[Application of Support Vector Machine Regression in Ozone Forecasting].

Area of Science:

• Environmental Science
• Atmospheric Chemistry
• Data Science

Context:

• Air quality monitoring in industrial areas is crucial.
• High ozone (O3) concentrations pose significant environmental and health risks.
• Accurate forecasting of O3 levels is essential for mitigation strategies.

Purpose:

• To develop and evaluate a Support Vector Machine Regression (SVMr) model for forecasting hourly, daily maximum, and 8-hour moving average ozone concentrations.
• To identify key meteorological variables and ozone precursors influencing ozone formation.
• To compare the performance of SVMr against multiple linear regression methods.

Summary:

• SVMr was employed to predict ozone concentrations using meteorological data and precursor observations during high ozone periods in Nanjing.
• The model achieved a high squared correlation coefficient (R2) of 0.84 for hourly ozone forecasts.
• Key predictors included ozone pre-concentrations, Ultraviolet Radiation B (UVB), and nitrogen dioxide (NO2) concentrations. Including precursors improved accuracy by 10%-28%.

Impact:

• SVMr demonstrated superior performance compared to multiple linear regression for ozone concentration forecasting.
• The study highlights the importance of incorporating precursor data for enhanced air quality prediction.
• Findings can inform air pollution control policies and public health advisories in industrial regions.