Data-Driven Resistivity Zonation: Integrating Inversion, Kriging, and Clustering to Unravel Subsurface Complexity for Groundwater Exploration

Data-Driven Resistivity Zonation: Integrating Inversion, Kriging, and Clustering to Unravel Subsurface Complexity for Groundwater Exploration | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Data-Driven Resistivity Zonation: Integrating Inversion, Kriging, and Clustering to Unravel Subsurface Complexity for Groundwater Exploration Gumilar Utamas Nugraha, Hendra Bakti, Rachmat Fajar Lubis, Asep Mulyono, and 2 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6174105/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 01 Aug, 2025 Read the published version in Discover Water → Version 1 posted 18 You are reading this latest preprint version Abstract This study integrates geoelectrical resistivity inversion, geostatistical analysis, and clustering techniques to improve subsurface characterization in Cikole, Lembang, West Bandung Regency. The primary objective is to enhance resistivity-based zonation for groundwater exploration by employing a dipole-dipole array with 10-meter electrode spacing, followed by 2D inversion to visualize resistivity distributions. Geostatistical interpolation using Kriging effectively captures spatial variability, while clustering methods (K-Means, DBSCAN, and Hierarchical Clustering) classify resistivity-based subsurface zones. The results reveal that low-resistivity zones ( 100 Ωm) are associated with compacted volcanic deposits. Among the clustering methods, Hierarchical Clustering provides the most geologically meaningful classification by preserving gradual resistivity transitions. The integration of clustering with geostatistical interpolation enhances subsurface interpretation, reducing uncertainty in hydrogeological assessments. However, DBSCAN's sensitivity to parameter selection limits its effectiveness in identifying multiple resistivity clusters. This study confirms that the combination of geoelectrical inversion, geostatistics, and clustering improves the accuracy of subsurface mapping, offering a data-driven approach applicable to complex geological environments. Future research should focus on optimizing DBSCAN parameters and incorporating additional geophysical methods to refine groundwater characterization. Resistivity Inversion Geostatistical Interpolation Clustering Analysis Groundwater Exploration Subsurface Zonation Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 Figure 10 Figure 11 1. Introduction Geophysics is a branch of Earth sciences that applies the principles of physics to study the Earth's internal and near-surface structures geophysics. It encompasses various techniques that measure physical properties such as seismic waves, gravitational fields, magnetic anomalies, and electrical conductivity to infer geological formations and subsurface conditions(Lillie 1988 ; Milsom 2003 ; Kirsch 2006 ; Pehme 2011 ). One of the most widely used geophysical methods is electrical resistivity, which involves measuring the resistance of subsurface materials to the flow of electrical current(Nugraha et al. 2020 , 2022a , b ). Electrical resistivity methods, commonly referred to as geoelectrical methods, are crucial for exploring subsurface structures, particularly in hydrogeology, engineering geology, and environmental studies(Herman 2001 ; Civita and De Maio 2004 ; Auken et al. 2009 ; Wang et al. 2015 ; Gance et al. 2016 ; Hasan et al. 2019 ). Among these techniques, Electrical Resistivity Tomography (ERT) has gained significant popularity due to its ability to produce high-resolution two-dimensional and three-dimensional images of subsurface resistivity distributions (Herman 2001 ; Heaney 2003 ). By injecting electrical current into the ground through multiple electrodes and measuring the resulting voltage differences, ERT enables the characterization of geological formations, detection of groundwater reservoirs, and identification of subsurface contaminants (Mazáč et al. 1985 ; Auken et al. 2009 ; Jardani et al. 2013 ). The advancement of inversion techniques and data processing methods has further improved the accuracy and applicability of ERT, making it an essential tool for subsurface investigations across various disciplines (Kuras et al. 2009 ; Brunet et al. 2010 ; Tso et al. 2017 ; Dietrich et al. 2018 ). Electrical resistivity tomography (ERT) is one of the most widely used geophysical methods for investigating subsurface structures(Giguère et al. 2008 ; Clément et al. 2009 ; Coulouma et al. 2013 ; Tsourlos et al. 2014 ; Bernatek-Jakiel and Kondracka 2016 ; Giampaolo et al. 2016 ; Martel et al. 2018 ; NuGraha et al. 2021 ; Nugraha et al. 2022a ). It is based on the measurement of the electrical resistivity of geological formations by injecting electrical current into the ground through a set of electrodes and recording the resulting potential differences(Frohlich and Parke 1989 ; Milsom and Eriksen 2011 ). The measured resistivity values are influenced by various subsurface parameters, including lithology, porosity, water content, and mineral composition(Telford et al. 1990 , 2012a , b ). ERT has been extensively applied in hydrogeological studies, environmental site assessments, engineering geology, and geotechnical investigations (Gance et al. 2016 ; Su et al. 2017 ; Benabdelouahab et al. 2018a ). In groundwater studies, ERT is used to delineate aquifers, identify contamination plumes, and assess groundwater recharge areas (Chandra et al. 2010 ; Mao et al. 2015 ; Benabdelouahab et al. 2018b ). In environmental studies, it helps in detecting pollutants and mapping subsurface waste deposits(Mao et al. 2015 ). In geotechnical applications, ERT assists in assessing soil stability, detecting fractures, and evaluating foundation conditions (Coulouma et al. 2013 ; Giampaolo et al. 2016 ). Despite its versatility, one of the main challenges in using ERT is the interpretation of resistivity values, which can be ambiguous(Florsch and Muhlach 2018 ). The same resistivity range can correspond to different geological materials under varying conditions (Florsch and Muhlach 2018 ). For example, dry sand and fractured rock filled with air may have similar resistivity values but vastly different geological implications. Similarly, clay and highly mineralized groundwater can exhibit overlapping resistivity signatures. Traditionally, ERT data interpretation has relied heavily on qualitative and semi-quantitative methods, where resistivity values are visually compared with known geological references or inferred from experience(NuGraha et al. 2021 ; Nugraha et al. 2022a ). While effective to some extent, this approach is subjective, lacks reproducibility, and does not fully utilize the spatial variability of resistivity data. The increasing complexity of subsurface investigations demands more advanced and systematic approaches to enhance ERT data interpretation. To address these limitations, geostatistical methods and machine learning-based clustering techniques can be integrated into the interpretation process. Geostatistics provides a framework for analyzing the spatial correlation of resistivity values and improving interpolation accuracy(Xu et al. 2016 ; Hörning et al. 2020 ; Liang et al. 2024 ). Meanwhile, clustering algorithms can automatically classify different subsurface zones based on resistivity distributions, reducing subjectivity in interpretation(Audebert et al. 2014 ; Xu et al. 2017 ; Shi and Wang 2024 ). One of the main challenges in Electrical Resistivity Tomography (ERT) interpretation is the difficulty in defining clear boundaries between subsurface zones(Slater et al. 2002 ; Zarroca et al. 2011 ; Kim et al. 2014 ; Martorana et al. 2017 ). This issue arises due to several factors, including noise and uncertainties in resistivity measurements, which result from instrumental errors, heterogeneous subsurface properties, and inversion artifacts(Heaney 2003 ; Telford et al. 2012b , a ; Binley 2015 ; Florsch and Muhlach 2018 ; Cardarelli and De Donno 2019 ). Additionally, the overlapping resistivity values of different geological materials make it challenging to distinguish between formations with similar electrical properties(Telford et al. 2012a ). Furthermore, there is a lack of standardized quantitative zoning methods, as most existing approaches rely on expert judgment rather than data-driven techniques. Given these challenges, there is a growing need for a more robust and systematic approach that can effectively process ERT data, identify geological patterns, and delineate distinct subsurface zones with higher confidence. Groundwater is one of the most vital natural resources, providing a crucial supply of drinking water, supporting agricultural irrigation, and sustaining various industrial activities. As surface water sources become increasingly stressed due to climate change and over-extraction, groundwater serves as a critical alternative for human consumption and economic development. Many regions rely heavily on groundwater to meet their water demands, making its sustainable management a top priority. However, the depletion and contamination of groundwater reserves pose significant environmental and socio-economic challenges, highlighting the urgent need for effective exploration and monitoring techniques. Among the various methods for groundwater exploration, Electrical Resistivity Tomography (ERT) has proven to be highly effective due to its ability to produce detailed subsurface resistivity distributions. ERT enables the identification of water-bearing formations, aquifer boundaries, and potential contamination zones by measuring electrical resistivity variations in the subsurface. This method is non-invasive, cost-effective, and provides high-resolution imaging, making it particularly useful for hydrogeological studies. By integrating ERT with advanced computational techniques, groundwater exploration can be significantly improved, leading to more accurate assessments of aquifer properties and groundwater potential. To optimize groundwater exploration, the integration of resistivity inversion, geostatistical interpolation, and clustering methods offers a powerful approach for subsurface characterization. Inversion techniques refine raw resistivity measurements into a geologically meaningful resistivity model, while geostatistical analysis provides spatial interpolation to enhance the accuracy of subsurface resistivity distribution. Clustering methods further improve subsurface zoning by classifying resistivity values into distinct geological units, reducing ambiguity in groundwater assessments. The combination of these techniques enhances the interpretation of ERT data, minimizing uncertainties and providing a more systematic framework for groundwater exploration and management. This study aims to address these challenges by integrating geostatistical interpolation and clustering techniques for automated subsurface zoning using ERT data, with a focus on