Time correlation used to improve time diversity gain of rainfall prediction

Abstract

Correlation of the rainfall rate with time depends on geography, environment, and the type of rainfall. In this letter, we consider the use of such a correlation to improve the time diversity gain of a system that predicts rainfall over Japan. From a rain may with 4 years of data, we selected 43 locations throughout mainland Japan, and from these data, we derived the cumulative time percentage, the diversity gain from a time diversity method using 10 to 120 min delays, and the correlation from a time series of rainfall. This information is used to improve the prediction of rainfall. Root-mean-square errors show an improvement in the diversity gain.