Terrestrial Biomass and the Effects of Deforestation on the Global Carbon Cycle Results from a model of primary production using satellite observations Christopher S. Potter nvironmental policymakers, particularly those who consider options for mitigating global warming, require accurate assessments of sources and sinks for atmospheric carbon dioxide. A case in point is the Kyoto Protocol of 1997, the first negotiated agreement for reducing net terrestrial emissions of carbon dioxide on an international scale. Emissions may be reduced by curbing air pollution sources, slowing "slash and burn" deforestation, and promoting regrowth of forest areas that have been logged. The goal of the Kyoto Protocol is to restrict average annual carbon emissions to a percentage of 1990 emissions. Successful implementation of the Kyoto Protocol depends on a reliable accounting of how much carbon is stored and released from the world's forests. Several major uncertainties have hindered the development of a more accurate carbon budget; these uncertainties have to do with global carbon cycling in terrestrial ecosystems, carbon storage in standing plant biomass, and the net effects of forest loss and secondary regrowth (Schimel et al. 1996a). During the 1980s, terrestrial exchange of carbon dioxide with the atmosphere may have accounted for carbon sources of 0.6-2.6 Pg/yr, mainly from land-use change in the tropics (IPCC 1994). The wide range in these Uncertainties in implementing international agreements for greenhouse gas emissions can be addressed with better quantified values for forest biomass and regional variability in terrestrial productivity source estimates from the Intergovernmental Panel on Climate Change, which exemplifies the difficulty in implementing agreements like the Kyoto Protocol, is due in large part to poorly quantified values for forest biomass and to variability in terrestrial productivity over scales that are relatively small (100-200 km) compared to global forest coverage. Uncertainties in the global carbon budget would be reduced to more acceptable levels with the development of improved techniques for estimating variability in carbon stocks over vast forested areas (IGBP 1998). One such technique is the use of ecosystem modeling of carbon pools S. Potter (e-mail: cpotter@ Christopher fluxes. and Global ecosystem modmail.arc.nasa.gov)is a researchscientistat NASA Ames Research Center, Ecosystem els are valuable tools in situations in Science and Technology Branch, Moffett which ground-based measurements of carbon pools are not adequate to Field, CA 94035. October 1999 realistically capture regional variability. A computer model of this type based on satellite measurements has been developed to simulate ecosystem carbon cycling (Potter and Klooster 1997, 1998). This model, the NASA-CASA (National Aeronautics and Space AdministrationCarnegie Ames Stanford Approach) model, is designed to estimate daily and seasonal patterns in carbon fixation, plant biomass, nutrient allocation, litter fall, soil nutrient mineralization, and carbon dioxide exchange, including carbon emissions from soils worldwide. Direct input of satellite "greenness" data from the Advanced Very High Resolution Radiometer (AVHRR) sensor into the NASACASA model can be used to accurately estimate global monthly net primary production (NPP), biomass accumulation, and litterfall inputs to soil carbon pools at a geographic resolution of 10 latitude and longitude. Soil fertility factors included in the NASA-CASA model control the allocation of new plant growth to aboveground tissues (leaf and wood) versus fine-root tissue allocation for acquisition of soil nutrients. In this article, I examine several different methods for estimating changes in terrestrial biomass sources of atmospheric carbon dioxide using a combination of global satellite observations, ecosystem model (such as NASA-CASA) predictions of aboveground biomass for the late 1980s, and data on country-by-country changes in global forest cover for the years 1990-1995 (FAO 1997). When the NASA-CASA model is used, 769 University of California Press is collaborating with JSTOR to digitize, preserve, and extend access to BioScience ® www.jstor.org (b) EcosystemProduction NutrientMineralization (a) Soil MoistureBalance andPlantFunctionalTypes NDVI PPT PE (c) Biogenic TraceGas Flux FPAR - TEMP T ,PPT N- . NPP PET SOLAR Soil Surface Mo Mo Mo:4? . . Heat & Water Flux MM" M3 M" MS oi Leaf Litter Root Litter reeze/Thaw Microbes Profile Layers M Sf(Temp f (Temp) oil Organic Matter f(Lit q) GNO Grass Shrub f(WFPS)I f(WFPS) Tree C02 -1CH4 N20 Mineral N Figure 1. Structureof the NASA-CASA model. Predictions are based on plant functional types, such as deserts, grasses, and overstory woody plants (shrubs and trees). Plant functional types determine net primary productivity (NPP) and nutrient cycling rates, which in turn control trace gas fluxes. (a) Typical soil water balance is shown as the shaded depth level in soil profile layers (M1-M3; see text for details). Soil water balance is a function of both precipitation (PPT) and potential evapotranspiration(PET). Freeze-thaw of the soil profile is a function of the seasonally accumulatedheat flux. (b) Climate controls on NPP are defined by the equation NPP = SrFPAR Emax T W, where Sr is solar irradiance(SOLAR),FPARis fraction of absorbedphotosynthetically active radiation (derived from NDVI, the Normalized Difference Vegetation Index), Emax is a model constant, T is air temperature(TEMP),and W is soil moisture balance (i.e., a function of PPT and PET). Controls on litter and soil carbon decomposition are a function of soil temperature(Temp),of the water-filledpore space (WFPS)of mineral soil, and of the ratio of litter nitrogen to lignin (Lit q). (c) Biogenic emission fluxes of soil trace gases include heterotrophic respiration (CO2),methane (CH4),and nitrous (N20) and nitric (NO) oxides. the analysis suggests that yearly net terrestrial losses of carbon dioxide from changes in the world's forest ecosystems are 1.2-1.3 Pg of carbon for the early 1990s. This estimate, which accounts for forest area regrowth and expansion sinks