McKeown, R., Dennis S. Ojima, T.G.F. Kittel, D.S. Schimel, W.J. Parton, H. Fisher, T. Painter
Abstract
Ecosystem and atmospheric scientists are concerned with understanding how biospheric characteristics of the land surface will be modified in response to changing climate and land use. Ecosystem properties such as vegetation structure, carbon (C) fluxes, water exchange, and nitrogen (N) feedbacks to ecosystem dynamics respond differently when perturbed by climate or land use changes. These differences are reflected in both spatial and temporal features of the land surface characterization. Using ecosystem models to simulate changes in ecosystem properties due to land use and climate change perturbations across different land cover classes provides a way to test the sensitivity of different ecosystems.
Our analysis uses the spatially explicit version of the CENTURY ecosystem model to simulate ecosystem dynamics within a region. Site information including monthly rainfall, temperature, N deposition, soil properties, vegetation type and land-use management were defined for each grid cell. Simulation results based on a coarse grid regional representation vary significantly from those based on a finer grid. Attempting to resolve these discrepancies with simple spatial averaging of the driving variables often leads to conditions not actually observed in the region of interest. This led us to develop techniques to statistically represent sub-grid variability to better parameterize CENTURY and to assess changes in ecosystem dynamics resulting from changes in environmental factors. The results indicate that spatially-weighted averaging of results based on combinations of ecosystem drivers and properties which actually occur provide a better fit to observed ecosystem dynamics.
Introduction
The dominant factors controlling ecosystem dynamics are climate, nutrient availability, and land use. Our ability to predict changes in ecosystem dynamics and human welfare relative to climate or land use changes is dependent on the development of analytical tools to integrate our current understanding of how these ecosystems behave relative to human and environmental factors. The analysis of this information will need to incorporate critical factors of the physical environment (climate, land cover and soil) as well as factors defining human interactions with the environment (economic, social and cultural). The importance of past and current climate and land use cannot be overlooked in assessing how these ecosystems have developed over time and how they may change in the future relative to new policies, technological advances, economic conditions, and environmental constraints.
We used a spatially explicit version of the CENTURY model to examine the combined effects of climate and land use change on net carbon exchange in the central region of the US. Here, we present results of simulations performed at 0.5-degree resolution using soils and vegetation datasets that include a statistical representation of finer resolution information. Results of simulations made with and without sub-grid heterogeneity and with potential versus current vegetation cover are compared.
Model Description
The CENTURY ecosystem model Version 4 [Parton et al., 1987, 1993; Metherell, 1992] is a general model of plant-soil nutrient cycling which has been used to perform simulations across systems (including grasslands, agricultural lands, forests, and savannas) in various geographic regions. CENTURY is composed of several submodels: a soil organic matter / decomposition submodel, a water budget model, a grassland / crop production submodel, a forest production submodel, and management and events scheduling functions. The model computes the flow of C and N through a system of compartments using a monthly timestep. The following variables are required as input.
CENTURY incorporates simplified representations of key processes relating to carbon assimilation and turnover, including the impact of cropping, tillage, harvest, fire, grazing, and storm disturbances on ecosystems [Ojima et al., 1990; Sanford et al., 1991; Holland et al., 1992; Metherell, 1992]. Therefore CENTURY also requires parameters which specify disturbance types and frequencies for each biome.
As mentioned above, CENTURY consists of linked submodels. The soil organic matter (SOM) submodel simulates the decomposition of plant residues by microbes. The resulting microbial products become the substrates for SOM formation. The SOM is divided into three fractions: an active soil fraction representing live microbes and microbial products (1 to 4 yr. turnover time); a protected fraction representing the organic matter which is more resistant to decomposition as a result of physical or chemical protection (20 to 40 yr. turnover time); and a fraction that is physically protected or chemically resistant and has a long turnover time (800 to 1200 yr.). The water budget model calculates monthly evaporation, transpiration, the water content of the soil layers and snow, and saturated flow of water between soil layers. The plant production submodels both assume that the monthly maximum plant production is controlled by moisture and temperature, and that maximum plant production rates decrease if there are insufficient nutrient supplies. The grassland / crop production submodel simulates plant production for herbaceous crops and plant communities. The forest submodel simulates the growth of deciduous or evergreen forest in juvenile and mature phases. To simulate a savanna or shrubland, CENTURY uses both the grassland and forest submodels with some additional functions to perform nutrient competition and shading effects.
