Flow diagram describing the snow data processing and soil modeling procedure. A spatial filter was used to predict the snowoccurrence probability for MODIS cloud-contaminated pixels. Based on the cloud-free MODIS SCE imagery, the snow depth of each snowcovered 1 km pixel was estimated from the snow depth of the neighboring MERRA-2 grid cells with weights predicted using another spatialfilter. The downscaled 1 km snow depth data were then used to drive a remote-sensing-based soil process model to simulate the zero-curtainperiod and ALT. Both the snow data processing and soil modeling were carried out at an 8-day time step.