Abstract:

In this paper, we develop SARIMA time series models to forecast PM2.5 in the upper northern region of Thailand using monthly average PM2.5 data obtained from Giovanni. The data are structured as a grid of 26 grids, divided into a training set (January 2016-December 2023) with 96 values and a testing set (January-December 2024) with 12 values. The results show that the SARIMA(1,0,0)(1,1,1)_{12} model for Phayao province yielded the lowest MAPE value at 10.43%. Therefore, the resulting models demonstrate high potential for air pollution forecasting and policy planning in high-risk areas, enabling effective responses to pollution problems.