Assessing the Value of Integrating Remote-Sensing-Based Snow Products into the NOAA National Water Model for Seasonal Water Supply Prediction in the Western U.S.

In the Western U.S., one of the largest uncertainties in subseasonal to seasonal water forecasts is snowpack. The NOAA National Water Model (NWM) includes a subseasonal (30-day, "long-range") forecast product, which predicts snowpack water storage/release and streamflow. This particular NWM product is currently underutilized by water managers and resource planners, in part due to lack of grounding in observations and high uncertainties in the estimates.

We hypothesize that assimilation of snow observations into the NWM long-range forecast will improve skill and the overall utility of water supply forecasts. We take advantage of the new spatially and temporally complete MODSCAG + MODDRFS product suite, which uses spectral unmixing to estimate fractional snow-covered area (fSCA) and grain size. Using grain size we estimate clean snow albedo, and using spectral differencing we estimate dust-radiative impacts on albedo. These remote-sensing derived products are augmented by snow water equivalent (SWE) reconstruction in key basins. We identified three experimental regions with varying hydro-climatic regimes - Upper Colorado River basin, California Sierra Nevada, and Middle Snake basin. These regions present different challenges to both snow remote sensing (clouds, canopy, contaminants) and the snow model (rain/snow partitioning, wind redistribution, contaminants).

We present a baseline comparison of 2001-2018 NWM retrospective snow estimates against the observational products for fSCA, albedo, and SWE. We include a parameter sensitivity analysis for the snow depletion curve, snow water retention, and albedo formulations. We then apply a particle filter data assimilation approach and quantify impacts on streamflow forecast performance for the 2018 water year in terms of timing and quantity metrics. Future work will build from these experiments by accounting for observational uncertainty, varying the particle filter screening metrics, and integrating the results into a framework that ranks the value of different snowpack observations (fSCA, albedo, SWE) in improving seasonal streamflow forecasts in different hydro-climatic regimes.