Particulate matter with aerodynamic diameters ≤1 μm (PM1) has a greater impact on the human health but has been less studied due to fewer ground observations. This study attempts to improve the retrieval accuracy and spatial resolution of satellite-based PM1 estimates using the new ground-based monitoring network in China. Therefore, a space-time extremely randomized trees (STET) model is first developed to estimate PM1 concentrations at a 1 km spatial resolution from 2014 to 2018 across mainland China. The STET model can derive daily PM1 concentrations with an average across-validation coefficient of determination of 0.77, a low root-mean-square error of 14.6 μg/m3, and a mean absolute error of 8.9 μg/m3. PM1 concentrations are generally low in most areas of China, except for the North China Plain and Sichuan Basin where intense human activities and poor natural conditions are prevalent, especially in winter. Moreover, PM1 pollution has greatly decreased over the past 5 years, benefiting from emission control in China. The STET model, incorporating the spatiotemporal information, shows superior performance in PM1 estimates relative to previous studies. This high-resolution and high-quality PM1 data set in China (i.e., ChinaHighPM1) can be greatly useful for air pollution studies in medium- or small-scale areas.