I. The Terrestrial Ecosystem Model (TEM) and Beyond

A central contribution of our IDS team over the last several years has been in our development of regional and global scale terrestrial ecosystem models (TEM at global scales; PnET at regional scales¹). This work has been central not only in our work but also in our Annual Reports. This year we are stressing the contributions in certain important areas or along certain themes. We a) highlight TEM²; b) discuss the corresponding efforts to develop a diagnostic NPP model based upon remotely sensed data that will be used to improve TEM and serve as a calibration devise for TEM and other models, c) sketch our recent work on the agricultural and trace gas components in an overall terrestrial model, and, as noted, d) present some recent work on hyper-spectral data..

I. 1. The Contributions from TEM. The Terrestrial Ecosystem Model (TEM) is a process-based model that has been designed to estimate the spatial and temporal distribution of the major carbon and nitrogen fluxes and pool sizes of the terrestrial biosphere at regional and global scales. It was initiated in the first years of our NASA EOS IDS investigation (e.g., Raich et al., 1991)³, and has continued to evolve through a series of versions. The first two versions of TEM were used to examine spatial patterns of net primary production in South America (Raich et al., 1991) and in North America (McGuire et al., 1992). The third version of TEM was used to examine the response of NPP to elevated temperature and carbon dioxide for temperate forests (McGuire et al., 1993); and to GCM predicted climate change for the terrestrial biosphere (Melillo et al., 1993). The carbon storage predictions of the third version were also evaluated for global terrestrial ecosystems (Melillo et al., 1995) and for grassland and conifer forests (McGuire et al., in press). In addition, the US Forest Service used NPP estimates from the third version of TEM as part of their 1995 national assessment of the effects of global climate change on forest productivity (Joyce et al., 1995; McGuire and Joyce, 1995; Perez-Garcia et al., in press).

Version 4.0 of TEM (McGuire et al., 1995b; McGuire et al., submitted) was developed to improve the dynamics of soil organic carbon along gradients of temperature, moisture and soil texture; and to incorporate the role of nitrogen in the response of forest net primary productivity to elevated atmospheric carbon dioxide (McGuire et al., 1995a). With the use of a new climate database (Cramer and Leemans, personal comm.), TEM 4.0 estimates a lower global NPP for contemporary conditions (Xiao et al., in press; Kicklighter et al., in preparation) than the third version of TEM. The new climate database represents a cooler, drier and sunnier world than the earlier climate database (see Pan et al., 1996). In addition, TEM 4.0 estimates of the global soil organic carbon pool (McGuire et al., 1995; Xiao et al., in press) are less than the estimates of the third version of TEM because TEM 4.0 only considers the pool of fast-cycling or "reactive" soil organic carbon instead of the total amount of carbon stored in the soils.

In a parallel effort, Braswell and members of Schimel's IDS team (see Schimel et al., in press) used the terrestrial ecosystem model "Century" to evaluate the relative roles of water and nitrogen limitation of net primary productivity, spatially and in response to climate variability. Within ecology, there has been considerable confusion and controversy over the large-scale significance of limitation of net primary production (NPP) by nutrients vs. biophysical quantities (e.g., heat, water, sunlight) with considerable evidence supporting both views. The Century model, run to a quasi-steady state condition, predicts "equilibration" of water with nutrient limitation, because carbon fixation and nitrogen fluxes (inputs and losses) are controlled by water fluxes, and the capture of nitrogen into organic matter is governed by carbon fixation. Patterns in the coupled water, nitrogen, and carbon cycles are modified substantially by ecosystem type or species-specific controls over resource use efficiency (water and nitrogen used per unit NPP), detrital chemistry, and soil water holding capacity. We also examined the coupling between water and nutrients during several temperature perturbation experiments. Model experiments forced by satellite-observed temperatures suggest that climate anomalies can result in significant changes to terrestrial carbon dynamics. The cooling associated with the Mt. Pinatubo eruption aerosol injection may have transiently increased terrestrial carbon storage. However, because processes in the water, carbon, and nitrogen cycles have different response times, model behavior during the return to steady state following perturbation was complex and extended for decades after 1- to 5-year perturbations. Thus consequences of climate anomalies are influenced by the climatic conditions of the preceding years, and climate-carbon correlation may not be simple to interpret.

