Global ground-level SO2: gap-free mapping and analyses of long-term trends, exposure, and inequality.

Sulfur dioxide (SO2) is a harmful air pollutant and an important precursor of sulfate aerosol and acid deposition, yet global long-term, ground-level SO2 fields remain unavailable at daily resolution with seamless spatial coverage. Here we develop a two-stage multi-source data fusion framework based on LightGBM to generate a global, gap-free dataset within 60°S-60°N of near-surface SO2 over land from 2 October 2004 to 31 December 2024 at 0.1°×0.1° resolution. Verification results show strong agreement with daily ground observations under sample-based 10-fold cross-validation (R = 0.91), while more stringent station- and year-based validation indicates reduced performance, particularly at finer temporal scales. The resulting product reveals pronounced global heterogeneity, with persistent hotspots throughout 2005-2024. Over this period, global near-surface SO2 generally declined, but trends diverged substantially across regions, with large decreases in China, North America, and Europe and more modest or heterogeneous changes across parts of South Asia, the Middle East, and Africa. Population-weighted exposure (the population-share-weighted mean ground-level concentration) decreased in most regions; however, inequality analysis using the Concentration Index (CI; the population-weighted covariance between SO2 exposure and countries' GDP per capita rank) indicates that since the early 2010s the remaining global SO2 burden has become increasingly concentrated in lower-income countries. This dataset provides a consistent basis for tracking long-term SO2 changes, identifying persistent exposure hotspots, and supporting regional-to-global analyses of population exposure and cross-country inequality, and is particularly useful at monthly-to-annual timescales.