This file was created by scanning the printed publication.
Errors identified by the software have been corrected;
however, some errors may remain.
Forest Cover and Natural Volatile Organic
Compound Emissions in North America 1
Christopher D. Geron 2
Abstract-Forest inventory data is important in deriving emission
estimates of biogenic volatile organic compound (BVOC) at hourly
to annual temporal and tens of square meter to global spatial
resolutions. We discuss methods used to adapt remotely sensed data
and forest inventories to BVOC emission estimation. Databases
employed include USDA Forest Service Forest Inventory and Analysis (FIA) and Canadian Ministry of Forests (British Columbia) data,
which we use to estimate canopy coverage at species level resolution. The plot level data is also used to speciate 1.1 kilometer gridded
remotely sensed classifications of vegetation cover, foliar mass, and
leaf area. Developing ecosystem-level emission rates for vegetation
categories in existing remotely sensed databases is also discussed.
We compare resulting emission and canopy cover estimates from
the different approaches at county levels. Due to assumptions made
ofthe composition of the forest cover-types, emission estimates can
vary by more than an order of magnitude for the different approaches. We discuss techniques to combine temporal and biophysical measures from remote sensing data with vegetation species
information from the survey data. Potential improvements to forest
inventories for these and similar applications relating to air pollution exposure are discussed.
Tropospheric or low level ozone (0 3 ) is a major constituent of smog and is responsible for billions of dollars in crop
and forest loss and respiratory health effects each year. To
aid in understanding and controlling this problem, the
North American Research Strategy for Tropospheric Ozone
(NARSTO) mandates that emission models and inventories
of tropospheric ozone precursors including volatile organic
compounds (VOC), the oxides of nitrogen (NOx ), and carbon
monoxide (CO) be developed for Canada, Mexico, and the
U.S.A. Sources include mobile (e.g. automobiles, aircraft),
point (e.g. fossil fuel powered utility plants), area (agricultural or small dispersed industrial sources such as refinishing, refueling, and printing operations), and biogenic or
natural sources. The largest source ofVOC in North America
is natural or biogenic in origin. In the U.S. biogenic VOC
(BVOC) emissions exceed those from all other sources combined. Emissions from forests are estimated to account for at
least 90% of these BVOC emissions, with crops and rangeland vegetation accounting for the balance (Lamb et al.
1993). Annual estimates of anthropogenic VOC for the U.S.
are approximately 20 tg yr-1, while corresponding estimates
ofBVOC are approximately 40 tg yr-1. Furthermore, BVOC
Ipaper presented at the North American Science Symposium: Toward a
Unified Framework for Inventorying and Monitoring Forest Ecosystem
Resources, Guadalajara, Mexico, November 1-6,1998.
2Christopher D. Geron is with the National Risk Management Research
Laboratory, U.S. Environmental Protection Agency, Research Triangle Park,
North Carolina 27711, USA. e-mail: cgeron@engineer.aeerl.epa.gov
126
tend to be more reactive than most anthropogenic VOC
and are therefore more important on a per molecule basis in
0 3 production. Here I present a brief overview on the use of
forest inventory and remote sensing classifications for use in
estimating BVOC emissions in North America.
Methods
Thelandcover characteristics data used to estimate BVOC
emissions in North America are constructed from several
independent databases. Forest density in the United States
(fraction of forest cover per 1.1 km grid cell) is taken from the
database of Zhu and Evans (1994). Since it is based on
optical measurements from satellite at two different (30 m
and 1.1 km) scales, this dataset provides estimates offorest
canopy fractions for areas usually not covered by other
sources, such as small woodlots, riparian zones, forested
urban regions, and the semi-arid western woodlands with
less dense canopy cover. The percentage of tree species
present within each grid cell is determined from the forest
survey data described in Geron et al. (1994). Additional data
for forests in the eleven westernmost conterminous United
States and Alaska, including those for the National Forests
and Parks, have also been compiled and added to this data.
This information yields crown cover percentages by species
level (compared to genus level coverage of Geron et al. 1994)
to be consistent with species level emission data discussed
later in this paper. Desert, rangeland, grassland, shrubland,
and agricultural areas by crop type, are allocated to the
remaining portions of the grid cells from the Agricultural
Census and the North American Land Cover Classification
(NALCC). Growing season peak foliar biomass densities are
derived from the methods of Geron et al. (1994). However,
for seasonal simulations, biomass density is adjusted by
multiplying the peak density by the ratio of NDVI for the
time and domain in question to NDVI for peak density.
Similar procedures are used to construct vegetation cover
databases from 1.1 km and Canadian forest inventory data,
although at this point no corresponding remotely sensed
forest density data exists for Canada. Therefore, total forest
area is restricted to the forest area estimated from the
Canadian forest inventory and the NALCC. Mexican land
and vegetation cover is also described using the NALCC.
