A novel bagging ensemble approach for predicting summertime ground-level ozone concentration

Area of Science:

• Environmental Science
• Atmospheric Chemistry
• Data Science

Background:

• Ground-level ozone (O3) pollution poses significant risks to human health and vegetation.
• Ozone formation is a complex photochemical process influenced by multiple interrelated factors.
• Accurate prediction of O3 concentrations is vital for air quality management.

Purpose of the Study:

• To develop and evaluate an ensemble bagging approach for modeling summer-time ground-level O3.
• To assess the feasibility of using meteorological parameters with ensemble models for O3 prediction.
• To compare the performance of base classifiers against ensemble methods in predicting O3 peaks.

Main Methods:

• Utilized an ensemble bagging approach with seven meteorological parameters as input.
• Employed Multilayer Perceptron, RTree, REPTree, and Random Forest as base learners.
• Validated model performance using error measures (IoAd, R2, PEP) and an independent test dataset.

Main Results:

• The bagged random forest model achieved a superior Nash-Sutcliffe coefficient of 0.93.
• Ensemble models demonstrated effectiveness in predicting ground-level O3 concentrations.
• The study highlighted the importance of predicting peak O3 concentrations for health and environmental impact.

Conclusions:

• Ensemble bagging, particularly with random forest, is a highly effective method for predicting ground-level O3.
• The developed models address a research gap in big data analysis for air pollutant prediction.
• Accurate O3 prediction models are essential for safeguarding public health and environmental quality.