A Novel Framework for Airshed Delineation and PM_2.5 Estimation across India Using Machine Learning and Spatial...

Area of Science:

• Environmental Science
• Atmospheric Chemistry
• Geospatial Analysis

Background:

• Air pollution, particularly fine particulate matter (PM2.5), presents a significant public health concern in India.
• Existing air quality analyses often overlook the complex spatial distribution of PM2.5 due to climatic, topographic, and anthropogenic influences, especially when confined to administrative boundaries.

Purpose of the Study:

• To develop an innovative spatial airshed delineation framework for enhanced air quality management in India.
• To improve the accuracy and localization of PM2.5 concentration modeling by integrating airshed characteristics.

Main Methods:

• Utilized clustering algorithms to integrate PM2.5 concentrations, meteorological data, and land characteristics for spatial airshed delineation.
• Developed a national-level machine learning model (Random Forest) using MERRA-2 reanalysis and ground-based data to estimate PM2.5.
• Incorporated the developed airsheds into the machine learning model to refine predictive performance.

Main Results:

• Identified seven major and five transitional airsheds across India, demonstrating multi-year consistency for standardized management.
• The integration of airshed-based clustering significantly enhanced the PM2.5 predictive model, increasing R² from 0.71 to 0.80 and reducing RMSE from 27.58 to 23.25 μg/m³.
• The framework facilitated the identification of dominant pollution sources within distinct regional airsheds.

Conclusions:

• The study provides a robust, data-driven framework for spatial airshed delineation and region-specific PM2.5 modeling.
• The developed airshed approach supports more accurate, actionable, and localized air quality management strategies in India.
• This methodology addresses limitations of analyses bound by administrative borders, offering a more holistic view of air pollution dynamics.