Regularized Extreme Learning Machine for Electrical Resistivity Tomography Inversion of Seawater Intrusion: A Sand Tank Experimental Verification

Seawater intrusion poses a significant threat to coastal aquifers and regional socioeconomic sustainable development. Accurate monitoring of this process remains challenging and requires interdisciplinary approaches to increase precision and reliability. This study adopts Electrical Resistivity Tomography (ERT) for monitoring seawater intrusion and establishes a geoelectrical inversion model based on the chaotic sparrow search algorithm (CSSA) optimized regularized extreme learning machine (RELM). The CSSA-RELM inversion algorithm was proposed to mitigate the uncertainty induced by random parameter initialization in RELM. The dataset comprises 402 homogeneous aquifer samples and 4382 heterogeneous aquifer samples, partitioned into an 80% training subset and a 20% independent test subset. The standalone RELM achieved its optimal performance with the hidden node number L = 75, regularization factor C = 2−3.5 and sigmoid activation function, yielding a mean training R2 = 0.8708 and a mean testing R2 = 0.8361. CSSA-RELM adopts this optimal set of RELM hyperparameters identified without further hyperparameter tuning. The CSSA is configured with a fixed population size of 80 and a maximum of 50 iterations. Under this setting, the CSSA-RELM model achieves a mean training R2 = 0.8895 and a mean testing R2 = 0.8514, showing better performance than standalone RELM. Sand tank experiments with ERT monitoring methods were conducted to observe the migration and characteristics of the seawater/freshwater interface during the evolution and restoration of seawater intrusion. Comparisons with both RELM and the least-squares method indicate that the CSSA-RELM delivers more satisfactory resistivity images and improved inversion accuracy, especially for complex subsurface configurations, relatively high measurement noise and sparse electrode layouts. The method provides a promising technical option for routine seawater/freshwater interface mapping and supports evidence-based management to safeguard the long-term sustainability of coastal groundwater resources.

Authors

Institutions

Publication Details

Journal
Sustainability
Published
2026-11-18
DOI
https://doi.org/10.3390/su182211756
Primary Topic
Geophysical and Geoelectrical Methods
Type
article
Field-Weighted Citation Impact
0.00

Funders

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

Regularized Extreme Learning Machine for Electrical Resistivity Tomography Inversion of Seawater Intrusion: A Sand Tank Experimental Verification

Shumei Zhu, Jia Xiong, Lusi Wei, Jun Ma et al.
2 links
Sustainability
Geophysical and Geoelectrical Methods
article

Regularized Extreme Learning Machine for Electrical Resistivity Tomography Inversion of Seawater Intrusion: A Sand Tank Experimental Verification

Shumei Zhu, Jia Xiong, Lusi Wei, Jun Ma, Zhifang Zhou, Qiang Zhang
article en

Abstract

Seawater intrusion poses a significant threat to coastal aquifers and regional socioeconomic sustainable development. Accurate monitoring of this process remains challenging and requires interdisciplinary approaches to increase precision and reliability. This study adopts Electrical Resistivity Tomography (ERT) for monitoring seawater intrusion and establishes a geoelectrical inversion model based on the chaotic sparrow search algorithm (CSSA) optimized regularized extreme learning machine (RELM). The CSSA-RELM inversion algorithm was proposed to mitigate the uncertainty induced by random parameter initialization in RELM. The dataset comprises 402 homogeneous aquifer samples and 4382 heterogeneous aquifer samples, partitioned into an 80% training subset and a 20% independent test subset. The standalone RELM achieved its optimal performance with the hidden node number L = 75, regularization factor C = 2−3.5 and sigmoid activation function, yielding a mean training R2 = 0.8708 and a mean testing R2 = 0.8361. CSSA-RELM adopts this optimal set of RELM hyperparameters identified without further hyperparameter tuning. The CSSA is configured with a fixed population size of 80 and a maximum of 50 iterations. Under this setting, the CSSA-RELM model achieves a mean training R2 = 0.8895 and a mean testing R2 = 0.8514, showing better performance than standalone RELM. Sand tank experiments with ERT monitoring methods were conducted to observe the migration and characteristics of the seawater/freshwater interface during the evolution and restoration of seawater intrusion. Comparisons with both RELM and the least-squares method indicate that the CSSA-RELM delivers more satisfactory resistivity images and improved inversion accuracy, especially for complex subsurface configurations, relatively high measurement noise and sparse electrode layouts. The method provides a promising technical option for routine seawater/freshwater interface mapping and supports evidence-based management to safeguard the long-term sustainability of coastal groundwater resources.

SustainabilityVol. 18(22)
Hohai University (CN), Pearl River Hydraulic Research Institute (CN), Ministry of Water Resources of the People's Republic of China (CN), Suzhou University (CN)
Science and Technology Planning Project of Guangdong Province
Clean water and sanitation
Geophysical and Geoelectrical Methods
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