The recently developed spatially explicit version of CENTURY uses gridded maps of site specific driving variables as input (Figure 1). A simulation using this version of CENTURY begins by accessing the site information (climate, soil texture, and land use classification parameters) for a particular cell. Once the model has information on the land use classification for a cell, the model accesses the schedule file associated with that land use. Information contained in the schedule file determines how CENTURY assigns management parameters, when events occur, and when the simulation ends.
CENTURY is an inherently transient model rather than an equilibrium model. In order to bring the C and N pools of simulated ecosystems to levels which are reasonable representations of existing ecosystems, CENTURY was run for at least 2000 years for each grid cell with prescribed disturbance regimes for specific biomes. Due to a lack of information about early site management, a generalized pattern for each system was modified so that the 2000 year run yields a representation similar to current conditions.
Geographical Data Bases
The domain for this model experiment was a region in the central grasslands of the United States and adjacent Rocky Mountains, including parts of Nebraska, Kansas, Colorado, and Wyoming. We made use of soil property (including soil texture and depth) and vegetation class data at two different scales of spatial resolution, 0.5-degree [from the VEMAP database, Kittel et al. 1995] and 10-km. The soil data were based on Kern's [1994, 1995] 10-km gridded Soil Conservation Service national-level (NATSGO) database. We used cluster analysis to group the 10-km sub-grid elements into 4 dominant soil types (modes) for each 0.5-degree cell. With this approach, soils properties for a 0.5-degree cell are represented by 1 or more dominant soil profiles rather than by an ``average soil profile'' which may not correspond to an actual soil in the region.
The vegetation classes were those used by VEMAP [VEMAP Members 1995, Kittel et al. 1995]. The classes were defined physiognomically in terms of dominant life-form and leaf characteristics including leaf seasonal duration, shape, and size [Running et al. 1994]. In the case of grasslands, the classes were defined physiologically with respect to dominance of species with the C3- versus C4-photosynthetic pathway. Distribution of these classes was based on a 10-km gridded map of Kuchler's [1964, 1975] potential natural vegetation [NGDC in press]. For the purpose of this exercise, we assumed that this distribution of potential natural vegetation is in equilibrium with current climate. Current vegetation was estimated using the EROS Data Center 1-km land cover data for the US [Loveland et al. 1991], derived from AVHRR NDVI data. (Figure 2)
To avoid choosing vegetation-soil combinations which may not exist for a grid cell, actual vegetation-soil pairs were identified using bivariate histograms [Kittel et al. 1996, (Figure 3)]. The combination considered dominant was the most frequently occurring soil texture (%sand, %clay, bulk density) and the most frequently occurring land cover on that soil texture for each cell. The land cover and soil information was always coupled with the soil modes selected first and then 4 land cover modes selected for each soil mode. Simulations were run using the 16 most frequently occurring vegetation-soil combinations and the results combined on an area -weighted basis. This allowed us to include sub-grid information in our simulations without explicitly increasing the resolution of the simulations.
Methods
CENTURY simulations were run at a 0.5-degree resolution. Site input for a single grid cell's simulation consists of monthly precipitation, mean monthly maximum temperature, mean monthly minimum temperature, and a combination of land cover and soil texture. Each of the site input parameters is stored in a separate gridded map. The simulation of ecosystem dynamics for a single cell is independent of all other cells. To reduce I/O, CENTURY runs a single cell's entire scenario (all timesteps) before moving onto the next cell.
To account for sub-grid heterogeneity we ran 16 full grid simulations using identical climate information, each driven by different map sets containing existing land cover - soils combination. For all output variables, we used information on the area covered by the existing land cover - soils combinations to compute an area-weighted value for each cell.
Results: Land Cover Change
Across the region, potential natural vegetation based on the Kuchler vegetation classification was dominated by shortgrass steppe (roughly 60% of the area). Approximately 15% of the potential natural vegetation was classified as forest. The remaining area was occupied by cool-season grasses and tree-grass associations. The current land cover dataset indicates that cropland conversion resulted in a 33% transfer of the total land area into wheat-fallow / grassland mixture (class 118) mostly from shortgrass steppe.
