Multi-Physics Data Assimilation Framework for Remotely Sensed Snowpacks to Improve Water Prediction

Recent advances in remote sensing of snow using Synthetic Aperture Radar have shown the potential for retrievals of Snow Water Equivalent (SWE) at high spatial resolution with good accuracy. These data can be integrated with physically based models to reconstruct spatial heterogeneity and reduce uncertainty in quantifying SWE. In this study, we present a Multi-Physics Data Assimilation Framework (MPDAF) to improve operational water prediction by assimilating snow measurements/retrievals or microwave data. This framework is demonstrated over Grand Mesa, Colorado during NASA's SnowEx’17 campaign. To illustrate the potential benefit of satellite-based time-series of SAR measurements, we investigate the value of data assimilation (DA) with window lengths determined by potential satellite revisit times and anticipated estimation error models. Daily assimilation of integral quantities like snow depth showed dramatic improvement in predicted snow depth, SWE and in vertical profile of snow density. Independent evaluation against pit measurements shows that assimilation of SWE retrievals from airborne SnowSAR backscatter measurements substantially reduced bias in snow depth (from −22% to 0%) and SWE (from −19% to 3%), also recovering spatial heterogeneity not resolved by weather forecasts. Assimilation impacts both snowpack microphysics and the forward simulation of volume backscatter in X and Ku bands with dependence on snow depth changes. The uncertainty in the forward estimates of backscatter is consistent with the synthetic measurement uncertainty based on pit data, thus demonstrating that MPDAF preserves end-to-end physical consistency among assimilated retrievals and forward simulations of backscatter measurements critical for operational retrievals.

Duke Scholars

Author Ana P. Barros Civil and Environmental Engineering