Quantifying source contributions to ambient NH₃ using Geo-AI with time lag and parcel tracking functions

Area of Science:

• Environmental Science
• Artificial Intelligence
• Geospatial Analysis

Background:

• Ambient ammonia (NH3) is a key precursor to particulate matter (PM) formation, necessitating accurate monitoring and source identification for effective pollution control.
• Limited real-time NH3 data from air quality stations hinder comprehensive understanding and reduction efforts.
• Existing methods for NH3 prediction and source apportionment face challenges in scope and real-time applicability.

Purpose of the Study:

• To develop and validate a novel geospatial-artificial intelligence (Geo-AI) model for predicting ambient ammonia (NH3) concentrations.
• To identify and quantify the contribution of various sources to NH3 concentrations using advanced analytical techniques.
• To provide an adaptable methodology for NH3 prediction and source analysis in regions with sparse monitoring data.

Main Methods:

• Development of a novel Geo-AI base model integrating machine learning algorithms and geographic predictor variables.
• Application of parcel tracking functions within the Geo-AI framework for detailed spatial analysis.
• Utilizing Shapley Additive Explanation (SHAP) with domain knowledge for source contribution analysis.

Main Results:

• The Geo-AI model achieved high accuracy in predicting hourly average NH3 values from 2016 to 2018, explaining up to 96% of the total variance.
• SHAP analysis identified water bodies, traffic, and agricultural emissions as the most significant factors influencing NH3 concentrations in Taichung.
• The methodology demonstrated adaptability for regions with limited NH3 monitoring infrastructure.

Conclusions:

• The developed Geo-AI model offers a pioneering and effective approach for NH3 prediction and source apportionment.
• Accurate NH3 data and source identification are critical for developing targeted policies to reduce particulate matter.
• This adaptable Geo-AI methodology serves as a vital tool for future environmental policy and regulation development.