The comparison of T-Mode and Pearson Correlation matrices in classfication of daily rainfall patterns in Peninsular Malaysia

Abstract

The aim of this study is to identify daily rainfall patterns of wet days linked to the topography of Peninsular Malaysia using two different configurations of points in the data. The data used in this study were obtained from 75 rain gauge stations in Peninsular Malaysia from the year 1975-2007. We only consider data for the period in which southwest monsoon occur from June until September yielding a total of 153 days.A typical classification approach in identifying daily rainfall patterns requires the use of configuration points of entities between the rows and column of the data based on correlation matrices. In this study, we compare effect on the cluster of daily rainfall patterns on two types of correlation matrices: T-mode correlation matrix and Pearson correlation matrix. These matrices are then used as inputs for Principal Component Analysis (PCA) to reduce the dimension of the dataset before clustering the rainfall patterns of wet days. We have found that although Tmode correlation matrix is popularly used in subtropical climate studies, it is unable to show clear classification in defining daily rainfall patterns in tropical climate data. Using Calinski and Harabasz Index, only two-rainfall pattern cluster can be identified on T-mode correlation matrix. On the other hand, Pearson correlation matrix showed three different rainfall patterns and each cluster are identified to be linked to certain topographic characteristics. These three clusters indicate that the rainfall pattern during the southwest monsoon experiencing the most heavy rain in the western part of the Peninsula, particularly in characterizing the rainfall pattern of the northwestern and western region of Peninsular Malaysia. These clusters are mapped out using ARCGIS software.