Ensemble-based reanalysis of the seasonal snowpack using multispectral satellite imagery
HAPpy Hour Seminar
2:00 pm MDT
Accurately estimating the distribution of seasonal snow water equivalent (SWE) in the world’s mountains has been coined the greatest unsolved problem in snow hydrology. Out of the many existing methods for SWE estimation the so-called reconstruction approach arguably holds the most promise in far-flung mountainous regions, at least retrospectively.
Reconstruction ingests information from the remotely-sensed depletion of fractional snow-covered area during the melt season into a snowmelt model. Traditional reconstruction, where SWE is built up in reverse from the observed melt out date to the day of peak SWE using snowmelt estimates, is plagued by uncertainties in the observed snow-cover depletion, the forcing data, and the snowmelt model. Ensemble-based reanalysis builds on the traditional reconstruction by trying to account for these uncertainties. In this approach, a prior ensemble of annual model realizations is updated through a batch assimilation of satellite-retrieved snow-cover data, resulting in a constrained posterior ensemble of snowpack trajectories.
Here, we present reanalyses of SWE both for Ny-Ålesund, a test site in the high-Arctic, and the Mammoth Lakes basin in California for several water years. For both sites, the availability of independent, accurate and representative measurements allows us to perform an extensive validation. We investigate the impact of assimilating snow-cover retrievals from multiple satellite senors. We compare the performance of three established data assimilation schemes: the ensemble smoother, the particle batch smoother, and an iterative ensemble smoother. Our results provide guidance for reanalysis efforts in remote cold regions around the world where the seasonal snowpack is an uncertain but vital component of the surface energy-and water balance.