During 1995 and 1996, we used TEM 4.0 in several regional studies to investigate the potential effects of climate change, elevated CO₂ and vegetation redistribution on net primary productivity and carbon storage of high latitude ecosystems (McGuire and Hobbie, in press), China (see abstract by Pan et al. for the IGBP SAC-IV meeting; Pan et al., in preparation; Xiao et al., submitted; Xiao et al., 1996b) and the United States (VEMAP Members, 1995). The results of these equilibrium simulations indicate that terrestrial ecosystems have the potential to act as a net carbon sink if atmospheric CO₂ is stabilized and that both ecosystem structure and function play an important role in the ability of terrestrial ecosystems to act as a long-term carbon sink. The TEM 4.0 results also indicate that an interaction between elevated CO₂ and climate change may play an important role in the overall response of NPP to climate change. In addition to these regional studies, we also used TEM 4.0 in several studies to examine various sources of uncertainty in developing regional estimates of net primary productivity and carbon storage.

Recently, a number of geographically-referenced data sets have become available that can be used to develop regional estimates of NPP and carbon storage. By using TEM 4.0 with various climate and soil texture data sets for the United States, Pan et al. (1996) found that differences in solar radiation data sets had the largest effect on TEM estimates of NPP for the conterminous United States. A similar comparison between VEMAP climate data sets (Kittel et al., 1995), developed for the Vegetation-Ecosystem Modeling and Analysis Project, and climate data sets developed specifically for the northeastern United States (Ollinger et al., 1995) also indicated that differences in solar radiation had the largest effect on TEM estimates of NPP in the northeastern US (see abstract by Jenkins et al. for the 1996 annual meeting of the Ecological Society of America; Jenkins et al., in preparation). Jenkins et al. (in preparation) also found that the use of different methods of representing vegetation in a grid cell (i.e., mosaic of vegetation types vs. dominant vegetation) had little effect on regional NPP estimates, but that NPP estimates among grid cells were more variable using the mosaic approach than the dominant vegetation approach.

To develop regional and global estimates of NPP and carbon storage, TEM normally uses geographically-referenced data sets organized at a spatial resolution of 0.5 degree longitude x 0.5 degree latitude to capture the spatial variability of environmental conditions across the region. Because a 0.5 degree grid cell covers a large area and environmental conditions are considered to be constant within the grid cell, NPP estimates might be improved using data sets with a finer spatial resolution. To examine this question, we applied TEM 4.0 to the northeastern US using climate, vegetation, and soil texture data sets organized at a spatial resolution of 2 km x 2 km (see abstract by Jenkins et al. for the 1996 annual meeting of the Ecological Society of America; Jenkins et al., in preparation), and we applied TEM 4.0 to the historical range of temperate forests in North America using data sets organized at a spatial resolution of 10 km x 10 km (see abstract by Nungesser et al. for the 1996 annual meeting of the Ecological Society of America). We found little differences between regional NPP estimates based on the coarser 0.5 degree data sets and the finer resolution data sets. However, large differences in NPP estimates could occur between spatial scales in areas that contained a diversity of vegetation types, such as in mountainous terrain or near ecosystem boundaries.

We have also used TEM 4.0 in five model comparison projects, two at the global scale and three at the regional scale:

In the Community Terrestrial Biosphere Model Project, seasonal variations in net ecosystem production (NEP) estimated by TEM across the globe were used in conjunction with the Max Planck ocean and atmospheric transport models to reproduce the seasonal/latitudinal signature of CO₂ in the atmosphere. These results were compared to similar seasonal CO₂ reproductions using other terrestrial biosphere models (Heimann et al., submitted). For CO₂ monitoring stations in the northern hemisphere, TEM 4.0, coupled to the ocean and atmospheric transport models, simulated the amplitude and phase of the seasonal atmospheric CO₂ cycle with reasonable accuracy. In the tropics, however, the model tended to predict larger seasonal exchanges of carbon than indicated by observations. A series of modeling experiments were developed to examine if these differences were due to model shortcomings in the phenology algorithms used, consideration of an inadequate rooting depth in the tropics or the lack of consideration of the effects of land use and vegetation fires on CO₂ exchanges when developing NEP estimates. The results of these modeling experiments proved to be inconclusive, but did provide information as to how NDVI data might be used to improve phenology and canopy allocation algorithms in terrestrial biosphere models (Sitch et al., in preparation; see also Section I. 2. of this Report).