in temperate and boreal forest zones, is based on the most recent global maps for observed climate, soils, plant cover, and changes in forest areas from natural and human forces. Modeling terrestrialproduction from remote sensing Major carbon fluxes between the atmosphere and terrestrial biosphere are often expressed in terms of net biomass accumulation from annual NPP. Several methods exist for estimating terrestrial NPP over large areas of the globe. One set of methods is based on ecophysiology mod770 els, which link carbon metabolism with water, energy, and nutrient cycling in plants (e.g., Kindermann et al. 1996, Schimel et al. 1996b, Churkina and Running 1998). Other methods use remotely sensed data to provide direct time-series data on properties of the vegetation cover, such as changes in surface "greenness," which is commonly expressed as leaf area index or canopy light absorption (e.g., Maisongrande et al. 1995, Potter et al. 1999) and which represents the status of the leaf canopy of any plant functional type (e.g., forest, savanna, grassland, tundra, or desert). Global NPP of vegetation can be predicted using the relationship between greenness reflectance properties and absorption of photosynthetically active radiation (PAR), assuming that net conversion efficiencies of PAR to plant carbon can be approximated for different eco- systems or are nearly constant across all ecosystems (Goetz and Prince 1998). A successful approach to estimating terrestrial plant production is based on the concept of vegetation greenness in the NASA-CASA formulation (Potter and Klooster 1999). Canopy greenness is measured using a Normalized Difference Vegetation Index (NDVI). NDVI is a unitless parameter (scaled from 0 to 1000) that is computed from the ratio of visible and near-infrared radiation reflected from the canopy as detected by the AVHRR satellite sensor. The AVHRR NDVI of greenness has been closely correlated with vegetation parameters such as the fraction of absorbed PAR (FPAR) and leaf area index (Running and Nemani 1988, Sellers et al. 1994, DeFries et al. 1995). Terrestrial NPP fluxes from the NASA-CASA model have been exBioScience Vol. 49 No. 10 tensively validated against both seasonal patterns of atmospheric carbon dioxide measurements at sampling stations around the world (Denning 1994) and multi-year estimates of NPP from field stations and tree rings (Malmstr6m et al. 1997). The monthly fraction of annual NPP flux, defined as net fixation of carbon dioxide by vegetation, is computed, in the NASA-CASA model, on the basis of light-use efficiency (Monteith 1972). Monthly production of plant biomass is estimated in the model as the product of surface solar irradiance, Sr(Bishop and Rossow 1991), FPAR from the AVHRR NDVI, and a light utilization efficiency term (Smax) that is multiplied by air temperature (T) and soil moisture balance (W) stress scalars: colates through to lower layers and may eventually leave the system as seepage and runoff. In the NASACASA model, freeze-thaw dynamics with soil depth operate according to the empirical degree-day accumulation method (Jumikis 1966, as described by Bonan 1989). Because it is based on plant production as the primary carbon and nitrogen cycling source, the NASACASA model is able to couple daily and seasonal patterns in soil nutrient mineralization plus soil heterotrophic respiration (Rh) of carbon dioxide with net nitrous oxide, nitric oxide, and methane emissions from soils worldwide. Net ecosystem production can be computed as NPP minus Rh fluxes, excluding direct effects of fire and other small-scale disturbances. The trace gas components of NPP = S FPAR TW the NASA-CASA model (Potter et max The Smax term is set at 0.56 g/MJ (car- al. 1996a, 1996b) are summarized in bon/PAR), a value that derives from Figures lb and 1c. The soil model calibration of predicted annual NPP component of NASA-CASA uses a to previous field estimates of NPP set of compartmental difference (Potter et al. 1993). The T stress term equations with a structure compais computed with reference to deri- rable to those of the CENTURY ecovation of optimal temperatures system model (Parton et al. 1992). for plant production. The Topt(Topt) set- First-order equations simulate exting varies with latitude and longi- changes of decomposing plant resitude, ranging from near 0 ?C in the due (metabolic and structural fracArctic to the mid-30s (?C) in low- tions) at the soil surface. They also latitude deserts. The W term is esti- simulate surface soil organic matter mated from monthly water deficits, fractions that presumably vary in which are based on a comparison of age and chemical composition. Acmoisture supply (precipitation and tive (microbial biomass and labile stored soil water) to evapotranspirasubstrates), slow (chemically protion water potential demands using tected), and passive (physically prothe method of Thornthwaite (1948). tected) fractions of the soil organic in the matter are represented in the model. NASA-CASA Algorithms model allow evapotranspiration to Along with moisture availability and be connected to water content in the litter quality, estimated soil temperasoil profile layers (Figure la; Potter ture in the M1 layer controls soil 1997). The model design includes organic matter decomposition. heat and moisture content computaInterannual results of global net tions for three soil layers: surface ecosystem production estimates from organic matter (M1), topsoil (0.3 m; the NASA-CASA model (Potter and M2), and subsoil to rooting depth Klooster 1999, Potter et al. 1999) (1-10 m; M3). These layers can differ suggest that land plants have been in soil texture, moisture-holding ca- absorbing more carbon dioxide in pacity, and carbon-nitrogen dynam- the Northern Hemisphere during the ics. Water balance for each layer of late 1980s than previously believedthe soil is modeled as the difference almost one-third of the annual amount between precipitation plus volumet- of carbon dioxide released from the ric percolation inputs, on the one burning of fossil fuels. When AHVRR hand, and monthly estimates of po- satellite data are used in the NASAtential