A NEW ALGORITHM APPROACH FOR THE RETRIEVAL OF SNOW DEPTH USING PASSIVE MICROWAVE SATELLITE DATA

Snow depth retrieval algorithms from passive microwave satellite data have been trained and tested using stationmeasured snow depth.Although locally accurate, station measurements are sparce in many remote areas of the globe.In addition, the over-reliance on ground truth data is not a good algorithm development practice.Unlike station-measured snow depth, snow reanalysis data such as those developed at European Center for Medium-Range Weather Forecasts (ECMWF) and NOAA's National Centers for Environmental Prediction (NCEP) are essentially complete.These are high quality global snow data produced by employing advanced blending techniques of land surface models and observations.To take advantage of snow reanalysis data, a new satellite snow depth retrieval concept has been conceived that will rely on reanalysis data for training new algorithms and station data for testing, evaluation and monitoring.Another change in our approach is the development of regression-based algorithms with spatially variable coefficients.This is a new approach and development paradigm that we are launching to develop improved satellite algorithms at NOAA/NESDIS across current and future satellite missions.Our experience has shown that it remains challenging to accurately model local terrain factors impacting the passive microwave signal by generic physical or statistical approaches.To this end, ECMWF and NCEP snow depth reanalysis data have been collected and matched up to satellite data from microwave instruments on board NOAA's Joint Polar Satellite System (JPSS) and European Space Agency's Metop-SG satellite platforms.The global matched up dataset is being analyzed to derive the set of snow depth regression coefficients.