Integrating a 1D Thermal Lake Model into a Global and Regional Climate Model: Model Evaluation and Regional Climate Simulation

Compared to solid ground, lakes tend to have decreased albedo, increased ground heat conductance, and increased effective ground heat capacity. These features alter local surface fluxes compared to nearby vegetation, which in turn alter the climate of the nearby atmosphere and surrounding land areas. Interest in feedbacks between lake behavior and climate change provides motivation for including lakes in global climate models, as does the desire to do effective regional downscaling of climate model predictions over regions with large lake area fraction, like the Great Lakes region. Finally, the initiation, warming, and expansion of Arctic thermokarst lakes could provide an important geophysical and biogeochemical feedback to climate warming. The Community Land Model (CLM) 3.5 currently uses a 1D Hostetler lake scheme. We have updated this model to improve the characterization of surface fluxes, eddy diffusivity, and convective mixing. We also link the lake model with the full snow physics found over other land surface types (including 5 snow layers, aerosol deposition, partial transparency of snow layers, and snow aging), add phase change & ice physics to the lake model, and include soil layers beneath lakes. These soil layers will be an important component of future thermokarst lake modeling, as thermokarst lakes tend to form regions of unfrozen soil (talik) beneath them that become active sites for anaerobic decomposition of pre-modern peat. We have also integrated the updated lake model into a modified version of the Weather Research and Forecasting (WRF) Model 3.0. We will present comparisons between predicted and observed thermal conditions, snow and ice depths, and surface energy fluxes at several lake sites, using local meteorological forcing or integrated regional atmospheric coupling. The thermal predictions are generally reasonable and show a marked improvement from runs performed with the baseline CLM 3.5 version of the lake model. Over Sparkling Lake, the lake model simulates the hourly lake water temperatures from 2002-2005 in very good agreement (<3 K error) with observations at all depths when the model is forced with the observed climatology. Over other relatively shallow (2 m - 50 m) lakes, thermal depth profiles are generally well reproduced, with maximum summer surface temperatures generally within ~3 K of the observations and mixing depths accurate to about 20% of lake depth. Snow depths, ice depths, and freezing duration are evaluated. Over very deep lakes like Lake Geneva, the model predicts insufficient mixing below about 30 m depth, which is a known deficiency of the Hostetler Lake Model, although the surface temperature is well replicated. We conclude that the model is suitable for inclusion into regional and global climate models for evaluating climate feedbacks, particularly for shallow thermokarst lakes. Finally, we present a comparison of the simulated historical surface climatology to observations for the Great Lakes region in WRF, both with and without the new lake model.