evapotranspiration plus drain- CASA simulation model of net ecoage output, on the other. Inputs from system production, the results show rainfall can recharge the soil layers increasing carbon dioxide accumuto field capacity. Excess water per- lation in vegetation over extensive October 1999 areas of Canada, Europe, and Russia during 1985-1988. This simulation modeling indicates precisely which forest and tundra areas on Earth act as temporary "sinks" for atmospheric carbon dioxide in response to warmer than average spring temperatures or lower than average summer drought stress. Many large-scale disturbances, such as conversion of land to agriculture, may be detected in the AHVRR data used to drive the model. Soil fertility effects on biomass allocation are included in the NASACASA model. For global simulations, the 1 grid resolution Soil Map of the World (FAO/UNESCO 1971, Zobler 1986) is classified in the NASACASA model according to three relative levels of soil fertility (low, medium, and high), following the scheme proposed by Esser (1990) and Bouwman (1990). For low-fertility soils, a -10% adjustment is made that allocates increasing root biomass from NPP for greater acquisition of soil nutrients (Wilson and Tilman 1991). For medium- and high-fertility soils, a +10% adjustment is made that allocates increasing stem and leaf biomass from NPP to support greater light-harvesting functions in the canopy (Gleeson and Tilman 1990, Redente et al. 1992, Lusk et al. 1997). These adjustments represent conservative effects of fertility on root allocation for forests. Carbon turnover resulting from tree mortality is expressed, in the NASA-CASA model, in terms of the mean residence time (t, in years) of carbon in the standing woody tissue pool, depending on the plant functional type. Allocation ratios (a, as percentage of NPP) and mean residence times for leaf and fine-root biomass are expressed in a similar manner, based on estimates from the literature (Table 1). These empirical values for t and a together determine the accumulation rates of plant biomass in living plant and soil pools across the model's 10 global grid. Global data drivers Complete AVHRR data sets for the 1980s have been produced from National Atmospheric and Oceanic Administration (NOAA) Global Area Coverage Level 1B data. These data consist of reflectances and bright771 Table 1. Allocation and residence time parameters for major plant functional types.a Plant functional typeb tc leaf croot a wood Cdleaf x root uwood Tundra 0.25 0.30 0.25 0.45 0.25 0.30 0.25 0.25 0.30 0.25 0.25 0.25 0.55 0.25 0.25 0.25 0.25 0.25 0.50 0.45 0.50 NAe 0.50 0.45 0.50 0.50 0.45 1.5 1.0 2.5 1.5 1.5 1.0 1.5 1.5 1.0 3.0 3.0 3.0 5.0 3.0 3.0 3.0 3.0 5.0 50 50 50 NA 40 40 50 50 25 0.25 0.25 0.50 1.5 2.0 25 High-latitude forest Boreal coniferous forest Temperate grassland Mixed coniferous forest Temperate deciduous forest Desert and bare ground Semi-arid shrubland Savanna and woody grassland Tropical evergreen rain forest sSources for information on parameter settings include Cannell (1982), Aber and Melillo (1991), Running and Gower (1991), Redente et al. (1992), and Lusk et al. (1997). bGlobal cover classes defined by DeFries and Townshend (1994) using Advanced Very High Resolution Radiometer (AVHRR) observations of seasonal plant greenness. coa,the proportional allocation constant of plant tissue pools. d,, the residence time (in years) of carbon in plant tissue pools. eNA, not applicable. ness temperatures for cloud cover derived from the five-channel crosstrack scanning AVHRR aboard the NOAA Polar Orbiter "afternoon" satellites (NOAA-7, NOAA-9, and NOAA-11). Monthly composite NDVI data sets remove much of the contamination due to cloud cover that is present in the daily AVHRR data sets (Holben 1986). Additional processing of the satellite imagery is nevertheless necessary to eliminate remaining artifacts. As part of the Global Inventory Monitoring and Modeling Studies (GIMMS) program (Los et al. 1994) of NASA Goddard Space Flight Center, Sellers et al. (1994) developed Fourier algorithms and solar zenith angle adjustments for interannual AVHRR data sets to further correct NDVI signals from global 10 data sets (averaged from 8 km values) for the 1980s. Fourier algorithm/solar zenith (FAS) processing removes many artifacts present in previous NDVI data sets, including cloud cover and aerosol interference. These GIMMS NDVI data show minimal correlations with equatorial crossing times of the NOAA satellites (Malmstr6m et al. 1997), which suggests that corrections have been made for orbital drifts and switches between satellites. For surface temperature and precipitation drivers, long-term (19311960) average values from Leemans and Cramer (1990) were used after being adjusted with monthly 1' climate anomalies for the period 19801988 (Dai and Fung 1993). Average monthly FAS-processed NDVI and solar irradiance data (Bishop and Rossow 1991) from the respective 1980s time series data sets (Potter and Klooster 1998) were used to Table 2. Ecosystem carbon estimates for vegetation classes from the NASA-CASA model and the field data analysis of Olson et al. (1983).a Vegetation class Total area (x 106 km2)b Average carbon in AGB (g/m2)c Total carbon (Pg) in AGBc Average carbon in AGB (g/m2)b Total carbon in AGB (Pg)b Average NPPd carbon (g. M-2. yr-1)b Total NPP carbon (Pg/yr)b Broadleaf evergreen forest Coniferous evergreen forest and woodland 15.4 8159 126.1 12622 195.0 1075.4 16.6 13.5 11381 153.5 6014 84.9 233.3 3.3 High-latitude deciduous forest and woodland Tundra Mixed coniferous forest and woodland Wooded grassland and shrubland 5.9 11918 70.1 5603 34.0 244.5 1.5 8.6 7.5 428 5321 3.8 40.2 1878 7386 17.7 56.9 72.6 386.2 0.7 3.0 24.3 4540 111.2 8193 202.4 782.4 19.3 22.1 17.7 14.6 3.6 476 190 747 7390 10.6 3.4 10.9 26.9 154 525 182 6715 3.5 9.4 2.7 24.7 172.5 21.4 365.0 397.4 4.0 0.4 5.4 1.5 11.5 145.4 971 11.1 567.8 1721 20.0 651.1 69.0 0.8 56.4 Temperate grassland Desert Cultivated Broadleaf deciduous forest and woodland Semi-arid shrubland Total aAboveground biomass (AGB) was estimated from extrapolation of field data reported by Olson et al. (1983) and from the NASA-CASA model. AGB is presented for both average values and totals by vegetation class. Average carbon (g/m2) is multiplied by area in each vegetation class to estimate total carbon in AGB (Pg). bFrom the NASA-CASA model. cFrom Olson et al. (1983). dNPP, net primary production. 