The NPP estimates of TEM 4.0 across the globe were also compared to approximately 17 other global models in a model inter-comparison workshop (often referred to as Potsdam '95) sponsored by the International Geosphere-Biosphere Program's (IGBP) Task Force on Global Analysis, Interpretation, and Modelling (GAIM). In addition to our IDS, participants included members of the Schimel-IDS, Sellers-IDS, and Running MODIS Teams. Although the models all used the same climate data and soil texture data for the inter-comparison, the modelling groups could not agree on a common geo-referenced data set to describe vegetation distribution across the globe. The global NPP estimates varied by a factor of 2 among the seventeen models (see abstract by Moore et al. for the First GAIM Science Conference; Moore et al., in preparation). Relatively low global NPP was estimated by TEM 4.0 compared to the other global models. Differences in the NPP estimates among the models could be attributed to differences in: a) the sensitivity of the various models to climate (Ruimy et al., in preparation; Schloss et al., in preparation); b) calculation of water balance (Churkina et al., in preparation); c) phenology (Fischer et al., in preparation; Kicklighter et al., in preparation) and d) vegetation distribution used by the modeling groups (Fischer et al., in preparation; Schloss et al., in preparation).

For the Vegetation-Ecosystem Modeling Analysis Project (VEMAP), we compared the NPP and carbon storage estimates of the conterminous United States for contemporary climate and three climate change scenarios among TEM 4.0, Century (from the Schimel EOS-IDS Team) and Biome-BGC (from the Running MODIS Team). These biogeochemistry models also used the changes in vegetation distribution predicted by three biogeography models to examine the coupled response of ecosystem structure and function to climate change. Although the biogeochemistry models estimated similar NPP and total carbon storage for the conterminous United States under contemporary conditions, the response of NPP and total carbon storage to climate change varied among the models (VEMAP Members, 1995). All models estimated increases in NPP with climate change and elevated CO₂, but TEM predicted the largest increases (up to 40%). The models differed substantially in their estimates of the response of total carbon storage to climate change with TEM predicting increases in carbon storage and Biome-BGC predicting decreases in carbon storage. All three models showed correlations among water use, nitrogen availability and primary production, but the models simulated spatial variability in ecosystem processes in substantially different ways (Schimel et al., in press). To understand better these different responses, we are currently pursuing a detailed comparison of some of the mechanisms in each of the models such as water balance calculations (Cienciala et al., in preparation; Hibbard et al., in preparation) or how elevated CO₂ affects NPP (Pan et al., submitted).

In addition to the comparison of global models described above, we are also conducting a formal comparison of TEM to two regional versions of PnET model. The original version of PnET was developed specifically for all forest types in the northeastern United States and has been used in our IDS to provide high-resolution estimates of the effect of climate change and atmospheric deposition on the carbon, nitrogen, and water balances of forests in the northeastern United States (Aber et al., 1995). A modified version of PnET has also been developed to provide similar carbon, nitrogen and water balance estimates for forests in the southeastern United States by Dr. Steve McNulty of the US Forest Service. In the northeastern United States, TEM and PnET provide similar regional estimates of NPP, model bias occurs at both the low and high end of the NPP range (see abstract by Jenkins et al. for the 1996 annual meeting of the Ecological Society of America; Jenkins et al., in preparation). Although regional estimates of NPP are also similar between TEM and PnET for the entire southeastern United States, TEM estimates a higher NPP for temperate deciduous forests and a lower NPP for warm temperate mixed forests (McNulty et al., in preparation). In addition, NPP estimates for specific 0.5 degree grid cells can differ substantially between the two models in the southeastern United States.

NASA EOS funding has also partially supported the application of TEM as part of an integrated assessment activity being conducted in conjunction with the MIT Joint Program on Global Change. This activity involves coupling TEM with a reduced form GCM, an atmospheric chemistry model and an economics model (Xiao et al., 1995; Xiao et al., 1996a; Prinn et al., 1996; Xiao et al., in press; Xiao et al., submitted; Prinn et al., in preparation). In one analysis of this activity, we found that the linkage of TEM 4.0 to a 2-dimensional climate model was useful for impact assessment and uncertainty analysis within the integrated assessment framework at the scales of the globe, economic regions and biomes (Xiao et al., in press).

