Machine learning models for estimating criteria pollutants and health-risk based air quality index over eastern coast coalmine complex belts

Pradeep Kumar¹Arti Choudhary¹P. K. Joshi²Dr Ram Pravesh Kumar²R. Bhatla^1*

• ¹Institute of Science, Banaras Hindu University, Varanasi, India
• ²Jawaharlal Nehru University, New Delhi, National Capital Territory of Delhi, India

Estimating criteria pollutants is crucial due to their continuous rise and impact on respiratory health. To mitigate the impact of air pollution on human health, it is essential to understand the concentration of air pollutants at specific locations. This study aims to evaluate variation and estimate the levels of criteria pollutants and their potential on human health risk in the vicinity of a coal-mine complex and thermal power plant situated in an eastern coastal state of India.The pre-existing hotspot regions Talcher and Brajrajnagar, host of many coal-fired power plants and cluster of coal mining blocks of coastal state Odisha, are considered. Talcher shows consistently higher levels of PM10, NO2, and SO2, reflecting a greater industrial impact.Brajrajnagar, while also impacted, exhibits comparatively lower pollutant concentrations. The observed seasonal trends highlight the necessity for targeted mitigation strategies to address pollution levels and associated health risks in these regions. Novel machine learning (ML) models, including Independent Component Regression (ICR), ElasticNet (ENET), and Boosted Tree (BT), are applied to estimate criteria pollutants. Statistical analyses highlight BT as the superior model, outperforming ENET and ICR in pollutant estimation particularly in Talcher.Taylor plots and statistical evaluations further validate the BT model's robustness in air pollutant estimation. Additionally, the study assesses the associated health risks posed to nearby populations of Talcher and Brajrajnagar. The analysis highlights significant spatial disparities in pollution levels, with Talcher consistently recording higher concentrations of PM10, NO2, and SO2 and poorer AQI compared to Brajrajnagar. Talcher also shows greater health risks, with pollutant exposure linked up to 6% higher risks for PM10, 5% for NO2, and up to 3% for SO2. The health risk-based air quality index (HAQI) reveals underestimation of health risks by the current AQI, emphasizing the need for improved metrics to address multipollutant exposure impacts.