The effect of distance-weighted source areas in geographic predictions of nutrient discharge from coastal-plain watersheds

M.E. Baker, D.E. Weller, and T.E. Jordan. Smithsonian Environmental Research Center

Geographic predictions of nutrient discharge typically use proportional summaries of source areas derived from land cover maps. Proportions lack information about the spatial distribution of source areas, so correlations with similar characterizations of watershed-wide or riparian sinks often confound statistical models of nutrient discharge and hamper the explicit interpretation of results. To explore whether characterizing the spatial arrangement of source areas improved predictions of nutrient discharge, we developed a series of maps representing simple measures of Euclidean distance to watershed outlets and stream networks as well as more complex expressions of topographic and hydrologic proximity. We compared the performance of distance-weighted cropland and development in statistical models of nutrient discharge from more than 320 study watersheds in Maryland’s Coastal Plain. Different distance weighting schemes had significant impacts on nutrient discharge predictions and the relative effectiveness of the weighting schemes varied depending on the land cover class considered. Our empirical results support the importance of spatial arrangement in characterizing watershed nutrient sources. In addition, different distance weights represent alternative hypotheses about patterns and potential mechanisms of material transport, so effective weighting schemes may help resolve the physical factors controlling variation of nutrient discharge within and among watersheds.