Abstract: In recent years, photochemical smog pollution frequently occurs in the eastern cities of China. O₃, as a landmark pollutant of photochemical smog, is increasingly becoming the primary pollutant affecting urban or regional air pollution. To explore the Beijing-Tianjin-Tangshan (BTT) region surface ozone pollution characteristic and changing trend of time and space, this study based on feedforward (back propagation, BP) neural network, an inversion model of surface ozone concentration was established combined with the nonlinear relationship between the ozone concentration data of ground monitoring stations and the total ozone column data of satellite and meteorological data of meteorological stations. The results show that the R², RMSE, and MAE of the reliability evaluation model are 0.888, 10.742, and 9.596, respectively, and the accuracy of the established neural network model is relatively high. From 2016 to 2019, the annual average ozone concentration in the Beijing-Tianjin-Tangshan region showed an increasing trend. In the four seasons, ozone concentration in Beijing, Tianjin, and Tangshan are the highest in summer and the lowest in winter. The results provided a technical reference for near-surface ozone estimation and have important practical significance for environmental monitoring.

Key words: Beijing-Tianjin-Tangshan (BTT) region, BP neural network, ozone inversion, spatio-temporal characteristic analysis, variation trend of ozone