groundwater exploration.The first objective is to apply geostatistical methods, particularly Kriging, to perform resistivity interpolation and construct a continuous spatial resistivity model. This will provide a smoother and more comprehensive representation of subsurface resistivity distribution. Kriging is a widely used geostatistical interpolation method that provides reliable spatial predictions and has been extensively applied across various Earth science disciplines, including hydrogeology, geophysics, and environmental studies(Javed et al. 2021 ; Balacco et al. 2023 ; Masoudi et al. 2024 ). In subsurface investigations, Kriging effectively models spatial variability by utilizing statistical relationships between measured data points, making it a powerful tool for mapping geological formations and predicting resource distributions (Singh and Verma 2019 ; Menke 2022 ; Singh et al. 2025 ). The robustness of Kriging in handling spatial uncertainty has led to its adoption in diverse geoscientific applications, such as groundwater potential assessment, mineral exploration, and soil property mapping (Singh and Verma 2019 ; Menke 2022 ; Singh et al. 2025 ). The second objective is to implement clustering techniques, including K-Means, DBSCAN, and Hierarchical Clustering, to classify subsurface zones based on variations in resistivity values. K-Means is a centroid-based algorithm that partitions data into a predefined number of clusters by minimizing the variance within each cluster (Liang and Yoon 2024 ). DBSCAN (Density-Based Spatial Clustering of Applications with Noise) identifies clusters based on the density of data points, making it particularly effective for detecting irregularly shaped geological formations and filtering out noise (Belyadi and Haghighat 2021 ; Ahmadi 2024 ; Liu et al. 2025 ). Hierarchical Clustering builds a nested structure of clusters using either an agglomerative (bottom-up) or divisive (top-down) approach, allowing for a more flexible and interpretable clustering process (Seo and Shneiderman 2003 ; Theodoridis and Koutroumbas 2009 ; Ahmadi 2024 ; Dong et al. 2024 ; Liu et al. 2025 ). These methods will help in identifying distinct geological formations and detecting patterns that are not easily recognizable through traditional qualitative interpretation. The third objective is to evaluate the effectiveness of these clustering approaches in distinguishing geological formations, ensuring that the results align with known geological features. Finally, the study aims to compare the clustering outcomes with geological references to assess their reliability in real-world groundwater exploration applications. By fulfilling these objectives, this research seeks to develop a quantitative and reproducible workflow for improving subsurface characterization using ERT data for groundwater exploration.. This research holds significant contributions to geophysical and environmental studies, particularly in groundwater exploration, by enhancing the accuracy of ERT interpretation, improving geological zoning, and expanding its applications to various hydrogeological challenges. By integrating geostatistical interpolation with clustering analysis, the proposed approach offers a more precise and systematic classification of subsurface resistivity data, which is crucial for identifying groundwater-bearing formations, delineating aquifers, and assessing water quality. The automatic classification of subsurface zones using clustering techniques reduces reliance on expert judgment, thereby increasing the reproducibility and objectivity of geological interpretations. Moreover, the methodology can be applied to practical groundwater exploration applications such as locating potential aquifers, detecting contamination plumes, and evaluating groundwater recharge zones. Beyond ERT, the framework developed in this study has the potential to be extended to other geophysical datasets, including seismic surveys, gravity measurements, and electromagnetic sounding, enabling a multi-method approach for more comprehensive groundwater resource assessments. This broader applicability underscores the value of integrating advanced data-driven techniques into groundwater exploration and environmental management. 2. Materials and Methods 2.1 Geology and Hydrogeology The oldest rock unit in the study area is a Tertiary sedimentary sequence of claystone interbedded with sandstone (Yuniardi et al., 2019). This unit is overlain by Pre-Sunda Volcanic Rocks (lava and pyroclastic deposits), followed by the Lower-Upper Pleistocene Sunda Andesite Volcanic unit, Upper Pleistocene Pyroclastic Sunda, early Holocene andesitic Tangkubanparahu, and Late Holocene Pyroclastic deposits. The stratigraphic sequence includes Alluvium (Qa), Colluvium (Qc), Sandy Tuff (Qyd), Breccia and Agglomerate (Qyb), Pumiceous Tuff (Qyt), Undifferentiated Young Volcanic Products (Qyu), Undifferentiated Old Volcanic Products (Qvu), Older Volcanic Products (Qob), Tjitalang Formation (Pt), and Kaliwangu Formation (Pk), composed of various volcanic and sedimentary deposits such as tuff, breccia, lahar, sandstone, conglomerate, marl, and limestone (Yuniardi et al., 2019; Nurfiani & Bouvet de Maisonneuve, 2018; Sabrian et al., 2021). Tangkubanparahu volcano is part of the Sunda-Tangkubanparahu volcanic system, characterized by seven volcanic phases and four distinct periods of volcanism. Two major caldera-forming eruptions produced extensive pyroclastic deposits, including the Cisarua and Manglayang ignimbrites. The volcanic facies influencing groundwater occurrence include: (1) Volcanic Proximal Facies (3050–3172 masl) of impermeable andesites, (2) Medial Volcanic Facies (500–2076 masl) with impermeable pyroclastic deposits and fractured andesitic lava hosting 142 springs (178 l/s), and (3) Distal Volcanic Facies (100–650 masl) composed of permeable volcanic deposits hosting 53 springs (700 l/s). Groundwater aquifers in the study area comprise moderately to highly productive units, with variable transmissivity, deep water tables, and high spring discharge (> 100 l/s), though well yields remain low (< 5 l/s) (Delinom, 2009, 2011). The study area, Cikole-Lembang, West Bandung Regency, lies within the Lembang Groundwater Basin (95.56 km²), bordered by the Bandung-Soreang, Ciater, and Sumedang groundwater basins. Lembang has a population of approximately 196,690, with tourism as a key economic sector, hosting 36 major destinations. The geological and hydrogeological characteristics of the study area have been comprehensively documented in Nugraha et al. ( 2023 ), providing an in-depth analysis of stratigraphy, aquifer properties, and groundwater flow dynamics. Their findings serve as a key reference for understanding groundwater resources in the region. 2.2 ERT Data Acquisition and Processing A comprehensive explanation of data acquisition techniques, processing, analysis, and interpretation of geoelectrical data, along with its hydrogeological implications, can be found in Nugraha et al. ( 2023 ). The present text provides only a summary of their study. This research focuses on an advanced geoelectrical analysis involving kriging and clustering techniques. These analyses utilize the resistivity output from the inversion process conducted by Nugraha et al. ( 2023 ).Their study employed a single Electrical Resistivity Tomography (ERT) traverse, which will be further discussed in the subsequent section. The ERT survey utilized a dipole-dipole configuration with a 10-meter electrode spacing over a total length of 550 meters. The dipole-dipole array is defined by its geometric factor ( \(\:K\) ), which determines the apparent resistivity ( \(\:{\rho\:}_{a}\) ) measurement at each point. The geometric factor for a dipole-dipole array is given by (Binley 2015 ; Florsch and Muhlach 2018 ; Cardarelli and De Donno 2019 ): $$\:K=\frac{\pi\:an(n+1)(n+2)}{2}$$ 1 Where \(\:a\) = electrode spacing (10 m) and \(\:n\) = dipole separation factor (varies from 1 to 6). The apparent resistivity ( \(\:{\rho\:}_{a}\) ) at each measurement point is calculated using Ohm’s Law (Herman 2001 ; Telford et al. 2012a ): $$\:{\rho\:}_{a}=K\cdot\:V/I$$ 2 Where \(\:V\) = measured voltage (V) and \(\:I\) = injected current (A) ERT data were acquired using the Advanced Geosciences Inc. (AGI) SuperSting R8, an 8-channel resistivity meter (Nugraha et al. 2023 ). Direct current was injected through current electrodes, while voltage responses were recorded by potential electrodes. The survey employed a dipole-dipole array with 56 electrodes at 10-meter spacing, inserted ~ 0.25 meters into the ground, selected for its high horizontal resolution in volcanic sediment aquifers (Neyamadpour et al., 2010). Measurements lasted 1.2 seconds, with two data stacks, or three if the standard deviation exceeded 2%. Data processing and inversion were performed using AGI EarthImager 2D (v2.4.0) (Nugraha et al. 2023 ). A finite-element model with Smooth Model Inversion ensured stability across datasets. Apparent resistivity values were inverted using a least-squares approach to derive the true subsurface resistivity distribution. The inversion process solves for the resistivity model ( \(\:\varvec{m}\) ) by minimizing the objective function (Herman 2001 ; Telford et al. 2012a ; Binley 2015 ; Florsch and Muhlach 2018 ): $$\:{\Phi\:}\left(\varvec{m}\right)=\left|\right|{\varvec{W}}_{\varvec{d}}({\varvec{d}}_{obs}-{\varvec{d}}_{calc})|{|}^{2}+\lambda\:|\left|{\varvec{W}}_{\varvec{m}}\right(\varvec{m}-{\varvec{m}}_{0}\left)\right|{|}^{2}$$ 3 Where \(\:{\varvec{d}}_{obs}\) = observed apparent resistivity data, \(\:{\varvec{d}}_{calc}\) = calculated resistivity data from the current model, \(\:{\varvec{W}}_{\varvec{d}}\) = data weighting matrix, \(\:\varvec{m}\) = model parameters (true resistivity values), \(\:{\varvec{m}}_{0}\) = reference model, \(\:{\varvec{W}}_{\varvec{m}}\) = model weighting matrix and \(\:\lambda\:\) = regularization parameter. The inversion algorithm iteratively updates the resistivity model by solving (Herman 2001 ; Telford et al. 2012a ; Binley 2015 ; Florsch and Muhlach 2018 ) : $$\:{\varvec{m}}_{k+1}={\varvec{m}}_{k}+({\varvec{J}}^{T}{\varvec{W}}_{\varvec{d}}^{T}{\varvec{W}}_{\varvec{d}}\varvec{J}+\lambda\:{\varvec{W}}_{\varvec{m}}^{T}{\varvec{W}}_{\varvec{m}}{)}^{-1}{\varvec{J}}^{T}{\varvec{W}}_{\varvec{d}}^{T}({\varvec{d}}_{obs}-{\varvec{d}}_{calc})$$ 4 where J is the Jacobian matrix of partial derivatives. The final result is a 2D resistivity model, which provides insights into the geological structures beneath the survey area. The default "surface" inversion setting was applied for consistency across sites (Vogelgesang et al., 2019). A fine-model inversion was implemented, refining the mesh in near-surface areas while coarsening with depth. Resistivity values below 1 Ohm-m and above 10,000 Ohm-m were excluded as potential erroneous data. The final resistivity model achieved an average root mean square (RMS) error of 2.92% and an average L2 norm ratio of 0.9. Additionally, negative pseudo-resistivity values and non-conforming relative data exceeding 50% were removed (Vogelgesang et al., 2020). The computational analysis was performed on a system running Windows 11 Home Single Language (version 23H2) with an Intel(R) Core(TM) i7-12700H "Alder Lake" 2.30 GHz processor, 16 GB DDR5 RAM (4800 MHz), NVMe Gen 4x4 SSD, and dual graphics comprising Intel(R) Iris(R) Xe integrated GPU and NVIDIA GeForce RTX 3060 Laptop GPU. The data processing and analysis for Geostatistical interpolation and clustering techniques for subsurface zoning were conducted using Python within the Jupyter Notebook (version 7.0.8) environment, as provided in Anaconda Navigator (version 2.5.2). 