772 BioScience Vol. 49 No. 10 generate ecosystem model results typical of the 1980s. a Biomass and production estimates for the 1980s For the 1980s, predicted worldwide NPP from the NASA-CASA model was estimated at 56.4 Pg/yr of carbon (Table 2). Predicted worldwide aboveground biomass (including leaf and wood carbon) was estimated at 651 Pg of carbon (Figure 2). Average carbon storage in aboveground biomass was predicted to be highest in broadleafevergreenforests-mostly thoseintropicalzones-at morethan 0 5000 10000 15000 20000 25000 12kg/m2,withlowerstorageinmixed forestsof temconiferous-deciduous carbonstocks peratezones.Predicted in woodbiomasswere20-100 times b greaterthanpredictedcarbonstocks in leaf biomassfor most forestecosystemareas(Figure2). Accuracyof themodelpredictions can be evaluatedby comparisonto observations. availableground-based For example, the NASA-CASA model value for global carbon in abovegroundbiomass, 651 Pg, is somewhathigherthanOlsonet al.'s (1983) estimateof 568 Pg, which was basedon a selecteddata set of biomassobservations(Table2). Althoughtherearefew otherestimates of standingbiomassforcomparison, 0 250 500 750 1000 maximumpredicted aboveground biomass from the NASA-CASA gCm-2 modelfor tropicalmoist forestsexceeds17 kg/m2of carbon.This figure, unlike the lower Olson et al. Figure 2. Results from the NASA-CASA model for aboveground biomass. (a) (1983) estimates,is in close agree- Predictedwood carbon. (b) Predictedleaf carbon. Global patterns are representament with measured amounts of tive of net primaryproduction (NPP)computedusing satellite and climate inputs to abovegroundbiomass from recent the NASA-CASA ecosystem model for the 1980s. fieldstudies(Kauffmanet al. 1995, Carvalhoet al. 1998).Bycontrastto partiallyto the model ratio setting landsurfaceconditionsobservedby thefairlysimilarestimatesof global (leafversusroot tissue)for biomass satelliteremotesensingfor a recent abovegroundbiomass,averagepre- allocationfromNPP.Themodelra- timeperiodof lessthan5 years,that dicted aboveground biomass from tio results in higher belowground is, thelate 1980s.NASA-CASApreallocationof bio- dictions also include climate conthe NASA-CASAmodel for conifer- thanaboveground ous evergreenforests and high-lati- massat the expenseof aboveground trols on plantproductionat 10 spatialresolution,effectsof soil fertility tude deciduous forests deviates no- biomassin grasslands. on carbon allocation to different the et al. the two different methfrom Olson Overall, (1983) tably bio- plant tissues, and AVHRR-based estimates,mainlybecauseof the rela- ods forestimatingaboveground tively low estimated annual produc- mass, as comparedin Table2, may mappingof 1980s globallandcover tion in these vegetation classes based representthepotentialextremesand and use patterns (DeFries and on the monthlyNDVI driversused in the variabilityof biomassstoragein Townshend1994). Anotheradvanthe model. Comparedto the Olson et manyterrestrialecosystems.Unlike tageof theNASA-CASAmodelestial. (1983) observations, the lower the extrapolatedvalues from site mates for carbon in aboveground NASA-CASA model estimate of measurements reportedby Olsonet biomassis thattheymakeit possible abovegroundbiomass for grasslands al. (1983), the NASA-CASAmodel to separateleaf fromstandingwood and cultivated areas is attributable estimateis designedto reflectactual and fine-root stocks, thereby acOctober1999 773 counting for effects of deforestation on net carbon dioxide fluxes to the atmosphere. By contrast, site measurements reported by Olson et al. (1983) are for total standing biomass only and are not specific to a recent time period of less than 5 years,which is importantwhere land use is changing rapidly. Carbonlossesfrom deforestation FAO (1997) providesworldwide statistics on the status of forest cover, circa 1995, and on recent (19901995) rates of annual change in forest cover and condition. Changes in forest area or condition can be documented as positive (e.g., through natural colonization of nonforest landor regrowthon degradedstands) or negative (e.g., through deforestation, forest fires, or unsustainable exploitationfor wood). Theestimates of forest cover change during 19901995 for more than 170 countries indicate a net loss of 56.3 million ha of forests worldwide (FAO 1997). This value represents a decrease of 65.1 million ha in developing countries that was partly offset by an increase of 8.8 million ha in developed countries. Although the rate of deforestationin developingcountries is still considered high (13.7 million ha/yr), the rate appears to have slowed somewhat since the 1980s (15.5 million ha/yr; FAO 1997). By combining the 10 resolution NASA-CASA model estimates for standing biomass (i.e., aboveground biomass) in forests worldwide (Figure2) with country-by-countryrates of change in forest area compiled by FAO (1997), detailed estimates can be made for global carbon losses from deforestation and gains from expansion of planted forest area or secondary regrowth following abandonment of other land uses. To derive rates of carbon flux to the atmosphere from forest cutting and burning, a biomass combustion factor of 48% total aboveground biomass, which has been reported from tropical field studies by Kauffman et al. (1995), was used as a converter for FAO annual deforestation rates over the period 1990-1995. The resulting estimate for global carbon flux, which was based on 774 FAO (1997) conversion rates for 1990-1995, suggeststhat the net terrestrial loss of carbon dioxide from changes in area of the world's forest ecosystemswas 1.15 Pg/yrof carbon (Figure 3). This estimate includes 1.44 Pg/yr of carbon lost to the atmosphere from deforestation sources, offset in part by 0.29 Pg/yr of carbon accumulatedin the terrestrial biosphere through forest area regrowthand expansion. In this estimate, countries with the