Nonlinear pollution source switching under extreme rainfall revealed by budget-closed graph-based apportionment

Area of Science:

• Environmental Science
• Hydrology
• Water Resource Management

Background:

• Climate change is increasing extreme precipitation events.
• Existing watershed management models struggle to account for sub-weekly pollutant pulses.
• Pollutant transport is significantly restructured by episodic high-flow events.

Purpose of the Study:

• To develop and validate a novel framework for accurate source apportionment of watershed pollutant loads.
• To investigate the impact of extreme precipitation on the balance between point-source (PS) and nonpoint-source (NPS) contributions.
• To establish a data-driven approach for real-time, event-triggered watershed management.

Main Methods:

• Developed the Budget-Closed Graph-based Source Apportionment (BC-GSA) framework, integrating a graph attention network with mass-balance constraints for exact load closure.
• Computed nonpoint-source (NPS) loads as the residual of observed flux minus upstream transport and point-source (PS) inventory.
• Applied BC-GSA to the Yiluo River Basin, coupling hydrological analysis with source apportionment across 15 storm events.

Main Results:

• BC-GSA eliminated typical 5-15% budget leakage by enforcing exact mass balance.
• Extreme rainfall in July 2021 caused a rapid shift in pollutant contributions, with NPS TP rising from 41.5% to 89.2% and TN from 57.5% to 86.5% within 48 hours.
• A strong correlation (r = 0.946, p < 0.001) was found between hydrological amplification and NPS surge magnitude, with a physical saturation ceiling at 93-99% NPS.

Conclusions:

• The BC-GSA framework provides a physically credible and robust method for estimating NPS loads, validated through multiple independent lines of evidence.
• Findings support a transition from static Total Maximum Daily Loads to adaptive, event-triggered control protocols based on real-time hydrological thresholds.
• Accurate, real-time source apportionment is crucial for effective watershed management under intensifying climate change impacts.