| Environmental modeling has long been a recognized part of environmental risk assessment. Traditionally environmental modeling, specifically water quality modeling has been done over small areas and single watersheds, using spatially aggregated subunits for data collection and processing. For an analysis of environmental risk factors on a regional scale, models need to account for uncertainties in the processes, and heterogeneity of the environment, leading to extensive data and analysis requirements. Under such circumstances, spatial empirical modeling can offer useful insights because it can deal with a combination of processes, and provide a broader perspective on the degree and significance of the impacts on ecological resources at a regional scale. Using data from the eastern cornbelt plains ecoregion of Ohio, empirical models were developed relating watershed anthropogenic activities and a set of stream condition variables to the biological quality of the stream segments. The effects of land use, soil characteristics, and in-stream physical, and chemical factors on the biological indicators have been studied using nested watersheds as geographic units. The effects of scale, spatial aggregation, spatial autocorrelation, and alternative distance decay formulations for chemical pollutants, have been documented.;Data for this model were integrated from various sources. In-stream biological, chemical, and physical habitat data, watershed, and ecoregion geography, and stream line files, were obtained from the United States and the Ohio Environmental Protection Agency databases. Land use was obtained by classifying Landsat Thematic Mapper images in conjunction with the Census Bureau's Topologically Integrated Geographic Encoding and Referencing system (TIGER) data. Soil characteristics on a regional scale were extracted from the United States Department of Agriculture (STATSGO) database. A geographic information system (GIS) database was built using Arc/Info software for integrating and summarizing the data. Statistical packages, SPSS, and S-PLUS were used for subsequent modeling. The results show different spatial autocorrelation patterns for groups of watersheds within the ecoregion. Future work will improve the model for prediction of ecological indicators using spatially autoregressive functions and reliable confidence intervals by integrating input spatial data accuracy estimates with empirical estimates. |
