, March 5, U.S. Department of Agriculture,Soil Conservation Source, Latham, MD.Monteith 1965Monteith, J. L. 1965. “Evaporation and Environment,”Symp. Soc. Exp. Biol., Vol 119, pp 205-234.Morel-Seytoux 1980Morel-Seytoux, H. J. 1980. “Application of InfiltrationTheory in Hydrologic Practice, HYDROWAR Program,”CEP80-81HJM2, Engineering Research Center, ColoradoState University, Fort Collins, CO.Morel-Seytoux and Verdin 1981Morel-Seytoux, H. J., and Verdin, J. P. 1981. “Extensionof the Soil Conservation Service Rainfall-runoff Methodologyfor Ungaged Watersheds,” U.S. Federal HighwayAdministration Office of Research and Development, No.FHWA/RD-81/060, Washington, DC.Musgrave 1955Musgrave, G. W. 1955. “How much of the Rain Entersthe Soil,” In Water: U.S. Department of AgriculturalYearbook, pp 15l-159, Washington, DC.Nash 1957Nash, J. E. 1957. “The form of the instantaneous unithydrograph,” International Association of Science andHydrology, Pub. 45, Vol 3, pp 114-121.Overton and Meadows 1976Overton, D. E., and Meadows, M. E. 1976. Stormwatermodeling, Academic Press, New York.Palmer 1946Palmer, V. J. 1946. “Retardance Coefficients for LowFlow in Channels Lined with Vegetation,” Transactions ofthe American Geophysical Union, Vol 27, No. II,pp 187-197.Pilgrim 1986Pilgrim, D. H. 1986. “Bridging the gap between floodresearch and design practice,” Water Resources Research,Vol 22, No. 9, pp 165S-176S.Pilgrim and Cordery 1975Pilgrim, D. H., and Cordery, I. 1975. “Rainfall temporalpatterns for design floods,” Journal Hydraulics Division,American Society of Civil Engineers, Vol 101, No. HY1,pp 81-95.Ponce 1983Ponce, V. M. 1983. “Development of Physically BasedCoefficients for the Diffusion Method of Flood Routing,”Final Report to the USDA, Soil Conservation Service.Lanham, MD.Ponce and Yevjevich 1978Ponce, V. M., and Yevjevich, V. 1978. “Muskingum-Cunge Method With Variable Parameters,” JournalHydraulics Division, American Society of Civil Engineers.Vol 104, No. HY12, pp 1663-1667.Ragan and Dura 1972Ragan, R. M., and Dura, J. O. 1972. “Kinematic WaveNomegraph for Times of Concentration,” Journal HydraulicsDivision, American Society of Civil Engineers,Vol 10, No. HY98, pp 1765-1771.A-4

EM 1110-2-141731 Aug 94Rallison and Miller 1981Rallison, R. E., and Miller, N. 1981. “Past, present andfuture SCS runoff procedure, in Rainfall-RunoffRelationships,” Proc. of the Inter. Symp. on rainfall-runoffModeling, May 18-21, Mississippi State University, WaterResource Publications, Littleton, CO, pp 355-364.Rawls and Brakensiek 1982Rawls, W. J., and Brakensiek, D. L. 1982. “EstimatingSoil Water Retention from Soil Properties,” Journal IrrigationDrainage Division. American Society of CivilEngineers, Vol 108, No. IR2, pp 166-171.Rawls and Brakensiek 1983Rawls, W. J., and Brakensiek, D. L., 1983. “A Procedureto Predict Green and Ampt Infiltration Parameters,” Proceedingsof the American Society of Agricultural EngineersConference on Advances in Infiltration, AmericanSociety of Agricultural Engineers, St. Joseph, MI,pp 102-112.Rawls and Brakensiek 1985Rawls, W. J., and Brakensiek, D. L. 1985. “Predictionof Soil Water Properties for Hydrologic Modeling,” Proceedingsof the American Society of Civil EngineersWatershed Management in the Eighties Symposium,American Society of Civil Engineers, New York,pp 293-299.Rawls and Brakensiek 1989Rawls, W. J., and Brakensiek, D. L. 1989. “Estimationof Soil Water Retention and Hydraulic Properties,” UnsaturatedFlow in Hydrologic Modeling, Theory and practice,H. J. Morel-Seyton, ed., NATO ASI series, series c:Mathematical and physical sciences, - Vol 275.Rawls, Brakensiek, and Savabi 1988Rawls, W. J., Brakensiek, D. L., and Savabi, R. 1988.Infiltration Parameters for Rangeland Soils,” JournalRange Management, Vol 42, pp 139-142.Rawls, Brakensiek, and Saxton 1982Rawls, W. J., Brakensiek, D. L., and Saxton, K. E. 1982.“Estimation of Soil Water Properties,” TransactionsAmerican Society of Agricultural Engineers, St. Joseph,MI, Vol 25, No. 5, pp 1316-2320.Rawls, Brakensiek, and Soni 1983Rawls, W. J., Brakensiek, D. L., and Soni, B. 1983.