Volume 94A (2016) Pages 17-29
The sampling downscaling (SmDS) in which a regional atmospheric model is integrated for sampled periods was performed for summertime Hokkaido. Selected are top two and bottom two years of the general circulation model projection onto the first singular value decomposition mode where heavy precipitation in southern Hokkaido is correlated with the moisture flux convergence in the synoptic field. The SmDS result integrated for the four years successfully reproduces the dynamical downscaling for 30 years, in terms of climatological precipitation and the 99-percentile value of daily precipitation. This indicates that SmDS can be applied to the environment where local precipitation is mostly controlled by synoptic climate patterns. A further statistical consideration in this study supports the notion. It is also demonstrated that SmDS selects a group of years where extreme events likely occur another group of years where they rarely occur.