2.3 Geostatistical Interpolation (Kriging) To refine resistivity estimates and improve spatial resolution, Ordinary Kriging was applied. Kriging is a spatial interpolation technique that predicts unknown values based on known sample points. Kriging requires a variogram model to define spatial correlation. The experimental variogram is calculated as (Singh and Verma 2019 ; Menke 2022 ; Balacco et al. 2023 ): $$\:\gamma\:\left(h\right)=\frac{1}{2N\left(h\right)}\sum\:_{i=1}^{N\left(h\right)}\left[Z\right({x}_{i})-Z({x}_{i}+h){]}^{2}$$ 5 Where \(\:\gamma\:\left(h\right)\) = variogram function for lag distance \(\:h\) , \(\:Z\left({x}_{i}\right)\) = resistivity at location \(\:{x}_{i}\) and \(\:N\left(h\right)\) = number of data pairs at distance \(\:h\) . A best-fit theoretical model (e.g., spherical, exponential, Gaussian) was selected based on the variogram shape. The resistivity at an unknown location ( \(\:{Z}^{*}\left(x\right)\) ) is estimated using: $$\:{Z}^{*}\left(x\right)=\sum\:_{i=1}^{N}{\lambda\:}_{i}Z\left({x}_{i}\right)$$ 6 where \(\:{\lambda\:}_{i}\) are Kriging weights, computed by solving: $$\:\sum\:_{j=1}^{N}{\lambda\:}_{j}\gamma\:({x}_{i},{x}_{j})+\mu\:=\gamma\:({x}_{i},x)$$ 7 where \(\:\mu\:\) is the Lagrange multiplier ensuring unbiased estimates. Using the optimized variogram, Kriging interpolation was performed to generate a smoothed resistivity distribution. This process helps in reducing inversion artifacts and providing a clearer visualization of geological structures. This study applies Ordinary Kriging (OK) to interpolate electrical resistivity data collected from geoelectrical surveys. The dataset consists of spatial coordinates ( X, Y ) and corresponding resistivity values ( Z ), which are used to construct a continuous resistivity model. The primary goal of this approach is to enhance subsurface characterization by generating a high-resolution resistivity distribution while optimizing computational efficiency. To achieve this, a structured interpolation grid is defined with 30 × 30 resolution points, reducing the computational burden compared to a finer mesh. The OrdinaryKriging function from the PyKrige library is utilized, with a spherical variogram model, which is commonly used for modeling subsurface properties with spatial continuity. The interpolation is performed in point mode, where only the 20 closest data points contribute to each prediction, ensuring computational efficiency and preventing overfitting. The interpolated resistivity values are then visualized using a contour plot, which provides a clear representation of the spatial resistivity variations. The original data points are overlaid on the contour plot to assess the accuracy of the interpolation, ensuring that key subsurface resistivity patterns are preserved. A viridis colormap is applied to highlight resistivity variations, while a coolwarm colormap is used for the original data points to distinguish measured values from interpolated estimates. The application of n_closest_points and the loop-based backend minimizes memory usage and improves interpolation performance, making it suitable for large-scale geophysical studies. The output is saved as a high-resolution image to facilitate further analysis and integration with other geophysical datasets. 2.4 Clustering Techniques for Subsurface Zoning To classify different subsurface zones based on resistivity variations, three clustering techniques were applied: K-Means, DBSCAN, and Hierarchical Clustering. These methods allow for a more systematic and data-driven approach to geological zoning. The dataset is first loaded from a CSV file containing geophysical survey results with columns X, Depth, and Resistivity. The data is structured into a NumPy array for computational efficiency. The three selected attributes are used as input features for clustering, ensuring that both spatial (X, Depth) and physical (Resistivity) parameters are considered in the classification. This preprocessing step ensures that clustering is performed based on the combined influence of location and resistivity variations. 2.4.1 K-Means Clustering K-Means is a partitioning algorithm that groups data points into K clusters based on their similarity(Menke 2022 ; Ahmadi 2024 ; Dong et al. 2024 ; Liu et al. 2025 ). The algorithm follows these steps: Select K initial centroids randomly, Assign each data point to the nearest centroid based on Euclidean distance, Recalculate cluster centroids iteratively until convergence. The optimal number of clusters (K) was determined using the Elbow Method and Silhouette Score. Clusters were determined by minimizing intra-cluster variance: $$\:J=\sum\:_{i=1}^{K}\sum\:_{j\in\:{C}_{i}}||{x}_{j}-{\mu\:}_{i}|{|}^{2}$$ 8 where \(\:{C}_{i}\) is the set of points in cluster ii, and µi\mu_i is the centroid. This study applies K-Means clustering to classify subsurface zones based on electrical resistivity data collected from geoelectrical surveys. The dataset comprises spatial coordinates (X, Depth) and Resistivity values, which serve as the input parameters for clustering. The primary objective is to identify distinct resistivity zones, which may correspond to different geological formations or hydrogeological conditions. The clustering process is conducted using the K-Means algorithm, a widely used unsupervised learning technique for partitioning data into a predefined number of clusters. In this study, the number of clusters is set to three (k = 3), representing three distinct resistivity zones. The algorithm iteratively assigns data points to clusters based on their similarity in resistivity values and spatial distribution, ensuring that subsurface zoning is objectively classified. The resulting cluster assignments are visualized using a scatter plot, where data points are color-coded according to their respective clusters. A viridis colormap is applied to enhance the visual distinction between zones, and a color bar is included to indicate the cluster labels. The X-axis represents horizontal distance, while the Y-axis corresponds to depth, providing a clear spatial representation of the identified zones. The final output, saved as "Zonation Based on Resistivity.png", provides valuable insights into the spatial distribution of subsurface resistivity variations. This clustering-based approach reduces subjectivity in geological interpretation and enhances the reproducibility of resistivity-based zoning. The ability to automatically segment resistivity data into meaningful zones improves subsurface characterization for applications such as groundwater exploration, contamination mapping, and geotechnical assessments. 2.4.2 DBSCAN (Density-Based Spatial Clustering of Applications with Noise) DBSCAN is a density-based clustering method that identifies high-density regions and separates outliers (Belyadi and Haghighat 2021 ; Liang and Yoon 2024 ). It operates as follows: Defines a neighborhood radius ( ε ) and minimum number of points (MinPts) and Groups points with high local density into clusters also Identifies low-density points as noise (outliers). DBSCAN is particularly useful for detecting anomalous geological zones, such as faulted or fractured areas. Unlike K-Means, DBSCAN (Density-Based Spatial Clustering of Applications with Noise) does not require a predefined number of clusters. Instead, it relies on two key parameters: eps (ε) — the neighborhood radius around a point, min_samples — the minimum number of points required to form a cluster DBSCAN identifies clusters based on density, making it effective for detecting irregular geological formations and removing noise from the dataset. It assigns core points to dense regions, border points to cluster edges, and noise points as outliers. In this study, eps = 10 and min_samples = 5 are chosen based on data scale and distribution. The resulting clusters are visualized in a scatter plot, with different colors indicating different density-based clusters. Unlike K-Means, DBSCAN can detect zones of varying shapes, making it useful for identifying complex resistivity structures. 2.4.3 Hierarchical Clustering Hierarchical clustering builds a tree-like structure (dendrogram) to show relationships between resistivity zones(Seo and Shneiderman 2003 ; Theodoridis and Koutroumbas 2009 ). The process includes: Computing a distance matrix between all data points, Iteratively merging the closest clusters using linkage criteria (e.g., Ward’s method) and Cutting the dendrogram at an optimal level to define geological zones. This method provides a hierarchical view of subsurface formations, offering insights into the relative relationships between different resistivity zones. Hierarchical clustering is performed using the Agglomerative Clustering method with Ward linkage, which minimizes the variance within clusters during merging. Unlike K-Means and DBSCAN, Hierarchical Clustering does not require a predefined number of clusters, and the final grouping is determined through a dendrogram visualization. The steps include: Computing the pairwise distance between all data points. Iteratively merging the closest clusters based on the Ward variance minimization criterion. Constructing a dendrogram, which shows the hierarchical relationships between clusters. The dendrogram helps determine the optimal number of clusters by identifying natural splits in the hierarchical tree. In this study, three clusters are selected for visualization. The results are plotted similarly to K-Means and DBSCAN, showing distinct resistivity-based zones. 2.5 Evaluation of Clustering Results To assess the performance of clustering techniques, the following evaluation metrics were used:Silhouette Score Measures how well each point fits within its assigned cluster. $$\:S=\frac{b-a}{max(a,b)}$$ 9 where a is intra-cluster distance and b is nearest-cluster distance. 