“Agricultural Management Effects on Soil Water Process,Part I: Soil Water Retention and Green-Ampt InfiltrationParameters,” Transactions American Society of AgriculturalEngineers, pp 1747-1752.Richards 1931Richards, L. A. 1931. “Capillary Conduction of Liquidsthrough Process Mediums,” Physical Vol 1, pp 318-333.Russell, Kenning, and Sunnell 1979Russell, S. O., Kenning, B. F. I., and Sunnell, G. J. 1979.“Estimating Design Flows for Urban Drainage,” JournalHydraulics Division, American Society of Civil Engineers,Vol 105, No. HY1, pp 43-52.Sabol 1988Sabol, G. V. 1988. “Clark Unit Hydrograph andR-parameter Estimation,” Journal Hydraulics Engineering,American Society of Civil Engineers, Vol 114, No. 1,pp 103-111.Sauer et al. 1983Sauer, V. B., Thomas, W. O., Stricker, V. A., and Wilson,K. V. 1983. “Flood Characteristics of Urban Watershedsin the United States,” Water-supply Paper 2207,U.S. Geological Survey, Washington, DC.Schomaker 1966Schomaker, C. E. 1966. “The Effect of Forest and Pastureon the Disposition of Precipitation, Maine FarmResearch,” Maine Agricultural Experiment Station, Universityof Maine, Orono, ME.Skaggs and Khaleel 1982Skaggs, R. N., and Khaleel, R. 1982. “Infiltration InHydrologic Modeling of Small Watersheds,” AmericanSociety of Agricultural Engineers Monog. No. 5, Am.Soc. Agr. Engrs., St. Joseph, MI, pp 4-166.Smart 1972Smart, J. S. 1972. “Channel networks,” Advances inHydroscience, Academic Press, Orlando, FL, Vol 8,pp 305-346.Smith 1976Smith, R. E. 1976. “Approximations for Vertical InfiltrationRate Patterns,” Transactions American Society ofAgricultural Engineers, Vol 19, No. 3, pp 505-509.Snyder 1938Snyder, F. F. 1938. “Synthetic Unit-graphs,” Trans.Amer. Geophys. Union, Vol 19, pp 447-454.A-5

EM 1110-2-141731 Aug 94Soil Conservation Service 1972Soil Conservation Service. 1972. National EngineeringHandbook, Section 4. U.S. Department of Agriculture,Washington, DC.Soil Conservation Service 1986Soil Conservation Service. 1986. “Urban Hydrology forSmall Watersheds,” Technical Release 55, U.S. Departmentof Agriculture, Washington, DC.Thorntwaite 1954Thorntwaite, C. W. 1954. “A reexamination of the conceptand measurement of potential evapotranspiration,”Publ. in Climat., V7.U.S. Army Corps of Engineers 1956U.S. Army Corps of Engineers. 1956. “SnowHydrology,” North Pacific Division, Portland, OR.U.S. Army Corps of Engineers 1979U.S. Army Corps of Engineers. 1979. “Introduction andApplication of Kinematic Wave Routing TechniquesUsing HEC-1,” Training Document No. 10, HydrologicEngineering Center, Davis, CA.U.S. Army Corps of Engineers 1982U.S. Army Corps of Engineers. 1982. “Hydrologic Analysisof Ungaged Watersheds Using HEC-1,” TrainingDocument No. 15, Hydrologic Engineering Center, Davis,CA.U.S. Army Corps of Engineers 1984aU.S. Army Corps of Engineers. 1984a. “ExpectedAnnual Flood Damage Computation, User’s Manual,”Hydrologic Engineering Center, Davis, CA.U.S. Army Corps of Engineers 1984bU.S. Army Corps of Engineers. 1984b. “HMR52, ProbableMaximum Storm Computation (Eastern U.S.), User’sManual,” Hydrologic Engineering Center, Davis, CA.U.S. Army Corps of Engineers 1986U.S. Army Corps of Engineers. 1986. “Accuracy ofComputed Water Surface Profiles,” Research DocumentNo. 26, Hydrologic Engineering Center, Davis, CA.U.S. Army Corps of Engineers 1987U.S. Army Corps of Engineers. 1987. “StreamflowSynthesis and Reservoir Regulation Model (SSARR):User Manual,” Program No. 724, North Pacific Division,Portland, OR.U.S. Army Corps of Engineers 1988U.S. Army Corps of Engineers. 1988. Proceedings of aSeminar on Flood Damage Reduction Reconnaissance -Phase Studies, Hydrologic Engineering Center, Davis,CA.U.S. Army Corps of Engineers 1989U.S. Army Corps of Engineers. 1989. Proceedings of aHydrology and Hydraulics Conference on Functional andSafety Aspects of Corps Projects, Hydrologic EngineeringCenter, Davis, CA.U.S. Army Corps of Engineers 1990aU.S. Army Corps of Engineers. 1990a. “HEC-1 FloodHydrograph Package User’s Manual,” Hydrologic EngineeringCenter, Davis, CA.U.S. Army Corps of Engineers 1990bU.S. Army Corps of Engineers. 1990b. “HEC-2 WaterSurface Profiles User’s Manual,” Hydrologic EngineeringCenter, Davis, CA.U.S. Army Corps of Engineers 1990cU.S. Army Corps of Engineers. 