[A-AS07] Extreme weathers and disasters in urban environments in East Asia and relationship to climate change

[AAS07-02] Development of Super-Sampling Downscaling for Summertime Precipitation in Kyushu Island, Japan

*Hajime Ohnishi¹, Masaru Inatsu¹ (1.Hokkaido University)

We have developed a cost-effective downscaling method nominated super-sampling downscaling to address model uncertainty in regional evaluation for climate change. Although the global model uncertainty was evaluated through multi-model experiments as in CMIP5, it would be virtually impossible to dynamically downscale all the global model results for computational cost. The super-sampling downscaling method generated a downscaling dataset of a CMIP5 model by sampling from a large ensemble of dynamical downscaling datasets, d4PDF, already calculated by a single global model. The sampling was based on probability distribution of weather patterns for a target domain in a target season classified by self-organizing maps (SOM) obtained by learning both d4PDF and CMIP5 global model datasets. We performed this super-sampling downscaling focusing on summertime precipitation in Kyushu Island, Japan. The generated downscaling for CMIP5’s historical climate was sampled from d4PDF historical experiment, and that for CMIP5’s RCM8.5 climate was sampled from d4PDF 4K experiment. The results suggested that many CMIP5 models showed a general trend of decrease in eastern Kyushu and increase in western Kyushu.(Figure 1.)