Predictive modeling of air quality in the Tehran megacity via deep learning techniques

Area of Science:

• Environmental Science
• Data Science
• Atmospheric Chemistry

Background:

• Air pollution poses significant risks to public health and environmental safety in urban areas.
• Accurate forecasting of air pollutant concentrations is crucial for effective environmental management and public health protection.

Purpose of the Study:

• To forecast concentrations of key air pollutants (CO, O3, NO2, SO2, PM10, and PM2.5) in Tehran from 2013-2023.
• To evaluate the effectiveness of deep learning (DL) models against conventional machine learning (ML) methods for air pollution forecasting.
• To identify key meteorological driving variables influencing pollutant concentrations.

Main Methods:

• Utilized deep learning models, including Gated Recurrent Units (GRUs), Fully Connected Neural Networks (FCNNs), and Convolutional Neural Networks (CNNs).
• Compared DL model performance against conventional machine learning methods using R-squared (R2), RMSE, MAE, and MSE metrics.
• Analyzed the impact of meteorological variables such as temperature, humidity, dew point, wind speed, and air pressure on pollutant concentrations.

Main Results:

• Deep learning models demonstrated superior performance compared to traditional ML methods in air pollution forecasting.
• FCNN and GRU models achieved high accuracy for predicting NO2, PM10, and PM2.5, with R2 values up to 0.9276.
• Temperature and humidity were identified as the primary meteorological factors influencing the variability of pollutant concentrations.

Conclusions:

• Deep learning models provide significant accuracy for air pollution forecasting, serving as vital tools for environmental management.
• The findings offer practical insights for policymakers to develop and implement efficient air quality control strategies.
• Accurate forecasting enables proactive measures to mitigate the adverse effects of air pollution on public health and the environment.