Spatial modeling and prediction of snow-water equivalent using ground-based, airborne, and satellite snow data

In this research we modify existing spatial interpolation methodologies so that we can use ground-based and remotely sensed (airborne and satellite) snow data to characterize the spatial distribution of snow-water equivalent (SWE) and obtain optimal gridded SWE predictions in the upper Mississippi River basin. We developed and tested the models using ground-based, airborne, and satellite snow data collected over North and South Dakota, Minnesota, Wisconsin, Iowa, Illinois, and Michigan between March 3 and 6, 1996. Using these data and the spatial models, we obtained optimal gridded predictions of SWE and the associated root mean square prediction errors over a 5 min by 5 min grid covering Minnesota and parts of Wisconsin, North and South Dakota, Iowa, Michigan, and Canada. Because we use an optimal interpolation technique and incorporate satellite areal extent of snow cover data, the predictions are expected to be more accurate than those that would be obtained from interpolation procedures currently used by the National Weather Service. Maps of the gridded snow water equivalent predictions and of the associated error estimates provide a means to investigate the spatial distributions of the predictions and of the associated error estimates. Our research enables hydrologists and others not only to examine these spatial distributions but also to generate optimal gridded predictions of snow-water equivalent that will aid flood forecasting and water resource management efforts.