highest individuallosses of carbon from deforestation are Brazil, with 0.25 Pg/yr, followed by Zaire and Indonesia, with losses of just under 0.1 Pg/yr each (Table 3). Estimated rates of carbon accumulationin areas of forest regrowth are highest in Canada, France, and the United States, each with annual gains of more than 0.02 Pg/yr.It shouldbe notedthat although the age structureof forestscan have a significantimpact on estimated carbon pools from regrowth, it is not feasible to consider these types of age-dependentfactors in this kind of global-scale analysis because of a lack of data from satellite sources. Compiled by global vegetation classes, the highest net carbon losses from deforestation are predicted to occur in tropical evergreen forests, where they reach 0.69 Pg/yr, and tropical savanna woodlands, where they reach 0.56 Pg/yr. Net carbon gains from regrowth and expansion of forest areas over boreal and other high-latitude forest zones are predicted to be 0.09 Pg/yr,whereas predicted carbon gains in biomass over all temperateforestzones total nearly 0.06 Pg/yrfor the period 1990-1995. As an important footnote to this study, FAO (1997) mentions that a highly reliableestimatefor changein forest area between 1990 and 1995 could not be made for the Russian Federation. Reported rates of change in forested areas of Russia are listed in the 1997 FAO forest resource assessment at +0.07% annually. However, other recent reports estimate that forest fires in Russia consume 7.3 million ha, or approximately 1.18%, of the Federation's boreal forest area each year (Conard and Davidenko 1996). This forest loss estimate is an order of magnitude greater than that indicated by official fire statistics, a discrepancy with major implications for global carbon dioxide emissions that has apparentlynot been fully consideredin other recent assessments of the carbon cycle (e.g., IGBP 1998). If a modified (and conservative) loss estimate of 0.05% for net annual change in forest area for Russia over the period 1990-1995 is adopted, the projectednet terrestrial loss of carbon dioxide from changes in the world's forest ecosystems would be adjusted to 1.24 Pg/yr of carbon for the early 1990s. A less conservative loss estimate of 1.2% for annual change in Russian forest areawould make the areacoveredby the formerUSSRthe highestnational deforestationsource of carbon dioxide in the world, contributingto global net carbon losses that would exceed 1.7 Pg/yrfrom changes in forest ecosystems worldwide. Implications and conclusions A major finding from the global modeling analysis presented in this article is that the potential for recent carbon gains from forest area regrowth or expansion are dwarfedby the continuing losses from deforestation and fires. Although temperate and boreal forests may be increasing somewhat in area and possibly in biomass per unit area, this global analysis suggests that the measurable gains are currently offset by worldwide deforestation losses of carbon to the atmosphere at a ratio of approximately5 to 1. At this time, there appears to be no practical means to alter the carbon balance equation except by mitigatingdeforestation and forest degradation in the tropical countries and, possibly, the Russian Federation. Mitigation measures to limit the conversion of the world's forest biomass to atmospheric carbon dioxide would involve expanding the amount of forest area under protection from logging, reducing the accidental spread of savanna and agricultural wildfires into tropical forests, and improving harvesting practices to reduce waste and prevent damage to remaining vegetation and to soils. The Kyoto Protocol specifies that developed countries and those with economies in transition should undertake measures to reach specific BioScience Vol. 49 No. 10 reduction targets for greenhouse gases, restricting average annual emissions to a percentage of 1990 emissions. These reduction targets are based on the assumption that all emission sources and sinks can be accurately counted in assessing national compliance.However, despite the improved information on biomass carbon losses from deforestation reportedin this article,the Kyoto Protocol'ssectoron LandUse Change and Forestrywill continue to be one of the more difficult sectors to define and quantify out of the Kyoto negotiations and beyond. For example, soils and belowground biomass stocks of carbon decompose more slowly than the standing biomass that is burned or removed during deforestation. Some disturbed soils with large stocks of dead wood-for example, in converted cattle pastures-can continue to release their stored carbon for several decades after forest clearing. Satelliteremote sensing alone cannot detect changes in thesesoil surfaceand belowground processes. The unique advantage of ecosystem modeling, which cannot be achieved through a small set of soil sampling measurements alone, is underscoredin this context. For instance, the global NASA-CASA modelestimatesthat the total amount of belowground (fine-root) biomass in forested ecosystems worldwide represents 50 Pg of carbon, which could eventually make its way into the atmosphere in the event that standing biomass stocks in forests are intensively cut or burned. This root biomass total must be added to the NASA-CASA model estimate of morethan 170 Pg of carbonstoredin deadlitterandsoil carbonpools (those pools with meanresidencetime of less than 25 years) in global forest ecosystems (Potter and Klooster 1997). Higher-resolution applications of the NASA-CASA ecosystem model can be made to estimate forest biomass and production at the state or national level, down to at least the 1 km2 grid size for AHVRR greenness inputs. This spatial resolution can capture many direct effects of deforestation on land cover and predicted biomass carbon stocks. NASA-CASA model simulations using 8 km resolution land cover and NDVI data October 1999 7. -150 -100 -50 0 50 gC m2yr-1 Figure 3. Global carbon flux resulting from annual changes in forest area. Predictions are based on FAO (1997) conversion rates for 1990-1995, combined with NASA-CASAmodel predictionsof standingplant biomass. Negative flux