During 1995 and 1996, we also developed a transient version of TEM (Version 4.1) such that stocks and fluxes of carbon and nitrogen in terrestrial ecosystems can fluctuate in response to inter-annual variability in atmospheric CO₂ concentration and climate. The development of TEM 4.1 required the synchronization of water balance calculations with the calculation of carbon and nitrogen fluxes (Melillo et al., in preparation). In earlier, equilibrium versions of TEM, water balance variables (e.g., soil moisture, actual evapotranspiration) were estimated by an intermediate Water Balance Model (WBM; Vorosmarty et al., 1989) and these estimates were used as inputs into TEM (see Pan et al., 1996). Version 4.1 is being used to examine inter-annual variation of net primary production (NPP), net ecosystem production (NEP) and terrestrial carbon storage in response to historical and future changes in atmospheric CO₂ concentration and global climate as part of the Community Carbon Model Linkage Project (Melillo et al., in preparation; Heimann et al., in preparation) and the MIT Global Change Joint Program Global System Model (Prinn et al., 1996; Prinn et al., in preparation). In addition, a subset of these results has been used to examine inter-annual variations of NPP, NEP and carbon storage in the conterminous United States (Tian et al., in preparation). A transient version of TEM will also be used in the upcoming model comparisons of NPP and NEP in Phase II of VEMAP.

New work will involve additional modifications to TEM 4.1 to improve the interaction of carbon, nitrogen and water fluxes between terrestrial ecosystems and the atmosphere; and between terrestrial and aquatic ecosystems. Proposed model modifications will include changes in how TEM simulates carbon, nitrogen and water dynamics within terrestrial ecosystems; and how TEM is coupled to the atmosphere and aquatic ecosystems. For example, we will modify TEM to explicitly simulate nitrification, denitrification, leaching and biological N fixation. The model will also be able to use atmospheric N deposition prescribed from spatially-explicit data sets or estimated from atmospheric chemistry models. In addition, work will continue on the development of a transient version of TEM that will include changes in both ecosystem structure and function in collaboration with Colin Prentice and his research team from Lund, Sweden. Other approaches will be investigated as we turn our attention toward a more "dynamic" terrestrial ecosystem model. This, in turn, leads us to landuse and landuse change.

Currently, TEM does not incorporate landuse and landuse change in its simulations of global carbon and nitrogen dynamics. To assess the effects of human activities on global carbon and nitrogen dynamics, we need to know: 1) where these activities are occurring; and 2) how these activities effect ecosystem dynamics. To determine where agricultural activities are occurring, we are currently developing a cropland distribution model that simulates the spatial distribution of potential croplands under abiotic constraints (see abstract by Tian et al. for the 4th Biennial International Conference for Ecological Economics; Xiao et al., in preparation). The distribution of potential croplands will eventually be used as one input into a land use model that defines the spatial distribution of actual cropland. Other inputs to the land use model will include socio-economic constraints (e.g., human population, GNP per capita, crop productivity per unit area, food and nutritional requirements for people). The land cover and land use models will allow us to generate a global land cover data set of natural vegetation, croplands, and urban areas for future and contemporary climate conditions. We have also started to examine the effects of human activities on carbon dynamics by reviewing recent analyses of the consequences of deforestation on the global carbon budget (Melillo et al., 1996) and examining the consequences of forest-to-pasture conversion on CH₄ fluxes in the Brazilian Amazon Basin (Steudler et al., 1996).⁴

Summary of recent results using TEM:

• TEM simulations have shown that an interaction between elevated CO₂ and climate change also plays an important role in the response of terrestrial ecosystems to future climate change. Enhanced temperatures, associated with climate change, increases nitrogen availability such that more atmospheric CO₂ can be incorporated into plants than would be predicted by elevated CO₂.
• TEM was used to quantify the response of "natural" terrestrial ecosystems (landuse excluded) to transient atmospheric CO₂ concentrations over the last 200 years. A terrestrial sink of 0.8 Pg C was found in 1990 due to CO₂ fertilization alone.
• TEM was also used to examine the response of terrestrial ecosystems to inter-annual variability in air temperature and precipitation over the last 90 years. Global net primary productivity (NPP) was found to be more sensitive to inter-annual changes in precipitation than air temperature.
• TEM and work related to TEM has been used in the 1995 IPCC assessment of climate change (Kicklighter et al., 1994; Melillo, 1995; McGuire et al., in press).
• TEM was used in the 1995 assessment of the USDA Forest Service's Global Change Program (Joyce et al., 1995; McGuire and Joyce, 1995; Perez-Garcia et al., in press).
• TEM has been used as part of an integrated assessment study conducted by the MIT Joint Program on Global Change (Prinn et al., 1996; Prinn et al., in preparation).

