Clustering the Rainfall Stations in Thailand Using Self-Organizing Map

Abstract

The purpose of this study is to classify the pattern of rainfall stations in Thailand over the period from the year 2000 to 2009 using Self-Organizing Map (SOM). The number of stations is collected based on the complete rainfall data during rainy season. The optimal cases of learning rate function and different two SOM array sizes are investigated. The Gaussian function is applied as the neighborhood function. The results show that linear function and the 2×2 SOM array size are most suitable for classifying the pattern of rainfall. The similar characteristic of daily rainfall in each station is clustered into four patterns. Thus, SOM can be classified the similar rainfall stations, efficiently.