1990c. “HEC-6: Scourand Deposition in Rivers and Reservoirs, User’s Manual,”Hydrologic Engineering Center, Davis, CA.U.S. Department of Agriculture 1986U.S. Department of Agriculture. 1986. “UrbanHydrology for Small Watersheds,” Technical Release 55,Soil Conservation Service, Washington, DC.U.S. Department of Commerce 1947U.S. Department of Commerce. 1947. “ThunderstormRainfall,” Hydrometeorological Report No. 5, WeatherBureau, Office of Hydrologic Director, Silver Springs,MD.U.S. Department of Commerce 1961U.S. Department of Commerce. 1961. “GeneralizedEstimates of Probable Maximum Precipitation and Rainfall-FrequencyDate for Puerto Rico and Virgin Islands,”Technical Paper No. 42, Weather Bureau, Washington,DC.U.S. Department of Commerce 1962U.S. Department of Commerce. 1962. “Rainfall-FrequencyAtlas of the Hawaiian Islands for Areas to200 Square Miles,” Technical Paper No. 43, WeatherBureau, Washington, DC.A-6

EM 1110-2-141731 Aug 94U.S. Department of Commerce 1978U.S. Department of Commerce. 1978. “PMP in East ofthe 105 th Meridian,” Hydrometeorological Report No. 51,National Oceanic and Atmospheric Administration, Officeof Hydrology, National Weather Service, Silver Springs,MD.U.S. Department of Commerce 1982U.S. Department of Commerce. 1982. “Application ofPMP East of the 105 th Meridian,” HydrometeorologicalReport No. 52, National Oceanic and AtmosphericAdministration, Office of Hydrology, National WeatherService, Silver Springs, MD.U.S. Department of Commerce 1988U.S. Department of Commerce. 1988. “PMP Betweenthe 103 rd Meridian and the Continental Divide,” HydrometeorologicalReport No. 55, National Oceanic andAtmospheric Administration, Office of Hydrology,National Weather Service, Silver Springs, MD.U.S. Geological Survey 1983U.S. Geological Survey. 1983. “Precipitation-RunoffModelling System (PRMS): User’s Manual,” WaterResources Investigations Report 83-4238, U.S. GovernmentPrinting Office, Region No. 8, 779-344/9323,Washington, DC.U.S. Water Resources Council 1967U.S. Water Resources Council. 1967. “A Uniform Techniquefor Determining Flood Flow Frequencies,” Bulletin15, Washington, DC.U.S. Water Resources Council 1976U.S. Water Resources Council. 1976. “Guidelines forDetermining Flood Flow Frequencies,” Bulletin 17,Washington, DC.U.S. Water Resources Council 1977U.S. Water Resources Council. 1977. “Guidelines forDetermining Flood Flow Frequencies,” Bulletin 17A,Washington, DC.U.S. Water Resources Council 1981U.S. Water Resources Council. 1981. “Guidelines forDetermining Flood Flow Frequencies,” Bulletin 17B,Washington, D.C.Viessman et al. 1977Viessman Jr., W., Knapp, J.W., Lewis, G.L., Harbaugh,T.E. 1977. Introduction to Hydrology, Harper and Row,New York.Waananen and Crippen 1977Waananen, A.O., and Crippen, J.R. 1977. “Magnitude ofFrequency of Floods in California,” Water ResourcesInvestigations 77-21, U.S. Geological Survey, Washington,DC.Ward 1967Ward, R.C. 1967. Principles of Hydrology, McGraw-Hill(Limited), London.Woolhiser 1975Woolhiser, D.A. 1975. “Simulation of UnsteadyOverland Flow. Unsteady Flow In Open Channels,”K. Mahmood and V. Yevjevich, ed., Water ResourcesPublications, Colorado State University, Fort Collins, CO.Wright-McLaughlin Engineers 1969Wright-McLaughlin Engineers. 1969. Urban StormDrainage Criteria Manual. Denver Region Council ofGovernments, Denver, CO.A-2. Related PublicationsBurnash 1985Burnash, R. J. 1985. “The Sacramento WatershedModel, Hydrologic Modelling for Flood Hazards,” presentedat Hydrologic Modelling for Flood Hazards Planningand Management, University of Colorado at Denver,California-Nevada River Forecast Center, NationalWeather Service, Sacramento, CA.Chow 1964Chow, V. T. 1964. “Snow and Snow Survey,” Handbookof Hydrology, Section 10, New York.Hjelmfelt 1986Hjelmfelt, A. T. 1986. “Estimating Peak Runoff FromField-Size Watersheds,” Water Resources Bulletin,American Water Resources Association, Vol 22, No. 2,pp 267-274.National Oceanic and Atmospheric AdministrationNational Oceanic and Atmospheric Administration(NOAA). Atlas 2, “Precipitation-Frequency Atlas of theWestern United States,” NOAA Atlas 2, Office of Hydrology,National Weather Service, Silver Springs, MD.National Oceanic and Atmospheric AdministrationNational Oceanic and Atmospheric Administration(NOAA), National Weather Service (NWS). HydrometeorologicalBranch of the Office of Hydrology,A-7

