The effect of sub-pixel areal distribution of snow on the estimation of snow depth from spaceborne passive microwave instruments

The mapping of estimated snow depth (SD) or snow water equivalent (SWE) from spaceborne passive microwave imagery is usually achieved by detecting a snow scattering signal from the land surface and then by calibrating the magnitude of the scattering with snow depth or snow water equivalent. Scattering is estimated from the brightness temperature difference between 19 GHz and 37 GHz vertical polarization channels of a microwave radiometer (or frequencies not too dissimilar to these). If a snow scattering signal is present, it is generally assumed that snow covers the entire area of a coarse spatial resolution passive microwave pixel (or footprint). For seasonal snowpacks, this is a reasonable assumption because at high latitudes in general, snow cover is often spatially continuous over wide areas during the winter season. However, for spatially discontinuous snowpacks (such as early winter snow, ephemeral snow covers or perhaps regions marginal to mid-winter continental snow covers), snow might be detected in a microwave pixel but it may be inaccurate to assume that snow covers the entire pixel; snow might be spatially localized but dominant enough radiometrically to trigger a snow scattering signal. In this paper we investigate the effect of sub-pixel scale fractional snow extent on the microwave detection of snow. Under cloud-free conditions determined by the Moderate Resolution Imaging Spectroradiometer (MODIS) MOD10_L2 product, we compare the snow scattering signal for various DMSP Special Sensor Microwave Imager pixels with the MODIS MOD10_L2 snow product (500 m×500 m spatial resolution). Using the 25 km×25 km EASE grid projection for the passive microwave imagery, comparisons are made between the microwave scattering signal and the percentage of MODIS pixels classed as 100% snow within the 25 km×25 km. We also compare microwave snow scattering signals with the degree of MODIS snow pixel clustering/dispersion within 25 km×25 km pixels. This research has important implications for the errors of passive microwave mapping of SD or SWE in discontinuous snow covered regions.