Abstract: The analysis of temporal and spatial evolution characteristics of PM2.5 concentration is helpful to recognize the development and status of atmospheric pollution. However, the accumulation time of PM2.5 concentration monitoring is short PM2.5 and concentration is affected by emission intensity and meteorological factors. Therefore, it is necessary to study the PM2.5 concentration inversion model based on other existing data. PM2.5 concentration was affected by emission intensity and meteorological factors. Taking Hebei Province as an example, the PM2.5 concentration model is built by integrating Global Navigation Satellite System (GNSS) precipitable water vapor (PWV), wind speed and air pollutants.Firstly, the correlation analysis of PM2.5 concentration with air pollutants, GNSS PWV and wind speed is carried out. Then air pollutants, GNSS PWV and wind speed are used as input, and PM2.5 concentration is used as output, and urban PM2.5 concentration model and regional PM2.5 concentration model are constructed by back propagation (BP) neural network. Finally, the reliability of PM2.5 concentration model is carried out. The results show that the prediction accuracy of PM2.5 concentration level is high and the relative error is low compared with the measured value of PM2.5 concentration. The PM2.5 concentration model, which combines GNSS PWV, wind speed and air pollutants, can be used to analyze the temporal and spatial evolution characteristics of regional PM2.5 concentration. It can be used for reference for government air pollution control, and also can be used to monitor the concentration of PM2.5.