Multi-scale analysis based on spatial Markov chain model provides insights into long-term and short-term SO2 control in China

Sulfur dioxide (SO 2 ) is a major atmospheric pollutant in China, and understanding its spatiotemporal dynamics is essential for effective air quality management. Using hourly SO 2 data from more than 2,000 urban monitoring stations during 2015–2022, this study examines SO 2 variation at daily, monthly, and annual scales. Spatial patterns and temporal trends are analysed using the Global Moran’s I Index and both traditional and spatial Markov chain models, with spatial proximity defined by inverse distance. Model predictions are validated using independent data from 2023.The results show clear seasonal variability at daily and monthly scales, with higher SO 2 concentrations in winter and lower levels in summer, alongside a sustained national decline at the annual scale. Significant spatial autocorrelation is observed across all time scales, particularly in coal-dependent and industrial regions. Incorporating spatial proximity substantially changes transition probabilities, demonstrating strong spatial spillover effects, especially at shorter time scales. Validation results confirm the robustness of the spatial Markov model. Overall, the findings underscore the importance of spatial dependence and temporal scale in analysing SO 2 dynamics and provide scientific support for coordinated regional air pollution control strategies in China. • The mean annual concentration of SO 2 in China has dropped by 58% since 2015. • Transfer probabilities among various SO 2 levels differ across time scales. • Urban SO 2 concentration dynamics are significantly affected by neighboring areas. • Spatial Markov chain is more suitable for analyzing SO 2 dynamics at short time scales. • Geospatial relationships are important in the prediction and management of air pollutants.