The identification of spatial and temporal torrential rainfall patterns based on robust principal

Principal component analysis (PCA) guided clustering approach is widely used in high dimensional data especially in identifying the spatial distribution patterns of daily torrential rainfall. Typically, a common method of identifying rainfall patterns for climatological investigation employed T mode based Pearson correlation matrix to extract the relative variance retained. However, rainfall data in Peninsular Malaysia involve skewed observations which only take positive values and are skewed towards higher values. Thus, applying PCA based Pearson correlation on rainfall data set could affect cluster partitions and generate extremely unbalanced clusters in a high dimensional space. Another issues that crop up in this study is classical clustering techniques have drawbacks where these techniques require certain assumptions which contradict with the characteristics of rainfall and these techniques divides the database of rainfall patterns into clusters with the assumption that every rainfall pattern belongs only to one specific cluster. Therefore, the main objective of this research is to include a robust spatial and temporal daily torrential rainfall pattern approach in the PCA and also to employ the biclustering method to daily torrential rainfall analysis. Firstly, a robust dimension reduction method in Principal Component Analysis (PCA) is used to rectify the issue of unbalanced clusters in rainfall patterns due to the skewed nature of rainfall data. Secondly, a series of two-way clustering methods known as biclustering that could be applied to daily torrential rainfall analysis in order to show clear classification in defining daily rainfall patterns in Peninsular Malaysia. As results, it has a substantial improvement with the robust PCA combined with bicluster analysis in terms of the average number of clusters obtained and its cluster quality. The outputs can be beneficial to those who would like to apply a more accurate regional future rainfall data in their studies.