26.02.2013 Views

Biogenic hydrocarbon emission estimates for North Central ... - ACD

Biogenic hydrocarbon emission estimates for North Central ... - ACD

Biogenic hydrocarbon emission estimates for North Central ... - ACD

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

Abstract<br />

Atmospheric Environment 34 (2000) 3419}3435<br />

<strong>Biogenic</strong> <strong>hydrocarbon</strong> <strong>emission</strong> <strong>estimates</strong><br />

<strong>for</strong> <strong>North</strong> <strong>Central</strong> Texas<br />

Christine Wiedinmyer�, I. Wade Strange�, Mark Estes�, Greg Yarwood�,<br />

David T. Allen��*<br />

�Department of Chemical Engineering, University of Texas at Austin, Austin, TX 87812, USA<br />

�Texas Natural Resource Conservation Commission, Austin, TX, USA<br />

�ENVIRON International Corporation, Novato, CA, USA<br />

Received 21 May 1999; accepted 23 September 1999<br />

<strong>Biogenic</strong> <strong>hydrocarbon</strong> <strong>emission</strong>s were estimated <strong>for</strong> a 37 county region in <strong>North</strong> <strong>Central</strong> Texas. The <strong>estimates</strong> were<br />

based on several sources of land use/land cover data that were combined using geographical in<strong>for</strong>mation systems. Field<br />

studies were per<strong>for</strong>med to collect species and tree diameter distribution data. These data were used to estimate biomass<br />

densities and species distributions <strong>for</strong> each of the land use/land cover classi"cations. VOC <strong>emission</strong>s <strong>estimates</strong> <strong>for</strong> the<br />

domain were produced using the new land use/land cover data and a biogenic <strong>emission</strong>s model. These <strong>emission</strong>s were<br />

more spatially resolved and a factor of 2 greater in magnitude than those calculated based on the biogenic <strong>emission</strong>s<br />

landuse database (BELD) commonly used in biogenic <strong>emission</strong>s models. � 2000 Elsevier Science Ltd. All rights<br />

reserved.<br />

Keywords: <strong>Biogenic</strong> <strong>emission</strong>s; Isoprene; Land cover; GIS; <strong>Biogenic</strong> <strong>emission</strong>s modeling<br />

1. Introduction<br />

Vegetation is a signi"cant source of atmospheric <strong>emission</strong>s<br />

of volatile organic compounds (VOCs) (Guenther<br />

et al., 1995; Geron et al., 1995; Benjamin et al., 1997;<br />

Lamb et al., 1993). The USEPA has estimated that in<br />

1995, biogenic VOC <strong>emission</strong>s in the United States were<br />

greater than anthropogenic VOC <strong>emission</strong>s (USEPA,<br />

1998). These biogenic <strong>emission</strong>s are distributed non-uni<strong>for</strong>mly<br />

and there<strong>for</strong>e are major components of the VOC<br />

inventory in some areas and are minor contributions in<br />

others (Lamb et al., 1993; Guenther et al., 1995). One area<br />

in which biogenic <strong>emission</strong>s are expected to be substantial<br />

is eastern and central Texas. The dense hardwood<br />

and coniferous <strong>for</strong>ests of eastern and central Texas are<br />

* Corresponding author. Tel.: #1-512-471-0049; fax: #1-<br />

512-471-1720.<br />

E-mail address: allen@che.utexas.edu (D.T. Allen).<br />

1352-2310/00/$ - see front matter � 2000 Elsevier Science Ltd. All rights reserved.<br />

PII: S 1 3 5 2 - 2 3 1 0 ( 9 9 ) 0 0 4 4 8 - 3<br />

predicted to have high <strong>emission</strong>s of biogenic <strong>hydrocarbon</strong>s<br />

and these <strong>emission</strong>s may contribute to the <strong>for</strong>mation<br />

and transport of ozone into populated areas within<br />

Texas. The goal of the work described in this paper was<br />

to prepare a comprehensive inventory of biogenic <strong>emission</strong>s<br />

in the region surrounding the Dallas/Ft. Worth<br />

urban area.<br />

The area chosen <strong>for</strong> study was a 37 county region in<br />

<strong>North</strong> <strong>Central</strong> Texas that includes the Dallas/Ft. Worth<br />

metroplex. The Dallas/Ft. Worth urban area fails to meet<br />

the National Ambient Air Quality Standards <strong>for</strong> ozone<br />

and has recently been declared a serious non-attainment<br />

region <strong>for</strong> ozone. Preliminary <strong>estimates</strong> have indicated<br />

that roughly 50% of the VOC <strong>emission</strong>s in the 4-county<br />

Dallas/Ft. Worth non-attainment region are due to vegetation<br />

(Estes et al., 1997). More accurate biogenic <strong>emission</strong>s<br />

<strong>estimates</strong> <strong>for</strong> the region are there<strong>for</strong>e requisite <strong>for</strong><br />

developing a sound plan <strong>for</strong> improving air quality.<br />

The current EPA land use database used to create<br />

biogenic <strong>emission</strong>s <strong>estimates</strong> (<strong>Biogenic</strong> Emissions Land<br />

Use Database, BELD) is a compilation of several sources


3420 C. Wiedinmyer et al. / Atmospheric Environment 34 (2000) 3419}3435<br />

of vegetation data <strong>for</strong> the country. The BELD relies<br />

especially on the detailed USDA Forest Inventory and<br />

Analysis data (FIA) (Kinnee et al., 1997). The FIA data<br />

contain species and <strong>for</strong>est density data <strong>for</strong> commercially<br />

valuable <strong>for</strong>ests in the eastern United States and <strong>for</strong> some<br />

<strong>for</strong>ests in the west. A large region encompassing Texas,<br />

Oklahoma and parts of Kansas represents a transition<br />

zone from moist to semi-arid ecosystems with few commercially<br />

valuable <strong>for</strong>ests. In this region, the FIA contains<br />

little to no data, and, there<strong>for</strong>e, the BELD is lacking<br />

detailed species and coverage data <strong>for</strong> much of Texas.<br />

This manuscript reports the "rst detailed <strong>estimates</strong> of<br />

species and land cover coverage data in <strong>North</strong> <strong>Central</strong><br />

Texas and these <strong>estimates</strong> have been used to predict<br />

biogenic <strong>emission</strong>s. The following sections describe the<br />

methodologies employed and the results of the inventory.<br />

2. Methods<br />

The 37 <strong>North</strong>-<strong>Central</strong> Texas counties investigated are<br />

shown in Fig. 1. In completing a biogenic <strong>emission</strong> inventory<br />

<strong>for</strong> the Dallas/Ft. Worth photochemical modeling<br />

domain, a land use/land cover (LULC) database was<br />

Fig. 1. Counties in <strong>North</strong> <strong>Central</strong> Texas Study domain.<br />

constructed, integrating existing vegetation and land<br />

use databases. Field surveys were per<strong>for</strong>med to collect<br />

species and biomass distribution data, and these data<br />

were used to estimate leaf biomass densities <strong>for</strong> each land<br />

use classi"cation in the database. The LULC database<br />

together with the <strong>estimates</strong> of leaf biomass densities from<br />

the ground surveys were then used to predict biogenic<br />

<strong>emission</strong>s <strong>for</strong> the region.<br />

2.1. Land cover database development<br />

The "rst step in estimating biogenic <strong>emission</strong>s was to<br />

develop a comprehensive vegetation coverage map <strong>for</strong><br />

the region. Several potentially useful databases were<br />

identi"ed, and the in<strong>for</strong>mation they provided was assessed<br />

<strong>for</strong> the modeling domain.<br />

The USGS land cover characteristics (LCC) database,<br />

obtained from the US Geological Survey (US Geological<br />

Survey, ca., 1976), was examined <strong>for</strong> the study domain.<br />

The LCC coverage has been used <strong>for</strong> some previous<br />

biogenic inventories (Kinnee et al., 1997). The USGS<br />

LCC coverage utilizes 205 vegetation land cover classi-<br />

"cations; 66 of these 205 classi"cations are found in the<br />

<strong>North</strong> <strong>Central</strong> Texas study domain.