3. Results and Discussion 3.1. Results 3.1.1 2D Resistivity Inversion Results The following is a summary of the ERT results from previous study conducted by Nugraha et al. ( 2023 ). The ERT survey was carried out along a 540-meter east-west oriented line with a dipole-dipole electrode configuration and a 10-meter electrode spacing. The inversion results using EarthImager 2D software indicate a maximum modelable depth of 103 meters, revealing three distinct resistivity zones: low, medium, and high(Fig. 5 ). The low resistivity zone (123–292 Ωm), represented by dark to light blue colors, is observed at depths of 0–25 meters across the entire section, with greater thickness on the eastern side. The medium resistivity zone (293–700 Ωm), colored green to yellow, extends from 13 to 77 meters and from 77 to 103 meters, lying beneath the low-resistivity layer. It is thicker in the east and thinner in the west. The high resistivity zone (701–3875 Ωm) is concentrated in the middle of the cross-section, extending from 27 to 77 meters, with a thickness of about 50 meters. This layer is thicker in the west and thins towards the east. Lithologically, the low and upper-medium resistivity zones correspond to the Tangkubanparahu Pyroclastic Fall 2 (Tjp2), composed of scoria, pumice, and lithic fragments of andesite and basalt, typically loose to consolidated, well-stratified, and weathered brownish-yellow. The high rainfall in the study area, particularly during the rainy season when measurements were conducted, likely contributes to the low resistivity values due to increased water infiltration. The high-resistivity layer is associated with Tangkubanparahu lava (Tl), which consists of basaltic lava flows containing plagioclase, pyroxene, and minor olivine. These lava flows originate from Tangkubanparahu’s main crater and are widely exposed in areas such as Cisarua, Sagalaherang, and northern Bandung. A medium-resistivity layer in the ERT cross-section is interpreted as Tangkubanparahu Pyroclastic Fall 1 (Tjp1), predominantly composed of pumice (1–5 cm), scoria, and basaltic-andesitic lithic fragments, with occasional volcanic bombs. This unit, characterized by thin layers up to 1 meter thick, has a resistivity of 300–700 Ωm and a thickness of 71 meters in the west, thinning towards the east. The Sunda Pyroclastic Flow Deposits (Sap) and Sunda Lava (S1) were not detected, likely due to the inversion depth limit of 103 meters, whereas these deposits are estimated to be deeper than 150 meters. 3.1.2 Comparative Analysis of Clustering Methods To further analyze the spatial distribution of resistivity values, three different clustering methods were applied: K-Means Clustering, DBSCAN (Density-Based Spatial Clustering of Applications with Noise) and Hierarchical Clustering (Fig. 6 ) 3. 1. 2.1 K-Means Clustering The K-Means clustering method was employed to classify subsurface resistivity distributions into distinct zones. The resulting zonation is depicted in Fig. 7 , where different colors represent different clusters. The K-Means algorithm effectively segmented the resistivity data into three distinct clusters. This clustering technique partitions the dataset based on similarity, ensuring that points within a cluster exhibit comparable resistivity values. The spatial contiguity of the resulting clusters suggests that regions with similar resistivity distributions are systematically grouped, thereby enhancing geological interpretation by delineating subsurface zones with consistent electrical properties. The clustering analysis categorized the resistivity data into three distinct groups, designated as Cluster 0, Cluster 1, and Cluster 2. These clusters correspond to low, moderate, and high-resistivity zones, as previously identified through resistivity inversion. Cluster 2 (yellow) predominantly occupies the shallow subsurface, extending from the surface to approximately 20 meters in depth. The low resistivity values observed in this cluster suggest the presence of water-saturated formations, which may indicate aquifers or weathered volcanic deposits. The extensive distribution of Cluster 2 in the uppermost layers underscores its significance for groundwater exploration, as these formations likely comprise porous and permeable materials conducive to groundwater storage and transmission. Cluster 1 (teal) is primarily observed between 20 and 40 meters depth, suggesting a transitional zone between the water-bearing formations above and the more compact lithologies below. This transition zone may correspond to increasing compaction with depth or a shift from unconsolidated to semi-consolidated deposits. Cluster 0 (dark purple) becomes more prominent at depths of approximately 30 meters and beyond. These clusters are associated with higher resistivity values, indicative of compacted or less permeable rock formations. The presence of Cluster 0 at greater depths aligns with high-resistivity zones, likely representing volcanic rock formations. These formations may function as barriers to groundwater flow due to their lower porosity and permeability. The clustering results provide a refined segmentation of the subsurface, facilitating enhanced geological interpretation. The classification of resistivity data into meaningful clusters contributes to a more comprehensive understanding of lithological variations at different depths. 3. 1. 2.2. DBSCAN Clustering The DBSCAN (Density-Based Spatial Clustering of Applications with Noise) algorithm was applied to the resistivity dataset but resulted in a single dominant cluster (Fig. 8 ). This suggests that the majority of data points were classified into the same category, limiting its ability to differentiate geological units. Unlike hierarchical clustering, which provided a structured classification, DBSCAN failed to identify distinct subsurface zones in this case. DBSCAN is particularly useful for detecting clusters of varying density and identifying noise points. However, in this analysis, no significant noise points were detected, indicating that the resistivity dataset is relatively homogeneous. The effectiveness of DBSCAN largely depends on parameter selection, particularly the neighborhood radius (epsilon) and the minimum number of points required to form a cluster. If these parameters are not optimally set, the algorithm may fail to detect meaningful sub-clusters. The results indicate that the chosen parameter settings may have been too restrictive, preventing the formation of distinct clusters. Adjusting these parameters could potentially improve DBSCAN’s ability to identify subsurface zones with different resistivity characteristics. 3. 1. 2.3. Hierarchical Clustering Hierarchical clustering provided a structured representation of subsurface resistivity distribution, successfully identifying three distinct clusters (Fig. 9 ). Unlike the K-Means method, hierarchical clustering does not require a predefined number of clusters, allowing for a more flexible and adaptive interpretation. The clustering results were further validated using a dendrogram, as shown in Fig. 10 . This visualization confirms the presence of natural resistivity groupings, supporting the notion that hierarchical clustering provides a geologically meaningful classification. To classify subsurface resistivity patterns, a hierarchical clustering analysis was conducted, and the resulting dendrogram is presented in Fig. 10 . The analysis provided a structured classification of resistivity data, delineating distinct subsurface zones based on their electrical properties. In the dendrogram, the y-axis represents the Euclidean distance, which serves as a measure of dissimilarity between clusters. Higher values on the y-axis indicate greater differences between clusters, whereas lower values suggest closely related data points. The x-axis displays individual resistivity data points, arranged according to their clustering structure, providing an intuitive understanding of similarities and differences in resistivity values across regions. Three primary clusters were identified in the analysis. Each cluster is represented by a distinct color in the dendrogram—blue, green, and orange—indicating separate resistivity groupings. The presence of these clusters suggests significant variations in subsurface electrical properties. A notable observation in the dendrogram is a major separation at approximately 30,000 on the y-axis. This significant dissimilarity, or highest linkage, indicates that the dataset comprises two dominant resistivity groups with fundamentally different characteristics. Near the lower part of the dendrogram, smaller branches indicate finer subdivisions within each major cluster. These sub-clusters highlight internal variability within the resistivity zones, suggesting localized changes in porosity, mineral composition, or degrees of weathering. The clustering results align closely with the resistivity zonation map presented in Fig. 6 , reinforcing the validity of the classification approach. In the resistivity zonation map, distinct regions were identified based on interpolation techniques. These regions exhibit clear boundaries, which correspond to the clustering patterns observed in the dendrogram. 3.1.3 Geostatistical Analysis Using Kriging Interpolation The Kriging interpolation method was applied to estimate the spatial distribution of resistivity values based on the measured data points. The result is shown in Fig. 11 , where the resistivity distribution is interpolated over the entire survey area. 3.7.1 Interpretation of Kriging Results Kriging interpolation provides a detailed representation of resistivity variations across the study area, offering insights into subsurface geological structures. The original resistivity measurement points, represented by white and blue dots, indicate sampled values at specific depths. The color gradient in the Kriging interpolation map reflects resistivity distribution. Low resistivity values appear in dark blue to purple, whereas high resistivity zones are depicted in yellow to red. This gradient helps distinguish different subsurface formations. A notable high-resistivity anomaly exceeding 2000 Ω.m is observed between X = 250–400 meters at depths of 30–50 meters. 3.2. Discussion 3.2.1. Clustering Methods 3. 2.1. 1 K-Means Clustering A key advantage of the K-Means algorithm is its computational efficiency, which makes it a widely adopted method for clustering large datasets. Its capability to produce clear segmentation facilitates a straightforward analysis of subsurface resistivity variations. However, K-Means operates under the assumption that clusters are spherical and exhibit equal variance. This assumption may introduce inaccuracies when applied to geological formations characterized by irregular geometries and complex boundaries. Geological structures often exhibit gradual resistivity transitions rather than discrete boundaries. Consequently, the assumption of spherical clusters may not adequately capture the inherent variability of subsurface formations, potentially affecting the reliability of interpretations based on K-Means clustering. The predominance of low-resistivity zones in the shallow subsurface supports the hypothesis that these layers contain saturated sediments or weathered volcanic deposits with high porosity, favorable for groundwater accumulation. At intermediate depths, Cluster 1 (teal) represents a transitional zone with moderate resistivity values, indicative of variable porosity and permeability, suggesting the presence of a semi-confined aquifer. The transition from Cluster 2 to Cluster 1 at depths of 20–40 meters marks a change in subsurface conditions, potentially corresponding to increasing compaction, reduced permeability, and altered groundwater flow dynamics. The deepest layers, classified as Cluster 0 (dark purple), are associated with high-resistivity values, likely representing dense, compacted volcanic formations or bedrock, which function as an aquitard, restricting vertical water movement. While K-Means clustering provides a useful classification of resistivity data, its reliance on predefined cluster numbers and its sensitivity to geometric assumptions suggest that integrating alternative clustering methods, such as hierarchical clustering, may enhance the robustness of geological interpretations. 