I. 2. Diagnostic Tools: Remote Sensing for Terrestrial Models. This section could be termed: Integrating remote sensing and physical modeling to estimate the status of terrestrial vegetation. We have been considering the information content of optical remote sensing data, focusing on the dependence of reflectance on the relative positions of sun and sensor (bi-directional effects) to exploit MISR and MODIS on AM-1. We have developed a method for obtaining biophysical information about ecosystems from satellite data that uses physical models of canopy light environment in an "inverse-mode". Factors that obscure the results of other methods actually provide useful information in this approach. This effort is presented by Braswell et al. (1996) and represents an important collaboration with the IDS team of Schimel et al.

In Braswell et al. (1995 and 1996), we present an algorithm for the retrieval of fractional APAR (fAPAR), albedo, and other parameters from AVHRR (advanced very high resolution radiometer) reflectance measurements by inverting a modified version of the SAIL (scattering by arbitrarily inclined leaves) canopy radiative transfer model. The model is inverted using an effective bi-directional reflectance factor (BDRF) distribution created by aggregating AVHRR data into cells of size comparable to those used in current terrestrial biosphere models (50x50 km). Successful inversion results over an area in central Africa are presented and compared with a vegetation index-based analysis and other satellite data. The procedure also provides unique information on phenology derived from timing of changes in leaf optical properties and canopy structure. Our methods are unique in that they explicitly incorporate a priori ecological knowledge in the choice of model parameters and constraints. This approach can eventually be employed at pixel resolution with the EOS sensors, MODIS (moderate-resolution imaging spectrometer) and MISR (multiangle imaging spectro-radiometer).

In a parallel effort, Aber and Martin's project (MODLERS) brings together 14 Long-Term Ecological Research (LTER) sites and NASA's MODIS Land (MODLAND) Science Team in an effort to validate Earth Observation System-era global data sets.⁵ Using extensive ground data sets, ecosystem models, and remotely-sensed imagery, each LTER site is developing local maps of landcover class (LCC), leaf area index (LAI), and aboveground net primary productivity (NPP) for a 100 km² area at a pixel size of 25m. The fine resolution site data will be aggregated to the 1 km scale for comparison to large scale estimates of LAI, NPP, LCC developed by the MODLAND Science Team, which will be based upon MODIS observations.

Our participation in this effort is focused primarily at Harvard Forest, with some work at the Hubbard Brook Experimental Forest anticipated in 1997. We have identified the sampling area for Harvard Forest and have delivered a number of the following products to the MODLERS team: high resolution (30m) digital elevation model, atmospherically and geometrically corrected TM and AVIRIS data, TM and AVIRIS derived species composition, PnET estimates of productivity modelled using AVIRIS derived foliar nitrogen and species composition (See next Section).

I. 3. Preparing for Hyper Spectral Information. We are well aware that the original hyper spectral EOS instrument (HIRIS) is not being flown; however, it seems likely that new instruments will be flown that provide rich spectral information with spatial information finer than MODIS.⁶ Several efforts by our IDS support this path.⁷

A leaf spectral library has been developed over the past 6 years using a laboratory spectrometer, and this spectral library has been used by a number of investigators.⁸ An ASD FieldSpecFR instrument will be used for the measurement of fresh foliage spectra, field calibration targets, and solar irradiance to improve our own AVIRIS data analysis in addition to enhancing this spectral database.

In order to link the lab with the field and with space, we have collaborated with the US Forest Service Northeast Experimental Station during the past year to begin extensive sampling within the White Mt. National Forest.⁹ At the Bartlett Experimental Forest (BEF), 48 plots with a long history of productivity data collection have been sampled. An additional 48 Forest Inventory and Analysis plots (USDA Forest Service) were sampled within the White Mt. National Forest. Data collected at these sites includes foliar chemistry, specific leaf weight, canopy leaf area profile, tree diameter, nitrogen mineralization and litterfall. This field data collection was coordinated with a successful AVIRIS overflight of BEF on 28 June 1996.¹⁰ In support of the AVIRIS overflight, we were able to use field spectrometer to measure sky irradiance and calibration targets. This field spectrometer data will be used to obtain the best possible atmospheric corrections for the AVIRIS data.

We are extending this work to agricultural systems by working with the USDA Water Quality Demonstration Project for North Carolina to compile AVIRIS and GIS data for the Herrings Marsh Run watershed. This 4x8 km area is a mix of forest and agricultural land with more than 10 different crop species planted in (1-20 ha) fields.

This overall effort to expand our ability to use future space-borne, higher spectral information¹¹ directly links with one of our major themes: To assess and understand better the effects of historical land use on forest productivity and the response of these terrestrial systems to the combined stresses of landuse (present and future), chemical changes, including atmospheric CO₂, and climate change.

