Identification of rainfall temporal patterns

Abstract

Temporal pattern for rainfall events is required in the design and evaluation of hydrologic safety for hydraulic structures. In this paper, a method of statistical cluster analysis is applied to identify event-based representative temporal rainfall pattern in fourteen stations in Johor. The 8 hour separation time of no rain is used to divide the 5 years rainfall data into individual rainfall event according to the monsoons. The analysis is implemented from the dimensionless mass curve ordinates as the attributes in statistical cluster analysis of the cumulative storm depth over the non-dimensionalized time. As a result, three representative rainstorm patterns are identified and classified under the two basic types of advanced-type (A) and central-peak type (C) for northeast and southwest monsoon. Meanwhile, only one basic type is identified to represent rainstorm pattern during the inter-monsoon that is advanced type (A). In addition, the rainfall pattern is dependent on rainfall depth and duration, season and geographical location by contingency table test. The rainfall information presently used for design in Malaysia is very dissimilar to the representative curves derived in this study. The identification of three representative rainfall temporal patterns according to the monsoon seasons in Johor can be used as a basis to stochastically generate the plausible rainfall hyetographs of the specified pattern in Johor.