Rainfall Analysis over Mauritius Using Principal Component Analysis

Abstract

Principal Component Analysis (PCA) is one of the most popular dimensionality-reduction techniques used for the purpose of classification. It has gained popularity because of its ease of use and optimal nature in a mean square error sense. This paper focuses on an analysis of PCA as a classification tool for rainfall images. Rainfall maps were generated using monthly precipitation data from 226 stations over Mauritius for the years 1992 to 1995 and also for the mean monthly rainfall for the years 1961 to 1990. Three different methods of applying PCA were then used on the rainfall maps and each method generated principal components reflecting the relevant percentage of variability. To interpret these components, a study of the various rainfall bearing climate systems and the wind-field active in Mauritius proved to be necessary to enable their association of the most significant principal components. The effect of altitude on rainfall has been found to have most dominant effect on rainfall. Other climate systems such as the ITCZ, cyclones, anticyclones, cold fronts and the perturbation of easterlies were also found to have specific rainfall patterns associated with them.