Developing a novel Temporal Air-quality Risk Index using LSTM autoencoder: A case study with South Korean air quality...

Area of Science:

• Environmental Science
• Health Risk Assessment
• Artificial Intelligence

Background:

• Environmental indices simplify complex pollution data but often have limitations.
• Existing indices like AQHI may ignore cumulative pollutant effects and temporal dynamics.
• Accurate environmental risk information is crucial for public health decision-making.

Purpose of the Study:

• To develop a novel deep learning framework for a more comprehensive air quality index.
• To address limitations of conventional indices by incorporating temporal dependencies and non-linear risks.
• To introduce the Temporal Air-quality Risk Index (TARI) for improved health risk assessment.

Main Methods:

• Utilized a long short-term memory (LSTM) autoencoder for capturing complex interactions among environmental factors.
• Developed a risk score (RS) to account for non-linear and continuous environmental risks.
• Applied a deep learning framework to analyze air quality data and assess health impacts.

Main Results:

• The proposed Temporal Air-quality Risk Index (TARI) demonstrated superior performance compared to existing indices (CAI, AQHI).
• TARI showed stronger correlations with disease prevalence, indicating improved sensitivity to health risks.
• The deep learning approach effectively captured cumulative and temporal effects of air pollutants.

Conclusions:

• TARI provides a more accurate and sensitive assessment of air quality health risks by considering pollutant interactions and temporal dynamics.
• Deep learning offers a flexible and robust framework for developing advanced environmental indices.
• This novel approach has potential applications for various environmental monitoring and health assessment systems.