Auto-Modal: Air-Quality Index Forecasting with Modal Decomposition Attention

Area of Science:

• Environmental Science
• Computer Science
• Atmospheric Science

Background:

• The air-quality index (AQI) is crucial for assessing air pollution levels.
• Accurate AQI prediction is difficult using traditional models like WRF-CMAQ due to uncertainties in meteorological data and emission inventories.

Purpose of the Study:

• To propose a novel Auto-Modal network with Attention Mechanism (AMAM) for enhanced hourly AQI prediction.
• To improve the accuracy and generalization ability of AQI forecasting models.

Main Methods:

• Developed a dual-input path Auto-Modal network with Attention Mechanism (AMAM).
• The first path utilizes bidirectional encoder representations from a transformer with historical meteorological and pollutant data.
• The second path serves as a baseline, incorporating predictions from the WRF-CMAQ model.

Main Results:

• The AMAM model demonstrated superior performance across all prediction lengths compared to existing state-of-the-art models.
• Experimental results validate the effectiveness of the proposed dual-input path architecture.
• The attention mechanism enhances the model's ability to capture complex relationships in the data.

Conclusions:

• The proposed AMAM offers a significant advancement in hourly AQI prediction accuracy.
• This novel approach effectively addresses the limitations of traditional AQI forecasting models.
• AMAM shows great potential for real-time air quality management and public health advisories.