Prediction of Particulate Matter (PM2.5) Concentrations over an Urban Region using Different Satellite

The accurate estimation of ground-level particulate matter concentrations (PM2.5) is essential for assessing air quality and its impact on human health and the environment. This study focused on estimating PM2.5 concentrations from January 2021 to June 2023 in the city of Lucknow, India. Various models, including Bivariate Linear Regression (LR), Multiple Linear Regression (MLR), and Artificial Neural Network (ANN) predicted PM2.5 concentrations at the station. Additionally, CALIPSO observations successfully demonstrated the vertical aerosol layer profile in the region. To improve accuracy, we incorporated Aerosol Optical Depth (AOD) data from both MODIS and VIIRS, along with meteorological parameters. The dataset was divided into two periods: 2017-2020 for estimation and January 2021-June 2023 for model training. Our findings revealed a positive correlation between model outputs, observed ground data, and meteorological parameters. For MODIS, LR, MLR, and ANN models had correlation coefficients (R) of 0.41, 0.57, and 0.66. Similarly, for VIIRS, the R-values were 0.33, 0.55, and 0.64, indicating promising agreement between model predictions and actual PM2.5 concentrations. These findings contribute to a better understanding of air quality dynamics and can support policymakers in implementing effective measures to mitigate the adverse effects of particulate matter pollution on public health and the environment. Data sets underwent three divisions: 80% for training, and 10% each for validation and testing. ANN displayed strong correlation coefficients (R) across datasets, achieving MODIS R-values of 0.74 and 0.73 for training and overall sets, and VIIRS R- values of 0.74 and 0.72. This study highlights the significant accuracy improvement in PM2.5 estimation by integrating meteorological, land use data, and satellite AOD. While LR and MLR methods yielded comparable outcomes, ANN emerged as a superior technique for long-term PM2.5 estimation, holding promise for air quality monitoring and guideline adherence in diverse regions.