An hourly and localized optimization method for soil fugitive dust emission inventory based on machine learning.

Area of Science:

• Environmental Science
• Atmospheric Chemistry
• Geosciences

Background:

• Soil fugitive dust (SFD) significantly impacts urban air quality and ecosystems in northern China.
• SFD exhibits high spatial-temporal variability and diverse emission sources.
• Current understanding of SFD's critical impact factors and quantitative methods is limited.

Purpose of the Study:

• To identify key factors influencing SFD emissions using interpretable machine learning.
• To determine the interaction and action thresholds of these impact factors.
• To enhance quantitative methodologies for SFD emission calculation.

Main Methods:

• Application of interpretable machine learning techniques to analyze SFD data.
• Identification of principal impact factors and their interactions.
• Optimization of the Wind Erosion Equation model with localized parameters.

Main Results:

• Seasonal variations in SFD impact factors were identified.
• Bare soil source strength, including area and moisture, was found to be a substantial factor.
• The optimized Wind Erosion Equation model demonstrated improved accuracy for hourly SFD emissions.

Conclusions:

• The study provides a refined understanding of SFD emission drivers.
• Optimized parameterization schemes enhance the accuracy of SFD modeling.
• Findings support the development of targeted SFD prevention and control policies.