Abstract
The climatological characteristics of daily precipitation over Japan in the Kakushin regional climate change experiments using a non-hydrostatic model with a horizontal resolution of 5 km (NHM5km) are investigated from June to October between 1990 and 1999. Comparisons with the results of a global 20-km-mesh atmospheric climate model (AGCM20km), which provides the boundary conditions for NHM5km, show that NHM5km can improve the reproducibility of precipitation characteristics, including the frequencies of intense precipitation and wet days, from those in AGCM20km. Compared with observations, AGCM20km shows -30% and +33% biases in terms of the 10-year mean values of regional maximum precipitation and wet days, respectively, in the region 31°N-38°N and 129.8°E-142.0°E over Japan from June to October; these biases are reduced to +9% and +12% in the case of NHM5km. The largest biases in AGCM20km, overestimation of wet days in July and August, are successfully reduced in NHM5km. Probability density distributions of daily precipitation amount are superior in NHM5km than in AGCM20km for all analysis months, being in strong agreement with a raingauge-based daily precipitation dataset for June-August. These features indicate that NHM5km also provides improved seasonal variations in precipitation characteristics, which are crucial for reliable climate change experiments.
