Convection-Permitting Regional Climate Simulations of Precipitation over Japan

We evaluated the performance of a convection-permitting regional climate model with a grid spacing of 2 km in simulating precipitation in the present climate of Japan. This model is called the non-hydrostatic regional climate model, hereafter referred to as NHRCM02. We compared the precipitation reproduced by NHRCM02 with that observed, and calculated errors such as root mean square error (RMSE) and systematic error (bias). We also examined the errors of precipitation simulated by the driving model for NHRCM02, hereafter referred to as NHRCM05, which has a grid spacing of 5 km and includes the Kain-Fritsch convective parameterization.

The RMSE for the annual precipitation averaged over Japan simulated by NHRCM02 was less than that obtained from the NHRCM05 simulation. The difference in RMSE between the two simulations was statistically significant at the 95% level, indicating that the model performance was improved by using a higher resolution model. On the other hand, the magnitude of bias for the annual precipitation from NHRCM02 was also reduced compared with that of the NHRCM05, but their difference was not statistically significant at the 95% level.

In order to evaluate the effects of improved representation of complex topography on the reproducibility of precipitation, we examined the relationship between the degree of terrain complexity and the improvement in simulated precipitation amounts. For each observation point, the terrain complexity was quantified by the standard deviation of model height within an area surrounding the target point. Then, we defined an index to determine the improvement in the simulated precipitation by calculating the difference in the magnitude of precipitation bias, between the NHRCM02 and NHRCM05 simulations, normalized by the observed precipitation amount.

The index of the improvement in the simulated precipitation roughly increased with growing terrain complexity. This suggests that the improved representation of complex topography in the NHRCM02 simulation resulted in well reproducibility of precipitation. However, not all improvements in simulated precipitation could be attributed to the topographical factor. Another factor contributed to the improvement of the model performance in precipitation: representations of cumulus convection.