3. 2.1. 2. DBSCAN Clustering One of DBSCAN’s strengths is its ability to identify clusters of arbitrary shapes, making it suitable for datasets with complex geological structures. However, its sensitivity to parameter tuning presents a challenge in applications where optimal values are not easily determined. The failure to detect multiple clusters in this study suggests that DBSCAN may not be the most suitable method for classifying resistivity data in the given geological setting. In contrast to hierarchical clustering, which effectively captured gradual resistivity variations, DBSCAN’s reliance on density thresholds may not be well-suited for datasets where resistivity transitions occur progressively rather than in distinct, high-density clusters. Future studies should explore parameter optimization techniques, such as automated epsilon selection, to enhance DBSCAN’s performance in geological applications. Additionally, integrating DBSCAN with other clustering methods could provide a more comprehensive analysis of resistivity data. 3. 2.1. 3. Hierarchical Clustering A key advantage of hierarchical clustering is its ability to represent the natural progression of resistivity variations. This characteristic is especially beneficial for geological studies, where subsurface changes typically occur gradually rather than abruptly. The hierarchical structure enables researchers to examine clustering results at different levels of granularity, making it useful for both broad-scale and detailed subsurface analyses. Hierarchical clustering has demonstrated its capability to reveal resistivity transitions, suggesting that it may be more suitable for geological applications compared to K-Means, which tends to impose rigid cluster boundaries. The clustering approach enhances the interpretation of resistivity data by revealing patterns that may not be immediately apparent in conventional interpolation maps. By grouping similar resistivity values, the analysis provides a more structured representation of subsurface properties. High-resistivity zones form a distinct and separate cluster in the analysis, suggesting that these areas are geologically unique and potentially represent dense volcanic rocks or compact lithological formations. High-resistivity zones are often associated with materials such as fresh basaltic lava flows or consolidated rock formations with minimal water content. Their classification as an independent cluster indicates their distinct physical properties. Conversely, the low-resistivity regions form a separate cluster, likely associated with water-saturated layers or clay-rich formations. These materials typically exhibit high conductivity, resulting in lower resistivity values. The presence of a well-defined low-resistivity cluster suggests a clear distinction between water-bearing units and surrounding geological formations, which is crucial for hydrogeological assessments and groundwater exploration. A significant implication of hierarchical clustering in geoelectrical studies is the clear separation between clusters, suggesting the presence of distinct lithological units within the study area, corresponding to different geological formations. The clustering results support previous interpretations derived from resistivity inversion techniques. The agreement between the method reinforces confidence in the identified resistivity zones. 3. 2.1. 4. Implications for Subsurface Resistivity Zonation The comparison of these clustering methods highlights their respective advantages and limitations in subsurface resistivity analysis. Each technique offers unique insights, but their applicability depends on the characteristics of the dataset. K-Means provides clear segmentation of resistivity zones, making it a suitable choice for cases where well-defined clusters are expected. However, its assumption of spherical clusters limits its effectiveness in complex geological formations. DBSCAN, while effective for detecting clusters of varying density, struggled to differentiate multiple clusters in this study. The sensitivity of its parameters suggests that further optimization is necessary for meaningful classification. Hierarchical clustering emerged as the most geologically meaningful approach, as it successfully captured gradual resistivity transitions rather than enforcing rigid boundaries. This aligns better with the nature of geological formations. 3.2.2. Geostatistical Analysis This feature (Fig. 11 ) is highlighted in yellow and suggests the presence of a compact volcanic formation or a dry, resistive layer. Conversely, low-resistivity zones (< 500 Ω.m) dominate the upper layers. These regions, shown in dark blue, likely indicate water-bearing formations or weathered volcanic deposits, which are commonly associated with high moisture content or clay materials. The spatial distribution of resistivity values suggests a stratified subsurface structure, where resistive and conductive layers alternate at different depths. The observed transitions provide valuable information for geological and hydrogeological interpretation. 3.2.2.1. Correlation with Resistivity Inversion and Clustering The high-resistivity anomaly identified through Kriging interpolation aligns with the resistivity inversion results presented in Fig. 5 . This consistency confirms the presence of a compact and resistive subsurface layer. Additionally, clustering analysis (Fig. 6 ) highlights this region as a distinct class. The agreement between clustering and Kriging results reinforces the uniqueness of this geological feature. A smooth transition is observed between resistive and conductive zones, which follows a gradual gradient. This pattern suggests that Kriging effectively captures spatial variations in resistivity, reducing abrupt discontinuities in the interpolation model.By integrating resistivity inversion, Kriging interpolation, and clustering, a more comprehensive understanding of subsurface characteristics is achieved. These combined methods validate the presence of geologically distinct units and improve interpretation accuracy. 3.2.2.2. Geostatistical Model and Variogram Analysis The Kriging interpolation was conducted using a Gaussian variogram model, which effectively describes the spatial continuity of resistivity data. This model ensures a geologically consistent interpolation result. The general form of the variogram equation used in the analysis is: $$\:\gamma\:\left(h\right)={C}_{0}+C(1-{e}^{-h/a})$$ 10 where \(\:\gamma\:\left(h\right)\) represents the semi-variance at distance \(\:h\) , \(\:{C}_{0}\) is the nugget effect accounting for measurement errors and small-scale variability, \(\:C\) is the sill, representing the maximum variance, and aa is the range, indicating the distance at which spatial correlation diminishes. By optimizing these variogram parameters, Kriging generates a smooth and geologically meaningful resistivity interpolation. The careful selection of parameters ensures that the interpolated resistivity values accurately reflect underlying geological structures. The nugget effect ( \(\:{C}_{0}\) ) accounts for random noise and microscale variations that cannot be captured by the main interpolation model. A low nugget value suggests high data quality and minimal measurement errors. The sill ( \(\:C\) ) represents the maximum variance observed in the dataset. In this case, the sill value helps determine the overall contrast between high- and low-resistivity zones, providing insights into geological heterogeneity. The range ( \(\:a\) ) is particularly important as it defines the spatial correlation limit. Beyond this distance, resistivity values become statistically independent, indicating distinct geological formations. The Gaussian variogram model was selected due to its smooth transition properties. This model is particularly suitable for subsurface resistivity data, where gradual geological changes are expected. 3.2.2.3. Key Insights from Kriging Interpolation The identification of a high-resistivity anomaly across multiple analytical methods suggests a geologically significant feature. This consistency strengthens the interpretation of subsurface structures. Kriging interpolation provides a continuous representation of resistivity variations, reducing uncertainties commonly associated with discrete measurement points. The method enhances spatial understanding by filling data gaps with statistically optimized values. The integration of Kriging results with clustering-based zonation confirms the existence of distinct hydrogeological units. This cross-validation improves confidence in geological interpretations. The combination of geostatistical and geophysical techniques allows for a more robust subsurface characterization. By leveraging multiple methods, the study provides a comprehensive assessment of resistivity variations.Overall, Kriging serves as an essential tool for geoelectrical interpretation, offering a refined approach to resistivity mapping and enhancing the reliability of subsurface models. 