EM 1110-2-141731 Aug 94Hydrometeorological Report (HMR) Nos. 51 and 55,U.S. Department of Commerce (1978 and 1988).National Oceanic and Atmospheric AdministrationNational Oceanic and Atmospheric Administration(NOAA), National Weather Service (NWS). HydrometeorologicalBranch of the Office of Hydrology, HydrometeorologicalReport (HMR) No. 52, U.S. Department ofCommerce 1982.National Oceanic and Atmospheric AdministrationNational Oceanic and Atmospheric Administration(NOAA), National Weather Service (NWS), HydrometeorologicalReports (HMR), NOAA Technical ReportNWS 25.Penman 1948Penman, H. L. 1948. “Natural Evaporation from OpenWater, Bare Soil and Grass,” Proc. Roy. Soc. (London),Ser. A, Vol 193, pp 120-145.Ponce 1986Ponce, V. M. 1986. “Diffusion Wave Modeling ofCatchment Dynamics,” Journal of Hydraulics Division,American Society of Civil Engineers, Vol 112, No. 8.Strelkoff 1980Strelkoff, T. 1980. “Comparative Analysis of FloodRouting Methods,” Hydrologic Engineering Center, Davis,CA.U.S. Army Corps of Engineers 1969U.S. Army Corps of Engineers. 1969. “Lower ColumbiaRiver Standard Project Flood and Probable MaximumFlood, Memorandum Report,” North Pacific Division,Portland, OR.U.S. Army Corps of Engineers 1987U.S. Army Corps of Engineers. 1987. “HECDSS, User’sGuide and Utility Program Manuals,” CPD-45,Hydrologic Engineering Center, Davis, CA.U.S. Army Corps of Engineers 1989U.S. Army Corps of Engineers. 1989. “Digest of WaterResources Policies and Authorities,” Engineer Pamphlet1165-2-1, Washington, DC.U.S. Department of Commerce 1961U.S. Department of Commerce. 1961. “PMP inCalifornia,” Hydrometeorological Report No. 36, NationalOceanic and Atmospheric Administration, Office ofHydrology, National Weather Service, Silver Springs,MD.U.S. Department of Commerce 1966U.S. Department of Commerce. 1966. “PMP in NorthwesternStates,” Hydrometeorological Report No. 43,National Oceanic and Atmospheric Administration, Officeof Hydrology, National Weather Service, Silver Springs,MD.U.S. Department of Commerce 1972U.S. Department of Commerce. 1972. “PMP inColorado River and Great Basin Drainage,” HydrometeorologicalReport No. 49, National Oceanic and AtmosphericAdministration, Office of Hydrology, NationalWeather Service, Silver Springs, MD.U.S. Department of Commerce 1980U.S. Department of Commerce. 1980. “SeasonalVariation of PMP East of the 105 th Meridian,” HydrometeorologicalReport No. 53, National Oceanic and AtmosphericAdministration, Office of Hydrology, NationalWeather Service, Silver Springs, MD.U.S. Water Resources Council 1983U.S. Water Resources Council. 1983. Economic andEnvironmental Principles and Guidelines for Water andRelated Land Resources Implementation Studies.U.S. Government Printing Office, Washington, DC.A-8

EM 1110-2-141731 Aug 94Appendix BHydrologic Engineering ManagementPlan for Flood Damage Reduction Feasibility-PhaseStudiesB-1. IntroductionThis generic HEMP is appropriate for hydrologic analysisassociated with a typical USACE flood damage reductionfeasibility study. The intent of the hydrologic engineeringanalysis would be to determine existing and future stagefrequencyrelationships at all key points in the study area,along with flooded area maps by frequency. This analysiswould be performed for without-project and for variousflood reduction components which are considered feasiblefor relief of the flood problem.B-2. Preliminary InvestigationsThis initial phase includes reviewing literature of previousreports, obtaining the available data, and requesting additionalinformation needed to perform the investigation.a. Initial preparation.(1) Confer with the other disciplines involved in thestudy to determine the objectives, the hydrologic engineeringinformation requirements of the study for other disciplines,study constraints, etc.(2) Scope study objectives and purpose.(3) Review available documents.(a)(b)(c)(d)(e)(f)Previous USACE work.U.S. Geological Survey (USGS) (or other Federalagency) reports.Local studies.Hydrologic engineering analyses from reconnaissancereport.Initial Project Management Plan.Other.(4) Obtain historic and design discharges, dischargefrequencyrelationships, high water marks, bridge designs,cross sections, and other data.