Development of a Diurnal Rainfall Pattern Generator Incorporating Characteristics of Long- and Short-Term Rainfall Intensity

Abstract

A diurnal rainfall pattern generator which can statistically reproduce observed characteristics of heavy rainfall in target areas was developed. The data are obtained as hourly data, and it is possible to generate a lot of heavy rainfall with different patterns of internal structure. We can set any given value as a generation period, and the generator composed of four procedures as follows: 1) to decide the monthly event frequency, 2) to generate total rainfall amounts for each event of daily rainfall, 3) to disaggregate the total rainfall amount into hourly data, and finally, 4) to rearrange disaggregated data series by performing autocorrelation correction. An important purpose of the final step is to correct the autocorrelation relation of disaggregated hyetographs. In this process, a hyetograph with autocorrelation relation was newly generated by each rearrangement. The hyetographs were used as a reference pattern to rearrange the hourly data. Parameters in each step were estimated by using long-term rainfall data collected from the target area to reproduce observed characteristics. Here, we set 100 years as the generation period and 305 events were generated. The data were compared with observed data to verify accuracy of the method. The verification results demonstrated that the characteristics of heavy rainfall were reproduced with good accuracy, so these data showed availability as input rainfall data for any analysis such as flood simulation.