Databases for Regional Biogeochemical Modeling I : Intercomparison of Land Use / Cover Datasets
Regional biogeochemical analyses are broad enough to represent large-scale patterns and processes, but specific enough to incorporate fine-scale, spatially explicit variables and are, therefore, an important tool for evaluating many aspects of ecosystem ecology and land management (Paustian et al. 1997). Recent international agreements, such as the United Nations Framework Convention on Climate Change and the Montreal Protocol, arose, in part, from regional biogeochemical modeling results, and require national-level assessments which are apt to be based, in part, upon regional biogeochemical model output (e.g. Cannell et al. 1999). Similarly, other environmental issues, such as effects of acid deposition (Butler and Likens 1991) and nitrogen deposition (Townsend et al. 1996), are often evaluated using regional biogeochemical models. The need for such data is a relatively recent phenomenon that is likely to broaden in the future.
Terrestrial vegetation regulates the flow of energy into terrestrial ecosystems and directly influences mass transfer of water, C, and nutrients and also exerts secondary influences on ecosystems biogeochemistry by affecting soil moisture, soil nutrient transformations, and microclimate. Understanding the physical structure, seasonal behavior, and chemical composition of vegetation is, therefore, of first order importance when evaluating regional biogeochemistry (Nemani and Running 1995). Thus various aspects of terrestrial vegetation have necessarily been incorporated into ecosystem models (Parton et al. 1987, Potter et al. 1993, Running et al. 1994) which have been used to estimate rates of primary production (Paruelo et al. 1997, Running 1990), trace gas fluxes (Davidson et al. 1998), and changes in ecosystem C pools (Burke et al. 1990, Paustian et al. 1995) at the regional scale.
Utilization of land cover data from a variety of sources of uncertain quality in regional biogeochemical models belies the importance of vegetation parameters to biogeochemical models. Currently available land cover data sets use various collection techniques and classification methods and have distinct spatial and temporal resolutions. While agreement between different sources of data can be quite good at local scales (Iverson et al. 1989, Lathrop and Bognar 1994, Mladenoff et al. 1997) or when one type of land predominates (Mladenoff et al. 1997, Turner et al. 1993, Woodcock et al. 1994), it is unclear how accurately regional-scale land cover classes are reproduced by different data sources. Differences between land cover data sources may influence regional-scale model output (Turner et al. 1996). While major ground-truth investigations are the best way to evaluate the accuracy of land cover data sets (e.g. Bauer et al. 1994), examining precision of a number of sources of land cover data through intercomparison can be insightful since sources of data are often chosen for reasons other than perceived accuracy. Evaluation of data sources has largely been limited to intercomparison of forested land area estimates (Iverson et al. 1989, Lathrop and Bognar 1994, Turner et al. 1993).
We evaluated four readily available land use datasets which cover a range of environments within the conterminous United States. Specifically, we compared National Resources Inventory data (USDA 1994) with three remotely sensed data sets at four spatial scales. We assessed the particular sensitivities of the data sets, the level of concurrence between the datasets at various spatial scales, and the overall utility of the datasets for different purposes. Three states with variable climate, topography, land use, and population were used as case studies to ensure that our results were applicable to a range of environments. Results of this study indicate how incorporation of different land cover datasets may influence regional land cover estimates and, thus, influence biogeochemical model results. We also assessed the benefits and detriments of various methods of data collection, which will be particularly enlightening for development of global scale datasets.
References | Results
This research is funded by EPA