Statistical Model for Air Quality Forecasting: A Case Study of Dust Particles No Larger Than 2.5 Microns (PM2.5) in Chiang Mai Province

Abstract

This research aims to develop a statistical model for forecasting PM2.5 levels in Chiang Mai Province. Three forecasting techniques were compared: the Box-Jenkins model, Winter’s exponential smoothing model, and the linear regression model. The analysis utilized monthly average PM2.5 data from the Pollution Control Department covering 144 observations from 2013 to 2024. The dataset was divided into two subsets: a training set of 120 months (2013–2022) for model development and a testing set of 24 months (2023–2024) for model performance evaluation. Results showed that Winter’s exponential smoothing model yielded the most accurate predictions, with the smallest mean absolute percentage error (MAPE) among the three methods. The best-performing model was then used to forecast monthly PM2.5 levels for the year 2025. The forecast suggested that PM2.5 levels from February to April are likely to exceed the standard threshold of 37.5 µg/m3 set by the Pollution Control Department, underscoring the importance of proactive air quality management during this period.