The clustering analysis and spatial interpolation of intense rainfall data

Abstract

The measurement of rainfall data during disaster periods is an important task. Unfortunately, some rainfall data will be missed owing to unpredictable factors. Therefore, this study first collected the hourly rainfall records from past disaster events. Next, a statistic cluster analysis method was used to analyse the correlation between the rainfall records in each station. Finally, a spatial interpolation computing method was applied within each cluster to predict reliable rainfall estimates for the areas that lacked past rainfall records. The cluster analysis results showed that selecting the nearby three to four rainfall stations for the spatial interpolation analysis simplified the calculation process. In particular, the result of grouped cluster analysis could enhance the accuracy of the rainfall estimation in the mountainous areas. This study established a reliable rainfall estimation method as a basis for future regional disaster analysis.