Synthetic Generation of Tropical Rainfall Time Series Using an Event-Based Method

Abstract

An algorithm for simulating tropical rainfall using an event-based method has been developed. The simulation process uses a one-step Markov chain model. The driving mechanism of the model is based on the statistics derived from rainfall recorded using the event-based method of recording proposed by Tan and Chiew. These statistics are expressed in terms of a transition matrix that can be used to generate the next event after the occurrence of the first. The model was applied to rainfall recorded in Singapore. Despite the relatively sparse source data, the results show that the synthetic rainfall time series reproduces the essential characteristics of tropical rainfall in terms of mean and total rainfall, and the internal characteristics of tropical rainfalls. Furthermore, this methodology is capable of generating plausible estimates of extreme annual rainfall.