Related Experiment Video
Updated: Feb 20, 2026
03:31
Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications
Published on: December 15, 2023
1.1K
Knowledge-Guided Deep Learning Framework for Source Apportionment and Mitigation of Perchlorate Contamination in
Kunting Xie1, Lingjun Bu1, Jiaxin Xiao1
1Hunan Engineering Research Center of Water Security Technology and Application, College of Civil Engineering, Hunan University, Changsha 410082, China.
Environmental Science & Technology
|February 18, 2026
Summary
Fireworks production causes severe perchlorate contamination. A new framework integrating hydrology and deep learning effectively predicts and mitigates this industrial pollution in river basins.
Area of Science:
- Environmental Science
- Hydrology
- Data Science
Background:
- Fireworks manufacturing is a major source of perchlorate contamination in aquatic environments.
- Existing methods fail to address complex industrial pollution dynamics and fragmented governance in river basins, hindering effective mitigation.
Purpose of the Study:
- To investigate perchlorate contamination patterns in China's fireworks production regions.
- To develop and validate a novel knowledge-guided framework for predicting perchlorate concentration dynamics at a basin scale.
- To differentiate and quantify contributions from point and nonpoint sources of perchlorate pollution.
Main Methods:
- Developed a knowledge-guided framework integrating process-based hydrology with deep learning (Mixture-of-Experts).
- Analyzed spatiotemporal perchlorate contamination patterns in key Chinese fireworks production zones.
- Utilized attention analysis to identify key drivers of nonpoint source pollution.
- Performed multiscenario simulations to evaluate mitigation strategies.
Main Results:
- Identified a significant perchlorate "accumulation corridor" linked to industrial zones.
- The integrated model accurately predicts long-term trends and episodic peaks, outperforming traditional data-driven methods.
- Point sources account for approximately 65% of total loads in the corridor.
- Nonpoint source pollution is amplified during the rainy season (1.7–2.8× higher fluxes than dry seasons).
- Fireworks factory density, precipitation, and soil properties are key drivers for nonpoint perchlorate sources.
Conclusions:
- The developed framework offers a transferable approach for basin-scale diagnosis of industrial contaminant risks.
- Coordinated control strategies targeting point, nonpoint, and upstream inputs can significantly reduce perchlorate exceedance.
- The study provides a decision-support tool for managing industrial contaminant risks, adaptable to local conditions and regulatory frameworks.
