This paper investigates the relationship between air pollutant concentrations and seven meteorological parameters (wind speed, stable energy, Froude number, atmospheric boundary layer height, potential temperature lapse rate, transport index and gradient Richardson number) based on the near ground air pollutant concentrations and the high temporal and spatial resolution meteorological data from the Weather Research and Forecast (WRF) model and develops a regression model for predicting ground level daily mean PM10 (particulate matter with aerodynamic diameter less than 10 μm) and NO2 (nitrogen dioxide) concentrations for urban Lanzhou. The results show that in urban Lanzhou, pollutant concentrations correlate better with atmospheric boundary layer height and potential temperature lapse rate, and the correlations between NO2 and meteorological parameters are better than that between PM10 and meteorological parameters. The developed regression model performs better for NO2 than for PM10. The fitting degree of the developed regression model is higher in urban area than that in rural area, leading to the better performance in urban area. The overall performance of the regression model is as good as widely used comprehensive air quality models. The method provides a scientific basis for urban air quality forecasting and air pollution study.