Simulated soil C levels in the potential natural vegetation simulation and the simulation of current conditions were markedly different In the area converted to croplands from the shortgrass steppe, regional soil C levels averaged approximately 2.9 kg/m2. Higher soil C in croplands prior to conversion (~3.8 kg/m2) compared to shortgrass steppe not converted(~2.7 kg/m2) suggests that initial soil quality and organic matter content may have played a role in conversion management decisions (Figure 4).
Primary plant productivity for the region also indicated that differences between regions converted to cropland and non-converted areas existed. The Net Primary Production (NPP) for potential vegetation in grassland areas that were converted to croplands were about 72% greater than those areas not converted. This difference was greater than the relative change in NPP with cropping (Figure 5).
Results: Sub-grid Heterogeneity
When single dominant current vegetation and soil classes were simulated over the region, soil C levels declined by about 25% for the grid cells where cropland conversion took place. Soil C for the entire region dropped by about 10%. In simulations that included sub-grid information for land cover and soil, we saw an even greater loss in soil C. The grid cells in which land cover conversions took place showed a 32% loss in soil C from potential natural vegetation. An overall loss of 22% was seen for the entire region.
Conclusions
Land cover (vegetation), whether natural or human induced, is strongly correlated to environmental factors such as soil texture. In compiling data used to drive simulations these factors should not be decoupled. The non-linear behavior of ecosystem dynamics with respect to land cover change would lead to uncertain results. Using bivariate dataset development techniques, we were able to simulate sub-grid heterogeneity and maintain dependencies inherent in the data without increasing the resolution of the simulations. Although the technique we used required multiple 0.5-degree simulations, it still required fewer simulations than running at the resolution which was represented statistically by the bivariate data. Additional analysis of factors which determine the selection of areas for particular uses that go beyond the biophysical determinants also need to be incorporated into the data. This information is more difficult to quantify. However, efforts are underway to evaluate the contextual nature of social and economic factors in selection of areas for different land uses.
Acknowledgments
Support for this research was provided by the National Aeronautics and Space Administration's EOS Program through an EOS Interdisciplinary Science grant to David S. Schimel, and the VEMAP sponsors (the Electric Power Research Institute, the National Aeronautics and Space Administration, and the U.S. Department of Agriculture Forest Service). We also acknowledge the National Science Foundation support of the Climate System Modeling Program of the University Corporation for Atmospheric Research. We thank Nan Rosenbloom for her assistance in preparing datasets and post-processing.
References
Holland,E.A., W. J. Parton, J. K. Detling, and D. L. Coppock, Physiological response of plant population to herbivory and their consequences for ecosystem nutrient flow, Am. Nat., 140(4), 685-706, 1992.
Kern, J. S., Spatial patterns of soil organic carbon in the contiguous United States, Soil Sci. Soc. Am. J., 58, 439-455, 1994.
Kern, J. S., Geographic patterns of Soil water-holding capacity in the contiguous United States, Soil Sci Soc. Am. J., 59, 1126-1133, 1995.
Kittel, T. G. F., N. A. Rosenbloom, T. H. Painter, D. S. Schimel, and VEMAP Modeling Participants, The VEMAP integrated database for modeling United States ecosystem/vegetation sensitivity to climate change. J. Biogeography, 22, in press, 1995a.
Kittel, T. G. F., D. S. Ojima, D. S. Schimel, R. McKeown, J.G. Bromberg, T.H. Painter, N. A. Rosenbloom, W. J. Parton, and F. Giorgi, Model-GIS integration and dataset development for assessing the vulnerability of terrestrial ecosystems to climate change, in Proceedings Volume, NCGIA Second International Conference on Integrating GIS and Environmental Modeling, GIS World Books, Ft. Collins, Colo., 1996.
Kuchler, A. W., Potential Natural Vegetation of the Conterminous United States, manual to accompany the map, Spec. Publ. 36, 143pp., Am. Geogr. Soc., New York, 1964.
Kuchler, A. W., Potential Natural Vegetation of the United States, 2nd ed., map 1:3, 168,000, Am. Geogr. Soc., New York, 1975.
Loveland, T. T., J. W. Merchant, D. O. Ohlen, and J. F. Brown, Development of a land-cover characteristics database for the conterminous U. S., Photo. Eng. Rem. Sensing 57:1454-1463, 1991.
Metherell, A. K., Simulation of soil organic matter dynamics and nutrient cycling in agroecosystems, Ph.D. Dissertation, Colorado State University, Fort Collins, 1992.
