Publications

Ebrahimi-Khusfi, Z; Taghizadeh-Mehrjardi, R; Kazemi, M; Nafarzadegan, AR (2021). Predicting the ground-level pollutants concentrations and identifying the influencing factors using machine learning, wavelet transformation, and remote sensing techniques. ATMOSPHERIC POLLUTION RESEARCH, 12(5), 101064.

Abstract
This study was conducted to evaluate the performance of the support vector regression (SVR) model with and without applying wavelet transformation for predicting the PM10, PM2.5, SO2, NO2, CO, and O3 in Isfahan metropolis, central Iran. Ground-based data, TerraClimate, and MODIS products were used to predict air pollution parameters. These factors were first trained using the SVR and Wavelet-SVR models, and their performance was then compared using the error evaluation statistics. Uncertainties were evaluated using local errors and the clustering method. The influencing factors were lastly determined using the permutation features importance method. The results indicated that the Wavelet-SVR model resulted in improving the performance of prediction compared to the SVR model. The mean prediction interval values were also decreased after applying the wavelet transformation on the SVR model. It was found that the dominant agents affecting the temporal changes of study pollutants are soil moisture and meteorological drought. Urban development and increased energy consumption were observed in the areas with the highest air pollution. Researchers and stakeholders can use these findings to assess air pollution hazards and to improve air quality and human living conditions in metropolises.

DOI:
10.1016/j.apr.2021.101064

ISSN:
1309-1042