High-resolution full-coverage ozone (O₃) estimates using a data-driven spatial random forest model in...

Area of Science:

• Environmental Science
• Atmospheric Chemistry
• Data Science

Background:

• The Beijing-Tianjin-Hebei (BTH) region faces severe ozone (O3) pollution, necessitating accurate concentration estimates for effective public health interventions.
• Existing high-resolution O3 estimation methods in BTH lack sufficient accuracy.
• Maximum daily 8-hour ozone concentration (MDA8O3) is a critical metric for assessing ozone pollution levels and health risks.

Purpose of the Study:

• To develop a highly accurate, full-coverage estimation model for MDA8O3 concentrations in the BTH region.
• To leverage both spatial homogeneity and heterogeneity of MDA8O3 data for improved estimation.
• To provide reliable data for informing environmental policies and targeted interventions against ozone pollution.

Main Methods:

• Incorporation of data-driven spatial weight matrices (DDWs) into a random forest (RF) model.
• Application of the DDW-RF model to estimate MDA8O3 concentrations at a 1 km x 1 km resolution.
• Validation using 10-fold cross-validation, yielding R² = 0.937 and RMSE = 13.919 μg/m³.

Main Results:

• The DDW-RF model achieved high accuracy in estimating MDA8O3 concentrations across the BTH region from 2014 to 2022.
• Spatial analysis revealed higher MDA8O3 concentrations in the southeast BTH region, particularly near the Bohai Rim, Shandong, and Henan.
• Temporal analysis indicated an initial increase and subsequent decrease in MDA8O3 from 2014-2021, with a slight rise in 2022.

Conclusions:

• The developed DDW-RF model provides accurate and comprehensive MDA8O3 estimates, crucial for identifying pollution hotspots.
• Findings highlight the need for increased attention and resources for areas adjacent to Bohai Rim, Shandong, and Henan.
• Recommendations include regulating factory operations, upgrading industrial practices, and mitigating ozone precursor formation to reduce O3 generation.