An Integrated Inversion Framework for Efficient Identification of Groundwater Contamination Source with Deep Learning Tools

Abstract A sensitivity-driven multistage combined search system was established to enhance the refined search capability and accuracy of inverse modeling for the simultaneous identification of groundwater dense non-aqueous-phase liquids (DNAPLs) contamination sources and contaminant transport (CTR) parameters. This system was assisted by a convolutional neural network-bidirectional long short-term memory-attention (CNN-BiLSTM-A) model, which was designed for the highly reliable surrogate modeling of numerical simulations for DNAPL transport. The CNN-BiLSTM-A model was embedded into the combined inversion system, thereby reducing the computational burden of inverse iterations. The results revealed that CNN-BiLSTM-A successfully approximated the inner mapping of forward numerical simulation, with a mean relative error (MRE) of 2.3351% and a determination coefficient (R2) of 0.9965 for the prediction of the contaminant concentrations. The proposed sensitivity-driven whale optimization algorithm (SD-WOA) employed swarm intelligence to achieve a multi-chain parallel search process, while embedding sensitivity-based multistage feedforward (MF) mechanism to enhance the differentiated search capacity for source characteristics and CTR parameters. The MF-SD-WOA approach significantly reduces inversion uncertainty and mitigates equifinality by stepwisely constraining the search objective, spatial scope, and search trajectories, with the identification mean relative error of 2.55%, while the maximum relative error remaining lower than 6%. Using the MF-SD-WOA estimates to calibrate the numerical model are critical to the improvements in contaminant distribution characterization.

Authors

Institutions

Publication Details

Journal
Water Resources
Published
2026-11-18
DOI
https://doi.org/10.1134/s0097807826601160
Primary Topic
Groundwater flow and contamination studies
Type
article
Field-Weighted Citation Impact
0.00

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

An Integrated Inversion Framework for Efficient Identification of Groundwater Contamination Source with Deep Learning Tools

Tongzhe Liu, Jinxiao Wang, Lili Zhang, Yu Wang et al.
2 links
Water Resources
Groundwater flow and contamination studies
article

An Integrated Inversion Framework for Efficient Identification of Groundwater Contamination Source with Deep Learning Tools

Tongzhe Liu, Jinxiao Wang, Lili Zhang, Yu Wang, Dongning Hu, Zeyu Hou, Shuai Gao
article en

Abstract

Abstract A sensitivity-driven multistage combined search system was established to enhance the refined search capability and accuracy of inverse modeling for the simultaneous identification of groundwater dense non-aqueous-phase liquids (DNAPLs) contamination sources and contaminant transport (CTR) parameters. This system was assisted by a convolutional neural network-bidirectional long short-term memory-attention (CNN-BiLSTM-A) model, which was designed for the highly reliable surrogate modeling of numerical simulations for DNAPL transport. The CNN-BiLSTM-A model was embedded into the combined inversion system, thereby reducing the computational burden of inverse iterations. The results revealed that CNN-BiLSTM-A successfully approximated the inner mapping of forward numerical simulation, with a mean relative error (MRE) of 2.3351% and a determination coefficient (R2) of 0.9965 for the prediction of the contaminant concentrations. The proposed sensitivity-driven whale optimization algorithm (SD-WOA) employed swarm intelligence to achieve a multi-chain parallel search process, while embedding sensitivity-based multistage feedforward (MF) mechanism to enhance the differentiated search capacity for source characteristics and CTR parameters. The MF-SD-WOA approach significantly reduces inversion uncertainty and mitigates equifinality by stepwisely constraining the search objective, spatial scope, and search trajectories, with the identification mean relative error of 2.55%, while the maximum relative error remaining lower than 6%. Using the MF-SD-WOA estimates to calibrate the numerical model are critical to the improvements in contaminant distribution characterization.

Water ResourcesVol. 53(6)
Jilin University (CN), Jilin Jianzhu University (CN), Geological Exploration Institute of Shandong Zhengyuan (CN)
Natural Science Foundation of Jilin Province
Clean water and sanitation
Groundwater flow and contamination studies
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon