Mapping and monitoring regional patterns of species richness from geographic information

Biological diversity has become a major scientific and political issue, producing an urgent need for inventory and monitoring programs. Remote sensing provides tools to satisfy part of this need, but there has been no scientific framework for guiding its application in biodiversity assessments. A research agenda is proposed to expand our knowledge of the role remote sensing might play in providing improved information on the spatial distribution of species richness and its ecological determinants, and the response of these ecological factors to global change. Many physical and biological factors that are correlated with species richness have been mapped with remote sensing, including landscape geometry, primary productivity, and evapotranspiration. Additional research is required to apply remote sensing methods to the assessment of biodiversity in the context of earth system science and global change programs.; Sensitivity of maps of predicted species richness to spatial scale and habitat map generalization and accuracy were examined by means of a geographic information system (GlS) sensitivity analysis. Wildlife-habitat relationships (WHR) models were integrated with a map of habitats to predict species number within uniform grid cells for two distinct ecoregions in Idaho. Patterns of richness varied unpredictably with size of the spatial sampling units because of the complex interaction of factors that affect richness. For statewide Gap Analysis, a range of grid sizes between 10-100,000 ha are recommended for the Rocky Mountain Forest ecoregion and 10-60,000 ha for the Intermountain Sagebrush ecoregion. Contiguous, non-overlapping grids provide adequate sampling density.; Another GIS sensitivity analysis ascertained the effects of the level of generalization (minimum mapping unit) and accuracy of the habitat map on the predicted distribution of richness in the southern Sierra Nevada, California. Predicted richness declines monotonically as the habitat map is generalized, due to reduction in the number of habitat types mapped in a quadrat. Misclassification had the opposite effect of predicting more species than the baseline model. Both factors produced changes in the grid cells predicted as having the most species. It is expected that these effects diminish as sampling unit size increases.