Robust Clustering on Spatial Torrential Rainfall Patterns
Peninsular Malaysia has a tropical climate that is characterized by three monsoons. The aim of this study is to identify the spatial distribution patterns of the daily torrential monsoon rainfall in Peninsular Malaysia by clustering the most relevant principal directions between its day correlations. In such climate, the daily rainfall variability between monsoons typically differ and its daily rainfall patterns are influenced by the different wet days. Thus, the clustering results on such data would tend to generate too few clusters that are imbalanced when the different rainfall days are given equal weights in highlighting the spatial rainfall patterns. In this study, we use a robust correlation measure by applying a clustering method on the principal component loadings of the daily torrential rainfall based on a weighted correlation matrix of a Tukey’s biweight M-estimate. The findings indicate ten distinct rainfall patterns that appear to display the dominant role extended by the complex topography and exchange monsoons of the peninsular.
KeywordsTukey’s biweight correlation Robust correlation k-means cluster analysis Principal component analysis Pearson correlation
- 2.Wickramagamage P (2010) Seasonality and spatial pattern of rainfall of Sri Lanka: exploratory factor analysis. Int. J. Climatol. 30, 1235–1245Google Scholar
- 3.Hardin, J., Mitani, A., Hicks, L., VanKoten, B.: A robust measure of correlation between two genes on a microarray. BMC Bioinform. 8, 220 (2007). https://doi.org/10.1186/1471-2105-8-220
- 4.Shaharudin, S.M., Ahmad, N.: Modeling, design and simulation systems. In: CCIS, vol. 752, pp. 216–224. Springer, Singapore (2017). https://doi.org/10.1007/978-981-10-6502-6_19
- 5.Compagnucci, R.H., Araneo, D., Canziani, P.O.: Principal sequence pattern analysis: a new approach to classifying the evolution of atmospheric systems. Int. J. Climatol. 21, 197–217 (2001)Google Scholar
- 6.Choulakian, V.: Robust q-mode principal component analysis in L1. Comput. Stat. Data Anal. 37, 135–150 (2001)Google Scholar