Machine learning-based models are effective and practical tools for analyzing complex relationships between air quality parameters and meteorological variables and identifying key factors influencing air quality. The aim of this study is evaluation of three tree-based models’ performance namely —Random Forest (RF), XGBoost, and CatBoost—and a multilayer perceptron (MLP) neural network model, using daily data of five meteorological varibles, including wind speed, temperature, air pressure, relative humidity, and rainfall and two air pollutants (O3 and PM2.5), to estimate the daily concentration of carbon monoxide during winter season in Kerman city, south of Iran.The models’ performance was assessed using statistical indices including the coefficient of determination (R²), root mean square error (RMSE), and mean absolute error (MAE). Results indicated that the CatBoost model had the highest accuracy in estimating CO concentrations, with a R² of 0.778, RMSE of 0.284 (ppb), and MAE of 0.209 (ppb) during the test phase. The performance of the XGBoost and RF models was relatively similar, with R² values of 0.747 and 0.728, respectively. The MLP model showed the lowest accuracy, with R², RMSE and MAE of 0.693, 0.308 and 0.236 pbb, respectively. These results confirm the superior skill of tree-based models in comparison to the neural network-based model for estimating CO concentration in study region.