Snow Ensemble Uncertainty Project (SEUP): Characterization of Snow Water Equivalent Uncertainty Using an Ensemble-Based Land Surface Modeling

Accurate quantification of trends in snow water storage is critical in national hydrologic analyses, water resource allocation, emergency management, and irrigation and infrastructure design. Model-derived estimates snow water equivalent (SWE) are commonly used as the basis, but contain large uncertainties due to errors in land surface model (LSM) structure, parameterization, and boundary conditions. In this study, we apply an ensemble-based land surface modeling approach to assess uncertainty in SWE across North America over the 2009-2017. The NASA Land Information System (LIS) is used to simulate four different LSMs of varying complexity at a 5km spatial resolution using three different forcing datasets. Uncertainty is assessed spatially by evaluating ensemble spread across regions with different topography, snow classes and land covers to characterize the areas of high and low uncertainty in SWE. The day of year with the highest ensemble spread is computed for temporal uncertainty assessment. Results indicate that the largest variability in SWE is found in regions with the deepest snow, particularly along the northern Pacific coast line, eastern Canada along the northern Atlantic coastline, and Northern Rocky Mountains. The time when SWE is most variable is near the peak or during melt season and moves from south to north. The results also show that the uncertainty in total North American SWE is governed by the LSM configuration differences, rather than the choice of forcing data. The result of this study establishes a benchmark of our current understanding of simulated snow estimates over North America, which can be used for the planning of future snow satellite missions, field campaigns, and snow model development.