Snow cover mapping using AVHRR images and GIS for snowmelt runoff prediction

A procedure is developed that links a geographic information system (GIS) and remotely sensed data to determine the extent of snow cover in a complex mountainous watershed. Snow cover extent is used in a hydrologic model to predict snowmelt and rainfall produced runoff. A series of programs (AMLs) are used in the GIS ARC/INFO to reclassify AVHRR satellite data for snow cover extent, and to delineate the watershed boundary and elevation zones. Four GIS data layers (obtained from AVHRR channels 1, 3, NDVI and derived channel 6) were used in the reclassification scheme, which identified thresholds for the identification of snow-covered or snow-free areas. The scheme correctly predicted the presence or absence of snow cover 72% of the time using one evaluation method, and 78% of the time using a different evaluation method. The low resolution of the AVHRR data (I-km pixel size) and the small area of analysis probably reduced the accuracy of the results. Other possible inaccuracies in this method may be due to temperature variations that would cause snow to either remain or melt sooner or later than expected, low or ice clouds mistaken for snow cover, snow cover under clouds mistaken for clouds, or errors in the climate data. The use of subpixel analysis, further validation of the spectral differences between snow, cloud and land in AVHRR data using higher resolution satellite images, or ground truthing could refine this procedure.