Table 1<br />

Texas Parks and Wildlife Department Vegetation Classi"cations<br />

in the <strong>North</strong> <strong>Central</strong> Texas study domain. The area fraction<br />

of each classi"cation within the <strong>North</strong>-central Texas study<br />

region is reported here<br />

Description Percent of total<br />

study area<br />

Crops 22.6<br />

Ashe Juniper Parks/ Woods 1.2<br />

Bluestem Grassland 7.4<br />

Cottonwood } Hackberry-Saltcedar<br />

Brush/Woods<br />

0.2<br />

Elm } Hackberry 3.1<br />

Live Oak } Ashe Juniper Parks 1.4<br />

Live Oak } Ashe Juniper Woods 0.3<br />

Live Oak}Mesquite } Ashe Juniper Parks 2.6<br />

Mesquite-Lotebush Brush 3.2<br />

Mesquite Brush 0.2<br />

Oak - Mesquite - Juniper Parks/Woods 8.1<br />

Other Grasslands 7.8<br />

Pine } Hardwood Forest (Loblolly Pine<br />

} Sweetgum) or Pine } Hardwood Forest<br />

(Shortleaf Pine } Post Oak }<br />

Southern Red Oak)<br />

2.4<br />

Post Oak Parks/Woods 2.4<br />

Post Oak Woods, Forest and Grassland 18.3<br />

Post Oak Woods/Forest 7.3<br />

Silver } Bluestem } Texas Wintergrass<br />

Grassland<br />

5.3<br />

Urban 2.3<br />

Willow Oak } Water Oak } Blackgum<br />

Forest<br />

0.2<br />

Water Oak } Elm } Hackberry Forest 1.4<br />

Water 2.4<br />

A second database considered in the work, the Texas<br />

Parks and Wildlife Department's (TP & WD) vegetation<br />

database, utilized a very di!erent set of land cover classi-<br />

"cations. The 21 Texas Parks and Wildlife Department<br />

classi"cations found in the study domain are listed in<br />

Table 1. The TP & WD vegetation coverage was determined<br />

to be the most useful <strong>for</strong> this study. One of the<br />

main advantages of the TP & WD vegetation coverage is<br />

that it emphasizes larger plant species in its vegetation<br />

divisions. While both the LCC and TP & WD coverages<br />

may be technically accurate descriptions of the domain,<br />

crops and grasslands de"ne much of the USGS LCC. In<br />

general, trees are larger emitters of VOCs to the atmosphere,<br />

and are thus more important than crops and<br />

grasslands when estimating biogenic <strong>emission</strong>s. Also, although<br />

over 60 USGS LCC classi"cations were present<br />

in the study domain, only four classi"cations accounted<br />

<strong>for</strong> over 83% of the total area. There<strong>for</strong>e, the USGS LCC<br />

had a high resolution of land covers, but the coverage<br />

actually gave less detailed vegetation in<strong>for</strong>mation than<br />

the TP & WD coverage. Finally, the TP & WD coverage<br />

C. Wiedinmyer et al. / Atmospheric Environment 34 (2000) 3419}3435 3421<br />

has been made speci"cally <strong>for</strong> the vegetation found in<br />

Texas by ecologists familiar with the state. In developing<br />

these mappings, Landsat (earth satellite) data from the<br />

period of 1976}1980 were updated in 1986 with organized<br />

methods to `ground-trutha the vegetation <strong>for</strong> the<br />

eastern two-thirds of the state (Texas Parks and Wildlife<br />

Department, 1996). Although the TP & WD data have<br />

not been updated since 1986, the data were the most<br />

descriptive and speci"c to the study domain. Qualitative<br />

map accuracy was assessed over the study domain<br />

through ground observations, con"rming the relative<br />

accuracy of the TP & WD land descriptions compared to<br />

the LCC data.<br />

While the TP & WD database was useful in characterizing<br />

rural regions, the urban regions in the domain<br />

required a di!erent approach. The in<strong>for</strong>mation in the<br />

United States Geological Survey's land use land cover<br />

database was investigated. However, these data were<br />

compiled from in<strong>for</strong>mation obtained in the 1970s, and it<br />

was determined that, based on visual observations made<br />

by automobile, more recent data were required to accurately<br />

characterize the urban areas. The <strong>North</strong> <strong>Central</strong><br />

Council of Governments (NCTCOG) developed detailed<br />

urban land use data <strong>for</strong> Tarrant, Dallas, and parts of the<br />

eight surrounding countries. This database was constructed<br />

in 1990 by updating the United States Geological<br />

Survey's land use/land cover (USGS LULC) database.<br />

These data were more recent and represented the current<br />

land use <strong>for</strong> the Dallas}Ft. Worth urban area more<br />

accurately than the USGS LULC data. The NCTCOG<br />

land coverage classi"cations used in this database are<br />

listed in Table 2.<br />

A "nal database used in characterizing the study domain<br />

was the US Department of Agriculture-National<br />

Agricultural Statistics Service (USDA-NASS) database<br />

on crop species. 1995 county crop distribution data were<br />

used (USDA, 1997).<br />

The consolidated land use/land cover mapping used in<br />

this study is shown in Fig. 2 and incorporated data from<br />

three sources: the TP & WD database, the NCTCOG<br />

database and the USDA NASS crop data. The<br />

NCTCOG land use/land cover database was used <strong>for</strong> its<br />

detailed and current description of the land use in the<br />

urban Dallas/ Ft. Worth area. The TP & WD vegetation<br />

coverage was used to represent the remainder of the<br />

domain. The USDA National Agricultural Statistics Service<br />

(NASS) database was used to further describe the<br />

regions designated as cropland in the TP & WD vegetation<br />

coverage. The "nal mapping was also qualitatively<br />

compared to Landsat imagery data. Although the satellite<br />

imagery did not classify vegetation species, the grasslands<br />

and croplands could be distinguished from <strong>for</strong>est.<br />

These same trends in the Landsat data were seen in the<br />

composite dataset with clear divisions between the crop<br />

and grasslands versus the <strong>for</strong>est and woodlands (Strange,<br />

1998).


3422 C. Wiedinmyer et al. / Atmospheric Environment 34 (2000) 3419}3435<br />

Table 2<br />

<strong>North</strong> <strong>Central</strong> Texas Council of Governments LULC categories<br />

<strong>for</strong> the Dallas/Ft Worth urban region<br />

Code Classi"cation Percent of<br />

NCTCOG LULC<br />

111 Single family 14.04<br />

112 multi-family 1.02<br />

113 Mobile home parks 0.74<br />

114 Group quarters 0.09<br />

121 O$ce 0.40<br />

122 Retail 1.54<br />

123 Institutional 1.38<br />

124 hotel/motel 0.04<br />

131 Industrial 2.62<br />

141 Trans./communication 0.18<br />

142 Roadways 1.34<br />

143 Utilities 0.75<br />

144 Airports 0.73<br />

171 Parks and recreation 2.71<br />

172 Land"ll 0.10<br />

173 Construction 0.98<br />

181 Flood control 0.20<br />

300 Vacant 66.93<br />

500 Water 4.21<br />

2.2. Measurements of leaf biomass densities in the land use<br />

and land cover classixcations<br />

The next step in the preparation of the <strong>emission</strong> inventory<br />

was to associate a leaf biomass density by species to<br />

each land cover classi"cation. This was accomplished by<br />

measuring or estimating leaf biomass densities <strong>for</strong> each<br />

land use/land cover classi"cation in the domain. Di!erent<br />

methodologies were used in the <strong>for</strong>ested and urban<br />

sections of the domain to accomplish this task. These<br />

methods are described below.<br />

2.2.1. Forested areas<br />

The rural areas of the domain were classi"ed using the<br />

TP & WD vegetation coverage. The accuracy of these<br />

classi"cations was assessed by observation of the region<br />

by automobile, and by comparing the TP & WD maps to<br />

LANDSAT images. The preliminary survey from an automobile<br />

was used both to evaluate the general accuracy<br />

of the mappings and to identify the areas in which detailed<br />

"eld surveys were to be per<strong>for</strong>med. Locations <strong>for</strong><br />

ground surveys were chosen based on accessibility and<br />

how well the vegetation at each location represented the<br />

surrounding plant communities. This was a qualitative<br />

assessment, and there<strong>for</strong>e a source of uncertainty that<br />

will be discussed later in this report.<br />

Leaf biomass densities and species distributions <strong>for</strong> the<br />

LULC classes were determined based upon speci"c<br />

ground-level surveys using methods recommended by<br />

Dr. L. Klinger (personal communications, 1997). Transects<br />

within representative areas were chosen <strong>for</strong> extensive<br />

surveys. These areas contained undisturbed plant<br />

communities that were representative of the surrounding<br />

vegetation. Each transect covered a total area of 1000 m�,<br />

divided into 10 plots of equal area. Within each plot, all<br />

trees with a diameter at breast height (dbh) (&1.4 m)<br />

greater than 4 cm were identi"ed, and the dbh of each<br />

was measured. The height of each tree and the percent<br />

cover of the dominant plant coverages were estimated.<br />

Saplings, those stems that reached breast height, but had<br />

a dbh of less than 4 cm, were assigned a dbh of 2 cm.<br />

The raw data of speciated sapling count and tree dbh<br />

were used to calculate biomass densities <strong>for</strong> each land<br />

coverage classi"cation assigned within the domain. The<br />

biomass of each tree in every plot sampled was calculated<br />

using methods de"ned by Geron et al. (1994). The crown<br />

width <strong>for</strong> trees with narrow or conical crowns was calculated<br />

