Prediction and Mapping of Air Pollution in Bandung Using Generalized Space Time Autoregressive and Simple Kriging

This study aims to obtain a future mapping of NO2 and SO2 pollutant levels in the city of Bandung and it's around. Two mathematical models are used in this study namely the Generalized Space-Time Autoregressive (GSTAR) and Simple Kriging models. The GSTAR model is an estimator model based on time and this study is used to predict future pollutant levels at specific location points. While the Simple Kriging model is a spatial model which in this study is used to interpolate or estimate pollutant levels around points previously estimated using the GSTAR model. The pollutant level estimation of Simple Kriging is then used to map the pollutant levels in Bandung in the next few years. Ljung-Box statistical test results show that the GSTAR model is feasible to use. The assumption of stationary data on the simple Kriging model is determined using the Augmented Dickey-Fuller (ADF) test, while the determination of the semivariogram model is determined using the RMSE calculation. The ADF test showed that stationary data was obtained for the first derivative, whereas the most appropriate theoretical semivariogram model is an exponential semivariogram. Pollutant mapping results in 2021-2024 show that the northern and southern Bandung regions have higher levels of pollutant levels than other locations, although significant increases only occurred for NO2 levels.