Prediction of elemental carbon and organic carbon using a hybrid deep learning model integrated temporal dependencies and meteorological features

Elemental Carbon (EC) and Organic Carbon (OC) are critical components of PM_2.5, with significant implications for air quality and public health. Traditional prediction models often fail to capture the non-linear dynamics of EC and OC, particularly in urban environments with high traffic and industrial emissions. This study addresses this gap by proposing a novel hybrid deep learning model that integrates Bidirectional Long Short-Term Memory (BiLSTM) networks with attention mechanisms to account for temporal dependencies, meteorological factors, and co-pollutants (PM_2.5, O₃). Using comprehensive air quality and meteorological data from Changshu City, we identify distinct diurnal and seasonal patterns of EC and OC, driven by traffic emissions, weather conditions, and secondary aerosol formation. The proposed model significantly outperforms traditional methods, achieving high prediction accuracy for both EC and OC concentrations. Key innovations include the integration of attention mechanisms to prioritize critical time steps and the incorporation of meteorological features and co-pollutants, which enhance the model's ability to capture complex pollutant interactions. The results demonstrate the model's robustness in real-time air quality forecasting, providing actionable insights for urban planning and pollution mitigation strategies. This research contributes to the field by offering a scalable and accurate tool for predicting EC and OC in dynamic urban environments, ultimately supporting efforts to improve public health and sustainable urban development.