Influence of Lake Ice Biases in Reanalysis Data on Downscaled Climate Simulations over the Great Lakes Region

This study investigates the unrealistic representation of lake ice in the European Centre for Medium-Range Weather Forecasts' fifth-generation reanalysis (ERA5) over the Laurentian Great Lakes and its influence on regional climate modeling. Focusing on the high-ice-cover winter of 2009, we find that ERA5 systematically overestimates lake ice extent, with near-complete coverage persisting into April. This excessive ice cover suppresses sensible and latent heat fluxes, producing anomalously cold and dry atmospheric conditions, with temperature biases up to 13°C and lake-effect precipitation underestimated by up to 60% downwind of Lake Michigan-Huron and more than 100% over the eastern side of Lake Superior in late January. In contrast, simulations with a fully coupled lake-atmosphere model initialized from ice-free conditions given sufficient spin-up produce more realistic ice evolution. The improved ice representation provides better surface boundary conditions and reduces biases in air temperature, wind speed, and lake-effect precipitation.