as<br />

Crwd (m)"0.47#0.166 * dbh (cm) (1)<br />

and<br />

Crwd (m)"1.13#0.205 * dbh (cm) (2)<br />

<strong>for</strong> trees with spreading, spherical crowns. The total<br />

crown area of each tree was assumed to be circular and<br />

was calculated using the crown width as the diameter:<br />

Crwn Area(m�)"�(Crwd/4)�. (3)<br />

If the total crown area within a plot exceeded the area of<br />

the plot (100 m�), the total crown area was scaled down<br />

to 100 m� proportionally by species.<br />

The leaf biomass of each tree was calculated as the<br />

product of the crown area of each tree multiplied by<br />

foliar mass. Foliar mass was assigned by genus:<br />

700gm�� <strong>for</strong> Pinus and other coniferous genera, and<br />

375gm�� <strong>for</strong> all deciduous trees (Geron et al., 1994).<br />

These foliar mass constants are used <strong>for</strong> biogenic <strong>emission</strong><br />

estimations <strong>for</strong> areas throughout the country and<br />

were assumed to be applicable to the vegetation found in<br />

this region.<br />

Other methods have been used to determine biomass<br />

values <strong>for</strong> trees (Miller and Winer, 1984). These methods<br />

use leafmass factors in units of leafmass per cubic meter<br />

of canopy volume. These methods were established from<br />

surveys done in urban environments, where the trees are<br />

not always in a <strong>for</strong>ested canopy and the tree crowns are<br />

di!erent than those in closed canopies. Because <strong>for</strong>ested<br />

areas dominated the majority of the study domain, the<br />

methods described above were utilized <strong>for</strong> this study.<br />

Figs. 3 and 4 show typical results from the transect<br />

surveys of this study. Fig. 3 shows a transect through an<br />

area designated as Post Oaks Woods and Forests. The<br />

total biomass density within each plot remains relatively<br />

constant, although there is some variation in species. In


C. Wiedinmyer et al. / Atmospheric Environment 34 (2000) 3419}3435 3423<br />

Fig. 2. Final map of the new Land Use and Land Cover in the <strong>North</strong> <strong>Central</strong> Texas domain, showing the spatial distribution of the<br />

major vegetation classi"cations in the domain. The location of the <strong>for</strong>est transects are shown.


3424 C. Wiedinmyer et al. / Atmospheric Environment 34 (2000) 3419}3435<br />

Fig. 3. Biomass density values calculated <strong>for</strong> a transect taken in Fair"eld State Park. This transect was located within an area designated<br />

as Post Oak Woods and Forests by the Texas Parks and Wildlife vegetation database. Plots 0 and 1 on the chart represent the same plot<br />

within the transect. The "eld measurements <strong>for</strong> this plot were repeated <strong>for</strong> quality assurance purposes.<br />

Fig. 4. Biomass densities calculated <strong>for</strong> a transect taken in<br />

Possum Kingdom State Park. This transect was located within<br />

a region designated as Ashe Juniper Parks and Woods by the<br />

Texas Parks and Wildlife vegetation database.<br />

contrast, the results shown in Fig. 4 exhibit the changes<br />

that can occur within a single transect.<br />

Each land use category within the domain was assigned<br />

a total biomass density value and biomass densities<br />

by species. This was accomplished by averaging the<br />

"nal biomass density values of all transects taken within<br />

a speci"c land use category. Some transects were representative<br />

of areas within several vegetation categories<br />

and were also applied in the "nal averaging <strong>for</strong> all applicable<br />

vegetation classi"cations. Table 3 shows the Texas<br />

Parks and Wildlife Classi"cations in which surveys were<br />

per<strong>for</strong>med and the number of transects taken in each of<br />

those categories. Although only small areas within<br />

a large domain were surveyed, the small survey areas<br />

were chosen because they represented the larger area of<br />

vegetation.<br />

Certain rural regions within the domain were not surveyed.<br />

These areas did not contain signi"cant amounts of<br />

leaf biomass, did not contain many VOC-emitting tree<br />

species, or did not contribute to a signi"cant amount of<br />

area within the domain. The land use/land cover classi-<br />

"cations of these areas were subjectively assigned leaf<br />

biomass densities and species distributions based on the<br />

species anticipated <strong>for</strong> the areas and by the values of<br />

biomass densities <strong>for</strong> the surrounding vegetation communities.<br />

As an example of how leaf biomass densities<br />

and species distributions were assigned to a vegetation<br />

class <strong>for</strong> which no ground survey was taken, consider the<br />

Bluestem Grass classi"cation.<br />

The Bluestem Grass classi"cation was not surveyed<br />

because it was expected to contain less than 10% woody<br />

canopy coverage. The areas designated by the Bluestem<br />

Grass classi"cation within the domain are primarily surrounded<br />

by areas designated as Oak, Mesquite and Juniper<br />

Parks and Woods and are expected to contain the<br />

same species of trees. Visual observations of areas designated<br />

as Bluestem Grasslands con"rmed these assumptions.<br />

The Parks and Woods classi"cation is expected to<br />

have anywhere from 11 to 100% woody canopy coverage,<br />

although the surveys found the woody coverage to<br />

be closer to the upper bounds. Because of the relationship


Table 3<br />

Texas Parks and Wildlife Department vegetation classi"cations<br />

and the number of <strong>for</strong>est transects taken within each classi"cation<br />

Texas Parks and Wildlife classi"cations No. of representative<br />

transects taken<br />

Ashe juniper parks and woods 4<br />

Bluestem Grass �<br />

Cottonwood-Hackberry-Saltcedar<br />

Brush/Woods<br />

�<br />

Crops �<br />

Elm Hackberry Parks and Woods 2<br />

Live oak and Ashe Juniper parks 2<br />

Live oak and Ashe Juniper Woods 2<br />

Live oak-Mesquite-Ashe Juniper parks �<br />

Mesquite-Lotebush Brush �<br />

Mesquite Brush �<br />

Oak, Mesquite, and Juniper<br />

parks and woods<br />

6<br />

Other �<br />

Pine Hardwood <strong>for</strong>ests 3<br />

Post Oak Parks and Woods 2<br />

Post Oak Woods, Forests and<br />

Grasslands<br />

2<br />

Post Oaks Woods and Forests 6<br />

Silver-Bluestem-Texas Wintergrass<br />

grassland<br />

�<br />

Urban Treated separately<br />

Willow Oak-Water Oak-Blackgum<br />

Forests<br />

�<br />

Water Oak-Elm-Hackberry Forest �<br />

Water �<br />

Note: Some transects are counted more than once, as they are<br />

representative of more than one vegetation class.<br />

�Classi"cations that were not surveyed using Belt survey<br />

methods.<br />

between the two classi"cations, the Bluestem Grass classi"cation<br />

was assigned the same species distribution as<br />

the Oak, Mesquite and Juniper Parks and Woods classi-<br />

"cation, with only 20% of its biomass densities. Yarwood<br />

et al. (1997) describe complete details of the biomass<br />

assignments.<br />

2.2.2. Urban areas<br />

Vegetation species and leaf biomass densities were<br />

assessed <strong>for</strong> the classi"cations within the urban areas of<br />

Dallas and Ft. Worth. Where possible, such as in city<br />

parks, transects were surveyed, using the same methods<br />

employed in the <strong>for</strong>ested areas. In most areas, however, it<br />

