We updated the lake model in CCSM4.0 to improve the characterization of lake surface energy fluxes, eddy diffusivity, and convective mixing, and added new snow physics, phase change and ice physics, and soil layers beneath the lake. We are also considering several options for simulating subgrid thermokarst lake dynamics, including modeling an ensemble of lakes of different size categories with dynamic subgrid area. We tested the new lake model against observations for a dozen lakes with diverse geometries and climates, and compared to predictions of the old lake model at several of these sites. Unlike the old lake model, thermal predictions from the new lake model match the vertical and seasonal patterns in observed lake water temperature for all lakes, with typical errors of about 3 K.
We performed sensitivity experiments to estimate the effect of modeled lake properties and processes on the predicted surface fluxes when forced with historical atmospheric reanalysis data. The modeled lake depth, optical properties, snow properties, and inclusion of phase change physics all have significant impacts on predictions of seasonal and regional surface fluxes; differences in seasonal latent heat flux compared to the new lake model were as large as 100 W m-2. Coupling the lake model into a regional climate model (WRF3-CLM3.5) altered surface air temperatures in the Great Lakes region by 1-3 K, either improving or worsening the model bias compared to observations, depending on the season. We tested the sensitivity of the coupled global climate to the new lake model in CCSM4.0 and to the effect of increasing the lake area to a more realistic value. Canada and Northern Europe experienced broad summer cooling of about 1K, and changing atmospheric transport caused significant changes in remote areas. We conclude that global climate model predictions would benefit from a more realistic representation of lakes.