A satellite-driven model to estimate long-term particulate sulfate levels and attributable mortality burden in China

•We build a satellite-driven machine learning model to predict particulate sulfate concentrations across China using atmospheric big data from 2005 to 2018.•Our daily and monthly mean sulfate predictions agree well with ground observations with an out-of-bag cross-validationR2 of 0.68 and 0.93, respectively.•Population-weighted national annual mean sulfate concertation was relatively stable before the enforcement of the Air Pollution Prevention and Control Action Plan in 2013, then significantly decreased by 28.7% from 2013 to 2018.•The national annual mean total non-accidental and cardiopulmonary deaths attributed to sulfate decreased by 40.7% and 42.3% from 2013 to 2018, respectively. Ambient fine particulate matter (PM2.5) pollution is a major environmental and public health challenge in China. In the recent decade, the PM2.5 level has decreased mainly driven by reductions in particulate sulfate as a result of large-scale desulfurization efforts in coal-fired power plants and industrial facilities. Emerging evidence also points to the differential toxicity of particulate sulfate affecting human health. However, estimating the long-term spatiotemporal trend of sulfate is difficult because a ground monitoring network of PM2.5 constituents has not been established in China. Spaceborne sensors such as the Multi-angle Imaging SpectroRadiometer (MISR) instrument can provide complementary information on aerosol size and type. With the help of state-of-the-art machine learning techniques, we developed a sulfate prediction model under support from available ground measurements, MISR-retrieved aerosol microphysical properties, and atmospheric reanalysis data at a spatial resolution of 0.1°. Our sulfate model performed well with an out-of-bag cross-validationR2 of 0.68 at the daily level and 0.93 at the monthly level. We found that the national mean population-weighted sulfate concentration was relatively stable before the Air Pollution Prevention and Control Action Plan was enforced in 2013, ranging from 10.4 to 11.5 µg m−3. But the sulfate level dramatically decreased to 7.7 µg m−3 in 2018, with a change rate of −28.7 % from 2013 to 2018. Correspondingly, the annual mean total non-accidental and cardiopulmonary deaths attributed to sulfate decreased by 40.7 % and 42.3 %, respectively. The long-term, full-coverage sulfate level estimates will support future studies on evaluating air quality policies and understanding the adverse health effect of particulate sulfa