was not possible to use the same methods as used in the<br />

rural regions. The methods used in the urban areas are<br />

described below.<br />

Twelve ground surveys were per<strong>for</strong>med in residential<br />

areas. For each survey, the distance from the back yard<br />

C. Wiedinmyer et al. / Atmospheric Environment 34 (2000) 3419}3435 3425<br />

to the front yard of an average house in the selected<br />

neighborhood was measured. A survey team then drove<br />

a speci"ed distance through the neighborhood. In this<br />

space, all trees in front yards were identi"ed, and their<br />

dbh estimated to the nearest 5 cm. The area of each<br />

transect was calculated as the width of the house and<br />

yards multiplied by the length driven. The back yards<br />

were assumed to contain the same species and size distributions<br />

as the front yards, since access to private property<br />

was limited. The accuracy of this assumption has not<br />

been assessed. Upon visual observations, many back<br />

yards were seen to have trees and the inclusion of biomass<br />

densities <strong>for</strong> back yard trees in the inventory was<br />

expected to be more accurate than neglecting to include<br />

this in<strong>for</strong>mation.<br />

Areas designated as retail, airport, institutional and<br />

industrial were treated in a slightly di!erent manner. The<br />

perimeter of the area being surveyed was measured by an<br />

automobile. All trees within the perimeter were identi"ed<br />

and their dbh were estimated to the nearest 5 cm. The<br />

survey area was calculated from the perimeter measurements.<br />

The leaf biomass of each tree surveyed in the urban<br />

areas was calculated using methods de"ned by Geron et<br />

al. (1994) as described in the previous section. The leaf<br />

biomass of each species in a transect was summed to<br />

acquire leaf biomass by species. These biomass values<br />

were divided by the area of the transect to obtain the<br />

biomass density of the transect <strong>for</strong> each species. The total<br />

biomass densities, by species, from all of the transects<br />

that fell within a certain land use classi"cation were<br />

averaged. These averaged biomass density values were<br />

assigned to the represented land use category. All urban<br />

regions were assigned biomass densities from the results<br />

of the ground surveys except <strong>for</strong> those assumed to be<br />

zero, i.e. roadways and areas under construction. Table 4<br />

shows the NCTCOG urban classi"cations and the<br />

number of transects completed in each class.<br />

2.3. <strong>Biogenic</strong> <strong>emission</strong>s modeling<br />

2.3.1. Model selection<br />

The BEIS-2 and BIOME models were considered <strong>for</strong><br />

use in estimating biogenic <strong>emission</strong>s based on biomass<br />

densities and leaf coverage. Both of these models have<br />

been used in developing biogenics <strong>emission</strong>s inventories<br />

(Radian, 1996; Wilkinson et al., 1996; Estes et al., 1997).<br />

The BIOME model has been approved <strong>for</strong> use in Texas<br />

by USEPA. The primary di!erences between BEIS-2 and<br />

BIOME are in (1) the canopy model, (2) the <strong>emission</strong>s<br />

factors, and (3) the ability to use locally derived leaf<br />

biomass, species composition, and land use data.<br />

2.3.2. Canopy models<br />

The canopy model depicts the behavior of sunlight and<br />

heat as they interact with the <strong>for</strong>est canopy. Since


3426 C. Wiedinmyer et al. / Atmospheric Environment 34 (2000) 3419}3435<br />

Table 4<br />

Urban land use classi"cations in the Dallas/Ft. Worth urban<br />

region and the number of surveys per<strong>for</strong>med in each classi"cation<br />

Urban classi"cations No. of representative<br />

transects taken<br />

Residential: single, multi-family, group<br />

quarters, and mobile homes<br />

12<br />

O$ce Not surveyed<br />

Retail 5<br />

Institutional 8<br />

Hotel/motel Not surveyed<br />

Industrial 2<br />

Trans./Comm. Not surveyed<br />

Roadway Not surveyed<br />

Utilities Not surveyed<br />

Airport 2<br />

Parking garages Not surveyed<br />

Parks and recreation 5<br />

Land"ll Not surveyed<br />

Under construction Not surveyed<br />

Flood control Not surveyed<br />

Vacant� Not surveyed<br />

�Vacant lands in the urban region were replaced in the "nal<br />

database by the biomass density values of the Texas Parks and<br />

Wildlife Department classi"cations that overlaid the vacant<br />

regions.<br />

biogenic VOC <strong>emission</strong>s depend on sunlight and temperature,<br />

it is particularly important to estimate these<br />

parameters accurately. Several canopy models have been<br />

used in biogenic <strong>emission</strong> models, from very simple<br />

`look-up tablea models that reduce the solar radiation<br />

and temperature by "xed amounts as the canopy is<br />

penetrated (Guenther et al., 1993,1994; Geron et al.,<br />

1994), to complex energy balance models that calculate<br />

the temperatures based on many meteorological variables<br />

(Vogel et al., 1995). In an important recent study,<br />

however, Lamb et al. (1996) concluded that there is little<br />

practical di!erence between the temperature and solar<br />

radiation pro"les developed by simple and complex canopy<br />

models. It is much more important to accurately<br />

estimate the species composition and leaf biomass density<br />

of the <strong>for</strong>est in question; compared to these parameters,<br />

the choice of canopy model does not a!ect the<br />

accuracy of the biogenic <strong>emission</strong>s estimate. There<strong>for</strong>e,<br />

the choice of a biogenic <strong>emission</strong>s model was not based<br />

on the canopy model used.<br />

2.3.3. Base-rate <strong>emission</strong> factors<br />

BEIS-2 and BIOME are accompanied by their own<br />

base-rate biogenic <strong>emission</strong> factor databases. The term<br />

`base-ratea indicates <strong>emission</strong>s at 303C and half of full<br />

sunlight (1000 �Em�� s��). However, it is possible to<br />

use either set of <strong>emission</strong> factors with each model. We<br />

chose to use the BEIS-2 <strong>emission</strong> factors, given that these<br />

factors have been estimated rigorously and consistently<br />

(Guenther et al., 1993,1994; Geron et al., 1994), and that<br />

BEIS-3 <strong>emission</strong> factors are in development.<br />

2.3.4. Base-rate <strong>emission</strong> factor data<br />

The BEIS-2 base-rate <strong>emission</strong> factors <strong>for</strong> tree genera<br />

were used to calculate the <strong>North</strong> <strong>Central</strong> Texas (NCT)<br />

biogenic <strong>emission</strong>s inventory. To use these <strong>emission</strong><br />

factors in BIOME, it was necessary to convert the <strong>emission</strong><br />

factors from units of <strong>emission</strong> #ux (�gm�� h��)<br />

to units of <strong>emission</strong>s per gram of biomass (�g C g biomass��<br />

h��). This conversion was e!ected using the<br />

unit leaf biomass densities in Geron et al. (1994) and<br />

Radian (1996). BEIS-2 <strong>emission</strong> factors were missing <strong>for</strong><br />

some genera in the NCT domain. These genera included<br />

Albizia, Callicarpa, Ficus, Firminia, Hibiscus, Koelreuteria,<br />

Lagerstroemia, Ligustrum, Myrica, Photinia, Pyrus,<br />

Pistacia, Rhus, Sapindus, Sophora, Xanthoxylum, and Zelkova.<br />

Fortunately, none of these genera were common in<br />

the domain; many were not native. For missing genera,<br />

the taxonomic family average <strong>emission</strong> factor technique<br />

was used. This technique has been used by Benjamin<br />

et al. (1997), Benjamin and Winer (1998) and by the State<br />

of Texas. For some genera, BEIS-2 <strong>emission</strong> factors<br />

were not available <strong>for</strong> any member of their taxonomic<br />

families. In these cases, <strong>emission</strong> factors were drawn from<br />

Wilkinson and Emigh (1995), or from Benjamin and<br />

Winer (1998), or from an average of the <strong>emission</strong> factors<br />

of genera in the same taxonomic order. Table 5 shows the<br />

base-rate <strong>emission</strong> factors assigned to these genera.<br />

Crop leaf biomass densities (Radian and VRC, 1994)<br />

were used to convert the BEIS2 <strong>emission</strong> #ux units to<br />

BIOME units of <strong>emission</strong> per gram of leaf biomass.<br />

2.3.5. Use of local data<br />

The version of BEIS-2 available when the biogenic<br />

model was chosen could not readily accept any land use,<br />

species composition, or leaf biomass density data other<br />

than the biogenic <strong>emission</strong>s landuse database (BELD).<br />

There<strong>for</strong>e, BIOME, which has been successfully used in<br />

the past with domain-speci"c data as input (Estes et al.,<br />

1997), was used.<br />

2.3.6. Solar radiation data<br />

The algorithms included in earlier versions of BEIS-2<br />

(EC/R Incorporated, 1995) were used <strong>for</strong> calculating<br />

a gridded solar radiation "eld <strong>for</strong> the domain, based<br />

upon latitude, time of year, time of day, and cloud cover<br />

(Iqbal, 1983). No direct ambient solar radiation measurements<br />

were available <strong>for</strong> biogenic <strong>emission</strong>s modeling of<br />

the two episodes that will be presented in the results<br />

section, there<strong>for</strong>e solar radiation was estimated. Cloud<br />

cover data were obtained <strong>for</strong> the speci"c episode days<br />

from the National Climatic Data Center (NCDC). This


Table 5<br />

Emission factors assigned to genera without a BEIS-2 <strong>emission</strong> factor<br />

