Bias Correction of Snow Depth by Using Regional Frequency Analysis in the Non-Hydrostatic Regional Climate Model around Japan

Abstract

The 5-km-mesh, Non-Hydrostatic Regional Climate Model was used to simulate snow depths in Japan and to project their changes in the future. The simulated snow depths had large biases, and bias corrections were required to project future snow depths accurately. We developed a new method of bias correction that is accurate and easily implemented for automatic use on a computer. Three classification methods of regional frequency analysis were tested in nine regions of Japan. The classification method based on the second order of L-moments (L-cv) was the best bias correction method among those tested. We checked that this bias correction was useful method for future climate projections by using the test sample estimate. Snow depth in the future climate was projected to decrease by about 50 cm, such that the average snow depth over Japan was about 30 cm in the future climate. The projected decrease in the maximum snow depth was large in the Nagano and Gifu regions and small in the Hokkaido region compared with other regions.