(a) Local agencies (city/county, highwaydepartments, land use planning, etc.).(b)State.(c) Federal (USGS, Soil Conservation Service,U.S. Bureau of Reclamation, etc.).(d)(e)(f)Railroads.Industries.Other.(5) Scope major hydrologic engineering analysisactivities.(6) Prepare detailed Hydrologic EngineeringManagement Plan.b. Obtain study area maps.(1) County highway maps.(2) USGS quadrangle maps.(3) Aerial photographs.(4) Others.c. Estimate location of cross sections on maps(floodplain contractions, expansions, bridges, etc.).Determine mapping requirements (orthophoto) in conjunctionwith other disciplines.d. Field reconnaissance.(1) Interview local agencies and residents along thestream and review newspaper files, etc. for historic flooddata (high water marks, frequency of road overtopping,direction of flow, land-use changes, stream changes, etc.).Document names, locations, and other data for futurereference.(2) Finalize cross-sectional locations/mappingrequirements.(3) Determine initial estimate of “n” values for lateruse in water surface profile computations.(4) Take photographs or slides of bridges, construction,hydraulic structures, and floodplain channels andoverbank areas at cross-sectional locations. ConsiderB-1

EM 1110-2-141731 Aug 94dictating notes to a hand-held tape recorder to get a completeand detailed record.e. Write survey request for mapping requirementsand/or cross sections and high water marks.B-3.Development of Basin ModelThis phase of the analysis involves the selection of historicevents to be evaluated, the development of runoffparameters from gauged data (and/or regional data fromprevious studies) to ungauged basins, and the calibrationof the basin model to historic flood events. This stepassumes at least some recording stream gauge data is inor near the study watershed.a. Calibration of runoff parameters.(1) Select historic events to be evaluated based onavailable streamflow records, rainfall records, highwatermarks, etc.).(2) From USGS rating curves and time versus stagerelationships for each event, develop discharge hydrographsat each continuously recording stream gauge.Estimate peak discharge from floodcrest gauges.(3) Develop physical basin characteristics (drainageareas, slope, length, etc.) for basin above each streamgauge.(4) Select computation time interval (∆t) for this andsubsequent analyses. The computation interval must:(a) Adequately define the peak discharge of hydrographsat gauges.(b) Consider type of routing and reach travel times.(c) Have three to four points on the rising limb of thesmallest subarea unit hydrographs of interest.(d) Consider types of alternatives and futureassessments.(5) Using all appropriate rain gauges (continuous anddaily), develop historic storm patterns that correspond tothe selected recorded runoff events for the basins abovethe stream gauges.(b) Temporal distribution--from weightings of nearbyrecording rain gauges.(6) Determine best estimates of unit hydrograph andloss rate parameters for each event at each stream gauge.(7) Make adjustments for better and more consistentresults between events at each stream gauge. Adjustmentsare made to:(a) Starting values of parameters.(b) Rainfall totals and patterns (different weightingsof recording rain gauges).(8) Fix most stable parameters and rerun.(9) Adopt final unit hydrograph and base-flowparameters for each gauged basin.(10) Resimulate with adopted parameters held constantto estimate loss rates.(11) Use adopted parameters of unit hydrographs, lossrates and base flow to reconstitute other recorded eventsnot used in the above calibration to test the correctness ofthe adopted parameters and to “verify” the calibrationresults.b. Delineation of subareas. Subareas are delineatedat locations where hydrologic data are required and wherephysical characteristics change significantly.(1) Index locations where economic damage computationsare to be performed.(2) Stream gauge locations.(3) General topology of stream system.(a) Major tributaries.(b) Significant changes in land use.(c) Significant changes in soil type.(d) Other.(4) Routing reaches.(a) Average subarea totals--isoheytal maps.B-2

