Bayesian Hidden Markov Modelling on East Java Montly Rainfall Data

Hidden Markov Model (HMM) is the development of a Markov chain with a state that cannot be observed directly (hidden). One of the HMM applications is rainfall modeling. Sometimes rain has intermittent characteristics, seasonal variations, and variance inhomogeneities, so this study develops HMM on spatial rainfall data with a Bayesian approach. In this study, there is a general linear model (GLM) that supports exogenous variables directly affecting the rainfall characteristics of precipitation at each site over time, while the Markovian transition between weather stations represents seasonal and sub- seasonal weather variability. The data used is monthly rainfall data from January 2018 to June 2021 at eleven weather stations in East Java supported by location and altitude information. Rainfall is classified into some states to build a model. The best model is obtained based on the smallest AIC, namely the model with four hidden states. The results of the analysis show that in the steady-state development, the potential for rainfall in East Java has the highest chance of 0.346.