RUTGERS UNIVERSITY :: CLIMATE LAB :: GLOBAL SNOW LAB
home             publications             data  requests             links             NASA MEaSUREs
PRODUCTS
DOCUMENTATION
MICROWAVE SATELLITE
Northern Hemisphere
Visible Satellite Charts
    Daily
    Weekly
    Monthly
    Monthly Departure
    Monthly Climatology
    Documentation

Graphs
    Snow Anomalies
    Monthly Anomalies
    Seasonal Extent

Tabular
Area of Snow Extent
    Weekly
    Monthly
    Rankings
    Download

North America
Blended Snow Charts
    3-Way Pentads
    Station Pentads
    Microwave Pentads

Microwave Satellite

Microwave maps of snow extent and depth over Northern Hemisphere lands are being produced from SSM/I and SMMR data as part of our current NOAA grant. A multi-step processing procedure was developed that involves first discriminating land cover type, then eliminating periods with wet snow before assessing snow extent and depth from the microwave data. That process is as follows:

1. Brightness temperature grids in the EASE projection at a 25km resolution (Armstrong and Brodzik, 1995) available from NSIDC are read for each day. Ascending and descending overpasses are averaged, but only nocturnal overpasses are included due to the influence of liquid water (melt) on the 19-37GHz frequency gradient (Derksen et al. 2000).

2. The processing begins with the land cover decision tree of Grody and Basist (1996), which uses the 19, 22, 37 and 85GHz frequencies, both polarizations, to attempt to distinguish between falling liquid precipitation, frozen soil, bare soil, open water and snow covered soil. The number of 25km EASE grid cells in each of these categories is counted and recorded within each 1° x 1° grid cell. An additional category for depth hoar is included based on a 37GHz polarization normalized difference (Singh and Gan 2000).

3. The third step of the processing includes eliminating any other periods with melting snow (i.e., the presence of liquid water in the snow). Even small amounts of liquid water (<0.5% by volume) dramatically increases the emissivity of the snowpack and reduces the frequency gradient. It is the frequency gradient (e.g., the difference in 19 and 37GHz emissivity) that is used to determine the presence of snow and the snow depth. We are unable to say anything meaningful about the snow depth when liquid water is present in the snowpack. The Walker and Goodison (1993) snow melt algorithm (19GHz polarization difference of less than 5k) is augmented with a melt algorithm based on work by Goodison and Walker (1994) (37GHz V > 241K or 37GHz polarization difference of less than 10K) and a threshold of the 37GHz polarization normalized difference (Singh and Gan 2000).

4. A mask is applied for land areas. Due to the complex microwave signature of the Greenland ice sheet, areas with exposed glacial ice or ice lenses and layers do not show up as snow covered. A mask is applied to show the Greenland ice sheet as snow covered (with a depth of 999cm). Snow depths are set to -90cm over the ocean.

5. For each EASE grid cell, a snow depth is calculated. In order to estimate the effect of vegetation, which reduces the frequency gradient of a given depth of snow, the Robinson maximum snow covered albedo data set is included. The maximum snow covered albedo is a proxy for the protruding (visible) vegetation under deep snow cover. An algorithm was developed by fitting the frequency gradient, the 19GHz polarization gradient and the maximum snow covered albedo to the available cooperative snow depth observations for North America during 1987-2000. The algorithm is: 1.046*(19V-37V)+0.172*A, where 19V is the 19GHz, vertically polarized brightness temperature in Kelvin, 37V is the 37GHz, vertically polarized brightness temperature and A is the maximum snow covered albedo.

6. For each 1° x 1° cell, the mean snow depth is calculated. If the amount of bare soil or open water or falling rain exceeds half the 1° x 1° cell, then the first of two flags is set to false. It is possible to report a snow depth while the algorithm is inconclusive regarding the presence of snow. It is also possible for the algorithm to conclude that snow is present, but a snow depth cannot be calculated. (This may even be reported as a negative snow depth, indicating shallow and/or patchy snow.) If any of the melt algorithms indicates a substantial portion of the 1° x 1° cell is experiencing melt, then the second flag (a "melt flag") is set to true. Finally, the total number of EASE grid cells with available data and the total number of EASE grid cells with "good" estimates of depth in the 1° x 1° cell is indicated. The columns in the data files are day, month, year, latitude, longitude (lower left corner of cell), mean snow depth (cm), snow flag, melt flag, total number of available EASE grid cells, number of EASE grid cells with "good" estimates of snow depth.