3.2.2.4. Comparison with Geostatistical Analysis The clustering results show a strong correlation with Kriging interpolation, further validating the spatial distribution of resistivity values. The areas identified as low-resistivity zones through Kriging were consistently classified within Cluster 2. Similarly, the resistivity inversion model reinforces the clustering-based zonation, demonstrating the robustness of this classification approach. The consistency between multiple analytical methods enhances confidence in the geological interpretation. The integration of geostatistical analysis and clustering techniques provides a comprehensive framework for subsurface characterization. By combining both approaches, uncertainties in resistivity interpretation can be minimized.The ability of clustering to segment resistivity data into meaningful groups complements the continuous nature of Kriging interpolation. While Kriging provides a smooth representation of resistivity variation, clustering introduces discrete classification, aiding geological zonation. This agreement between methods suggests that the clustering-based classification effectively captures key hydrogeological features, reinforcing the validity of resistivity-based subsurface modeling. Integrating clustering techniques with geostatistical interpolation offers another avenue for improving subsurface modeling. Kriging interpolation provides a continuous representation of resistivity variations, while clustering introduces discrete classification that highlights key geological boundaries. By combining these approaches, researchers can develop a more comprehensive resistivity model. This hybrid methodology would allow for a better understanding of subsurface structures, aiding in groundwater exploration and geological mapping. The integration of clustering and interpolation can also help validate resistivity zonation results. If clustering-based classifications align with Kriging-generated resistivity distributions, confidence in the geological interpretation increases. Such cross-validation techniques ensure that subsurface models remain consistent across different analytical frameworks. For practical applications, an optimized clustering-interpolation framework could assist in hydrogeological assessments, mineral exploration, and environmental studies. The ability to delineate distinct resistivity zones with greater precision would be particularly beneficial for groundwater resource management. Future studies should focus on refining these methodologies, testing their applicability across diverse geological settings. By systematically evaluating different clustering strategies and integrating them with interpolation techniques, researchers can enhance the accuracy and reliability of geoelectrical subsurface models. 3.2.3 Integrated Interpretation of Inversion, Geostatistics, and Clustering for Groundwater Characterization The integration of geoelectrical resistivity inversion, geostatistical analysis, and clustering methods provides a comprehensive framework for characterizing subsurface hydrogeological conditions. The combined approach enhances the interpretation of resistivity variations, delineates groundwater-bearing formations, and improves the reliability of subsurface zonation. This multi-method approach ensures that geological complexity is adequately captured, reducing uncertainties in groundwater exploration and resource management. 3.2.3.1 Insights from Resistivity Inversion Resistivity inversion results reveal a stratified subsurface structure consisting of three primary zones: low, medium, and high-resistivity regions. Low-resistivity zones (< 30 Ωm) dominate the shallow subsurface, extending to depths of approximately 25 meters, and are interpreted as water-bearing formations. These zones correspond to unconsolidated pyroclastic deposits with high porosity and permeability, suggesting their potential as groundwater reservoirs. The presence of these low-resistivity regions is significant as they likely represent recharge areas where surface water infiltrates into deeper aquifers. Medium-resistivity zones (30–100 Ωm) appear as transitional layers, indicative of increasing compaction and reduced permeability with depth. These layers may represent partially weathered volcanic deposits or semi-consolidated sediments that still allow some degree of water movement. The identification of such transition zones is crucial for hydrogeological assessments, as they influence groundwater storage and transmission capabilities. High-resistivity zones (> 100 Ωm) represent compact volcanic formations, likely acting as aquitards that restrict groundwater flow. These formations are interpreted as dense basaltic lava flows or compacted pyroclastic deposits that have undergone lithification. The presence of these high-resistivity zones at greater depths suggests the existence of impermeable layers that confine deeper aquifers, potentially leading to artesian conditions. 3.2.3.2 Geostatistical Analysis and Spatial Interpolation Kriging interpolation effectively captures the spatial variability of resistivity values, providing a smoothed resistivity distribution that reduces inversion artifacts. The interpolated results align well with the inversion model, reinforcing the presence of distinct hydrogeological units. The variogram analysis reveals a moderate spatial correlation, indicating the heterogeneous nature of the subsurface. The application of Kriging interpolation allows for improved visualization of resistivity patterns, aiding in the identification of potential groundwater reservoirs and confining layers. A notable feature observed in the Kriging model is the presence of lateral variations in resistivity values, which may correspond to geological discontinuities such as faults or fractures. These structural features can play a crucial role in groundwater movement, as faults may act as either conduits or barriers depending on their nature. Identifying such features through geostatistical analysis enhances the ability to predict groundwater flow paths and assess the potential for groundwater extraction. 3.2.3.3 Clustering-Based Subsurface Zonation The clustering analysis further refines resistivity-based zonation by grouping similar resistivity values into distinct classes. K-Means clustering successfully segments the resistivity data into three main clusters, corresponding to low, moderate, and high-resistivity zones. However, its assumption of spherical clusters introduces limitations in capturing gradual geological transitions. The abrupt boundaries imposed by K-Means may not fully represent the continuous nature of subsurface resistivity variations, leading to potential misclassification of transitional zones. DBSCAN, while effective in detecting density-based clusters, struggled to differentiate multiple zones due to the relatively uniform resistivity distribution. The sensitivity of DBSCAN to parameter selection suggests that further optimization is needed to improve its performance in geological applications. Adjusting the neighborhood radius (epsilon) and the minimum points required to form a cluster could enhance its ability to detect meaningful resistivity groupings. Hierarchical clustering provided the most geologically meaningful classification by preserving gradual resistivity transitions and highlighting natural boundaries between hydrogeological units. The hierarchical dendrogram allows for multi-scale analysis, enabling researchers to examine clustering results at different levels of granularity. This flexibility is particularly useful for identifying both broad-scale lithological trends and localized variations within aquifer systems. 3.2.3.4 Hydrogeological Implications The integrated analysis confirms that the low-resistivity cluster corresponds to water-saturated formations, making it a key target for groundwater exploration. These formations are likely to host unconfined aquifers that contribute to regional groundwater recharge. The extent and thickness of these low-resistivity zones provide valuable information for designing well placement strategies and estimating sustainable yield. The transition from low to medium-resistivity zones suggests a gradual compaction of volcanic deposits, influencing groundwater flow and storage. This transition is significant because it marks the boundary between shallow, unconfined aquifers and deeper, semi-confined units. Understanding this boundary is essential for groundwater management, as it determines the connectivity between different aquifer layers and affects groundwater extraction potential. The high-resistivity zones function as impermeable layers, restricting vertical water movement and forming potential confined aquifers. These layers may serve as protective barriers that prevent contamination from surface pollutants, thereby enhancing the quality of deeper groundwater reserves. Identifying the extent of these confining units is crucial for assessing the vulnerability of groundwater systems to external influences. 3.2.3.5 Methodological Contributions By integrating resistivity inversion, geostatistical analysis, and clustering techniques, this study offers a robust framework for improving subsurface characterization. The combination of these approaches reduces uncertainty in hydrogeological assessments and enhances the delineation of groundwater reservoirs. The use of multiple methods allows for cross-validation, ensuring that the identified subsurface features are consistently represented across different analytical techniques. Future research should explore the optimization of clustering parameters and incorporate additional geophysical methods, such as seismic and electromagnetic surveys, to further refine subsurface models. Integrating borehole data and groundwater chemistry analysis could provide additional validation, strengthening the hydrogeological interpretations derived from resistivity-based studies. This multi-method approach demonstrates the effectiveness of combining traditional geophysical techniques with data-driven clustering and spatial analysis, providing a more reliable and systematic classification of groundwater-bearing formations in complex geological settings. The findings of this study contribute to the advancement of groundwater exploration methodologies, offering insights that can be applied to other volcanic terrains and similar hydrogeological environments. 4. Conclusion This study integrates geoelectrical resistivity inversion, geostatistical interpolation, and clustering-based zonation to improve subsurface characterization for groundwater exploration. The results confirm that low-resistivity zones (< 30 Ωm) correspond to water-bearing formations, highlighting their significance in groundwater recharge and storage. Geostatistical interpolation using Kriging effectively refines resistivity distributions, reducing noise and enhancing spatial interpretation. Among the clustering techniques, Hierarchical Clustering provides the most geologically meaningful classification, as it captures gradual resistivity transitions and natural zonation. The findings demonstrate that integrating clustering results with geostatistical interpolation enhances the accuracy of subsurface interpretation. The combined approach strengthens confidence in hydrogeological assessments, particularly in volcanic terrains where complex lithological variations impact groundwater distribution. Future studies should focus on refining clustering methodologies, particularly optimizing DBSCAN parameters to improve its performance in resistivity classification. Additionally, incorporating multiple geophysical techniques such as seismic refraction and magnetotellurics could further improve subsurface characterization. Borehole data and groundwater chemistry analysis would also provide valuable validation for resistivity-based interpretations. Overall, this study highlights the effectiveness of combining geophysical inversion, geostatistical modeling, and machine learning-based clustering for subsurface resistivity analysis. The methodological framework