Dynamical Downscaling Using CGCM Ensemble Average for Seasonal Prediction

This study introduces a new attempt to downscale the coupled general circulation model (CGCM) ensemble prediction dynamically more efficiently than conventional method. For the purpose, the ensemble seasonal prediction skill of dynamically downscaled precipitation over South Korea is evaluated by comparing two experiments. The first experiment (EXP1) involves conventional ensemble forecasts. Five ensemble members (EMs) are downscaled dynamically with initial and lateral boundary conditions obtained from the outputs of the five CGCM EMs. Then the results of each EM are averaged for ensemble prediction utilizing a simple composite method. The second experiment (EXP2) is the same as EXP1, but the initial and lateral boundary conditions are obtained by arithmetically averaging the outputs of the five CGCM EMs. Therefore, five integrations are carried out for the EXP1, but only one integration is performed for the EXP2. The results show the EXP2 simulates closer to the observed precipitation than EXP1. This improvement is attributed to the strongly simulated upper zonal wind that can influence the vertically integrated moisture flux convergence. EXP2 shows comparable or even better performance in simulating the interannual variability of summer precipitation than EXP1. Unlike conventional methods such as EXP1, EXP2 provides a prediction in a single integration, and the prediction is similar to or even better than the one obtained conventionally. Hence EXP2 can be a powerful means to drastically reduce the prediction time by reducing the number of integration to just one.