Genus Data source <strong>for</strong> <strong>emission</strong><br />

factor<br />

Isoprene <strong>emission</strong><br />

factor<br />

(�g/g-biomass/h)<br />

Monoterpene <strong>emission</strong><br />

factor<br />

(�g/g-biomass/h)<br />

Albizia Benjamin and Winer, 1998 1.37 0.53 1.9<br />

Callicarpa Verbenaceae family average 0.13 0.11 1.9<br />

Ficus Benjamin and Winer (1998) 8.61 0.08 1.9<br />

Firminia None found NA NA NA<br />

Hibiscus Malvales order average 0.1 0.1 1.9<br />

Koelreuteria Benjamin and Winer (1998) 16.23 0 1.9<br />

Lagerstroemia Benjamin and Winer (1998) 0 0 NA<br />

Ligustrum Benjamin and Winer (1998) 0 0 1.9<br />

Myrica Benjamin and Winer (1998) 6.76 0.79 1.2<br />

Photinia Rosaceae family average 0.13 0.11 1.9<br />

Pistacia Benjamin and Winer (1998) 0 3.39 1.9<br />

Pyrus Benjamin and Winer (1998) 0 0 1.9<br />

Rhus Benjamin and Winer (1998) 0 0 1.9<br />

Sapindus Sapindales order average 0.1 0.57 1.9<br />

Sophora Benjamin and Winer (1998) 1.37 0.53 1.9<br />

Viburnum Benjamin and Winer (1998) 0 0.08 NA<br />

Xanthoxylum Wilkinson and Emigh (1995) 0 1.49 1.9<br />

Zelkova Benjamin and Winer (1998) 0 0 1.9<br />

cloud cover database was used to calculate solar radiation<br />

<strong>for</strong> both the core and the regional domains. For the<br />

June 1995 episode described in the Results section, cloud<br />

cover observations were available <strong>for</strong> every hour; <strong>for</strong> the<br />

July 1996 episode described in the Results section, however,<br />

observations were available <strong>for</strong> every hour on 30<br />

June, but only every 6 h beginning on 1 July. This change<br />

in the observation schedule occurred nationwide, because<br />

the National Weather Service changed its method<br />

of observing and reporting cloud cover data beginning<br />

on 1 July, 1996. For the 1996 episode, after 1 July the<br />

observations were interpolated in time as well as space,<br />

so that during daylight hours, each observation was used<br />

<strong>for</strong> 6 h (the hour of observation, plus 3 h be<strong>for</strong>e and 2 h<br />

after the time of observation).<br />

The amount of transmittance through the cloud cover<br />

was calculated by considering cloud thickness, sky coverage,<br />

and cloud height (Iqbal, 1983; Pierce and Waldru!,<br />

1991). Although these data are the best available, this<br />

method is based on observations that are not continuous<br />

in space or time, and are interpolated between weather<br />

stations. A comprehensive satellite photo study was beyond<br />

the scope of this study, and it would be di$cult to<br />

estimate transmittance through the clouds based on observations<br />

of cloud tops (McNider et al., 1995). As a sensitivity<br />

check, the biogenic <strong>emission</strong>s model was run with<br />

no cloud cover. Solar radiation increased greatly in the<br />

absence of clouds, and there<strong>for</strong>e isoprene <strong>emission</strong>s <strong>estimates</strong><br />

were substantially higher than <strong>emission</strong>s <strong>estimates</strong><br />

calculated with cloud cover. Another quality check was<br />

C. Wiedinmyer et al. / Atmospheric Environment 34 (2000) 3419}3435 3427<br />

OVOC <strong>emission</strong><br />

factor<br />

(�g/g-biomass/h)<br />

per<strong>for</strong>med by creating contour plots of solar radiation to<br />

ensure the solar radiation "eld was reasonable. The "elds<br />

generated by the UAM-BEIS2 solar radiation algorithm<br />

were reasonable: sunset and sunrise occurred in west and<br />

east, respectively, and maximum radiation occurred at<br />

solar noon.<br />

2.3.7. Temperature data<br />

Temperature data were obtained from the NCDC.<br />

A gridded temperature "eld was created from interpolated<br />

temperature observations measured at stations<br />

across the domain. A qualitative assessment of contour<br />

plots of the temperature "eld indicated no anomalies.<br />

2.3.8. Other meteorological parameters needed<br />

<strong>for</strong> input to BIOME<br />

The wind "eld data needed <strong>for</strong> BIOME's canopy<br />

model were obtained from the output of the systems<br />

applications international mesoscale model (SAIMM)<br />

meteorological model, which is based on the Colorado<br />

State University Mesoscale Model (Mahrer and Piekle,<br />

1977,1978). Likewise, speci"c humidity data were also<br />

taken from the SAIMM meteorological modeling.<br />

3. Results and discussion<br />

3.1. Final database (coverage)<br />

A "nal Land Use and Land Cover mapping was created<br />

from the combination of databases discussed in


3428 C. Wiedinmyer et al. / Atmospheric Environment 34 (2000) 3419}3435<br />

Table 6<br />

Final total and oak biomass density values <strong>for</strong> each Texas Parks and Wildlife vegetation category in study domain. The standard<br />

deviation in the biomass densities derived <strong>for</strong> each category is reported<br />

Vegetation category Total biomass<br />

density (g m��)<br />

Standard<br />

deviation<br />

Total oak biomass<br />

density (g m��)<br />

Post Oak Woods, Forests and Grasslands 156 70 134 87<br />

Post Oak Woods and Forests 325 108 171 67<br />

Post Oaks Parks and Woods 251 82 230 108<br />

Live Oak and Ash Juniper Woods 339 343 44 41<br />

Live Oak and Ash Juniper Parks 153 27 9 12<br />

Oak, Mesquite, and Juniper Parks and Woods 326 181 33 73<br />

Elm-Hackberry Parks and Woods 394 37 226 129<br />

Pine Hardwood Forests 496 66 82 69<br />

Ash Juniper Parks and Woods 355 235 50 88<br />

Bluestem Grass 65 � 13 �<br />

Cottonwood-Hackberry-Saltcedar Brush/Woods 156 � 134 �<br />

Mesquite Brush 125 � � �<br />

Mesquite Lotebush Brush 125 � � �<br />

Silver-Bluestem-Texas Wintergrass Grassland 63 � 54 �<br />

Water Oak-Elm-Hackberry Forest 371 65 170 68<br />

Willow Oak-Water Oak-Blackgum Forests 371 65 170 68<br />

Standard<br />

deviation<br />

�Signi"es classi"cations in which no "eld surveys were taken: biomass density assignments are based on surveys taken in other<br />

classi"cations.<br />

Table 7<br />

Final biomass densities assigned to the urban land use classi"cations in the Dallas/Ft. Worth urban area<br />

Urban land use category Total biomass<br />

density(g m��)<br />

Standard<br />

deviation<br />

Biomass density<br />

of oaks (g m��)<br />

Standard<br />

deviation<br />

Institutional 11 7 3 3<br />

Vacant� 46 � 9 �<br />

Commercial 15 15 8 11<br />

Industrial 3 4 2 3<br />

Parks and Recreational 78 39 14 50<br />

Residential 42 25 10 6<br />

�Vacant lands in the urban region were replaced in the "nal database by the biomass density values of the Texas Parks and Wildlife<br />

Department classi"cations that overlaid the vacant regions.<br />

previous sections (Fig. 2). This "nal mapping assigns land<br />

use classi"cations to the entire study domain. These<br />

classi"cations are particular to the <strong>North</strong> <strong>Central</strong> region<br />

in Texas and describe the land use and vegetation cover<br />

more speci"cally than any other land cover database<br />

available <strong>for</strong> the region. This improved land use database<br />

was used as the basis <strong>for</strong> biogenic <strong>emission</strong>s <strong>estimates</strong>.<br />

Each of the land use and land cover classi"cations<br />

shown in Fig. 2 was assigned biomass densities and<br />

species distributions. Tables 6 and 7 show the total<br />

biomass and density values assigned to each of these<br />

classi"cations. The total oak biomass density of each<br />

classi"cation is also included. This parameter is important<br />

when estimating biogenic <strong>hydrocarbon</strong> <strong>emission</strong>s because<br />

oak species are very common in the domain, and<br />

the oak trees are the largest sources of biogenic isoprene<br />

<strong>emission</strong>s. The magnitude of oak biomass density assigned<br />

to a classi"cation is highly correlated with the<br />

amount of isoprene <strong>emission</strong>s expected from an area<br />

designated by that classi"cation. More detailed species<br />

assignments are reported elsewhere (Yarwood et al.,<br />

1997).<br />

3.2. Comparison of leaf biomass <strong>estimates</strong> with the BELD<br />

The completed vegetation database <strong>for</strong> <strong>North</strong> <strong>Central</strong><br />

Texas can be compared to the biogenic <strong>emission</strong>s landuse<br />

database (BELD). The BELD, as described in Kinnee<br />

et al. (1997), is the most current national vegetation<br />

database <strong>for</strong> estimating biogenic <strong>emission</strong>s. This<br />

database assigns species distributions and vegetation<br />

coverage by county <strong>for</strong> the contiguous United States.


