Snow Depth Spatial Distribution Using Microwave Remote Sensing at the Puna Tsang River Basin in Bhutan

Spatial distribution of snow amount derived from satellite observations has been previously achieved for flat region. But for mountainous regions, spatial distribution of snow amount has not been addressed because remote sensing instruments are very sensitive to the effect of the terrain slope; and because there is no available data for validation. This study focuses on the estimation of snow amount using a microwave radiative transfer model (RTM) in mountain region. AMSR-E satellite observations of brightness temperature (Tb) at 18.7GHz and 36.5GHz frequencies are compared to calculated values of Tb in Lookup Tables generated by the RTM model. The model uses a snow algorithm to derive the snow depth and temperature spatial distribution over the target region. This snow algorithm has been previously validated in a flat region using in-situ recorded snow-depth data. In this study, the local slope in mountainous terrain, where the local incidence angle is different than the 55 degree incidence angle from the satellite, is taken into account. The local incidence angle is calculated from the scalar product between the radiometer scanning vector and the surface normal vector of the local slope. The terrain DEM is used to calculate the slope and aspect of each terrain grid. Then, with the geolocation of the satellite as it passes over, the local incidence angle is computed. AMSR-E data resolution is about 25x25 km but at this resolution we can not meaningfully express the topographic terrain. Therefore, a DEM resolution of 1x1 km is used. To overcome the difference of spatial resolution between the satellite observation and the terrain grid, the approach is to estimate the Tb for the 18.7GHz and 36.5GHz frequencies with the local incidence angle for each terrain grid. Then, an averaged Tb for each footprint is computed from the weighted average of the Tb of each terrain grid based on the count of occurrence of the same local incidence angle. Then, the averaged Tb is compared with the observed Tb from the satellite. This strategy is applied since we assume a uniform snow depth and snow temperature within each satellite observation footprint. The lookup tables are generated from the brightness temperature by inputting the snow depth and snow temperature into the RTM model. For a range of local incidence angles, snow depths, and snow temperatures a lookup table is prepared. That is, one 18.7GHz Tb and one 36.5GHz Tb are calculated for each combination of snow depth and temperature for each terrain grid. Then, the calculation is reversed to obtain the snow depth and temperature by inputting the observed Tb from the satellite. The snow algorithm compares the observed Tb to the calculated average Tb in the lookup table, and estimates the corresponding snow depth and temperature for each observation footprint. The target region for this study is the Puna Tsang River Basin in Bhutan. Because there is no available data for validation, we validated the model results at this basin with output of snow depth from a hydrological model. The model Water & Energy Budget based Distributed Hydrological Model with improved Snow physics (WEB-DHM-S), is used to evaluate the RTM model performance. WEB-DHM-S outputs of stream discharge and snow cover area are previously validated with measured flow discharge and observed snow cover area by MODIS.