Implementation of bias correction scheme on KMA's operational global ensemble prediction system

Simple bias correction method is preliminarily implemented on the KMA operational global ensemble prediction system as a statistical post-processing scheme to reduce the biases in the ensemble forecasts. As a first experiment to remove the biases in the 1st moment of the ensemble distribution, decaying average bias assessment method is used. Experiments are conducted with several different weight factors (w) to find the optimal value. The result shows that 2% weight for the recent errors worked better than other weight factors, which coincides well with previous results of other studies.

The ensemble biases with respect to analysis fields for selected forecast variables are assessed daily using weight factor of 2% and removed from the ensemble forecasts for the period from January to December 2006. The preliminary results show that the ranked probability skill score (RPSS) and Brier skill score are increased and the RMSE is significantly decreased when the bias correction scheme is implemented. The effect of this statistical post-processing scheme is especially prominent for the first few days of forecast and during summer.