Estimating county-level digital economy development in china using satellite data: insights from explainable spatial machine learning

The digital economy is reshaping growth and societal change. However, conventional data lack spatial and temporal detail. Based on the concept of digital accessibility, this study used satellite data to examine the spatial patterns of digital economic development across China’s counties from 2018 to 2020. Baseline relationships were established using ordinary least squares regressions. Moran’s I and geographically weighted regression were utilised to detect spatial effects. Random forest and geographical random forest (GRF) models were also applied to capture nonlinearities. The GRF achieved high accuracy (R²: 0.94–0.96) and showed stronger explanatory power in western China. To prioritise key factors, SHapley Additive exPlanations highlighted urbanisation intensity and urban expansion as the most influential, with strong positive effects. Overall, the findings demonstrate that satellite data combined with spatial machine learning offer a powerful tool for monitoring and guiding policies aimed at reducing uneven digital transformation at the subnational level.