Automated Monitoring of Snow Cover over North America with Multispectral Satellite Data

Abstract

Current National Oceanic and Atmospheric Administration (NOAA) operational global- and continental-scale snow cover maps are produced interactively by visual analysis of satellite imagery. This snow product is subjective, and its preparation requires a substantial daily human effort. The primary objective of the current study was to develop an automated system that could provide NOAA analysts with a first-guess snow cover map and thus to reduce the human labor in the daily snow cover analysis. The proposed system uses a combination of observations in the visible, midinfrared, and infrared made by the Imager instrument aboard Geostationary Operational Environmental Satellites (GOES) and microwave observations of the Special Sensor Microwave Imager (SSM/I) aboard the polar-orbiting Defense Meteorological Satellite Program platform. The devised technique was applied to satellite data for mapping snow cover for the North American continent during the winter season of 1998/99. To assess the system performance, the automatically produced snow maps were compared with the NOAA interactive operational product and were validated against in situ land surface observations. Validation tests revealed that in 85% of cases the automated snow maps fit exactly the ground snow cover reports. Snow identification with the combination of GOES and SSM/I observations was found to be more efficient than the one based solely on satellite microwave data. Comparisons between the automated maps and the NOAA operational product have shown their good agreement in the distribution of snow cover and its area coverage. The accuracy of the automated product was found to be similar to and sometimes higher than the accuracy of the operational snow cover maps manually produced at NOAA.