Snow surveys were performed within 1- and 2-km2 grids during the accumulation and ablation periods in the Sierra Nevada of California in 2006, 2008, and 2009, and provide ground-truth for the MODIS Snow Covered Area and Grain size/albedo (MODSCAG) model. In addition, LANDSAT Thematic Mapper using the TMSCAG model at a 30-m resolution was used to compare results with MODIS and the snow surveys. The availability of accurate, fractional snow covered area (fSCA) from MODIS at a 500-m resolution offers unprecedented detail on snow accumulation and melt across variable terrain. fSCA from MODIS provides the only daily basin-wide snow coverage estimation for the western United States, where snow is a dominant component of precipitation and runoff. However, MODSCAG produces the projected fractional snow cover, which does not include snow under the tree canopy. A proposed modification to the method is to divide the observed fractional snow by the canopy opening, i.e. 1.0 minus the forest fraction. A comparison of mean fSCA during the accumulation season across eight 300-m elevation bands, in the 1500-3900 m range, in the Tuolumne and Merced River basins showed that canopy with openings less than 50% obscure significant snow cover. In snow surveys performed within 1- and 2- km2 grids in forested terrain with canopy openings of 50% or less, surveys measured complete snow cover, but MODIS detected fSCA across the study area of 0.3-0.4, comparable to the vegetation gap fraction. During the onset of ablation in a forested area with canopy openings of 0.6- 0.9, snow surveys measured nearly complete snow cover, and MODIS detected a mean fSCA across a 1-km2 study area of 0.83 (range 0.4-0.9). During the ablation period, when snowcover becomes highly variable, a ground survey in an area with canopy openings of 0.4-0.6 provided a snowcover estimate of 0.5, versus 0.1 for fSCA from MODIS. When compared to the 30-m fractional fSCA from the TMSCAG model there were mixed results. When snow surveys measured complete snow cover, LANDSAT showed from 15-20% more to 9% less fSCA than MODIS. However, when snowcover was highly variable during the ablation period and ground truth showed snow cover of 0.5, LANDSAT detected a mean fSCA of 0.3 (range 0-0.9). Even though the TMSCAG model pushes the limits of the scale-dependent biases to smaller basins, and terrain with canopy openings less than 0.5, the LANDSAT analysis confirms that under-canopy snow cannot be captured directly by an optical sensor. Although forest cover introduces potential biases, using MODSCAG in basin-scale hydrologic modeling one can ignore these biases at larger scales and lighter forest cover; while in smaller forested basins, or within elevation bands, where vegetation obscures significant snow cover, adjustments are necessary for mass balance.
Rice, R., Bales, R.C., Kirchner, P.B., Saksa, P.C., Rittger, K.E., Painter, T.H., Dozier, J. (2010): A Comparison of the Fractional MODIS and LANDSAT Thematic Mapper with Ground-Based Snow Surveys in the Sierra Nevada . Fall meeting, American Geophysical Union, December 2010. Abstract C33E-0589..