Most metropolitan areas experience increasing health and environmental pressure from air pollutants. Air pollutant concentrations are mostly influenced by meteorological factors, pollutant sources, and topography. The vertical atmospheric conditions can affect the surface air pollution concentration. In this paper, to evaluate the relationship between the vertical meteorological characteristics and air pollution concentrations (CO, O3, PM2.5, PM10 and SO2) four different models including multi-linear regression (MLR), multi-linear perceptron (MLP), generalized regression neural network (GRNN), and adaptive Neuro-Fuzzy inference system (ANFIS) were developed. The vertical temperature difference, relative humidity, U and V wind components in different altitudes were extracted from the operational archive of the European Center for Medium Range Weather Forecast (ECMWF 2019) from March 2011 to December 2017. The variance inflation factor (VIF) and tolerance tests were utilized to remove the independent variables, which made multi-collinearity. Results showed that among the studied air pollutant variables, variation of PM10 concentration was more influenced by vertical meteorological characteristics obtained using the GRNN model with R2 = 0.960, root Mean square error (RMSE) = 0.019 and mean absolute error (MAE) = 0.005 in tearing phase and R2 = 0.940, RMSE = 0.054 and MAE = 0.019 in testing phase. The results of ANFIS model for PM2.5 concentration were acquired highest among the other air pollutant variables in tearing phase (R2 = 0.819, RMSE = 0.036 and MAE = 0.025) and in testing phase (R2 = 0.960, RMSE = 0.001 and MAE = 0.001). The results illustrated that GRNN and ANFIS models outperform other models for air pollutants prediction using vertical atmospheric conditions in the Sanandaj, Iran.