Daily concentrations of air pollutants greatly affect air quality as an increase in industry and urbanization deteriorate the environment. For forty cities in Turkey, eleven variables are recorded to investigate the determinants of air pollution presumably by regressing the air pollutants PM10, NOx, and NO2. The temperature, wind, human factors such as population, vehicles, manufacturer, and suchlike are used to create a nonlinear air quality model in Turkey due to the multivariate nature of the data. A comparison of the nonparametric models of the concentration of these pollutants, using multivariate adaptive regression splines (MARS), was obtained to estimate the dependence between air pollutants and various factors. Finally, a model for PM10 concentration shows that climate effects are the most significant variables, whereas the predicted models for NOx and NO2 indicate that human factors, such as the number of manufacturers and the number of vehicles, are significant variables. In conclusion, the predicted models are easy to interpret and have advantages of capacity to produce the contributions of the factors for each pollutant model. It is advisable for researchers to examine and determine the suitability of their data sets using nonlinear models when atypical observations and high correlations exist in the data.