Forecasting O₃ and NO₂ concentrations with spatiotemporally continuous coverage in southeastern China using a Machine...

Area of Science:

• Atmospheric Chemistry and Air Quality Monitoring
• Environmental Science and Public Health

Background:

• Ozone (O3) and nitrogen dioxide (NO2) are key air pollutants contributing to smog, with adverse effects on health, ecosystems, and agriculture.
• Accurate spatiotemporal forecasting of O3 and NO2 is crucial for effective mitigation and public health protection.
• Current forecasting methods like chemical transport models (CTMs) and time series analysis have limitations in accuracy and continuous coverage.

Purpose of the Study:

• To develop a novel forecasting model for spatiotemporally continuous O3 and NO2 concentrations.
• To improve upon existing forecasting methods by integrating the random forest algorithm with NASA's GEOS-CF product.
• To provide reliable, near-real-time forecasts up to five days in advance for southeastern China.

Main Methods:

• Integration of the random forest algorithm with NASA's Goddard Earth Observing System "Composing Forecasting" (GEOS-CF) product.
• Development of a forecasting framework designed for spatiotemporally continuous air pollutant predictions.
• Validation using both overall and spatial cross-validation techniques.

Main Results:

• The integrated model significantly outperformed the baseline GEOS-CF model across all validation metrics.
• Substantial reduction in forecast errors compared to the initial GEOS-CF data.
• Demonstrated capability for accurate, near-real-time O3 and NO2 forecasts with continuous spatiotemporal coverage.

Conclusions:

• The developed random forest and GEOS-CF integrated model offers a superior approach for forecasting O3 and NO2.
• This method provides accurate, continuous spatiotemporal air quality forecasts, addressing limitations of existing techniques.
• The findings support enhanced decision-making for air pollution mitigation and public health advisories.