Calibration of a distributed snow model using MODIS snow covered area data

Spatial ground-based observations of snow are often limited at the watershed-scale, therefore the snow modeling component of a hydrologic modeling system is often calibrated along with the rainfall–runoff model using watershed discharge observations. This practice works relatively well for lumped modeling applications when the accuracy of sub-watershed processes is generally not of concern. However, with the increasing use of distributed models, realistic representation of processes, such as snow areal depletion, become more important. In this study, we test the use of snow covered area (SCA) data from the Moderate Resolution Imaging Spectroradiometer (MODIS) on board the Terra satellite for calibration of four key parameters in the distributed US National Weather Service (NWS) SNOW17 model in the North Fork of the American River basin in California, USA. Three tests are conducted; two rely solely on MODIS SCA data and one includes discharge in the calibration procedure. The three calibrations are compared to the use of parameters obtained from the NWS California Nevada River Forecast Center (CNRFC). The calibration approach that utilizes both MODIS SCA and discharge data produces the most accurate spatial (gridded) SCA and basin discharge simulations but not the best SCA summary statistics. In general it was found that improvement in simulated SCA when averaged and evaluated by elevation zone using standard summary statistics, does not necessarily coincide with more accurate discharge simulations.