Utilizing a novel snow reanalysis dataset from Landsat to evaluate National Water Model simulations of snow water equivalent

Snow accumulation and its melt dominate water resources in mountainous areas, with regions
such as the western United States deriving more than 75% of the total freshwater available
annually from snowmelt. Rain-on-snow floods, in some cases amplified by frozen soils, have
been responsible for some of the largest recorded floods in the West, as well as the North Central
States and the Northeast. Despite their importance snowpacks remain poorly quantified (especially, but not only, in topographically complex regions), leaving hydrological models such
as the National Water Model (NWM) poorly constrained with significant uncertainty in both
their physical and statistical parameterizations. Remote sensing can alleviate some of these
limitations and recently-produced datasets from optical satellite sensors have provided long-term
and high-resolution maps of snow water equivalent (SWE) over the Sierra Nevada.
We demonstrate the use of a novel SWE dataset derived from Landsat observations
to validate and assess the representation of SWE in the NWM. Apart
from directly evaluating the output of the NWM in terms of SWE, we diagnose model errors
and characterize them both statistically as well as in terms of the model parameterizations. The
results from the diagnostic and sensitivity analyses are then used to derive error models and
improved parameterizations for NWM that are tested and compared with the baseline NWM
version.