I. 4. The Agricultural and Trace Gas Component in Terrestrial Systems. The basic approach is to incorporate the DNDC model (developed by our colleague C. Li) into our terrestrial framework to address agricultural regions and to restructure TEM (as part of a larger restructuring to handle other forms of landuse change) to support trace gas studies. This effort is being lead jointly by Changsheng Li and Steve Frolking on the agricultural side and by Aber, Melillo, Frokling, and Moore on the ecosystem side; Skole leads the effort to determine actual landuse and land cover change, all of whom are co-investigators on our IDS. The effort in agricultural systems and grasslands has made significant strides in the last year.

Frolking participated in a workshop on improving the OECD/IPCC/IEA methodologies for generating national assessments of nitrous oxide emissions from agricultural soils and contribution to the workshop report (Mosier et al. 1996), which has now had both scientific and country review under the OECD/IPCC process.

Frolking completed an initial comparison of simulations of N₂O fluxes by the DNDC model with field data from the Central Plains Experimental Range in northeastern Colorado (draft report only at this point). This work has both illuminated a potential problem in the DNDC model, and set the groundwork for a larger model/data inter-comparison under the US Trace Gas Network Project (TRAGNET) banner, which will take place over the next between November 1996 and April 1997. Frolking was a co-organizer of a TRAGNET workshop held 15-18 October 1996 in Fort Collins, CO. At this workshop a number of trace gas flux measurement databases (CH₄, N₂O, and NO) were assembled and reviewed, and the framework was established for the initial TRAGNET model inter-comparison. This work is a natural extension of our inter-comparisons (Potsdam and VEMAP) in which we explore TEM against other models for natural ecosystems. Such inter-comparison are extremely valuable for identifying weaknesses in current structures as well as illuminating potential paths for future development (i.e. dynamic vegetation models)

The models participating in the TRAGNET model inter-comparison are DNDC/TEM (UNH/MBL.), CENTURY/NGAS (Colorado State), CASA (NASA-Ames), and EXPERTN (GSF Research Centre for Environment and Health, Germany). The initial sites selected for comparison and model evaluation are a dry, shortgrass steppe in Colorado, a fertilized meadow in Scotland, and a fertilized, cultivated agricultural field (barley and winter wheat) in Germany.

In a closely linked project (e.g. Holland et al., in press), we have been investigating the effect of the widespread mobilization of nitrogen into the atmosphere from industry, agriculture, and biomass burning and its subsequent deposition has the potential to alleviate nitrogen limitation of productivity in terrestrial ecosystems, and may contribute to enhanced terrestrial carbon uptake.

In a previous collaboration on nitrogen deposition (Townsend et al., 1996), a perturbation model was developed to quantify the terrestrial carbon storage resulting from human-induced nitrogen deposition. This effort has shown that carbon uptake is sensitive to the spatial distribution and quantity of deposited nitrogen.

To further examine the role of atmospheric chemistry and transport which can differ among chemical transport models, NOy deposition fields from five different 3-D chemical models (GCTM, GRANTOUR, IMAGES, MOGUNTIA, and ECHAM) were used to drive nitrogen deposition in simulations to evaluate the importance of the spatial distribution of nitrogen deposition for carbon uptake and to better quantify its magnitude and uncertainty. Because of differences in atmospheric sources of NOy, transport, resolution, and representation of chemistry, each of these models predicts distinct spatial patterns of nitrogen deposition on the global land surface; these differences lead to distinct patterns of carbon uptake that vary between 0.7 and 1.3 Gt. C y^-1 globally. Addition of NHx to NOy deposition increased C storage to between 1.5 and 2.0 Gt. of total terrestrial carbon storage annually. Less than 10% of the nitrogen was deposited on forests which were most able to respond with increased carbon storage because of the wide C:N ratio of wood as well as its long lifetime. However, the ability of terrestrial ecosystems to fix and store carbon may be compromised by the deleterious effects of nitrogen saturation, and by feedbacks operating through atmospheric chemistry, especially O₃ damage¹², as O₃ and N deposition tend to be spatially correlated. We estimate that N stimulation of terrestrial carbon uptake could result in a global carbon sink of 1.4 to 2.0 Gt. C y^-1, while the "missing terrestrial sink" is quite similar in magnitude.

In addition, because of the importance of soil organic matter to both the question of the carbon cycle (e.g., potential pool for storage) and to the issue of trace gas production, we have participated in a model inter-comparison with long-term soil carbon datasets that took place in Rothamsted, UK (Li et al. 1996; Smith et al. 1996).