P1.3
INTERCOMPARISON OF ALTERNATIVE VEGETATION DATABASES
FOR REGIONAL AIR QUALITY MODELING
Thomas Pierce*1 and Jonathan Pleim1
NOAA/ARL/Atmospheric Modeling Division, Research Triangle Park, North Carolina
Ellen Kinnee and Lara Joyce
DynCorp Inc., Research Triangle Park, North Carolina
1.	INTRODUCTION
Vegetation cover data are used to characterize
many air quality model processes, including the
calculation of heat, moisture, and momentum fluxes with
the Penn State/NCAR Mesoscale Meteorological Model
(MM5) and the estimate of biogenic emissions with the
Biogenic Emissions Inventory System (BEIS) (Byun and
Ching, 1999). This paper compares vegetation cover
and isoprene emissions estimated from three
contemporary databases: (1) the North American Land
Cover Characteristics (NALCC) version 2 database, (2)
the Biogenic Emissions Landcover Database (BELD3),
and (3) the National Land Cover Database (NLCD).
Each of these databases may be suitable for use with
regional-scale air quality models.
2.	DESCRIPTION AND ANALYSIS
The NALCC database, which is released by USGS
and supported by the National Center for Atmospheric
Research (NCAR) for use with MM5, consists of 1-km
resolved land cover classes derived from Advanced
Very-High Resolution Radiometer (AVHRR) satellite
data (USGS, 2001). BELD3 provides vegetation data to
the Biogenic Emissions Inventory System (EPA, 2001).
It combines the NALCC data with U.S. Forest Service
and U.S. Department of Agriculture databases so that
tree and crop cover (by species) are resolved to 1-km.
The NLCD was released by the Multi-Resolution Land
Characteristics Consortium (USGS, 2002). It is based
on Landsat-TM data and available at ~30-m resolution.
These databases were used to examine vegetation
cover distributions across the contiguous United States
and a 10s km2 domain centered over Nashville,
Tennessee. To enable comparisons of vegetation cover
and because of differences in vegetation classification
systems, vegetation classes from each database were
aggregated into seven broad categories: forests,
agriculture, urban, grass/shrubland, water, wetlands,
and barren.
The fractional cover of each of the seven categories
is shown in Figure 1. For the Nashville domain, the
focus of the remainder of this paper, agricultural area
estimates range from 24 to 39%, while forested area
estimates range from 52 to 72%.
* Corresponding author address: Thomas Pierce, EPA,
E243-04, RTP, NC 27711; email: pierce.tom@epa.gov.
1On assignment to the U.S. Environmental Protection
Agency, National Exposure Research Laboratory.
For the Nashville domain, the commonly-used
NALCC database indicates more forest area and less
crop area than the other two databases. This is
somewhat surprising because the NALCC
woodland/cropland class was mapped into the
agriculture category. Overall, forest and crop cover
percentages agree reasonably well between NLCD and
BELD3.
o Grass/shrubland
Agriculture
Wetlands
¦5	Water
E
o 	
T3
a Grass/shrubland
JZ
If)
rc	Urban
Agriculture
Total forest
Total forest
¦ NLCD
~	BELD3
~	NALCC
0.2 0.3 0.4 0.5 0.6
Fractional area
Figure 1. Fractional vegetation areas for the contiguous
U.S. and for a model grid centered over Nashville, TN
3. INFLUENCE ON ISOPRENE EMISSSIONS
It is hypothesized that differences among the three
vegetation databases will influence emission estimates
of isoprene, an important hydrocarbon for tropospheric
chemistry. Isoprene is emitted primarily from deciduous
trees, especially oaks. Using emission factors from
BEIS3 (Pierce et al., this issue), normalized isoprene
emissions were computed for each 1 km grid cell within
a 10s km2 domain centered over Nashville, Tennessee.
For the BELD3 and NLCD datasets, isoprene emission
factors were multiplied by the areas of individual tree

-------
species as derived from county-level tree species
distributions from the US Forest Service and the
forested fractions indicated with either the BELD3 or
NLCD datasets. For the NALCC dataset, the BEIS3
isoprene emission factor was related to the USGS
classification assigned to each 1 km grid cell.
Isoprene emissions computed along a transect
extending from south-central Missouri to western North
Carolina are shown in Figure 2. In general, estimates
obtained with the BELD3 and NLCD databases closely
agree (avg= 6349 vs. 6387, r2= 0.62). Although the
mean NALCC value is only slightly higher than these
two values (avg=6517), the NALCC-derived estimates
are not as closely correlated with the NLCD as BELD3
(r2= 0.24). The lack of agreement is particularly
noticeable in two areas: the Ozark Plateau (x=0-175 km)
and east of Nashville (x=540-640 km). As shown in
Figure 2, the isoprene emission estimates resulting from
NALCC are at least 20% lower than the BELD3 and
NLCD datasets in the Ozark Plateau, while they are
-100% larger along the 100-km transect east of
Nashville.
simulating other meteorologically-related processes that
depend on the characterization of vegetation data.
5.	REFERENCES
Byun, D. and J. Ching (eds.), Science Algorithms of the
EPA Modeis-3 Community Muitiscaie Air Quality
(CMAQ) Modeling System, U.S. Environmental
Protection Agency, 600/R-99/030, 1999.
U.S. Geological Survey (2001) North American Land
Characteristics Database. Available online:
http://edcdaac.usas.gov/alcc/na int.html [July 16,
2001],
	(2002) National Land Cover Database. Available
online: http://landcover.usas.aov/natllandcover.html
[February 7, 2002],
U.S. Environmental Protection Agency (2001) Biogenic
Emissions Landcover Database. Available online:
http://www.epa.aov/asmdnerl/bioaen.html
[December 5, 2001],
6.	DISCLAIMER AND ACKNOWLEDGEMENTS
4. FINAL THOUGHTS
The relative distribution of forest and agriculture
cover contained in the popular NALCC database is
shown here to differ from two other contemporary
databases for the mixed agricultural/forested region of
the Tennessee Valley. For estimating biogenic
emissions, it is shown that isoprene emissions can vary
by a factor of two depending on the source of vegetation
data. We therefore urge caution if using broadly-defined
vegetation classes (like those found in the NALCC data)
to derive biogenic emissions, and we recommend that
other databases, such as the NLCD, coupled with tree
species distribution information be considered for
This paper has been reviewed in accordance with
the U.S. Environmental Protection Agency's peer and
administrative review policies. Mention of products or
trade names does not constitute endorsement or
recommendation of their use.
The authors appreciate the assistance of Chris
Geron (EPA-NRMRL), Ross Lanetta (EPA-NERL-LCB),
and Tanya Otte (NOAA-ARL-ASMD)
FOR FURTHER INFORMATION
http://www.epa.gov/asmdnerl/biogen.html
Figure 2. Comparison of normalized BEIS3 isoprene emissions
derived from three vegetation databases
Cross-section: (36.6N.92.7W) to (35.7N.81.4W)

21000

18000
-C
15000
(M

L

CD
12000
3

0)
c
9000
(D

Q.

O
(/)
6000

3000
•NALCC
¦NLCD 	BELD3
0 100 200 300 400 500 600 700 800 900 1000
Distance (km)

-------