Runtime Bias Correction of Parameterized Lakes in Climate Models

Abstract A novel approach is presented for improving atmosphere‐lake heat and evaporative fluxes in global and regional climate models. The method involves deriving and applying runtime corrections to lake surface water temperature (LSWT) and lake ice, to reduce biases in the climatological annual cycle of these quantities. The bias correction procedure is developed for a particular single‐column thermodynamic lake model but could, in principle, be applied to other lake schemes, including fully prognostic models with hydrodynamics. Its application in a regional climate model (RCM) employing the North‐American CORDEX domain is presented and analyzed. The bias correction is found to improve the climatological seasonal cycle, spatial distributions, and interannual variability of LSWT and ice over the Great Lakes (GLs) water system. These improvements reduce significant biases in the seasonal cycle of evaporation over the GLs. Temperature and precipitation fields from the RCM are used to drive an offline hydrological model to derive runoff into each GL. This runoff is then combined with RCM‐derived precipitation and evaporation over the lakes to generate time series of component derived Net Basin Supply (NBS) for each lake. The lake bias correction generally improves each component of NBS for each of the GLs during the historical period compared to the uncorrected model. Accurate predictions of NBS are essential for optimal management of Great Lakes water levels and for mitigating flooding both within the lakes and in downstream areas affected by regulated outflows, such as the Ottawa Valley.