EM 1110-2-141731 Aug 94(5) Location of existing physical works (reservoirs,diversions, etc.) and potential location of alternate floodreduction measures to be studied.c. Subarea rainfall-runoff analysis of historic events.(1) Subarea rainfall.(a) Average subarea rainfall--from isohyetal maps.(b) Temporal distribution--weighting in accordancewith information from nearby recording raingages.(2) Average subarea loss rates.(a) From adopted values of optimization analyses.(b) From previous studies of similar basins in theregion.(c) Others.(3) Unit Hydrograph Parameters.(a) From relationships based on calibration results atstream gauges and physical basin characteristics.(b) From previous regional study relationships of unithydrograph parameters and physical basin characteristics.(c) From similar gauged or known basins.(d) From judgment, if no data is available.d. Channel routing characteristics.(1) Modified puls from water surface profile computations(HEC-2).(2) Optimized from stream gauge data (HEC-1).(3) Adopted parameters from previous studies, experience,etc.e. Reservoir routing (if reservoirs are present).This type of routing must be performed where storage hasa significant effect on reach outflow values, with reservoirsbeing the most notable example. However, onemust also apply these techniques where physical featureswarrant; such as, roads crossing a floodplain on a highfill, especially where culverts are used to pass the flowdownstream.(1) Develop area-capacity data (elevation areastoragerelationships).(2) Develop storage-outflow functions based onoutlet works characteristics.f. Generate hydrographs. Including the routinginformation of paragraph B-4, generate historic runoffhydrographs at locations of interest by combining androuting through the system for each flood.B-4.Hydraulic StudiesThese studies are used to determine water surface profiles,economic damage reaches, and modified puls channelrouting criteria.a. Prepare water surface profile data.(1) Prepare cross sections (tabulate data from eachsection).(a) Make cross sections perpendicular to flow.(b) Ensure sections are typical of reaches upstreamand downstream of cross section.(c) Develop effective flow areas. If modified pulsrouting criteria is to be determined from water surfaceprofile analyses, the entire section must be used (for storage)with high “n” values in the noneffective flow areas.(2) Refine “n” values from field reconnaissance andfrom analytical calculation and/or comparison with “n”values determined analytically from other similar streams.(3) Bridge computations--estimate where floodsevaluated will reach on each bridge and select either:(a) Normal bridge routine.(b) Special bridge routine.(4) Develop cross sections above and below bridgesto model effective bridge flow (use artificial levees orineffective flow area options, as appropriate).b. Proportion discharges. Proportion dischargesbased on hydrologic analyses of historic storms and plotpeak discharge versus river mile. Compute a series ofwater surface profiles for a range of discharges. Analysisshould start below study area so that profiles willB-3

EM 1110-2-141731 Aug 94converge to proper elevations at study limits. May wantto try several starting elevations for the series of initialdischarges.c. Check elevations. Manually check all differencesin water surface elevations across the bridge that aregreater than 3 ft.d. Obtain rating curves. The results are a series ofrating curves at desired locations (and profiles) that maybe used in subsequent analyses. Additional results are aset of storage versus outflow data by reach which, alongwith an estimate of hydrograph travel time, allow thedevelopment of modified puls data for the hydrologicmodel.B-5.Calibration of Models to Historic EventsThis study step concentrates on “debugging” the hydrologicand hydraulic models by recreating actual historicevents, thereby gaining confidence that the models arereproducing the observed hydrologic responses.a. Check hydrologic model.(1) Check historic hydrographs against recorded data,make adjustments to model parameters, and rerun themodel.