Modelling health implications of extreme PM2.5 concentrations in Indian sub-continent: Comprehensive review with longitudinal trends and deep learning predictions

Air pollution poses a critical global challenge, disproportionately impacting public health and the environment in developing nations like India. The review suggests rapid urbanisation, industrialization, and increased energy consumption have worsened air quality, where average annual PM2.5 concentration far exceeds World Health Organisation (WHO) guidelines of 5 μg/m3, leading to high mortality and increased disability-adjusted life years. Indoor air pollution from biomass burning exacerbates the issue, affecting millions of populations in India who rely on traditional fuels. Despite strides in air quality monitoring through National Mission on Air Pollution (NMAP), challenges such as uneven data coverage and limited ground stations for entire country, especially in rural areas, and outdated emission standards hamper effective policy implementation. Therefore, this study utilizes MERRA-2 reanalysis and Global Burden of Disease datasets, this study analysed disease-related mortality influenced by pollution extremes [MPM2.5 (Mean Annual Pollution through PM2.5), PM2.5D (Polluted days through PM2.5), MAPM2.5 (Maximum 1-day pollution amount), and PM2.599p (Heavily polluted regions)]. Single and multilinear regression analyses were conducted between pollution extremes and disease-related mortality, followed by a Convolution Neural Network (CNN) to predict mortality by disease, state, and gender based on pollution extremes. The study revealed significant spatiotemporal variation in PM2.5 concentrations across India, with northern states exceeding air quality guidelines and PM2.5 levels more than doubling in the Indo-Gangetic Plains between 1980-1990 and 2010–2020. Regression analysis showed correlation between PM2.5 and neurological disorders and chronic respiratory diseases, while respiratory infections and tuberculosis had the weakest correlation. Further a dense CNN model improved predictive accuracy, achieving R2 values between 0.84 and 0.94 across states, diseases, and genders. The study will provide a valuable insight to air quality and health monitoring programme (AQHMP) through suggesting stricter pollution standards, expanded rural monitoring, sector-specific policies, improved emission inventories, and advanced technologies with AI&ML and remote sensing for better data and reduced health risks.