In 2023, the average PM2.5 3 concentration in India was 54.4 μg/m3, while WHO recommended not more than 5 μg/m3. \(\hbox {PM}_{2.5}\) , due to its smaller size, poses a significant threat to human health. Therefore, this study aims to forecast long-term \(\hbox {PM}_{2.5}\) concentrations across multiple regions of India, enabling authorities to implement necessary measures well in advance to address the \(\hbox {PM}_{2.5}\) issue. For this study, three diverse regions in Gujarat, India—Vatva, Maninagar, and Sector 10—characterised by industrial zone, high population density, and green space, respectively, were considered. The average monthly concentration from February 2019 to December 2023 was used to train the forecasting models, while the data from 2024 was used as a test dataset. The categorical plot and ANOVA test indicate that \(\hbox {PM}_{2.5}\) concentrations exhibit seasonal variation across all three regions, with elevated levels observed during the winter and post-monsoon seasons and lower levels during the monsoon season. Holt-Winters and SARIMA models were employed to forecast long-term \(\hbox {PM}_{2.5}\) concentrations. Analysis of accuracy metrics, including root mean square error (RMSE) and mean absolute error (MAE), revealed that Sector 10 demonstrated the best model performance, with the Holt-Winters model marginally outperforming SARIMA in terms of computational efficiency and accuracy, with RMSE and MAE values of 10.45 and 8.36, respectively. Following the evaluation of model accuracy, the models were retrained on the entire dataset and employed to forecast \(\hbox {PM}_{2.5}\) concentrations for 2025. Results indicate that future \(\hbox {PM}_{2.5}\) concentration may decrease in Vatva, whereas no change is observed for Sector 10 and Maninagar.
