Fractional Snowcover Estimates from the Earth Observing System (EOS) Terra and Aqua Moderate Resolution Imaging Spectroradiometer (MODIS)

The Moderate Resolution Imaging Spectroradiometer (MODIS) on the NASA Earth Observing System (EOS) Terra and Aqua missions has shown considerable capability for mapping snowcover. The current MODIS approach being used applies the Normalized Snow Difference Index (NDSI). The NDSI takes the difference between 500 meter observations at 1.64 micrometers (MODIS band 6) and 0.555 micrometers (MODIS band 4) divided by the sum of these observations to determine whether MODIS pixels are snowcovered or not in mapping the extent of snowcover. NSDI is an effective means of mapping snowcover because of the spectral contrast of snow in these two bands. For many hydrological and climate studies using remote sensing of snowcover, it is desirable to determine if the MODIS snowcover observations could not be enhanced by providing the fraction of snowcover in each MODIS observation (pixel). Pursuant to this objective, studies have been conducted to assess whether there is sufficient signal in the NDSI parameter to provide useful estimates of fractional snowcover in each MODIS 500 meter pixel. To do this high spatial resolution (30 meter) Landsat snowcover observations were used and co-registered with MODIS 500 meter pixels. The NDSI approach was used to assess whether a Landsat pixel was or was not snow-covered. Then the number of snow-covered versus the total Landsat pixels in a MODIS pixel was used to determine the fraction of snow-cover within each MODIS pixel. These results were then used to develop statistical relationships between the NDSI value for each 500 meter MODIS pixel and the fraction of snow-cover in the MODIS pixel. Such studies were conducted for three different areas and conditions covered by Landsat scenes in Alaska, Russia, and the Quebec Province in Canada. The statistical relationships indicate that a 10 percent accuracy can be attained. Independent tests of the relationships were accomplished by taking the relationship of fractional snow-cover to NDSI from two areas (e.g., Alaska and Quebec) and testing it on the remaining area (e.g., Russia). Again the results showed that fractional snow-cover could be estimated with an uncertainty of less than 10 percent. The variability in the statistical relationships for the three areas led to small changes in the accuracy of the results (1-2% for mean error, mean absolute error and standard deviation). The results have been shown to have advantages over other published fractional snow-cover algorithms applied to MODIS data. Most recently the fractional snow-cover algorithm has been applied using 500-meter observations over the state of Colorado for a period spanning 25 days. The results exhibit good behavior in mapping the spatial and temporal variability in snow-cover over that 25-day period. Overall, these studies indicate that robust estimates of fractional snow-cover can be attained using the NDSI parameter over areas extending in size from watersheds relatively large compared to MODIS pixels to global land cover. Other refinements to this approach as well as different approaches are being examined for mapping fractional snow-cover using MODIS observations.