(2) If no stream gauges exist, check discharges atrating curves developed from water surface profiles athigh water marks. Consider accuracy of gauged dischargemeasurements, + or - 5 percent or worse.b. Adjust models to correlate with high water marksby ±1 ft (rule of thumb--may not be applicable for allsituations).c. Adopt hydrologic and hydraulic model parametersfor hypothetical frequency analysis.B-6. Frequency Analysis for Existing Land-UseConditionsThe next phase of the analysis addresses how often specificflood levels might occur at all required points in thestudy watershed. This operation is usually done throughuse of actual gauge data (when available) to performstatistical frequency analyses and through hypotheticalstorm data to develop the stage-frequency relationships atall required points.a. Determine and plot analytical and graphical frequencycurves at each stream gauge. Adopt stage/discharge frequency relations at each gauge. Limit frequencyestimate to no more than twice the data length(i.e., 10 years of data should be used to estimate floodfrequencies no rarer than a 20-year recurrence intervalevent).b. Determine hypothetical storms.(1) Obtain hypothetical frequency storm data fromNOAA HYDRO 35, NWS TP40 and 49, or from appropriateother source. Where appropriate, develop the standardproject and/or the probable maximum storm.(2) Develop rainfall pattern for each storm, allowingfor changing drainage area within the watershed model.c. Develop corresponding frequency hydrographthroughout the basin using the calibrated hydrologicmodel.d. Calibrate model of each frequency event to knownfrequency curves. Adjust loss rates, base flow, etc. Thefrequency flows at ungauged areas are assumed to correlateto calibrated frequency flows at gauged locations.e. If no streamflow records or insufficient recordsexist to develop analytical frequency curves, use the followingprocedure:(1) Obtain frequency curves from similar nearbygauged basins.(2) Develop frequency curves at locations of interestfrom previous regional studies (USGS, Corps of Engineers,state, etc.).(3) Determine frequency hydrographs for each eventfrom hydrologic model and develop a corresponding frequencycurve at the locations of interest throughout thebasin.(4) Plot all the frequency curves (including othermethods if available) and based on engineering judgementadopt a frequency curve. This curve may actually benone of these but simply the best estimate based on theavailable data.(5) Calibrate the hydrologic model of each frequencyevent to the adopted frequency curve. The frequencyB-4

EM 1110-2-141731 Aug 94curve at other locations may be determined from thecalibrated model results, assuming consistent peak flowfrequencies.f. Determine corresponding frequency water surfaceelevations and profiles from the rating curves developedby the water surface profile evaluations.B-7.Future Without-Project AnalysisWhere hydrologic and/or hydraulic conditions areexpected to significantly change over the project life,these changes must be incorporated into the H&H analysis.Urbanization effects on watershed runoff are theusual future conditions analyzed.a. From future land use planning informationobtained during the preliminary investigation phase, identifyareas of future urbanization or intensification of existingurbanization.(1) Types of land use (residential, commercial, industrial,etc.).(2) Storm drainage requirements of the community(storm sewer design frequency, on site detention, etc.).(3) Other considerations and information.b. Select future years in which to determine projecthydrology.(1) At start of project operation (existing conditionsmay be appropriate).(2) At some year during the project life (often thesame year as whatever land-use planning information isavailable).c. Adjust model hydrology parameters for all subareasaffected by future land-use changes.(1) Unit hydrograph coefficients reflecting decreasedtime-to-peak and decreased storage.(2) Loss-rate coefficients reflecting increased imperviousnessand soil characteristics changes.