Cooperative Institute for Research in Environmental Sciences

Remote Sensing of the Mountain Snowpack: Integration of Observations and Models to Support Water Resource Management and Ecosystem Science

Noah Molotch (1,2), Jeff Deems (2), Bin Guan (3), Leann Lestak (4), Dominik Schneider (1), Michael Durand (5), Thomas Painter (3), and Jeff Dozier (6)

(1) Geography/INSTAAR, University of Colorado, (2) WWA/CIRES, (3) Jet Propulsion Laboratory - CAL TECH, (4) Institute of Arctic and Alpine Research (5) Earth Sciences, The Ohio State University (6) Donald Bren School of Environmental Science and Management, University of California, Santa Barbara

The impacts of climate change on water sustainability in mountainous regions is inherently linked to changes in mountain snow accumulation and snowmelt timing which sustains agricultural and municipal water demands for 60 million people in the U.S. and one billion people globally. Hence, accurate estimates of the volume of snowpack water storage are critical for supporting water resource planning and management. While snow extent is one of the earliest observed land surface variables from space, hydrologic applications of these data have been limited as the variable of interest for water management is snow water equivalent (SWE) which is not remotely observable at the fine-scale resolution needed in the mountains. Since the early 1980’s, several works have illustrated the connection between runoff volume and snow cover depletion patterns as observed from satellite. In this regard, we present a series of experiments which illustrate that patterns of snow cover depletion can be coupled to spatially distributed snowmelt models to reconstruct the spatial distribution of SWE. In this regard, we present a proof-of-concept for a global product, providing daily estimates of snow water equivalent at 500-m scale for the observation record of the Moderate Resolution Imaging Spectroradiometer (MODIS). Estimates of the reconstructed SWE are validated against observed SWE from extensive snow surveys across the Sierra Nevada and Rocky Mountains with adequate spatial sampling, and compared to the operational Snow Data Assimilation System (SNODAS) SWE product produced by the U.S. National Weather Service. Snow survey SWE is underestimated by 4.6% and 36.4%, respectively, in reconstructed and SNODAS SWE, averaged over 17 surveys from sites of varying physiography. Corresponding root-mean-square errors are 0.20 m and 0.25 m, respectively, or 2.2 and 2.6 mean standard deviation of the snow survey SWE. Comparison between reconstructed and snow sensor SWE suggests that the current snow sensor network in the U.S. inadequately represents the domain SWE due to undersampling of the mid-lower and upper elevations. Correlation with full natural flow is better with reconstructed SWE than with ground-based snow sensors, or with SNODAS SWE on average; particularly late in the snowmelt season after snow stations report zero values but snow persists at higher elevations. These results indicate that inclusion of remotely sensed snow cover depletion patterns dramatically improves estimates of snow distribution in mountainous regions. Example applications for improving water resource management and understanding ecosystem response to water availability will be shown.