|
Abstract:
|
Regressions between streams and their watersheds' land uses confront several challenges, including complex spatial correlation structures of stream networks, highly multivariate responses for stream biology, and uncertain regressors derived from digital maps. Another challenge is to identify the areas within watersheds whose land uses best predict stream responses. GIS tools can be used to carve out candidate areas of influence, and the best-predicting candidate might be revealed by variable selection methods. Alternatively, I suggest summing the influences of all cells in the watershed's land use grid coverage for use as a regressor. Cell influence is modeled as a decaying function of cell distance upstream and upslope, and estimated influence decay rates describe the watershed areas of greatest influence. The model is illustrated for 52 small watersheds in Oregon's Willamette Valley.
|
