Monitoring the Underlying Probability of Daily Rainfall Occurrence in West Java over Space and Time by Means: A Bayesian Spatiotemporal Model Approach

Abstract

Monitoring the underlying probability of daily rainfall occurrence is needed for many applications, such as in agriculture, hydrology and ecology, epidemiology, maritime and tourism and society in general. We use Climate Hazard InfraRed Precipitation with Station Data (CHIRPS) to obtain the estimated probability of daily rainfall occurrence. The model has been used based on Kitagawa (1987). We used the Kitagawa model with introducing the spatial and spatiotemporal dependencies in the Binomial smoothing model using Bayesian spatiotemporal model approach by means of the Integrated Nested Laplace Approach (INLA) and applied this model for estimating the underlying probability of daily rainfall occurrence in West Java, Indonesia. Eight different model were tested and model 8 was found as the best model based on Bayesian and classical criterions. We found the circular temporal trend of random walk order one and spatiotemporal interaction type IV are the two most important components in explaining the change of probability of daily rainfall occurrence. The daily precipitation chances reached 90% from November up to February for every year.