Fig. 5. The percent <strong>for</strong>est and oak coverage <strong>for</strong> the BELD3.0 data and the database created <strong>for</strong> this study in <strong>North</strong> central Texas.<br />

Although the BELD applies very speci"c data to<br />

many areas of the nation, especially some eastern <strong>for</strong>ests,<br />

the data <strong>for</strong> <strong>North</strong> <strong>Central</strong> Texas were created from the<br />

USGS LCC database. As reported in Section 2.1, these<br />

C. Wiedinmyer et al. / Atmospheric Environment 34 (2000) 3419}3435 3429<br />

data are not the most accurate <strong>for</strong> the region. Fig. 5<br />

shows the percent <strong>for</strong>est and percent oak of the BELD3.0<br />

and the database created <strong>for</strong> this study <strong>for</strong> the north<br />

central Texas domain.


3430 C. Wiedinmyer et al. / Atmospheric Environment 34 (2000) 3419}3435<br />

As can be seen in Fig. 5, the BELD database assigns<br />

the majority of the NCT domain 0}15% <strong>for</strong>est coverage,<br />

and of that coverage, few oaks are reported, except in the<br />

eastern portion of the domain, where the FIA contains<br />

detailed in<strong>for</strong>mation. As presented in this study, many<br />

land use classi"cations were found to have signi"cant<br />

<strong>for</strong>est coverage, and oaks are often signi"cant components.<br />

In 6 of 16 vegetation classi"cations, including<br />

those classi"cations that cover the greatest areas of the<br />

domain, oak trees are the dominant species. Oak trees<br />

produce large <strong>emission</strong>s of isoprene; it is there<strong>for</strong>e essential<br />

to accurately quantify the oak distribution when<br />

building a biogenic <strong>emission</strong>s inventory. The BELD<br />

lacks speci"c in<strong>for</strong>mation about oak and other emitting<br />

tree species <strong>for</strong> <strong>North</strong> <strong>Central</strong> Texas since it is based on<br />

the FIA data, which contains little species and coverage<br />

data <strong>for</strong> the region. Because the BELD reports few oak<br />

species, and <strong>estimates</strong> <strong>for</strong>est coverage in the NCT domain<br />

spatially di!erent than this new land cover data,<br />

biogenic <strong>emission</strong>s <strong>estimates</strong> produced from the BELD<br />

may be underestimated and spatially inaccurate.<br />

3.3. <strong>Biogenic</strong> VOC <strong>emission</strong>s <strong>estimates</strong> by landuse<br />

and compound<br />

Table 8 presents the biogenic VOC <strong>emission</strong>s <strong>for</strong> two<br />

episode days, 21 June, 1995 and 3 July, 1996. The <strong>emission</strong>s<br />

are categorized by VOC species and by broad land<br />

use category. These data indicate that the biogenic compound<br />

with the largest <strong>emission</strong>s in the DFW core domain<br />

is isoprene; isoprene contributes 90% of the mass of<br />

biogenic VOC <strong>emission</strong>s on 21 June, 1995, and 90% on<br />

3 July, 1996. Isoprene <strong>emission</strong>s are the largest category<br />

of biogenic VOC <strong>emission</strong>s <strong>for</strong> both the urban and rural<br />

land use categories, but monoterpenes are the largest<br />

compound class emitted by croplands. <strong>Biogenic</strong> VOC<br />

<strong>emission</strong>s <strong>for</strong> the July episode are notably larger than <strong>for</strong><br />

the June episode, probably due to the higher temperatures<br />

during the July episode.<br />

Fig. 6 shows biogenic VOC <strong>emission</strong>s <strong>for</strong> <strong>North</strong> <strong>Central</strong><br />

Texas, as calculated by BEIS-2 with the BELD<br />

database <strong>for</strong> 7/3/96. Fig. 7 shows the biogenic <strong>emission</strong>s<br />

<strong>estimates</strong> <strong>for</strong> the same day, 7/3/96, calculated with the<br />

BIOME model and BEIS2 <strong>emission</strong> factors. This <strong>emission</strong><br />

estimation was produced with the new landuse<br />

database produced by this study. The total biogenic<br />

<strong>emission</strong>s estimated with use of BEIS2 and the BELD<br />

are on the order of 4000 metric tons d��, whereas the<br />

BIOME <strong>estimates</strong> produced with the new landuse<br />

database and BEIS2 <strong>emission</strong>s factors is on the order of<br />

9600 metric tons d��. The new <strong>emission</strong> <strong>estimates</strong> are<br />

more than twice that of the BEIS2 <strong>estimates</strong>. The spatial<br />

distribution of the biogenic <strong>emission</strong>s throughout the<br />

domain also di!ers between the two model runs. The two<br />

models used di!erent landuse and leaf biomass data and<br />

di!erent canopy models. Both model runs used identical<br />

Table 8<br />

<strong>Biogenic</strong> VOC <strong>emission</strong>s <strong>for</strong> the Domain (t/d) from BIOME and<br />

the new north central Texas land use data<br />

Land use Monoterpenes OVOC Isoprene<br />

21-Jun-95<br />

Urban 1 4 41<br />

Non-crop Rural 117 465 5076<br />

Cropland 6 4 3<br />

Total 124 473 5120<br />

3-Jul-96<br />

Urban 2 6 63<br />

Non-crop Rural 163 657 7496<br />

Cropland 9 6 4<br />

Total 174 669 7563<br />

base <strong>emission</strong> factors, temperature and solar radiation<br />

algorithms, and meteorological data. The di!erence in<br />

canopy models may have contributed to the di!erences<br />

in the <strong>emission</strong> results, but would not have contributed<br />

to the change in the spatial resolution of the <strong>emission</strong><br />

<strong>estimates</strong>.<br />

The highest biogenic VOC <strong>emission</strong>s in the non-attainment<br />

counties were seen in Denton County, which contained<br />

substantial acreage of Post Oak Woods, Forests<br />

and Grasslands. Relatively low <strong>emission</strong>s were observed<br />

across the non-attainment counties, however, relative to<br />

the rural areas. The highest biogenic VOC <strong>emission</strong>s<br />

in the domain were seen in Ellis, Navarro, and Limestone<br />

counties, which are dominated by Elm-Hackberry<br />

Forests (Table 6 shows that Elm-Hackberry Forest classi"cation<br />

has a high percentage of oak leaf biomass).<br />

Emissions were also high in Jack and Wise counties,<br />

which are dominated by the Post Oak Parks and Woods<br />

classi"cation. The latter result is signi"cantly di!erent<br />

from BEIS-2 results; the BELD land use database does<br />

not include substantial oak <strong>for</strong>ests in those counties<br />

(Yarwood et al., 1997). The BEIS-2/BELD results do not<br />

show the biogenic VOC hot spots seen in the <strong>emission</strong>s<br />

predicted by this work. In the BEIS-2/BELD results, the<br />

highest biogenic VOC <strong>emission</strong>s are seen in areas farther<br />

east and south from the DFW area. The di!ering geographic<br />

distribution could be signi"cant when the biogenic<br />

modeling results are used in photochemical<br />

modeling. The lowest biogenic VOC <strong>emission</strong>s in the<br />

domain were seen in McLennon, Hill, Falls, and Bell<br />

counties, which are dominated by crops.<br />

3.4. Sensitivity analysis<br />

Signi"cant uncertainties occur throughout the process<br />

of creating a biogenic <strong>emission</strong> inventory. Uncertainties<br />

in <strong>emission</strong> factors, activity coe$cients, canopy models<br />

and meteorological data are beyond the scope of this


Fig. 6. <strong>Biogenic</strong> VOC <strong>emission</strong>s <strong>for</strong> north-central Texas on 7/3/96, calculated with the BEIS2 model, using the BELD land use database.<br />

Units are tons/day of Carbon Bond IV <strong>hydrocarbon</strong>s. Grids are 16�16 km (256 km�). Note that the colors are scaled to match those in<br />

Fig. 7, which contains 4�4 km grid cells. Each color represents the same range of <strong>emission</strong> densities in both tileplots.<br />

paper, but have been discussed elsewhere (Benjamin<br />

et al., 1997; Lamb et al., 1996; Lamb et al., 1993; Geron<br />

et al., 1994; Guenther et al., 1999). The analysis of uncertainties<br />

presented in this paper will focus on the<br />

C. Wiedinmyer et al. / Atmospheric Environment 34 (2000) 3419}3435 3431<br />

uncertainties in <strong>emission</strong> <strong>estimates</strong> due to variability in<br />

the primary data collected in this work } <strong>for</strong>est tree<br />

diameter distributions and resulting leaf biomass density<br />

<strong>estimates</strong>.


