Spatio-temporal Variations and Forecast of PM2.5 concentration around selected Satellite Cities of Delhi, India using ARIMA model

Abstract

Air pollution presents serious threats to society around the world, especially in India. Among various ambient air pollutants, particulate matter (PM_2.5 & PM₁₀) have drawn significant attention from researchers owing to its adverse health impacts. Therefore, the accurate prediction of particulate matter 2.5 (PM_2.5) is essential for effective air pollution management and the prevention of respiratory diseases. The present study aims to systematically monitor and forecast the concentration of PM_2.5 in selected satellite cities of Delhi, an area that has been relatively underexplored despite its high pollution levels. In such data scarce zone, the estimation and prediction of PM_2.5 have been done using an autoregressive integrated moving average (ARIMA) model. The model's predictive accuracy and stability were validated with correlation coefficient (R), root mean square error (RMSE), mean absolute error (MAE), and relative prediction error (RPE). The results indicate that ARIMA model predicted PM_2.5 with sufficient accuracy for the current research area, demonstrating superior values of R (0.90), R² (0.82) and lower RPE (16.84), RMSE (18.28), MAE (16.89). The findings of the study indicate that the ARIMA model is a reliable method to predict PM2.5 concentrations, with acceptable accuracy. However, the ARIMA model depends on historical time series data to find trend and predict future conditions, assuming that the series remains static. Subsequently, it cannot consider the external factors that might cause alterations in the series. Such assumption limits its ability to effectively model cause-and-effect relationships. This approach is helpful for policy formulation and governance.