DEVELOPING AND CALIBRATING A STOCHASTIC RAINFALL GENERATOR MODEL FOR SIMULATING DAILY RAINFALL BY MARKOV CHAIN APPROACH

Rainfall is an important parameter in tropical humid regions for which paddy production systems depend. A significant portion of paddy water requirements is supplied by natural rainfall. Several studies have predicted changes in rainfall patterns and in the amount of rain that may be obtainable in future owing to climate change. There is increased concern about future water availability for an important crop such as rice. Need to develop new water management tools for sustainable production is inevitable, but such tools require long-term climate data that is credible and consistent with the time. This study concerns itself with evaluating a stochastic weather generator (WGEN) model for simulating daily rainfall series. The model is assessed using long-term historical rainfall data obtained from a rice growing irrigation schemes in Malaysia. The model is based on a first-order two-state Markov chain approach which uses two transition probabilities and random number to generate rainfall series. Selected statistical properties were computed for each station and compared against those retrieved from the model after model training and testing. The results obtained from these comparisons are quite satisfactory giving confidence about the performance and future outputs from the model. The model has shown good skill in describing the rainfall occurrence process and rainfall amounts for the area. The model will be adapted in a subsequent study for downscaling and simulating effective daily rainfall series corresponding to future climate scenarios.