Modeling of atmospheric particulate matters via artificial intelligence methods

Abstract

Nowadays, pollutants continue to be released into the atmosphere in increasing amounts with each passing day. Some of them may turn into more harmful forms by accumulating in different layers of the atmosphere at different times and can be transported to other regions with atmospheric events. Particulate matter (PM) is one of the most important air pollutants in the atmosphere, and it can be released into the atmosphere by natural and anthropogenic processes or can be formed in the atmosphere as a result of chemical reactions. In this study, it was aimed to predict PM10 and PM2.5 components measured in an industrial zone selected by adaptive neuro-fuzzy inference system (ANFIS), support vector regression (SVR), classification and regression trees (CART), random forest (RF), k-nearest neighbor (KNN), and extreme learning machine (ELM) methods. To this end, in the first stage of the study, the dataset consisting of air pollutants and meteorological data was created, the temporal and qualitative evaluation of these data was performed, and the PM (PM10 and PM2.5) components were modeled using the R software environment by artificial intelligence methods. The ANFIS model was more successful in predicting the PM10 (R-2 = 0.95, RMSE = 5.87, MAE = 4.75) and PM2.5 (R-2 = 0.97, RMSE = 3.05, MAE = 2.18) values in comparison with other methods. As a result of the study, it was clearly observed that the ANFIS model could be used in the prediction of air pollutants.