values (-) indicate loss of carbon to the atmospherefrom deforestation, whereas positive flux values (+) indicate gains of carbon from the atmosphere resulting from forest regrowth or expansion. Gridded 10 flux values are based on the assumption that rates of forest carbon loss or gain occur in proportionto the geographicdistribution of standing biomass amount on a national level. made using the eddy correlation approach (Goulden et al. 1996, Malhi et al. 1998), in which fast-response measurements of the vertical component of wind speed and humidity allow for calculation of forest moisture and carbon fluxes based on the covariance of these meteorological variables. Together with reliable imagesof dominantforestcovertypes from Landsator other relativelyhigh resolution mapping sources, these eddy correlationtower flux measurements of net ecosystem carbon exchange may be extended with confidenceto regionalscalesandcompared to ecosystem model predictions. From regional to global scales, a network of gas sampling stations would provide records over time for atmosphericcarbon dioxide concentrationsthat could be used with models of global circulation to track terrestrial locations and fluxes of net ecosystem production. ecosystem models. A promisingnew method for charA series of tower-based measurements of net ecosystem carbon ex- acterizingforest structureis the Vegchange over several seasons can be etation Canopy Lidar (VCL), an ac- drivershave generatedregional budgets for NPP and trace gas fluxes over the southeastern United States (Davidson et al. 1998) and for the entire nation of Brazil (Potter et al. 1998). A common limiting factor for more extensive (e.g., pan-tropical) application of the model at 1-8 km resolution is the lack of compatible high-quality maps for monthly climate and soil data inputs. Severalmethods can complement global ecosystem models driven by satellite remotely sensed data to aid in defining limits on carbon stocks and narrowing the uncertainties in net terrestrial sources of carbon in the atmosphere. At the forest stand level, intensive sampling of plant, litter, and soil carbon pools can be combined with geographic information systemsfor extrapolationacross the landscape.These same field measurements can be used to calibrate 775 Table 3. Averageestimatedchange in forest abovegroundbiomass (AGB)over the years 1990-1995 for countries with the highest carbon flux rates. Annualrateof AGB in forestsa Country Brazil Zaire Indonesia Angola Bolivia Venezuela Paraguay Malaysia Burma Thailand Mexico Philippines Colombia Tanzania Zambia India Poland Italy Portugal Norway United Kingdom Australia Greece Ireland Canada France Former USSR United States (Pg carbon) change (% of forest area)b Predictedchange in AGB (Pg/yr carbon) 113.56 31.50 21.74 12.86 11.22 10.65 4.54 4.66 7.16 3.69 10.15 2.66 11.37 -0.45 -0.66 -0.94 -1.01 -1.13 -1.08 -2.48 -2.29 -1.33 -2.48 -0.88 -3.24 -0.48 5.58 -0.95 -0.025 6.25 6.75 0.55 1.09 0.47 1.33 1.08 19.62 0.48 0.62 39.43 3.20 98.35 37.65 -0.81 +0.01 +0.14 +0.09 +0.87 +0.34 +0.56 +0.04 +2.43 +2.80 +0.07 +1.13 +0.07 +0.28 -0.024 +0.001 +0.001 +0.001 +0.004 +0.005 +0.006 +0.008 +0.012 +0.017 +0.028 +0.036 +0.067c +0.106 -0.247 -0.099 -0.098 -0.063 -0.061 -0.055 -0.054 -0.051 -0.046 -0.044 -0.043 -0.041 -0.026 aFrom the NASA-CASA model. values indicate loss of carbon to the atmosphere from deforestation, whereas positive values indicate gains of carbon from the atmosphere resulting from forest regrowth or expansion. Data from FAO (1997). cThis value for States of the Former USSR is based on annual rates of change of +0.07% for forest areas in Russia (FAO 1997). By including reported rates of forest fire in Russia (Conard and Davidenko 1996), this estimated change in forest carbon converts to a minimum value of -0.024 Pg/yr of carbon for States of the Former USSR. bNegative tive remote sensing method using lidar (light detection and ranging) technology. VCL is the first selected Earth System Science Pathfinder mission for NASA's Earth Science Enterprise (Dubayah et al. 1997). Lidar systems have been used extensively for small-scale observations taken from airborne platforms. The Shuttle Laser Altimeter provided the first spaceborne vehicle for global-scale lidar observations. With launch anticipated within the next few years, the VCL will sample extensively over the earth's closed-canopy forests, collecting elevation data for canopy tops as well as for the underlying terrestrial topography. The sensor should provide direct estimates of degraded forest areas, areas of regrowth, and areas of intact forest structure for quantifying changes in biomass. In the full assessment of human impacts on carbon emissions, several noteworthy sources of uncer776 tainty remain in the type of global modeling analysis presented in this article for biomass and deforestation effects on terrestrial carbon fluxes. For example, FAO's forest resource assessments before 1997 have been questioned as giving somewhat conservative estimates of deforestation rates (Houghton 1994). As an improvement, the forest cover data for developing countries contained in FAO (1997) are based on national-level assessments that were prepared on different dates and that, for reporting purposes, have been adjusted to the standard reference years of 1990 and 1995. This adjustment was made using a model function developed to correlate changes in forest cover area over time with ancillary variables, including population change and density, initial forest cover, and ecological zone of the forest area under consideration. Further improvements are being added with each successive FAO report. For example, new national-level assessments were submitted by several governments to FAO as updated inputs to the model for the calculation of the forest cover area for 1995 and the recalculation of the 1990 estimate. Specifically, forest inventory information was revised for several countries, including Brazil, Bolivia, Mexico, Cambodia, and the Philippines. Despite improved methods of deforestation rate assessment, certain destructive land-use practices are rarely reported and therefore not included in