Compositional Constraints are Vital for Atmospheric PM_2.5 Source Attribution over India

India experiences some of the highest levels of ambient PM_2.5 aerosol pollution in the world. However, due to the historical dearth of in situ measurements, chemical transport models that are often used to estimate PM_2.5 exposure over the region are rarely evaluated. Here, we conduct a novel model comparison with speciated airborne measurements of fine aerosol, revealing large biases in the ammonium and nitrate simulations. To address this, we incorporate process-level changes to the model and use satellite observations from the Cross-track Infrared Sounder (CrIS) and the TROPOspheric Monitoring Instrument (TROPOMI) to constrain ammonia and nitrogen oxide emissions. The resulting simulation demonstrates significantly lower bias (NMB_Modified: 0.19; NMB_Base: 0.61) when validated against the airborne aerosol measurements, particularly for the nitrate (NMB_Modified: 0.08; NMB_Base: 1.64) and ammonium simulation (NMB_Modified: 0.49; NMB_Base: 0.90). We use this validated simulation to estimate a population-weighted annual PM_2.5 exposure of 61.4 μg m^-3, with the RCO (residential, commercial, and other) and energy sectors contributing 21% and 19%, respectively, resulting in an estimated 961,000 annual PM_2.5-attributable deaths. Regional exposure and sectoral source contributions differ meaningfully in the improved simulation (compared to the baseline simulation). Our work highlights the critical role of speciated observational constraints in developing accurate model-based PM_2.5 aerosol source attribution for health assessments and air quality management in India.