Estimation of Extreme Rainfall Patterns Using Generalized Linear Mixed Model for Spatio-temporal data in West Java, Indonesia

Abstract

Abstract Based on geographical location, Indonesia is crossed by the equator which has a tropical climate with a high diversity of rainfall. This can lead to an increase or decrease in extreme rainfall that has the potential to cause hydrometeorological disasters such as floods, winds and landslides. Estimation of rainfall is not only influenced by the local topography in each region, but there are also spatial and temporal influences. Therefore, this study aims to estimate the pattern of extreme rainfall and map the rain observer station based on the status of extreme rainfall on spatio-temporal data in the West Java province, Indonesia. Estimator of extreme rainfall patterns uses the generalized linear mixed model (GLMM) which consists of spatial and temporal random components with INLA (integrated nested Laplace approximation) inference. Response data in the form of daily rainfall in 73 observer stations, within 2 years period of observation. Average rainfall data is assumed to have gamma distribution, rainfall above the average is identified using the Bernoulli distribution then identified (extreme) rainfall above average is assumed to have generalized pareto distribution. From the rainfall patterns produced there are a number of conclusions, including that the average rainfall in West Java experiences a consistent downward trend in 13-th week to 32-nd week, which indicates that there is a decrease in rainfall intensity at the beginning of dry season in early April until the end of dry season at the end of August. While in the case of extreme rainfall, the decrease in rainfall intensity consistently occurs from the first week in January to 32-nd week at the end of August. The resulting extreme rainfall mapping states that more extreme rainfall occurs in the north to southeast region of West Java.