3432 C. Wiedinmyer et al. / Atmospheric Environment 34 (2000) 3419}3435<br />

Fig. 7. <strong>Biogenic</strong> VOC <strong>emission</strong>s <strong>for</strong> north central Texas on 7/3/96, calculated with the BIOME model, using the new land use and<br />

vegetation data. Units are tons/day of Carbon Bond IV <strong>hydrocarbon</strong>s. Grids are 4�4 km (16 km�). Note that the colors are scaled to<br />

match those in Fig. 6, which contains 16�16 km grid cells. Each color represents the same range of <strong>emission</strong> densities in both tileplots.<br />

The various methods used to determine biomass<br />

densities and species distributions <strong>for</strong> each vegetation<br />

classi"cation involved a series of assumptions and approximations.<br />

The transect location selection and the<br />

averaging techniques used to assign species distributions<br />

and biomass densities to each land use classi"cation<br />

are subject to sample bias and uncertainty. Because<br />

these steps were per<strong>for</strong>med with qualitative reasoning,


Table 9<br />

<strong>Biogenic</strong> (Carbon Bond IV) <strong>hydrocarbon</strong> <strong>emission</strong>s <strong>for</strong> primary episode days, as calculated by BIOME using new <strong>North</strong> <strong>Central</strong> Texas<br />

biogenics database and BEIS2 <strong>emission</strong> factors<br />

CB-IV <strong>emission</strong>s by county (t/d)<br />

Date Dallas Tarrant Denton Collin Total DFW Core domain total<br />

6/21/95 54 85 136 19 294 6643<br />

6/22/95 58 93 152 22 325 7288<br />

7/3/96 80 128 213 31 453 9635<br />

statistically based uncertainties are di$cult to produce.<br />

Nevertheless, a preliminary assessment of uncertainties<br />

in the <strong>emission</strong> <strong>estimates</strong> can be per<strong>for</strong>med.<br />

To assess the e!ects of the choice of transect location<br />

and averaging techniques used to assign species distributions<br />

and biomass densities <strong>for</strong> the classi"cations, lower<br />

and upper bounds of total biomass densities were calculated<br />

<strong>for</strong> two important vegetation classi"cations in<br />

the domain. Two transects were surveyed in the Post<br />

Oak Woods, Forests and Grasslands classi"cation. The<br />

biomass densities and species distributions from each of<br />

the two transects were averaged and used to calculate the<br />

biogenic <strong>emission</strong>s <strong>estimates</strong> shown in Fig. 7. One transect<br />

had a much lower total biomass density than the<br />

other: 106 versus 206 g m��, with Oak biomass densities<br />

of 72 and 196 g m��, respectively. This variability was<br />

intended, since the two transects combined were expected<br />

to accurately describe the land use classi"cation. The<br />

variation of Oak and total biomass densities between the<br />

two surveys suggest an uncertainly of a factor of approximately<br />

2.<br />

Six transects were per<strong>for</strong>med at locations designated<br />

by the Post Oak Woods and Forest classi"cation. These<br />

transects had varying species and biomass densities<br />

which, when combined, were assumed to be representative<br />

of the Post Oak Woods and Forest classi"cation.<br />

Of these six transects, the total biomass calculated<br />

from the data of each ranged from 143 to 420 g m��. The<br />

percentage of Oaks in the six transects ranged from 35 to<br />

100%. This suggests an uncertainty of a factor of 3 <strong>for</strong><br />

total and Oak biomass densities within each landuse<br />

category.<br />

This method of determining upper and lower bounds<br />

of the <strong>emission</strong> <strong>estimates</strong> <strong>for</strong> vegetation classi"cations<br />

may not produce the true extremes. Although the lowest<br />

and highest total biomass values were chosen to determine<br />

bounds on the <strong>estimates</strong>, this may not lead to true<br />

<strong>emission</strong> bounds; nevertheless these bounds give an indication<br />

of the potential magnitude of the uncertainties,<br />

especially at small scales associated with <strong>for</strong>est canopy<br />

cover.<br />

C. Wiedinmyer et al. / Atmospheric Environment 34 (2000) 3419}3435 3433<br />

4. Conclusions<br />

A GIS land cover/ land use database <strong>for</strong> a 37 county<br />

domain in <strong>North</strong> <strong>Central</strong> Texas was developed. This<br />

database includes vegetation species distributions and<br />

biomass densities, speci"c to the vegetation in Texas. The<br />

methods employed in this study may be applied to other<br />

areas within Texas. Similar studies could also be conducted<br />

outside the state but will require the use of a rural<br />

vegetation database other than that provided by the<br />

Texas Parks and Wildlife Department. The USGS LCC<br />

data could be used <strong>for</strong> this purpose, but will require more<br />

extensive ground surveying.<br />

The vegetation database created in this work is an<br />

improvement over the data reported in the BELD <strong>for</strong> the<br />

<strong>North</strong> <strong>Central</strong> Texas region. This new database reports<br />

greater overall biomass densities and higher densities of<br />

oak and other emitting tree species that the BELD does<br />

not report. This improved biomass characterization can<br />

lead to an improvement in the biogenic <strong>emission</strong>s <strong>estimates</strong><br />

<strong>for</strong> the region.<br />

A new biogenic <strong>emission</strong>s inventory was created <strong>for</strong><br />

a modeling episode in 1996 using this in<strong>for</strong>mation. The<br />

<strong>emission</strong>s inventory <strong>for</strong> biogenic sources <strong>for</strong> use in<br />

photochemical models of the <strong>North</strong> <strong>Central</strong> Texas region<br />

is more speci"c and accurate than any data previously<br />

available. Total <strong>emission</strong>s <strong>for</strong> the region were estimated<br />

to increase over 100% from BEIS2/BELD <strong>estimates</strong> <strong>for</strong><br />

the same region. The spatial variations in the <strong>emission</strong>s<br />

inventory also changes with the use of the new landuse<br />

database. New `hot-spotsa of biogenic <strong>emission</strong>s in the<br />

domain were located.<br />

A detailed assessment of the uncertainties of the<br />

methods used to determine the vegetation species distributions<br />

and biomass densities should be per<strong>for</strong>med. The<br />

choice of location, direction and averaging of the transects<br />

can greatly a!ect the biogenic <strong>emission</strong>s <strong>estimates</strong>.<br />

A qualitative analysis of the uncertainty in transect<br />

selection and averaging suggests that an uncertainty of a<br />

factor of 2}3 can be applied to the biomass densities<br />

assigned to each land use classi"cation. But, the di!erences


3434 C. Wiedinmyer et al. / Atmospheric Environment 34 (2000) 3419}3435<br />

in <strong>for</strong>est and oak coverage between the BELD3 and the<br />

new land cover data are greater than this factor in many<br />

areas of the domain. In addition, <strong>emission</strong> factor and<br />

foliar mass data should be collected <strong>for</strong> tree species in the<br />

region to verify the values used in this study.<br />

Acknowledgements<br />

The authors thank Alex Guenther, Lee Klinger, Bill<br />

Baugh and Chris Geron <strong>for</strong> useful discussions and practical<br />

advice on conducting "eld surveys. The authors also<br />

thank Pryanka Bandypadha, Patrick Gri$th, Doug<br />

Goldman and Todd Barkman, who participated in the<br />

"eld surveys. Dr. David Maidment provided valuable<br />

expertise with the Geographical In<strong>for</strong>mation Systems.<br />

David Jacob from the Texas Natural Resource Conservation<br />

Commission assisted in the modeling. The authors<br />

thank the anonymous reviewers of this manuscript <strong>for</strong><br />

their advice and helpful comments. This work was supported<br />

by contract �UTA97-0302 with the Texas Natural<br />

Resource Conservation Commission.<br />

References<br />

Benjamin, M., Winer, A., 1998. Estimating the ozone-<strong>for</strong>ming<br />

potential of urban trees and shrubs. Atmospheric Environment<br />

32 (1), 53}68.<br />

Benjamin, M., Sudol, M., Vorsatz, D., Winer, A., 1997. A spatially<br />

and temporally resolved biogenic <strong>hydrocarbon</strong> <strong>emission</strong>s<br />

inventory <strong>for</strong> the Cali<strong>for</strong>nia South Coast Air Basin.<br />

