Optimal Estimation of Snow Related Parameters in Noah Land Surface Model Using an Evolutionary Algorithm

Snow processes in the land surface models (LSMs) include the snow cover fraction, snow albedo, and snow depth — all interacting with the atmospheric conditions. Most LSMs include parameters based on empirical relations, resulting in uncertainties in model solutions. In addition, such parameters often reflect only the local characteristics where the empirical relations are made. Therefore, the empirical parameters need to be optimized when they are applied to different regions. This study seeks the optimal snow-related parameters over South Korea where heavy snowfall events occur in the winter. The optimization is conducted using a micro-genetic algorithm (micro-GA) and the in situ and satellite observations for the snow depth, snow cover fraction, and snow albedo. The micro-GA is one of the evolutionary algorithms to search for the best potential solution based on natural selection and the survival of fitness. To represent the regional empirical parameters using the single-column model (e.g., Noah LSM), we selected the representative stations over South Korea to cover various vegetation types. Next, we identify which snow-related parameters can be optimized and suggest the optimal parameters using the micro-GA over South Korea. As a result, the Noah LSM simulations, using the optimized parameters, reduced the biases by 45.1% and 32.6 % for the snow depth and snow albedo, respectively, and the root mean square errors by 17.0 %, 8.2 %, and 5.6 % for snow depth, snow cover fraction, and snow albedo, respectively.