Prediction of air pollutants for air quality using deep learning methods in a metropolitan city

Abstract

Air quality forecasting is very difficult in metropolitan areas due to emissions, high population density, and uncertainty in defining meteorological areas. The use of incomplete information during the training phase and the poor model selection to be used restrict the air quality estimation. In this study, predictions of PM10 and SO2 air pollutants in 2022 were made by using Long Short-Term Memory Networks (LSTM), Recurrent Neural Network (RNN), and Multilayer Perceptron (MLP) by revising the error term of traditional methods and completing the missing data. The data of Basaksehir district of Istanbul province, where industrialization and population are very concentrated, were obtained from the national air quality monitoring network. When PM10 and SO2 estimation results obtained are compared with the real values, 15.15 real data belonging to PM10, is estimated as 15.11 in LSTM. Likewise, 4.65 real data belonging to SO2, is estimated as 5.18 in LSTM. As a result of the application, LSTM predicts PM10 and SO2 better than the MLP and RNN models. The results were compared with other studies in the literature, and the proposed LSTM deep learning architecture performed well compared to studies using data sets and location information under similar conditions.