Comparing Measurements, Simulations, and Forecasts of Snow Water Equivalent Across the Great Lakes Basin

Basin-scale estimates of the water budget of the North American Great Lakes are based on a geographically broad (and, in some areas, relatively sparse) monitoring network that spans the United States-Canadian international border, and a limited ensemble of models. Of the various components of the Great Lakes water budget, snow water equivalent (and its contribution to runoff) represents one that is estimated by a regional rainfall-runoff simulation model (the NOAA large basin runoff model, or LBRM) and by a data assimilation model (via the NOAA National Operational Hydrological Remote Sensing Center Snow Data Assimilation System). Importantly, both products are employed in regional operational water budget and water level forecasts, including those developed by the US Army Corps of Engineers, the New York Power Authority, and Ontario Power Generation.

While these forecasts are periodically evaluated for skill based on a comparison between water level projections and observations, we know of no study that has either compared LBRM simulations of SWE to corresponding NOHRSC estimates, or explored the potential benefits of assimilating NOHRSC estimates into the LBRM and propagating those benefits into water level-based management decisions. To address this gap in research and operational knowledge, we compare simulated and "observed" SWE for select sub-basins in the Great Lakes region. We refer to the NOHRSC-SNODAS product as an "observed" estimate of SWE because it combines airborne and surface measurements with satellite derived snow information and model simulations.

Our findings indicate general agreement between LBRM-simulated and observation-based estimates of SWE, particularly with respect to the timing of most individual events and the timing of peak SWE. However, we find discontinuities in the timing and duration of snowmelt, the magnitude of the peak runoff, and the overall cumulative seasonal total runoff. Finally, we propagate these estimates of SWE into runoff simulations, and compare those to observed runoff to assess the relative benefits of alternatives for assimilating SWE observations into operational hydrological forecasting.