(3) Routing coefficients reflecting decreased traveltimes through the watersheds hydraulic system.d. Operate the hydrology model and determinerevised discharge-frequency relationships throughout thewatershed for future without project conditions.B-8.Alternative EvaluationsFor the alternatives jointly developed with the members ofthe interdisciplinary planning team, modify the hydrologicand/or hydraulic models to develop the effects of eachalternative (individually and in combination) on floodlevels. Alternatives can include both structural (reservoirs,levees, channelization, diversions, pumping, etc.) ornonstructural (flood forecasting and warning, structureraising or relocation, flood proofing, etc.). Considerablyless hydrologic engineering effort is necessary for modelingnon-structural alternatives compared to structural.a. Consider duplicating existing and future withoutprojecthydrologic engineering models for individualanalysis of each alternative or component.b. Model components. Most structural componentsare usually modeled by modifying storage outflow relationshipsat the component location and/or modifyinghydraulic geometry through the reach under consideration.The charts given in Chapter 3 of EM 1110-2-1416 containmore information on the analysis steps for each of thefollowing alternatives:(1) Reservoirs--adjust storage-outflow relationshipsbased on spillway geometry and height of dam.(2) Levees--adjust cross-sectional geometry based onproposed levee height(s). Evaluate effect of storage lossbehind levee on storage-outflow relationships and determinerevised discharge-frequency relationships downstream and upstream, if considered significant.(3) Channels--adjust cross-sectional geometry basedon proposed channel dimensions. Evaluate effect of channelcross section and length of channelization on floodplain storage, modify storage-outflow in reach, and determinerevised downstream discharge-frequency relationships,if considered significant.(4) Diversions--adjust hydrology model for reductionof flow downstream of the diversion and to identify wherediverted flow rejoins the stream (if it does).B-5

EM 1110-2-141731 Aug 94(5) Pumping--adjust hydrology model for variouspumping capacities to be analyzed.c. Evaluate the effects of potential components onsediment regime.(1) Qualitatively--for initial screening.(2) Quantitatively--for final selection.d. Consider nonstructural components.(1) Floodproofing/structure raises--elevations ofdesign events primarily.(2) Flood forecasting--development of real-timehydrology model, determination of warning times, etc.e. Perform alternate evaluation and selection.Alternative evaluation and selection is an iterativeprocess, requiring continuous exchange of informationbetween a variety of disciplines. An exact work flow orschematic is not possible for most projects, thus paragraphB-8 could be relatively straightforward for one ortwo components or quite complex, requiring numerousreiterations as more cost and design information is knownas project refinements are made. Paragraph B-8 is usuallythe area of the HEMP requiring the most time and costcontingencies.B-9.Hydraulic DesignThis paragraph and paragraph B-8 are partly intertwined,as hydraulic design must be included with the sizing ofthe various components, both to operate hydrologic engineeringmodels and to provide sufficient information fordesign and costing purposes. Perform hydraulic designstudies commensurate with the level of detail of thereporting process.a. Reservoirs--dam height, spillway geometry, spillwaycross section, outlet works (floor elevation, length,appurtenances, etc.), scour protection, pool guide takingline, etc.b. Levees--levee design profile, freeboardrequirements, interior drainage requirements, etc.c. Channels--channel geometry, bridge modifications,scour protection, channel cleanout requirements,channel and bridge transition design, etc.d. Diversions--similar to channel design, also diversioncontrol (weir, gates, etc.)e. Pumping--capacities, start-stop pump elevations,sump design, outlet design, scour protection, etc.f. Nonstructural--floodproofing or structure raiseelevations, flood forecasting models, evacuation plan, etc.B-9. Prepare H&H Report in Appropriate Level ofDetailThe last step will be to thoroughly document the resultsof the technical analyses in report form. Hydrologic andhydraulic information presented will range from extensivefor feasibility reports to very little for most FDM’s.a. Text.b. Tables.c. Figures.B-6