Evaluation of the Precision of Return Period calculated by GSMaP data

Abstract:

A heavy rain by Typhoon1326 was generated over IzuOshima Island on October 17, 2013. As for nine rain gauges in IzuOshima Island, precipitation amount exceeded the threshold of the special warning for 3 hours. However, the special warning was not announced because it did not satisfy "heavy rainfall over the threshold value is recorded at more than 10 points". It is suggested that the special warning at islands is hard to be announced. Therefore we propose to apply GSMaP which can be used not only over land but over ocean for the special warning. We calculated the threshold of the special warning by using GSMaP data and compared it with that by Japan Meteorological Agency (JMA).

GSMaP has the following characteristics.

・GSMaP estimates precipitation every hour with the resolution of 0.1 degrees grid.

・GSMaP is not good at orographic rainfall , but it is good at rainfall over the Tropics and ocean.

・GSMaP covers almost all over the world between 60 degrees north and south.

The standard of JMA is as follows.

・The amount of rainfall in 3 hours exceed 50-year value at 10 points in the spread of the prefecture degree.

・The amount of rainfall in 48 hours exceed 50-year value at 50 points in the spread of the prefecture degree.

We calculated the amount of 50-year rainfall by GSMaP and compared it with the threshold value of JMA. But there are only 9-year GSMaP data used in this study. The number of sample is too little to calculate return period. We have also added the data at the surrounding grids. We tried to increase the number of samples in this way. GSMaP has several problems. One of that’s GSMaP mistakes snow for rainfall. So we grasped the characteristics of such case. And we made the mask to take such case. When we calculate with the mask, the abnormal values has decreased. We improved stability of calculation return period with these ways.