Disparity in exposure to PM2.5 species in India reveals the importance of considering PM2.5 composition in epidemiology studies

Ambient PM_2.5 exposure remains the most critical environmental risk to public health in India; however, exposure to PM_2.5 species remains poorly quantified due to the absence of a systematic network for measuring PM_2.5 composition. This has limited our understanding of the differential health impacts of PM2.5 species. Here, we developed a novel high-resolution (1-km x 1-km) dataset of concentrations of six aerosol species (BC, OC, sulfate, nitrate, ammonium, dust) and assessed exposure inequality by integrating sociodemographic data from the National Family Health Surveys (NFHS-4 and NFHS-5) across urban and rural India, focusing on gender and wealth.

We trained machine learning models to predict the mass fractions of six PM_2.5 species derived from a chemical transport model (CTM), using four predictor variable types: (1) Multi-angle Imaging Spectro-Radiometer-retrieved size- and shape-segregated AODs, (2) sectoral emissions, (3) meteorology, and (4) geospatial variables. These predicted mass fractions were combined with satellite-derived PM_2.5 data to estimate monthly mass concentrations across South Asia. The model shows robust performance against ground-based observations (R² = 0.61; RMSE = 4.23 mg/m³). 

Population-weighted exposure to BC, OC, sulfate, nitrate, ammonium, and dust in India for the NHFS-4 (NFHS-5) was estimated to be 4.22 (4.19), 10.51 (11.02), 5.99 (5.85), 6.97 (7.07), 5.65 (5.58), and 15.97 (16.40) mg/m³, respectively. Exposure was highest in low-SDI states (the Indo-Gangetic Plain and Central India), driven by persistent reliance on biomass and solid fuels. Middle-SDI states achieved the largest reductions in overall exposure from NHFS-4 to NHFS-5, likely due to high clean-fuel conversion rates. In these regions, the urban poor faced a disproportionate relative burden from OC exposure (Z-score of -1.001 in NFHS-5), suggesting that air quality improvements primarily benefited the wealthy. While relative disparities narrowed in urban clusters between the two survey rounds, they widened in rural clusters. Women in rural regions were consistently exposed to elevated levels of carbonaceous aerosols (BC and OC), highlighting the gendered impacts of residential air pollution. In high-SDI urban areas, the relative disparity shifted from positive to negative. This shows that vulnerability patterns change as states progress to higher levels of development. These results underscore the dynamic nature of human exposure throughout economic changes. The study further emphasises the essential need for gender- and wealth-stratified exposure tracking to ensure that national clean air programs do not neglect disadvantaged people as they progress.