Mining Public Datasets for Modeling Intra-City PM_2.5 Concentrations at a Fine Spatial Resolution

Area of Science:

• Environmental Science
• Data Science
• Public Health

Background:

• Air quality models are crucial for assessing health impacts of pollutants at fine spatiotemporal scales.
• Current models often require expert-selected, area-specific data for emissions and dispersion, limiting scalability.
• This necessitates more automated and adaptable approaches for building accurate air quality models.

Purpose of the Study:

• To develop and validate a data mining approach for automatically generating air quality models.
• To utilize publicly available OpenStreetMap (OSM) data for predicting fine particulate matter (PM2.5) concentrations.
• To quantify the influence of diverse geographic features on air quality.

Main Methods:

• A data mining approach was employed, leveraging OpenStreetMap data.
• The model automatically generates air quality predictions for PM2.5 concentrations.
• The approach quantifies the impact of various geographic features (e.g., commercial buildings, parking lots) on air quality.

Main Results:

• The expert-free model achieved accurate PM2.5 concentration predictions.
• The approach successfully incorporated diverse geographic features representing pollution sources.
• Quantified the impact of geographic features on air quality from various distances.

Conclusions:

• The developed data mining approach offers an accurate, expert-free method for air quality modeling.
• This technique enhances traditional models by automatically integrating geographic data from OSM.
• Enables scalable, context-specific spatiotemporal air pollution models for health impact studies, especially for sensitive populations.