Calibration and validation of lake surface temperature simulations with the coupled WRF-lake model

A one-dimensional (1-D) physically based lake model was coupled to the Weather Research and Forecasting (WRF) model version 3.2 developed by the National Center for Atmospheric Research to dynamically simulate physical processes of lakes and their effects on weather and climate at local and regional scales. Our study area is focused on the Great Lakes. This coupled model realistically reproduces the lake surface temperature (LST) at a buoy station in a shallow lake (Lake Erie) while generating strong LST biases ranging from -20 to 20 A degrees C at a buoy station in a deep lake (Lake Superior). Through many sensitivity tests, we find that the biases in the deep lake LST simulations result from the drastic underestimation of heat transfer between the lower and upper parts of the lake through unrealistic eddy diffusion. Additional tests were made to calibrate the eddy diffusivity in WRF-Lake. It is found that when this parameter is multiplied by a factor ranging from 10(2) to 10(5) for various lake depths deeper than 15 m, the LST simulations for the deep lake buoy station show good agreement with observations, and the bias range reduces to +/- 4 A degrees C. Essentially, the enlarged eddy diffusivity strengthens heat transfer within the lake columns in the deep lake, which is significantly underestimated in the lake model without calibration. Validation simulations with the calibrated eddy diffusivity were carried out for the whole of Lake Superior and Lake Erie. The LST simulations still have a substantial bias reduction when compared with those produced with the original eddy diffusivity, indicating that the calibrated parameter is representative. In addition, the improved 1-D lake model with WRF reasonably reproduces the remotely sensed LST geographic distribution.