developed here contributes to the advancement of groundwater exploration and geological mapping in complex subsurface environments, offering a robust, data-driven approach for hydrogeological assessments. Declarations Author contribution : All authors contributed to the study conceptualization. Methodology: G.U.N; Formal analysis and investigation:A.M, H.B and R.F.L; Writing - original draft preparation: G.U.N; Writing - review and editing: A.M, R.F.L, H.B and Y.S; Resources: R.F.L, H.B and Y.S; Supervision: R.F.L , H.B, Y.U and Y.S Funding : No Funding are available. Data availability Available under request Ethics approval Not applicable Consent to participate Not applicable Consent to publication Not applicable Competing interest The authors declare no competing interests. References Ahmadi M (2024) Chapter 5 - Clustering. In: Ahmadi M (ed) Artificial Intelligence for a More Sustainable Oil and Gas Industry and the Energy Transition. Elsevier, pp 183–239 Audebert M, Clément R, Touze-Foltz N, et al (2014) Time-lapse ERT interpretation methodology for leachate injection monitoring based on multiple inversions and a clustering strategy (MICS). J Appl Geophy 111:320–333. https://doi.org/https://doi.org/10.1016/j.jappgeo.2014.09.024 Auken E, Guérin R, de Marsily G, Sailhac P (2009) Hydrogeophysics. Comptes Rendus - Geoscience 341:795–799. https://doi.org/10.1016/j.crte.2009.09.003 Balacco G, Fiorese GD, Alfio MR (2023) Assessment of groundwater nitrate pollution using the Indicator Kriging approach. Groundw Sustain Dev 21:100920. https://doi.org/https://doi.org/10.1016/j.gsd.2023.100920 Belyadi H, Haghighat A (2021) Chapter 4 - Unsupervised machine learning: clustering algorithms. In: Belyadi H, Haghighat A (eds) Machine Learning Guide for Oil and Gas Using Python. Gulf Professional Publishing, pp 125–168 Benabdelouahab S, Salhi A, Himi M, et al (2018a) Using resistivity methods to characterize the geometry and assess groundwater vulnerability of a Moroccan coastal aquifer. Groundw Sustain Dev 7:293–304. https://doi.org/10.1016/j.gsd.2018.07.004 Benabdelouahab S, Salhi A, Himi M, et al (2018b) Using resistivity methods to characterize the geometry and assess groundwater vulnerability of a Moroccan coastal aquifer. Groundw Sustain Dev 7:293–304. https://doi.org/10.1016/j.gsd.2018.07.004 Bernatek-Jakiel A, Kondracka M (2016) Combining geomorphological mapping and near surface geophysics (GPR and ERT) to study piping systems. Geomorphology 274:193–209. https://doi.org/10.1016/j.geomorph.2016.09.018 Binley A (2015) Tools and Techniques: Electrical Methods. Elsevier B.V. Brunet P, Clément R, Bouvier C (2010) Monitoring soil water content and deficit using Electrical Resistivity Tomography (ERT) - A case study in the Cevennes area, France. J Hydrol (Amst) 380:146–153. https://doi.org/10.1016/j.jhydrol.2009.10.032 Cardarelli E, De Donno G (2019) Advances in electric resistivity tomography: Theory and case studies. Elsevier Inc. Chandra S, Dewandel B, Dutta S, Ahmed S (2010) Geophysical model of geological discontinuities in a granitic aquifer: Analyzing small scale variability of electrical resistivity for groundwater occurrences. J Appl Geophy 71:137–148. https://doi.org/10.1016/j.jappgeo.2010.06.003 Civita M, De Maio M (2004) Assessing and mapping groundwater vulnerability to contamination: The Italian “combined” approach. Geofisica Internacional Clément R, Descloitres M, Günther T, et al (2009) Influence of shallow infiltration on time-lapse ERT: Experience of advanced interpretation. Comptes Rendus - Geoscience 341:886–898. https://doi.org/10.1016/j.crte.2009.07.005 Coulouma G, Lagacherie P, Samyn K, Grandjean G (2013) Comparisons of dry ERT, diachronic ERT and the spectral analysis of surface waves for estimating bedrock depth in various Mediterranean landscapes. Geoderma 199:128–134. https://doi.org/10.1016/j.geoderma.2012.07.026 Dietrich S, Carrera J, Weinzettel P, Sierra L (2018) Estimation of Specific Yield and its Variability by Electrical Resistivity Tomography. Water Resour Res 54:8653–8673. https://doi.org/https://doi.org/10.1029/2018WR022938 Dong Q, Chen X, Huang B (2024) Chapter 15 - Cluster analysis. In: Dong Q, Chen X, Huang B (eds) Data Analysis in Pavement Engineering. Elsevier, pp 289–301 Florsch N, Muhlach F (2018) Direct Current Electrical Methods. Everyday Applied Geophysics 1 27–103. https://doi.org/10.1016/b978-1-78548-199-4.50002-7 Frohlich RK, Parke CD (1989) The Electrical Resistivity of the Vadose Zone — Field Survey. Groundwater 27:524–530. https://doi.org/https://doi.org/10.1111/j.1745-6584.1989.tb01973.x Gance J, Malet JP, Supper R, et al (2016) Permanent electrical resistivity measurements for monitoring water circulation in clayey landslides. J Appl Geophy 126:98–115. https://doi.org/10.1016/j.jappgeo.2016.01.011 Giampaolo V, Capozzoli L, Grimaldi S, Rizzo E (2016) Sinkhole risk assessment by ERT: The case study of Sirino Lake (Basilicata, Italy). Geomorphology 253:1–9. https://doi.org/10.1016/j.geomorph.2015.09.028 Giguère R, Fradette L, Mignon D, Tanguy PA (2008) ERT algorithms for quantitative concentration measurement of multiphase flows. Chemical Engineering Journal 141:305–317. https://doi.org/10.1016/j.cej.2008.01.011 Hasan M, Shang Y, Akhter G, Jin W (2019) Application of VES and ERT for delineation of fresh-saline interface in alluvial aquifers of Lower Bari Doab, Pakistan. J Appl Geophy 164:200–213. https://doi.org/https://doi.org/10.1016/j.jappgeo.2019.03.013 Heaney M (2003) Electrical Conductivity and Resistivity Herman R (2001) An introduction to electrical resistivity in geophysics. Am J Phys 69:943–952. https://doi.org/10.1119/1.1378013 Hörning S, Gross L, Bárdossy A (2020) Geostatistical electrical resistivity tomography using random mixing. J Appl Geophy 176:104015. https://doi.org/https://doi.org/10.1016/j.jappgeo.2020.104015 Jardani A, Revil A, Dupont JP (2013) Stochastic joint inversion of hydrogeophysical data for salt tracer test monitoring and hydraulic conductivity imaging. Adv Water Resour 52:62–77. https://doi.org/10.1016/j.advwatres.2012.08.005 Javed Z, Chand S, Ahmad M (2021) Spatial statistical analysis of groundwater quality based on inverse distance weighting and ordinary kriging in District Sheikhupura, Pakistan. Desalination Water Treat 235:324–332. https://doi.org/https://doi.org/10.5004/dwt.2021.27653 Kim JH, Tsourlos P, Yi MJ, Karmis P (2014) Inversion of ERT data with a priori information using variable weighting factors. J Appl Geophy 105:1–9. https://doi.org/10.1016/j.jappgeo.2014.03.003 Kirsch R (2006) Groundwater geophysics: a tool for hydrogeology Kuras O, Pritchard JD, Meldrum PI, et al (2009) Monitoring hydraulic processes with automated time-lapse electrical resistivity tomography (ALERT). Comptes Rendus - Geoscience 341:868–885. https://doi.org/10.1016/j.crte.2009.07.010 Liang Y, Xia R, Yeh T-CJ, et al (2024) Characterizing preferential infiltration of loess using geostatistical electrical resistivity tomography. Eng Geol 340:107692. https://doi.org/https://doi.org/10.1016/j.enggeo.2024.107692 Liang Y, Yoon J-Y (2024) 2 - Fundamentals of machine learning. In: Yoon J-Y, Yu C (eds) Machine Learning and Artificial Intelligence in Chemical and Biological Sensing. Elsevier Science, pp 23–70 Lillie RJ (1988) Whole Earth Geophysics. An Introductory Textbook for Geologists and Geophysicists. 192 Liu Z, Gu Z, Liu P (2025) Chapter 9 - Clustering analysis. In: Liu Z, Gu Z, Liu P (eds) Transportation Big Data. Elsevier, pp 287–315 Mao D, Revil A, Hort RD, et al (2015) Resistivity and self-potential tomography applied to groundwater remediation and contaminant plumes: Sandbox and field experiments. J Hydrol (Amst) 530:1–14. https://doi.org/10.1016/j.jhydrol.2015.09.031 Martel R, Castellazzi P, Gloaguen E, et al (2018) ERT, GPR, InSAR, and tracer tests to characterize karst aquifer systems under urban areas: The case of Quebec City. Geomorphology 310:45–56. https://doi.org/10.1016/j.geomorph.2018.03.003 Martorana R, Capizzi P, D’Alessandro A, Luzio D (2017) Comparison of different sets of array configurations for multichannel 2D ERT acquisition. J Appl Geophy 137:34–48. https://doi.org/10.1016/j.jappgeo.2016.12.012 Masoudi MJ, Ashrafzadeh A, Khaledian M, Janatrostami S (2024) Assessment of groundwater quality for agricultural purposes in Qazvin Province, northwestern Iran: A fuzzy inference and indicator Kriging approach. Environmental and Sustainability Indicators 24:100528. https://doi.org/https://doi.org/10.1016/j.indic.2024.100528 Mazáč O, Kelly WE, Landa I (1985) A hydrogeophysical model for relations between electrical and hydraulic properties of aquifers. J Hydrol (Amst) 79:1–19. https://doi.org/https://doi.org/10.1016/0022-1694(85)90178-7 Menke W (2022) Chapter 10 - Interpolation, Gaussian process regression, and kriging. In: Menke W (ed) Environmental Data Analysis with MatLab® or Python (Third Edition). Academic Press, pp 319–348 Milsom J (2003) Field Geophysics, 3rd ed. Milsom J, Eriksen A (2011) Field Geophysics NuGraha GU, Alam BYCSSS, Nur AA, et al (2021) Vertical Electrical Sounding Exploration of Groundwater in Kertajati, Majalengka, West Java, Indonesia. Indonesian Journal on Geoscience 8:. https://doi.org/10.17014/ijog.8.3.359-369 Nugraha GU, Bakti H, Lubis RF, et al (2022a) Aquifer vulnerability in the Coastal Northern Part of Lombok Island Indonesia. Environ Dev Sustain 24:1390–1410. https://doi.org/10.1007/s10668-021-01459-0 Nugraha GU, Gaol KL, Hartanto P, Bakti H (2020) Aquifer Vulnerability: Its Protection and Management—A Case Study in Pangkalpinang City, Indonesia. International Journal of Geophysics 2020:8887914. https://doi.org/10.1155/2020/8887914 Nugraha GU, Nur AA, Pranantya PA, et al (2022b) Analysis of groundwater potential zones using Dar-Zarrouk parameters in Pangkalpinang city, Indonesia. Environ Dev Sustain. https://doi.org/10.1007/s10668-021-02103-7 Nugraha GU, Nur AA, Sudrajat Y, et al (2023) Sub-surface configuration in the northern part of Lembang groundwater basin recharge area. Appl Water Sci 13:. https://doi.org/10.1007/s13201-023-02004-2 Pehme PE (2011) Groundwater Geophysics: A Tool for Hydrogeology. Environmental and Engineering Geoscience. https://doi.org/10.2113/gseegeosci.17.1.96 Seo J, Shneiderman B (2003) Interactively Exploring Hierarchical Clustering Results. In: BEDERSON BB, SHNEIDERMAN BEN (eds) The Craft of Information Visualization. Morgan Kaufmann, San Francisco, pp 334–340 Shi Z, Wang C (2024) Near-surface imaging by joint inversion of ERT and seismic traveltime data with guided FCM clustering. J Appl Geophy 222:105332. https://doi.org/https://doi.org/10.1016/j.jappgeo.2024.105332 Singh P, Singh PN, Srivastava S, et al (2025) Chapter 11 - Appraisal of spatial interpolation techniques in predicting soil organic carbon using earth observation datasets. In: Petropoulos GP, Da Silva Fuzzo DF, Triantakonstantis D, et al. (eds) Earth Observation for Monitoring and Modeling Land Use. Elsevier, pp 265–283 Singh P, Verma P (2019) Chapter 5 - A Comparative Study