Scale effects in determining snowmelt from mountainous basins using a distributed approach for snow water equivalence and radiation, and a point snowmelt model.

摘要

Rates of snowmelt distributed across Emerald Lake watershed, an alpine basin located in the Sierra Nevada, California, were estimated for water year 1987 using a point snowmelt model applied to regions that were classified based on distributed snow water equivalence and net solar radiation (NSR). A 5-m resolution digital elevation model (DEM) and a 5-m classified digital terrain model of snow water equivalence (SWE) were resampled to coarser resolutions (25-m, 30-m, 50-m, and 100-m) using the nearest neighbor approach. These images were used to define other snowmelt physical parameters and the initial state of the snowpack before melting. Topographic parameters calculated at 50-m and 100-m resolution exhibited significant differences in their histogram distribution as compared to the 5-m DEM. The most important were variations in slope, aspect, sky view factor, and terrain configuration factor, which influenced radiation calculations and the definition of distributed parameters for snowmelt calculations. Elevations, however, did not change significantly from one resolution to the other. The distribution of topographic parameters modeled at 25-m and 30-m, remained almost unchanged. Four, seven and ten classes of snow water equivalence and net solar radiation were combined using a band interleave process to determine the maximum number of combined classes. The point snowmelt model was then applied to these areas, which shared similar SWE and NSR characteristics, to obtain hourly melt rates. Modeled snowmelt rates were compared to the total daily discharge observed at the outlet of Emerald Lake watershed. There was good agreement for resolutions 5-, 25-, 30-, and 50-m but not for the 100-m DEM, as modeled net solar radiation was too high and water was released from the basin too early. Model performance using three tests (Nash-Sutcliffe criteria, sum of squares of the deviations and the sum of the absolute differences between observed discharge and computed melting) showed that the 30-m resolution DEM with combined classes of 7 SWE and 7 NSR provided the best snowmelt performance for this distributed approach. Finally, fractional snow cover area at one month intervals were estimated, showing that this approach offers the potential to model spatially distributed snow covered area in alpine regions.