Full-coverage 1 km daily ambient PM<sub>2.5</sub> and O<sub>3</sub> concentrations of China in 2005–2017 based on a multi-variable random forest model

Abstract. The health risks of fine particulate matter (PM2.5) and ambient ozone (O3) have been widely recognized in recent years. An accurate estimate PM2.5 O3 exposures is important for supporting risk analysis environmental policy-making. aim our study was to construct random forest models with high-performance daily average concentration maximum 8 h (O3-8 hmax) China 2005–2017 at a spatial resolution 1 km × km. model variables included meteorological variables, satellite data, chemical transport output, geographic socioeconomic variables. Random based on 10-fold cross-validation established, temporal validations were performed evaluate the performance. According sample-based division method, daily, monthly yearly estimations from test datasets gave model-fitting R2 values 0.85, 0.88 0.90, respectively; these 0.77, 0.77 0.69 O3-8 hmax, respectively. their lagged can significantly affect both hmax estimations. During 2005–2017, exhibited an overall downward trend, while experienced upward trend. Whilst patterns barely changed between 2005 2017, trend had characteristics. dataset accessible public https://doi.org/10.5281/zenodo.4009308 (Ma et al., 2021a), shared Chinese Environmental Public Health Tracking (CEPHT, 2022) available https://cepht.niehs.cn:8282/developSDS3.html.