Bias correction of rainfall simulation in east asia using a statistical-dynamic method

This study presents a bias correction method based on the singular value decomposition analysis (SVDA) for improving rainfall simulation and prediction in the East Asia through the annual cycle. The dynamical model simulates better regional mean precipitation distribution in boreal winter and spring. The location of larger interannual rainfall variability is also better captured by the model from December to May. Using SVDA technique, the corresponding model and observed precipitation variability are found and used to adjust the model prediction. Additional regression method is used to compare the forecast skill enhancement by different statistical correction procedures. The averaged skill enhancement based on the rainfall anomaly pattern correlation (APC) over East Asia is 0.1 except boreal summer. This is added to the APC of direct dynamical model result that is higher in winter (~0.4) and lower in summer (near zero). Nevertheless, the reduction of root mean square error (RMSE) is evident even in summer. The mean decrease in RMSE is around 0.2 mm/day. The failure of this statistical bias correction method in boreal summer is suspected to be related to that the model has the difficulty in simulating reasonable rainfall pattern and its interannual variability in East Asia.