Abstract
Underestimation of PM(2.5) emissions from the agricultural sector persists as a major difficulty for air quality studies, partly because of underutilization of high-resolution observation platforms for constructing a global emissions inventory. Coarse-resolution products used for such purposes often miss fine-scale burnt areas created by stubble-burning practices, which are primary sources of agricultural PM(2.5) emissions. For this study, we used the high-resolution Sentinel-2 observations to examine the spatiotemporal variability of burnt areas in Punjab, a major hotspot of agricultural burning in India, during the post-monsoon fire season (October-December) in 2022-2024. The results highlight the Sentinel-2 capability of detecting more than 34,000 km(2) of burnt areas (approx. 68% of Punjab's total area) as opposed to the less than 7000 km(2) (approx. 12% of Punjab's total area) detected by MODIS. The study also reveals, in unprecedented detail, multi-annual spatial and temporal shifting of burning events from northern to central and southern Punjab. This detection discrepancy has led to marked disparities in estimated monthly emissions, with approximately 217.3 million tons of PM(2.5) emitted in October 2022 compared to 8.7 million tons found by EDGAR v.8.1. This underscores higher-resolution observation systems intended to support construction of a global PM(2.5) emissions inventory.