Performance comparison of LUR and OK in PM2.5 concentration mapping: a multidimensional perspective

Area of Science:

• Environmental Science
• Geospatial Analysis
• Air Quality Monitoring

Background:

• Sparse PM2.5 monitoring sites limit traditional air quality mapping accuracy.
• Existing point-based evaluation methods for Land Use Regression (LUR) and Ordinary Kriging (OK) are insufficient.
• A multidimensional approach is needed to assess PM2.5 mapping model performance.

Purpose of the Study:

• To evaluate Land Use Regression (LUR) and Ordinary Kriging (OK) for PM2.5 concentration mapping in Houston.
• To introduce and apply information entropy, an area-based statistic, for model performance evaluation.
• To compare the effectiveness of LUR and OK using both point-based and area-based metrics.

Main Methods:

• Utilized Land Use Regression (LUR) modeling and Ordinary Kriging (OK) interpolation for PM2.5 mapping.
• Employed traditional point-based statistics (error rate, RMSE) for validation.
• Integrated information entropy, an area-based statistic, for multidimensional performance assessment.

Main Results:

• Point-based validation showed subtle differences in accuracy between LUR (error rate 6.13%) and OK (7.01%).
• Area-based validation using information entropy revealed LUR (7.79) captured more spatial variation than OK (3.63).
• LUR modeling demonstrated superior refinement of PM2.5 spatial distribution compared to OK interpolation.

Conclusions:

• Land Use Regression (LUR) modeling is more effective than Ordinary Kriging (OK) for detailed PM2.5 spatial distribution mapping.
• Integrating point- and area-based statistics enhances the multidimensional evaluation of air pollution models.
• This study advocates for a comprehensive statistical approach in air quality mapping research.