assessments of tropical forest inventories-namely, selective logging and the accidental spread of agricultural fires into forests. For example, in the Amazon rain forest, logging companies substantially reduce the living biomass of forests through the harvest and associated damage of trees (Nepstad et al. 1999). This damage allows sunlight to penetrate to the forest floor, where it dries out the organic debris created by logging. The incidence of ground fires can increase dramatically when the combined effect of severe droughts provoke forest leaf shedding and greater flammability. Satellites are generally unable to detect these "hidden" processes of forest impoverishment. Eventually, the effects of logging and drought can lead to large-scale forest fires in Amazonia and many other tropical countries, particularly during El Nifio events, which intensify the ecological imand portance of deforestation weather-related carbon emissions because of severe drought conditions. The need to accurately inventory and monitor carbon storage over entire nations will likely gain economic importance with the expected establishment of an international system for trading carbon credits. Accounting for carbon pools and fluxes on a global level will be possible only by bringing to bear and improving all available methods of ecosystem monitoring and inventory, including remote sensing, groundbased sampling, tower flux measurements, ecosystem modeling, statistiand geographic cal analysis, information systems. Errors from each of the methods must be reduced to the greatest extent possible by BioScience Vol. 49 No. 10 systematic comparisons of their independent estimates. Acknowledgments Thanks to Steven Klooster for continued development of the NASACASA model computer program. Three anonymous reviewers provided helpful comments on a previous version of the manuscript. This research was supported by the NASA Ames Research Center Basic Research Council (Code No. 274-5271-28). References cited Aber JD, Melillo JM. 1991. Terrestrial Ecosystems. Philadelphia: Saunders College Publications. Bishop JKB, Rossow WB. 1991. Spatial and temporal variability of global surface solar irradiance. Journal of Geophysical Research 96: 16839-16858. Bonan GB. 1989. A computer model of the solar radiation, soil moisture and soil thermal regimes in boreal forests. Ecological Modelling 45: 275-306. Bouwman AF. 1990. Global distribution of major soils and land cover types. Pages 33-59 in Bouwman AF, ed. Soils and the Greenhouse Effect. New York: John Wiley & Sons. Cannell MGR. 1982. World Forest Biomass and Primary Production Data. London: Academic Press. Carvalho JAJr, Higuchi N, Araujo TM, Santos JC. 1998. Combustion completeness in a rainforest clearing experiment in Manaus, Brazil. Journal of Geophysical Research 103: 13195-13199. Churkina G, Running S. 1998. Contrasting climatic controls on the estimated productivity of global terrestrial biomes. Ecosystems 1: 206-215. Conard SG, Davidenko EP. 1996. Fire in for Siberian boreal forests-Implications global climate and air quality. Paper presented at the International Symposium on Air Pollution and Climate Change Effects on Forest Ecosystems; 5-9 February 1996; Riverside, CA. Dai A, Fung IY. 1993. Can climate variability contribute to the missing CO2 sink? Global Biogeochemical Cycles 7: 599-609. Davidson EA, Potter CS, Klooster SA, Schlesinger P. 1998. Model estimates of regional nitric oxide emissions from soils of the southeastern United States. Ecological Applications 8: 748-759. DeFries R, Townshend J. 1994. NDVI-derived land cover classification at global scales. International Journal of Remote Sensing 15: 3567-3586. DeFries R, et al. 1995. Mapping the land surface for global atmosphere-biosphere models: Toward continuous distributions of vegetation's functional properties. Journal of Geophysical Research 100: 2086720882. Denning AS. 1994. Investigations ofthe trans- October 1.999 port, sources, and sinks of atmospheric CO2 using a general circulation model. PhD dissertation. Colorado State University, Fort Collins, CO. Dubayah R, Blair B, Bufton J, Clarke D, JiJai J, Knox R, Luthcke S, Prince S, Weishampel J. 1997. The Vegetation Canopy Lidar Mission. Land Satellite Information in the Next Decade. Vol. II. Washington (DC): American Society for Photogrammetry and Remote Sensing. Esser G. 1990. Modelling global terrestrial sources and sinks of CO2 with special reference to soil organic matter. Pages 247-262 in Bouwman AF, ed. Soils and the Greenhouse Effect. New York: John Wiley & Sons. [FAO] Food and Agriculture Organization of the United Nations. 1997. State of the World's Forests 1997. Rome: Food and Agriculture Organization- of the United Nations. [FAO/UNESCO] Food and Agricultural Organization/United Nations Educational, Scientific and Cultural Organization. 1971. Soil Map of the World. 1:5,000,000. Paris: United Nations Educational, Scientific and Cultural Organization. Gleeson SK, Tilman D. 1990. Allocation and the transient dynamics of succession on poor soils. Ecology 71: 1144-1155. Goetz SJ, Prince SD. 1998. Variability in light utilization and net primary production in boreal forest stands. Canadian Journal of Forest Research 28: 375-389. Goulden MJ, Munger JW, Fan SM, Daube BC, Wofsy SC. 1996. Exchange of carbon dioxide by a deciduous forest: Response to interannual climate variability. Science 271: 1576-1577. Holben BN. 1986. Characteristics of maximum-value composite images from temporal AVHRR data. International Journal of Remote Sensing 7: 1417-1434. Houghton RA. 1994. The worldwide extent of land-use change. BioScience 44: 305313. [IPCC] Intergovernmental Panel on Climate Change. 1994. Climate Change 1994: Radiative Forcing of Climate Change and an Evaluation of the IPCC IS92 Emission Scenarios. Cambridge (UK): Cambridge University Press. [IGBP] International Geosphere Biosphere Programme Terrestrial Carbon Working Group. 