Unveiling Temporal and Spatial Variations in the Daily PM2.5 Over Indian Cities
摘要
Air pollution significantly threatens the environment, the economy, and human health. Among the air pollutants, PM2.5 (diameter of particulate matter is 2.5 micrometers or smaller) is a notable threat to human health. Analysing the temporal and spatial variations in daily PM2.5 concentrations is crucial for developing pollution control measures and regulations. However, due to the absence of long-term observed daily PM2.5, the temporal and spatial variations in the daily PM2.5 concentrations over India have not been explored. Therefore, this study aims to estimate the daily PM2.5 concentration from 2000 to 2020 using a long short-term memory (LSTM)-based machine learning approach across fourteen stations in India and analyse the temporal and spatial variations in daily PM2.5 concentrations. For a country like India, where air pollution significantly burdens public health and the economy, this work provides an essential tool for improving data-driven environmental policies. The inputs of the LSTM model for the estimation of daily PM2.5 are satellite-derived aerosol optical depth (AOD), rainfall, atmospheric pressure, atmospheric temperature, and wind in the horizontal and vertical directions. The results indicate that the LSTM model performed well across different monitoring stations, with high R2 (coefficient of determination) values (0.87–0.96), relatively low RMSE (root mean square error) values (1.05–3.36 µg/m3) and relatively low MAE (mean absolute error) values (0.86–2.35 µg/m3). The temporal analysis revealed an overall increase in the PM2.5 concentration across all 14 stations. Among the 14 stations, a maximum increase of 161.67% in the monthly average daily PM2.5 concentration is observed at S12 (Faridabad) during December. Furthermore, trend analysis using the Mann‒Kendall trend test revealed a steep increasing trend in the PM2.5 concentration during the winter months. Most of the stations in the North Indian state showed steep increasing trends in all months compared to other stations. In addition, trend analysis on the PM2.5 air quality subindex (AQI) revealed that the AQI categories “satisfactory” and “good” consistently exhibited decreasing trends, while the AQI categories “severe” and “very poor” displayed persistent increasing trends across all stations. The findings can be effectively used by policymakers, regulatory bodies, and researchers working towards improving air quality in India.