Seasonal and Regional Characteristics in the Frequency Distribution of Precipitation

Abstract

The frequency distribution of precipitation changes from inverse J-shaped type, through the exponential, to normal type with increase in a length of lumped period. The seasonal and regional characteristics of the change with lumped period from 1 hour to 30 days are investigated for AMeDAS precipitation data sets. The gamma distribution is employed as a distribution model fitted to the observed data and the method of maximum-entropy as a method of the parameter estimation. The shape of a frequency distribution can be expressed quantitatively by the information content due to logarithmic geometric mean. The following findings are drawn from the frequency analysis: 1) Lumped period in which a frequency distribution is the exponential one for winter-half-year is rather short on the Japan Sea side and rather long on the Pacific Ocean side. The regional difference is understood by considering that main precipitation-induced disturbance is different for each region and each disturbance has its own time-series characteristics. 2) As for variation of information content due to logarithmic geometric mean in short-term, there is latitudinal difference in the Japanese Islands, which corresponds with difference between precipitation characteristics represented by rain shower and those by continuous rain.