Driving factors and predictive model of dissolved N₂O concentrations in a complex aquatic network

Area of Science:

• Environmental Science
• Ecology
• Biogeochemistry

Background:

• Inland water networks are significant sources of nitrous oxide (N₂O), a potent greenhouse gas.
• The complexity of these systems hinders accurate N₂O emission quantification.

Purpose of the Study:

• To develop an integrated framework for causal inference and nonlinear predictive modeling of N₂O emissions.
• To identify key drivers of N₂O variability in China's Taihu Basin.

Main Methods:

• Combined structural equation modeling (SEM), machine learning (ML), and SHapley Additive exPlanations (SHAP).
• Utilized a framework to bridge causal inference with nonlinear predictive modeling.

Main Results:

• Nitrate-nitrogen (NO₃⁻-N) and water temperature (WT) were dominant factors influencing N₂O variability.
• Dissolved organic carbon (DOC) plays a dual role: a macroscopic sink and a microscopic catalyst in N₂O production.
• A parsimonious model with four routine parameters achieved satisfactory predictive capability (test R² = 0.54) for regional assessment.

Conclusions:

• The developed framework offers a scalable and transferable methodology for understanding complex aquatic ecosystems.
• Provides a robust tool for basin-scale N₂O estimation and targeted greenhouse gas mitigation strategies.