Modelling Particulate Matter (PM₁₀) Variations During Transboundary Haze Events Using a Modified Quantile Regression...

Area of Science:

• Environmental Science
• Atmospheric Chemistry
• Data Science

Background:

• Transboundary haze in Southeast Asia causes significant environmental and health issues.
• Accurate forecasting of particulate matter (PM10) is crucial for managing haze events.

Purpose of the Study:

• To develop and evaluate novel air quality forecasting models for PM10 concentrations during transboundary haze in Malaysia.
• To enhance predictive accuracy by integrating advanced feature selection techniques with quantile regression.

Main Methods:

• Development of hybrid models: Quantile Regression with Relief-based ranking (QR-Relief), correlation-based selection (QR-correlation), and Principal Component Analysis (QR-PCA).
• Evaluation of model performance using metrics like Mean Absolute Error (MAE), Normalized Absolute Error (NAE), and Root Mean Square Error (RMSE).
• Validation using an independent dataset from 2019 to confirm real-world applicability.

Main Results:

• The hybrid models (QR-Relief, QR-correlation, QR-PCA) significantly outperformed traditional Quantile Regression and Multiple Linear Regression models.
• Feature selection techniques demonstrably improved the predictive reliability of the air quality models.
• Validated models showed strong performance, confirming their utility for real-world haze event prediction.

Conclusions:

• Advanced feature selection integrated with quantile regression offers a robust framework for accurate PM10 forecasting.
• The developed models provide valuable decision-support tools for environmental and public health management during haze events.
• This methodological advancement supports the creation of effective early warning systems for transboundary haze.