Atmospheric Environment 31 (18), 3087}3100.<br />

EC/R Incorporated, 1995. Urban Airshed Model (UAM) <strong>Biogenic</strong><br />

Emission Inventory System Version 2 (BEIS-2) User's<br />

Guide. EPA Contract No. 68-D3-0034; prepared <strong>for</strong> Richard<br />

A. Wayland, Air Quality Modeling Group, US EPA, September<br />

28, 1995.<br />

Estes, M., Jacob, D., Jarvie, J,. MacKay, J., Smith J., 1997.<br />

<strong>Biogenic</strong> <strong>emission</strong>s modeling <strong>for</strong> Southeastern Texas. Proceedings<br />

of a Specialty Conference Cosponsored by the Air<br />

and Waste Management Association and the U.S.EPA, New<br />

Orleans, Louisiana.<br />

Geron, C.D., Pierce, T.E., Guenther, A.B., 1995. Reassessment of<br />

biogenic volatile organic compound <strong>emission</strong>s in the Atlanta<br />

area. Atmospheric Environment 29 (13), 1569}1578.<br />

Geron, C.D., Guenther, A.B., Pierce, T.E., 1994. An improved<br />

model <strong>for</strong> estimating <strong>emission</strong>s of volatile organic compounds<br />

from <strong>for</strong>ests in the eastern United States. Journal of<br />

Geophysical Research 99, 12773}12791.<br />

Guenther, A., Hewitt, N., Erickson, D., Fall, R., Geron, C.,<br />

Graedel, T., Harley, P., Klinger, L., Lerdau, M., McKay, W.,<br />

Pierce, T., Scholes, B., Steinbrecher, R., Tallamraju, T.,<br />

Taylor, J., Zimmerman, P., 1995. A global model of natural<br />

volatile organic compound <strong>emission</strong>s. Journal of Geophysical<br />

Research 100 (D5), 8873}8892.<br />

Guenther, A.B., Zimmerman, P., Wildermuth, M., 1994. Natural<br />

volatile organic compound <strong>emission</strong> rate <strong>estimates</strong> <strong>for</strong> US.<br />

woodland landscapes. Atmospheric Environment 28,<br />

1197}1210.<br />

Guenther, A.B., Zimmerman, P.R., Harley, P.C., Monson, R.,<br />

Fall, R., 1993. Isoprene and monoterpene <strong>emission</strong> rate<br />

variability: model evaluations and sensitivity analyses. Journal<br />

of Geophysical Research 98, 12609}12617.<br />

Iqbal, M., 1983. An Introduction to Solar Radiation. Academic<br />

Press, San Diego, Cali<strong>for</strong>nia, pp. 107}277.<br />

Kinnee, E., Geron, C., Pierce, T., 1997. United States land use<br />

inventory <strong>for</strong> estimating biogenic ozone precursor <strong>emission</strong>s.<br />

Ecological Applications 7 (1), 46}58.<br />

Klinger, L., 1997. Personal communication.<br />

Lamb, B., Pierce, T., Baldocchi, D., Allwine, E., Dilts, S., Westberg,<br />

H., Geron, C., Guenther, A., Klinger, L., Harley, P.,<br />

Zimmerman, P., 1996. Evaluation of <strong>for</strong>est canopy models<br />

<strong>for</strong> estimating isoprene <strong>emission</strong>s. Journal of Geophysical<br />

Research 101, 22787}22798.<br />

Lamb, B., Gay, D., Westberg, H., Pierce, T., 1993. A biogenic<br />

<strong>hydrocarbon</strong> <strong>emission</strong> inventory <strong>for</strong> the USA using<br />

a simple <strong>for</strong>est canopy model. Atmospheric Environment<br />

27A, 1}18.<br />

Mahrer, Piekle, 1978. A test of an upstream spline interpolation<br />

technique <strong>for</strong> the advective terms in a numerical mesoscale<br />

model. month Weath Review 106, 818}830.<br />

Mahrer, Piekle, 1977. A numerical study of the air #ow<br />

over irregular terrain. Control of Atmospheric Physics 50:<br />

98}113.<br />

McNider, R.T., Song, J.A., Kidder, S.Q., 1995. Assimilation of<br />

GOES-derived solar insolation into a mesoscale model <strong>for</strong><br />

studies of cloud shading e!ects. International Journal of<br />

Remote Sensing 16, 2207}2231.<br />

Miller, P.E., Winer, A.M., 1984. Composition and Dominance in<br />

Los Angeles basin urban vegetation. Urban Ecology 8, 29}54.<br />

Pierce, T., Waldru!, P., 1991. PC-BEIS: a personal computer<br />

version of the biogenics <strong>emission</strong> inventory system.<br />

Journal of Air and Waste Management Association 41,<br />

937}941.<br />

Radian, 1996. <strong>Biogenic</strong> Sources Preferred Methods, Final Report.<br />

Prepared <strong>for</strong> Area Sources Committee of the Emissions<br />

Inventory Improvement Program, May, 1996.<br />

Radian and Valley Research Corporation, 1994. Final Report,<br />

<strong>Biogenic</strong>s Emissions Factors Project, Texas Natural Resource<br />

Conservation Commission. Prepared by Radian Corporation<br />

and Valley Research Corporation; prepared <strong>for</strong><br />

TNRCC, March 7, 1994.<br />

Strange, I.W., 1998. M.S. Thesis. University of Texas at Austin,<br />

Austin, Texas.<br />

Texas Parks and Wildlife Department, 1996. The Vegetation<br />

Types of Texas; Introduction. Available on the Internet at<br />

http://www.tpwd.state.tx.us/admin/veg/intro.html.<br />

U.S. Department of Agriculture, NASS data, 1997. Data from<br />

the USDA National Agriculture Statistics Service. Available<br />

on the internet at `http://www.usda.gov/nass/a<br />

U.S. EPA, 9/3/98. National Air pollutant Emissions Trends,<br />

1990}1998. Available on the internet at http://www.epa.gov/<br />

oar/emtrnd/.<br />

U.S. Geological Survey, ca. 1976. Land Use/Land Cover<br />

databases, available on the internet at `http://edcwww.cr.<br />

usgs.gov/doc/edchome/ndcdb/ndcdb.htmla.<br />

Vogel, C., Baldocchi, D., Luhar, A., Rao, K.S., 1995. A comparison<br />

of a hierarchy of models <strong>for</strong> determining energy balance


components over vegetation canopies. Journal of Applied<br />

Meteorology 34, 2182}2196.<br />

Wilkinson, J.G., Emigh, R.A., Pierce T.E., 1996. An intercomparison<br />

of biogenic <strong>emission</strong>s <strong>estimates</strong> from BEIS2 and<br />

BIOME: Reconciling the di!erences. The Emission Inventory:<br />

Key to Planning, Permits, Compliance, and Reporting,<br />

Proceedings of Air and Waste Management Specialty Conference,<br />

September 4}6, 1996, New Orleans, Louisiana, pp.<br />

985}1009.<br />

C. Wiedinmyer et al. / Atmospheric Environment 34 (2000) 3419}3435 3435<br />

Wilkinson, J.G., Emigh, R.A., 1995. The Emissions Modeling<br />

System (EMS-95) User's Guide prepared by Alpine Geophysics<br />

<strong>for</strong> the Texas Natural Resource Conservation Commission.<br />

March 13, 1995.<br />

Yarwood, G., Allen, D., Guenther, A., Wiedinmyer, C., Baugh,<br />

B., Quigley, C., Strange, W., 1997. Leaf Biomass Density<br />

Data <strong>for</strong> <strong>North</strong> <strong>Central</strong> Texas. Prepared <strong>for</strong> the Texas<br />

Natural Resource Conservation Commission. October 15,<br />

1997.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!