1998. The terrestrial carbon cycle: Implications for the Kyoto Protocol. Science 280: 1393-1394. Jumikis AR. 1966. Thermal Soil Mechanics. New Brunswick (NJ): Rutgers University Press. Kauffman JB, Cummings DL, Ward DE, Babbitt R. 1995. Fire in the Brazilian Amazon: Biomass, nutrient pools, and losses in slashed primary forests. Oecologia 104: 397-408. Kindermann J, Wiirth G, Kohlmaier GH. 1996. Interannual variation of carbon exchange fluxes in terrestrial ecosystems. Global Biogeochemical Cycles 10: 737755. Leemans R, Cramer WP. 1990. The IIASA database for mean monthly values of temperature, precipitation and cloudiness of a global terrestrial grid. Laxenberg (Austria): International Institute for Applied Systems Analysis. Report WP-41. Los SO, Justice CO, Tucker CJ. 1994. A global lx1 NDVI data set for climate studies derived from the GIMMS continental NDVI data. International Journal of Remote Sensing 15: 3493-3518. Lusk CH, Contreras O, Figueroa J. 1997. Growth, biomass allocation and plant nitrogen concentration in Chilean temperate rainforest tree seedlings: Effects of nutrient availability. Oecologia 109: 49-58. Maisongrande P, Ruimy A, Dedieu G, Saugier B. 1995. Monitoring seasonal and interannual variations of gross primary productivity, net primary productivity, and net ecosystem productivity using a diagnostic model and remotely sensed data. Tellus Series B Chemical and Physical Meteorology 47: 178-190. Malhi Y, Nobre AD, Grace J, Kruijt B, Pereira MGP, Culf A, Scott S. 1998. Carbon dioxide transfer over a central Amazonian rain forest. Journal of Geophysical Research 103: 31593-31612. Malmstr6m CM, Thompson MV, Juday GP, Los SO, Randerson JT, Field CB. 1997. Interannual variation in global scale net primary production: Testing model estimates. Global Biogeochemical Cycles 11: 367-392. Monteith JL. 1972. Solar radiation and productivity in tropical ecosystems. Journal of Applied Ecology 9: 747-766. Nepstad DC, et al. 1999. Large-scale impoverishment of Amazonian forests through logging and fire. Nature 398: 505-508. Olson JS, Watts JA, Allison LJ. 1983. Carbon in Live Vegetation of Major World Ecosystems. Oak Ridge (TN): Oak Ridge National Laboratory. Environmental Sciences Division Publication Number 1997. ORNL-5862. Parton WJ, McKeown B, Kirchner V, Ojima D. 1992. CENTURY Users Manual. Fort Collins (CO): Natural Resource Ecology Laboratory, Colorado State University. Potter CS. 1997. An ecosystem simulation model for methane production and emission from wetlands. Global Biogeochemical Cycles 11: 495-506. Potter CS, Klooster SA. 1997. Global model estimates of carbon and nitrogen storage in litter and soil pools: Response to change in vegetation quality and biomass allocation. Tellus Series B Chemical and Physical Meteorology 49: 1-17. . 1998. Interannual variability in soil _ trace gas (CO2, N20, NO) fluxes and analysis of controllers on regional to global scales. Global Biogeochemical Cycles 12: 621-637. Detecting a terrestrial biosphere sink.1999. for carbon dioxide: Interannual ecosystem modeling for the mid-1980s. Climatic Change 42: 489-503. Potter CS, Randerson J, Field CB, Matson PA, Vitousek PM, Mooney HA, Klooster SA. 1993. Terrestrial ecosystem production: A process model based on global satellite and surface data. Global Biogeochemical Cycles 7: 811-841. Potter CS, Davidson EA, Verchot L. 1996a. Estimation of global biogeochemical controls and seasonality in soil methane consumption. Chemosphere 32: 2219-2246. Potter CS, Matson PA, Vitousek PM, 777 rritle Setv a new course guide for Cultural Usesof Plants: A Guideto Learning aboutEthnobotany high school science teachers :+ MeetstheNationalScienceEducation Gabriell DeBear Paye Standards activities andtestsstressing laboratory SIncludes ISBN 0-89327-422-4 science,for eachunit inquiry-based .: Stressesinterdisciplinary assignmentsthatpro- October 1999 motehands-on withan emphasis on learning, research skills honingthestudent's THE NEWYORKBOTANICAL GARDENPRESS 200th Street and Kazimiroff Boulevard Bronx, NY 10458-5126 Tel: (718) 817-8721 0 Fax: (718) 817-8842 * WebSite:www.nybg.org E-mail:[email protected] Davidson EA. 1996b. Process modeling of controls on nitrogen trace gas emissions from soils world-wide. Journal of Geophysical Research 101: 1361-1377. Potter CS, Davidson EA, Klooster SA, Nepstad DC, de Negreiros GH, Brooks V. 1998. Regional application of an ecosystem production model for studies of biogeochemistry in Brazilian Amazonia. Global Change Biology 4: 315-334. Potter CS, Klooster SA, Brooks V. 1999. Interannual variability in terrestrial net primary production: Exploration of trends and controls on regional to global scales. Ecosystems 2: 36-48. Redente EF, Friedlander JE, Mclendon T. 1992. Response of early and late semiarid seral species to nitrogen and phosphorus gradients. Plant and Soil 140: 127-135. Running SW, Gower ST. 1991. FORESTBGC, A general model of forest ecosystem processes for regional applications. II. Dynamic carbon allocation and nitrogen budgets. Tree Physiology 9: 147-160. Running SW, Nemani RR. 1988. Relating seasonal patterns of the AVHRR vegetation index to simulated photosynthesis and transpiration of forests in different climates. Remote Sensing of Environment 24: 347-367. Schimel DS, Alves D, Enting I, Heimann M, Joos F, Raynaud D, Wigley T. 1996a. CO2 and the Carbon Cycle. Pages 76-86 in Houghton JT, ed. Climate Change 1995. Cambridge (UK): Cambridge University Press. Schimel DS, Braswell BH, McKeown R, Ojima DS, Parton WJ, Pulliam W. 1996b. Climate and nitrogen controls on the geography and timescales of terrestrial biogeochemical cycling. Global Biogeochemical Cycles 10: 677-692. Sellers PJ, Tucker CJ, Collatz GJ, Los SO, Justice CO, Dazlich DA, Randall DA. 1994. A global 1x1 NDVI data set for climate studies. Part 2: The generation of global fields of terrestrial biophysical parameters from the NDVI. International Journal of Remote Sensing 15: 3519-3545. Thornthwaite CW. 1948. An approach toward rational classification of climate. Geographical Review 38: 55-94. Wilson SD, Tilman D. 1991. Components of plant competition along an experimental gradient of nitrogen availability. Ecology 72: 1050-1065. Zobler L. 1986. A world soil file for global climate modeling. Washington (DC): National Aeronautics and Space Administration. NASA Technical Memo 87802. 778 BioScience Vol. 49 No. 10
© Copyright 2026 Paperzz