Efforts are also underway to assess species composition for
Mexican forests. Emission factors are formulated on a leaf
biomass basis for individual tree species in mg-carbon g-1
(foliage dry weight) hr-1 and on an areal basis for mixed
vegetation categories (e.g. western conifers) in mg-carbon m-2
(land surface area) hr-1. In-depth detail on these approaches
can be found in Geron et al. (1994), Guenther (1997),
Guenther et al. (1994), Guenther et al. (1995), and Kinnee
et al. (1997).
USDA Forest Service Proceedings RMRS-P-12. 1999
Results and Discussion
In the U.S. forest inventory database, there are currently
over 450,000 plots (see approximate locations in Fig. 1) and
more than 4.5 million tree records. Although there are over
320 tree species in the FIA databases of the U.S., a large
proportion of estimated BVOC emissions can be accounted
for by relatively small number of species. This is illustrated
in Table 1 where tree species of extensive range, abundance,
and high emission factors dominate the relative BVOC
emission potential, which is a function ofthese factors. In the
eastern U.S., BVOC emissions are controlled by Quercus,
Populus, Pinus, and Picea species, which account for an
estimated 83% of hourly BVOC emissions under standard
conditions of PAR = 1000 mmol m-2 S-l (approximately half
of full sunlight) and leaf temperature = 30°C. Pinus, Picea,
Pseudotsuga, Abies, Quercus, andPopulus species dominate
BVOC emission pot en tial (95%) in the Western U. S . (West of
102° Longitude). Less than 2 percent of the estimated forest
canopy area is composed of species for which no current
BVOC emission information exists. Areas with highest
emissions are those with high components of conifers,
(which are high monoterpene and methyl butenol emitters)
Quercus (oaks), Picea (spruces) and Populus species (which
are high isoprene emitters). Estimated geographical distribution ofBVOC emissions in North America under standard
conditions of PAR = 1000 mmol m- 2 S-l and leaf temperature = 30°C is illustrated in Figure 2. Due to current lack of
forest inventory data for Mexican and most of Canadian
forests in the landcover characteristics database, emission
uncertainties in these areas are roughly an order ofmagnitude. Uncertainties in the U.S. are much lowerCwithin ± 50%
for isoprene ± 100% for other compounds), largely due to use
of the detailed forest inventory.
Substituting the 1.1 km forest density data and the western U.S. forest inventory (used in the BEIS3 model) in place
of the Biogenic Emissions Landuse Database (BELD) of
Kinnee at al. (1997) used in BEIS2 (version 2 ofthe Biogenic
Emission Inventory System) yields significant changes in
many regions of the western United States. Forest genus
classification in the BELD for the western U.S. was estimated by assigning fixed crown cover percentages to species
listed in each cover type (Kinnee et al. 1997). In western
Washington, Lamb et al. (In Prep.) found that this resulted
• Plot location
SO'IlJ'Ce: Forest hwerdory and Analysjs.
EastwideJV'V EStY4de Fores[ InVEfLtory Dabbase
British C o11J1'1lbia Mmis1ry of ForestS
Figure 1.-Approximate locations of forest inventory survey plots compiled to date used to estimate BVOC emissions in North America.
USDA Forest Service Proceedings RMRS-P-12. 1999
127
Table 1.-Crown cover and emission potentials (under standard conditions) by forest tree genera and region within the U.S. Area is
horizontal projected crown area in 106 hectares. %Crn is proportion of total crown area. REP is relative emission potential, which
is equal to the product of the basal total BVOC emission rate, biomass density, crown area, and a canopy adjustment factor (to
account for shading within the canopy) for each genus relative to the total REP for the region. Cum is cumulative REP for the
region.
Genus
Quercus
Populus
Pinus
Picea
Liquidambar
Abies
Acer
Nyssa
Thuja
Carya
Area
32.1
4.18
20.4
4.18
3.78
2.08
12.6
2.56
1.31
6.17
Eastern U.S.
%Crn
0.248
0.032
0.158
0.032
0.029
0.016
0.097
0.020
0.010
0.048
REP
Cum
0.499
0.141
0.094
0.093
0.059
0.023
0.019
0.010
0.009
0.007
0.499
0.640
0.733
0.826
0.886
0.908
0.927
0.937
0.946
0.953
Western U.S. (including Alaska)
%Crn
Area
REP
Genus
Pinus
Picea
Pseudotsuga
Abies
Quercus
Populus
Tsuga
Juniperus
Thuja
Alnus
10.1
12.7
8.41
5.60
3.15
2.17
2.43
3.78
0.98
1.32
0.184
0.231
0.153
0.102
0.057
0.039
0.044
0.069
0.018
0.024
0.278
0.234
0.138
0.136
0.104
0.062
0.013
0.012
0.011
0.003
Cum
0.278
0.512
0.651
0.787
0.890
0.953
0.966
0.977
0.988
0.991
,
-->.
.~~~;~~~~;£~ ~
....
Total Em:issians (mg C/m2fh)
00
1
Do-!
111-2
.2-4
.4-8
.8-10
I
Figure 2.-Hourly estimated BVOC emissions from North American Landscape types un_H
standard conditions of PAR=1 000 mmol m-2 S-l (approximately half of full sunlight) and leaf
temperature= 30°C.
