Assessing the utility of passive microwave data for Snow Water Equivalent (SWE) estimation in the Sutlej River Basin of the northwestern Himalaya

Tuesday, 16 December 2014
Ty Brandt, University of California Santa Barbara, Santa Barbara, CA, United States, Bodo Bookhagen, University of Potsdam, Potsdam, Germany and Jeff Dozier, University of California, Mammoth Lakes, CA, United States
Since 1978, space based passive microwave (PM) radiometers have been used to comprehensively measure Snow Water Equivalent (SWE) on a global basis. The ability of PM radiometers to directly measure SWE at high temporal frequencies offers some distinct advantages over optical remote sensors. Nevertheless, in mountainous terrain PM radiometers often struggle to accurately measure SWE because of wet snow, saturation in deep snow, forests, depth hoar and stratigraphy, variable relief, and subpixel heterogeneity inherent in large pixel sizes. The Himalaya, because of their high elevation and high relief—much above tree line—offer an opportunity to examine PM products in the mountains without the added complication of trees. The upper Sutlej River basin— the third largest Himalayan catchment—lies in the western Himalaya. The river is a tributary of the Indus River and seasonal snow constitutes a substantial part of the basin’s hydrologic budget. The basin has a few surface stations and river gauges, which is unique for the region. As such, the Sutlej River basin is a good location to analyze the accuracy and effectiveness of the current National Snow and Ice Data Center’s (NSIDC) standard AMSR-E/Aqua Daily SWE product in mountainous terrain. So far, we have observed that individual pixels can “flicker”, i.e. fluctuate from day to day, over large parts of the basin. We consider whether this is an artifact of the algorithm or whether this is embedded in the raw brightness temperatures themselves. In addition, we examine how well the standard product registers winter storms, and how it varies over heavily glaciated pixels. Finally, we use a few common measures of algorithm performance (precision, recall and accuracy) to test how well the standard product detects the presence of snow, using optical imagery for validation. An improved understanding of the effectiveness of PM imagery in the mountains will help to clarify the technology’s limits.