NGDC. NOAA/EPA Global Ecosystems Database Project, Global Ecosystems Database Version 1.0, Disk B. National Geophysical Data Center, Boulder, Colo., Digital data, (in press).
Ojima, D. S., W. J. Parton, D. S. Schimel and C. E. Owensby, Simulated impacts of annual burning on prairie ecosystems, in Fire in North American Tallgrass Prairies, edited by S. L. Collins and L. L. Wallace 175 pp., Univ. of Oklahoma Press, Norman, Okla. and London, England, 1990.
Ojima, D. S., D. S. Schimel, W. J. Parton, and C. E. Owensby, Short- and long-term effects of fire on N cycling in tallgrass prairie. Biogeochemistry, 24:67-84, 1994.
Ojima, D. S., K. A. Galvin, and B. L. Turner II, The global impact of land-use change. BioScience, 44:300-304, 1994.
Parton, W. J., D. S. Schimel, C. V. Cole, and D. C. Ojima, Analysis of factors controlling soil organic matter levels in Great Plains grasslands, SSSAJ 51(5):1173-1179, 1987.
Parton, W. J., C. V. Cole, J. S. B. Stewart, D. S. Ojima, and D. S. Schimel, Simulating regional patterns of soil C, N, and P dynamics in the U.S. Central grassland region, in Ecology of Arable Land, edited by M. Clarholm and L. Bergstrom, pp. 99-108, Kluwer Academic Publishers, 1989.
Parton, W. J., et al., Observations and modeling of biomass and soil organic matter dynamics for the grassland biome worldwide, Global Biogeochem. Cycles, 7(4), 785-809, 1993.
Riebsame, W. E., W. J. Parton, K. A. Galvin, I. C. Burke, L. Bohren, R. Young, and E. Knop, Integrated modeling of land use and cover change. A conceptual scheme for applying and integration strategy to agricultural land use on the U.S. Great Plains, BioScience 44(5):350-356, 1994.
Running, S. W., T. R. Loveland, and L. L. Pierce, A vegetation classification logic based on remote sensing for use in global biogeochemical models, Ambio, 23(1), 77-81, 1994.
Sanford, Jr., R. L., W. J. Parton, D. S. Ojima, and D. J. Lodge, Hurricane effect on soil organic matter dynamics and forest production in the Luquillo Experimental Forest, Puerto Rico: Results of simulation modeling, Biotropica, 23:364-372, 1991.
VEMAP Participants, Vegetation/Ecosystem modeling and analysis project(VEMAP): Comparing biogeography and biogeochemistry models in a continental-scale study of terrestrial ecosystem responses to simulate change and CO2 doubling, Global Biogeochem. Cycle, 9, 407-438, 1995
Author Information
Rebecca McKeown
Associate Scientist: National Center for Atmospheric
Research (NCAR)
Research Associate: Natural Resource Ecology Laboratory
(NREL)
Colorado State University, Ft. Collins, CO 80523
Phone: (970)
491-1623
FAX: (970) 491-1965
beckym@NREL.colostate.edu
Dennis Ojima
Research Scientist: NREL
Colorado State University,
Ft. Collins, CO 80523
Phone: (970) 491-1976
FAX: (970) 491-1965
dennis@NREL.colostate.edu
Timothy Kittel
Deputy Project Scientist: Climate System Modeling Program
(CSMP
Research Associate: NREL
University Corporation for Atmospheric
Research
P.O. Box 3000, Boulder, CO 80307
Phone: (303) 497-1611
FAX:
(303) 497-1695
kittel@ncar.ucar.edu
Thomas Painter
Graduate Student: Dept. of Geography
University of
California, Santa Barbara
painter@crseo.ucsb.edu
Hank Fisher
University Corporation for
Atmospheric Research
P.O. Box 3000, Boulder, CO 80307
Phone: (303)
497-1617
FAX: (303) 497-1695
fisherh@ncar.ucar.edu
David Schimel
Scientist, Section Head: Climate and Global Dynamics
Division (CGD)
Senior Scientist: NREL
University Corporation for
Atmospheric Research
P.O. Box 3000, Boulder, CO 80307
Phone: (303)
497-1610
FAX: (303) 497-1695
schimel@ncar.ucar.edu
William Parton
Senior Scientist: NREL
Colorado State University,
Ft. Collins, CO 80523
Phone: (970) 491-1987
FAX: (970) 491-1965
billp@NREL.colostate.edu
