We
have developed a statistical downscaling method for estimating probabilistic
regional climate projection using multi general circulation models (GCMs). A
regression model was established so that the combination of weight of GCMs reflects
the characteristics of the variation of observation at each grid point. Cross
validation was conducted to validate the stochastic model.
We
applied this method to the monthly surface air temperature and precipitation for
the present climate and future projection using CMIP5 dataset. Temperature
increase was remarkable in the northern part of Japan and cold season in the
end of 21
st century. The probability of temperature increase
exceeding 4 K around Kanto-region was over 60 % for RCP8.5, which was
approximately 30 % for RCP4.5. Winter precipitation decrease was notable in the
Japan-Sea side, that may crucially affect the snow amount and water resources
in this region. As for summer precipitation, precipitation increase in the
Pacific side of the southern Japan was projected with relatively high
probability. In order to adapt the warmer climate, this stochastic model plays
an important role in estimating various risks associated with the climate
change.
View full abstract