Identifying interactive effects of spatial drivers in soil heavy metal pollutants using interpretable machine...

Area of Science:

• Environmental Science
• Geochemistry
• Data Science

Background:

• Accurate identification of spatial drivers is crucial for effective soil heavy metal (SHM) management.
• Understanding the complex interactions of these drivers remains a significant challenge.

Purpose of the Study:

• To develop a comprehensive framework for identifying and elucidating the interactive effects of spatial drivers on SHM.
• To analyze SHM pollution (Hg, Cd, Zn) in the Pearl River Delta (PRD) using integrated analytical techniques.

Main Methods:

• Integration of Geodetector, CatBoost, and SHapley Additive exPlanations (SHAP) techniques.
• Analysis of fourteen environmental factors influencing SHM distribution.
• Focus on interactive effects of spatial drivers on heavy metal pollution.

Main Results:

• SHM distribution is influenced by combined individual factors and interactions within the source-flow-sink process.
• Interactions between Hg and Cd, and interactions involving Zn, showed enhanced spatial explanatory power.
• Spatial heterogeneity significantly impacts environmental factor contributions to SHM concentrations, with fluctuating effects when interactions are considered.

Conclusions:

• The SHAP method provides understandable insights into heavy metal pollution, enabling comparison of factor importance and directional impacts.
• Interactive factors play a crucial role in SHM pollution, necessitating consideration in management strategies.
• Findings offer valuable insights into spatial heterogeneity of SHM pollution for developing accurate management strategies.