Forecasting Flood Using Markov Chain Model

Abstract

Johor is currently experiencing rainfall irregularities and frequent flood event, thus, study on forecasting flood event is crucial. It is quite impossible to elude the event, but people can prepare the hit of the event. Hence, Markov Chain Model was applied to forecast floods to minimize and mitigate the flood risks. The long historical rainfall data were collected for eight (8) stations in Johor from Department of Irrigation and Drainage (DID) Malaysia. In identifying the historical of wet event, the rainfall data of 30 years was obtained, then the Standardized Precipitation Index (SPI) was applied for 6-month, 9-month, and 12-month of time scales. The results from the time scales showed the significant upward trend for all stations. Meanwhile, to forecast the flood, the Markov Chain Model was applied as it uses the transition probability matrix to determine the probability of flood occurrence. Initial state and initial month were used for the prediction of 1 to 4 months ahead. The results showed the prediction for 1 month ahead obtained the highest value around 0.955. The highest value obtained for 2 months ahead was 0.914, while 3 months and 4 months ahead were 0.832 and 0.844, respectively. Overall, the results prove that the Markov Chain Model has a potential to forecast wet events.