Improving satellite-based estimation of surface ozone across China during 2008?2019 using iterative random forest model and high-resolution grid meteorological data

China is faced with increasing ozone pollution due to rapid economic development and urbanization. Although the ground monitoring network provides continuous real-time ozone measurements, its practical applications are limited due to sparse spatial distribution. The monitoring network coupling with various data and the machine learning algorithms is a promising approach to estimate surface ozone concentrations. However, previous studies on ozone estimation in China are restricted to small study scale, low spatial resolution and low predictive ability. The study aims to 1) improve the accuracy of surface ozone estimates across China using an iterative random forest (RF) model, more recent ground monitoring data and high-resolution grid meteorological data, and 2) estimate the daily max 8-h average ozone concentrations across China during 2008?2019 at a spatial resolution of 0.0625?. The iterative RF model showed that the sample-based and site-based cross-validation (CV) R2 were 0.84 and 0.79, respectively, indicating higher accuracy than the single RF model and previous studies. Daily max 8-h average ozone data product across China was estimated during 2008?2019 with an improved spatial resolution of 0.0625?. The newly generated ozone data product shows great potential in future studies to assess the short-term and long-term health effect of ozone pollution.