128
USDA Forest Service Proceedings RMRS-P-12. 1999
in isoprene emissions which were 5 to 10 times higher than
measured mixed layer fluxes. The BEIS3 landcover data
yielded isoprene emissions that were within a factor of
two ofthe flux estimates, even though total canopy coverage
for the state of Washington was only 8% less than that from
BELD. Canopy coverage of high isoprene emitters in the
BEIS3 (using the western U.S. forest inventory data) database was only 1-2% for the region on average. Conversely,
BEIS3 isoprene estimates for other western regions may be
considerably higher than that yielded by BELD. These
include portions of California, Colorado, and Utah.
Forest density estimates in the agricultural midwestern
states is also somewhat higher than that from the BELD,
since the Zhu and Evans (1994) data yield estimates of tree
cover on small woodlots, transportation and stream corridors, and other noncommercial forest lands. This may result
in isoprene emissions being from 10 to 50% higher than
BEIS2 emissions. Forest density, and therefore emissions,
in the remainder of the eastern U.S. remains largely unchanged due to changes in the landuse databases, since the
forest extent estimated from the Forest Density data is
similar to that estimated from the forest inventory data
compiled thus far.
Our current plot coverage of forest inventory data appears to be fairly thorough in terms of regional coverage of
major forest types. However, coverage of commercially less
valuable forest ecosystems can be sparse. For instance in the
semi-arid forest scrubland of central Texas and Oklahoma,
little forest survey data currently exists. This introduces
considerable uncertainty (roughly an order of magnitude)
in BVOC emission estimates for landscapes in this region.
Vegetation coverage in desert-chapparral, riparian, and
urban environments is also sometimes sparse or lacking.
Using remotely sensed indices allows us to estimate greenness of vegetation coverage in these areas, but supporting
plot level (ground based) measurements of species abundance and canopy characteristics would improve trace gas
flux estimates from these systems considerably. This would
also aid researchers in determining which vegetation
species to examine for trace gas emissions.
Recommendations
Caution must be exercised when using generalized forestl
landcover classes for estimating BVOC emissions. Since
individual tree species within a forest ecotype can have
drastically different BVOC emission factors (qualitatively
and quantitatively), we recommend that forest inventory
data be used to determine amount of crown cover by species
within areas where these emissions may be important.
USDA Forest Service Proceedings RMRS-P-12. 1999
However, current forest inventories often stress areas of
commercial importance, and neglect areas where timber
harvesting may not be economical or feasible. These areas
are often very important for trace gas exchange, and include
urban forests, wetland forests and riparian zones, and
marginally productive (but extensive) semi-arid forests
such as those in Central Oklahoma, Texas (see Fig. 1),
much of Mexico and portions of Canada. Finally, most
parameters relating to foliage amount must be estimated
from attributes such as tree diameter, height, and commercial or wood volumetric indices of stocking. If measures
of canopy closure from aerial photos (such as that included in the British Columbia inventory) or ground measurement of intercepted solar radiation could be associated
with each forest inventory plot, then regional models and
inventories of forest trace gas exchange would be greatly
improved.
Acknowledgments
The authors wish to thank the USDA Forest Service
FIA units for their assistance in obtaining and interpreting
the Forest Inventory and Analysis Data. Richard Woods of
the British Columbia Ministry of Forest Resources was
especially helpful in compiling and interpreting the British
Columbia Forest Inventory data. Cliff Stanley and Ellen
Kinnee of DynTel Inc. was provided invaluable database
management in this effort. Pat Meredith provided expert
editorial support.
Literature Cited
Geron, C. D., A B. Guenther, and T. E. Pierce, An improved model
for estimating emissions of volatile organic compounds from
forests in the eastern United States. J. Geophys. Res. 99, 12,77312,791, 1994.
Guenther, AB., C. N. Hewitt, D. Erickson, R. Fall, C. D. Geron, T.
Graedel, P. Harley, L. Klinger, M. Lerdau, W. A McKay, T. E.
Pierce, B. Scholes, R. Steinbrecher, R. Tallamraju, J. Taylor, and
P. R. Zimmerman, A global model of natural volatile organic
compound emissions. J. Geophys. Res. 100:8873-8892, 1995.
Guenther, AB., P.R. Zimmerman, and M. Wildermuth, Biogenic
volatile organic compound emission rate estimates for U.S.
woodland landscapes. Atmos. Environ. 28, 1197-1210, 1994.
Lamb, B., D. Gay, H. Westberg, and T. Pierce, A biogenic hydrocarbon emission inventory for the U.S. using a simple forest
canopy model. Atmos. Environ. 27A: 1673-1690,1993.
USEPA, National Air Pollutant Emission Trends, 1900-1994, Office
of Air Quality Planning and Standards, EPA-4541R-95-011, Research Triangle Park, NC 27711, October, 1995.
Zhu, Z. and D.L. Evans, U.S. Forest types and predicted percent
forest cover from AVHRR data. Photogram. Eng. and Rem. Sens.
60:525-531, 1994.
129