of Spatial Interpolation Technique (IDW and Kriging) for Determining Groundwater Quality. In: Venkatramanan S, Prasanna MV, Chung SY (eds) GIS and Geostatistical Techniques for Groundwater Science. Elsevier, pp 43–56 Slater L, Binley A, Versteeg R, et al (2002) A 3D ERT study of solute transport in a large experimental tank. J Appl Geophy 49:211–229. https://doi.org/10.1016/S0926-9851(02)00124-6 Su L jun, Xu X qian, Geng X yu, Liang S qing (2017) An integrated geophysical approach for investigating hydro-geological characteristics of a debris landslide in the Wenchuan earthquake area. Eng Geol 219:52–63. https://doi.org/10.1016/j.enggeo.2016.11.020 Telford WM, Geldart LP, Sheriff RE (1990) Applied Geophysics - Solid Earth Geophysics Telford WM, Geldart LP, Sheriff RE (2012a) Electrical Properties of Rocks and Minerals. In: Applied Geophysics Telford WM, Geldart LP, Sheriff RE (2012b) Methods Employing Natural Electrical Sources. In: Applied Geophysics Theodoridis S, Koutroumbas K (2009) Chapter 13 - Clustering Algorithms II: Hierarchical Algorithms. In: Theodoridis S, Koutroumbas K (eds) Pattern Recognition (Fourth Edition). Academic Press, Boston, pp 653–700 Tso CHM, Kuras O, Wilkinson PB, et al (2017) Improved characterisation and modelling of measurement errors in electrical resistivity tomography (ERT) surveys. J Appl Geophy 146:103–119. https://doi.org/10.1016/j.jappgeo.2017.09.009 Tsourlos P, Papadopoulos N, Papazachos C, et al (2014) Efficient 2D inversion of long ERT sections. J Appl Geophy 105:213–224. https://doi.org/10.1016/j.jappgeo.2014.03.022 Wang TP, Chen CC, Tong LT, et al (2015) Applying FDEM, ERT and GPR at a site with soil contamination: A case study. J Appl Geophy 121:21–30. https://doi.org/10.1016/j.jappgeo.2015.07.005 Xu S, Sirieix C, Marache A, et al (2016) 3D geostatistical modeling of Lascaux hill from ERT data. Eng Geol 213:169–178. https://doi.org/https://doi.org/10.1016/j.enggeo.2016.09.009 Xu S, Sirieix C, Riss J, Malaurent P (2017) A clustering approach applied to time-lapse ERT interpretation — Case study of Lascaux cave. J Appl Geophy 144:115–124. https://doi.org/https://doi.org/10.1016/j.jappgeo.2017.07.006 Zarroca M, Bach J, Linares R, Pellicer XM (2011) Electrical methods (VES and ERT) for identifying, mapping and monitoring different saline domains in a coastal plain region (Alt Empordà, Northern Spain). J Hydrol (Amst) 409:407–422. https://doi.org/10.1016/j.jhydrol.2011.08.052 Additional Declarations No competing interests reported. Cite Share Download PDF Status: Published Journal Publication published 01 Aug, 2025 Read the published version in Discover Water → Version 1 posted Editorial decision: Revision requested 22 Apr, 2025 Reviews received at journal 22 Apr, 2025 Reviewers agreed at journal 21 Apr, 2025 Reviews received at journal 19 Apr, 2025 Reviewers agreed at journal 18 Apr, 2025 Reviewers agreed at journal 18 Apr, 2025 Reviewers agreed at journal 16 Apr, 2025 Reviewers agreed at journal 16 Apr, 2025 Reviewers agreed at journal 16 Apr, 2025 Reviewers agreed at journal 16 Apr, 2025 Reviewers agreed at journal 05 Apr, 2025 Reviewers agreed at journal 05 Apr, 2025 Reviewers agreed at journal 04 Apr, 2025 Reviewers agreed at journal 04 Apr, 2025 Reviewers invited by journal 02 Apr, 2025 Editor assigned by journal 28 Mar, 2025 Submission checks completed at journal 28 Mar, 2025 First submitted to journal 06 Mar, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6174105","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":446616633,"identity":"7f9f1ea6-3b19-4f25-9352-b9d0931fd1f5","order_by":0,"name":"Gumilar Utamas Nugraha","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA3klEQVRIiWNgGAWjYHAD5gNAQkKGGKWMDRCaLQGkhYcULTwGYJKgenkHHvMHP3Ps7Bmkez6/ulFjwcPAfvjoBnxaDA/wGDb2bktObJA5u8065xjQYTxpaTfwamngMWzg3cacwCCRu804hw2oRYLHjKCWxr/b6u0ZJHKeGef8I0KLPAOPYTPvtsOMDRI5zI9z24jQYsDMVjhbdtvxxDaJNDPm3D4JHjZCfpFvb97w8e22ant+ieTHn3O+1cnxsx8+ht+Ww1AGGxBJQBn4gXwDgs38gZDqUTAKRsEoGJkAADRbQD6TIlQXAAAAAElFTkSuQmCC","orcid":"","institution":"National Research and Innovation Agency","correspondingAuthor":true,"prefix":"","firstName":"Gumilar","middleName":"Utamas","lastName":"Nugraha","suffix":""},{"id":446616634,"identity":"7b1df2da-33b9-44a4-82d9-e5f2c96c378a","order_by":1,"name":"Hendra Bakti","email":"","orcid":"","institution":"National Research and Innovation Agency","correspondingAuthor":false,"prefix":"","firstName":"Hendra","middleName":"","lastName":"Bakti","suffix":""},{"id":446616635,"identity":"e7a68bf6-10d9-4c14-bd4c-c13ecf7a267f","order_by":2,"name":"Rachmat Fajar Lubis","email":"","orcid":"","institution":"National Research and Innovation Agency","correspondingAuthor":false,"prefix":"","firstName":"Rachmat","middleName":"Fajar","lastName":"Lubis","suffix":""},{"id":446616636,"identity":"99622584-e316-4705-8620-637902f0b581","order_by":3,"name":"Asep Mulyono","email":"","orcid":"","institution":"National Research and Innovation Agency","correspondingAuthor":false,"prefix":"","firstName":"Asep","middleName":"","lastName":"Mulyono","suffix":""},{"id":446616637,"identity":"18ac6f67-f908-44d9-8e66-868fb2055000","order_by":4,"name":"Yuniarti Ulfa","email":"","orcid":"","institution":"Geology and Mining Politechnics (Polgeta-AGP)","correspondingAuthor":false,"prefix":"","firstName":"Yuniarti","middleName":"","lastName":"Ulfa","suffix":""},{"id":446616638,"identity":"4adbdbd3-d1f4-4618-bb63-9ec15abcbde1","order_by":5,"name":"Yayat Sudrajat","email":"","orcid":"","institution":"National Research and Innovation Agency","correspondingAuthor":false,"prefix":"","firstName":"Yayat","middleName":"","lastName":"Sudrajat","suffix":""}],"badges":[],"createdAt":"2025-03-07 01:53:16","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6174105/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6174105/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1007/s43832-025-00261-7","type":"published","date":"2025-08-01T16:21:40+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":81256897,"identity":"650958f1-6829-4592-94aa-7e7d0ea5febd","added_by":"auto","created_at":"2025-04-24 05:06:57","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":2812729,"visible":true,"origin":"","legend":"\u003cp\u003eThe geology of Sunda and Tangkuban Parahu volcanic complex, West Java.\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-6174105/v1/840cf5d2e1d835acecb38de7.png"},{"id":81256888,"identity":"8faa3cd1-7d65-4d70-b807-e0c90a89f14b","added_by":"auto","created_at":"2025-04-24 05:06:57","extension":"jpeg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":148492,"visible":true,"origin":"","legend":"\u003cp\u003eStudy Area ERT Measurement(Nugraha et al. 2023)\u003c/p\u003e","description":"","filename":"2.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-6174105/v1/8e3f9ab7cb2d971add9c3f14.jpeg"},{"id":81256890,"identity":"7cf9409e-1c31-4f24-8ebf-898dc5d87c0b","added_by":"auto","created_at":"2025-04-24 05:06:57","extension":"jpeg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":684950,"visible":true,"origin":"","legend":"\u003cp\u003eElectrical Resistivity Tomography (ERT) Measurement. Dipole-dipole Configuration with 10 meters electod spacing. 550 m survey line (Nugraha et al. 2023)\u003c/p\u003e","description":"","filename":"floatimage3.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-6174105/v1/8f3a75813ec6d94225013cba.jpeg"},{"id":81256895,"identity":"4c0e26b4-427d-4193-9b53-ec1f5700a905","added_by":"auto","created_at":"2025-04-24 05:06:57","extension":"jpeg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":597070,"visible":true,"origin":"","legend":"\u003cp\u003eScatter Plot of Injected Current \u0026amp; Measure Voltage\u003c/p\u003e","description":"","filename":"floatimage4.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-6174105/v1/ba3b8859b1156d8eaa15d531.jpeg"},{"id":81256887,"identity":"e2ebad23-799c-427d-a1b0-a4e45ae7e1a3","added_by":"auto","created_at":"2025-04-24 05:06:57","extension":"jpeg","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":113765,"visible":true,"origin":"","legend":"\u003cp\u003eElectrical Resistivity Tomography (ERT) Inversion Result\u003c/p\u003e","description":"","filename":"floatimage5.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-6174105/v1/2f2cb5b237c5d79b2b4f1dca.jpeg"},{"id":81258246,"identity":"56c0b557-d84d-4309-8d07-fa88c808b4cb","added_by":"auto","created_at":"2025-04-24 05:30:57","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":145085,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eComparative Results of Different Clustering Methods\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"floatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-6174105/v1/aba72991a3bc7c8267b6c7f2.png"},{"id":81258244,"identity":"2c7789f3-8a8c-4a65-97f4-3acf87254a05","added_by":"auto","created_at":"2025-04-24 05:30:57","extension":"jpg","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":97420,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eK-Means clustering result\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"7.jpg","url":"https://assets-eu.researchsquare.com/files/rs-6174105/v1/f67a70afa19f17763b30478d.jpg"},{"id":81256894,"identity":"f0f37bdc-46fa-497b-8276-5e821e49dfa3","added_by":"auto","created_at":"2025-04-24 05:06:57","extension":"jpg","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":88685,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eDBScan Clustering Result\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"8.jpg","url":"https://assets-eu.researchsquare.com/files/rs-6174105/v1/5670ea0f4ee52dd852174b36.jpg"},{"id":81256893,"identity":"8a7a3461-c22e-4d58-9fc4-5dbc723cec7f","added_by":"auto","created_at":"2025-04-24 05:06:57","extension":"jpg","order_by":9,"title":"Figure 9","display":"","copyAsset":false,"role":"figure","size":90661,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eHierarchical Clustering\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"9.jpg","url":"https://assets-eu.researchsquare.com/files/rs-6174105/v1/796d854a6d0878e9ce4a6aad.jpg"},{"id":81256891,"identity":"48bd3453-dd6a-45cc-95cd-5168afc525a1","added_by":"auto","created_at":"2025-04-24 05:06:57","extension":"png","order_by":10,"title":"Figure 10","display":"","copyAsset":false,"role":"figure","size":73938,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eHierarchical Clustering Dendrogram\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"floatimage10.png","url":"https://assets-eu.researchsquare.com/files/rs-6174105/v1/2f7ea098310ed7e377f065a8.png"},{"id":81256906,"identity":"d3901679-e551-4f78-9321-b93fba7f6192","added_by":"auto","created_at":"2025-04-24 05:06:57","extension":"png","order_by":11,"title":"Figure 11","display":"","copyAsset":false,"role":"figure","size":464600,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eResistivity Kriging Interpolation\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"floatimage11.png","url":"https://assets-eu.researchsquare.com/files/rs-6174105/v1/4dd2c42bec2ed6fd5dfba6d9.png"},{"id":88268367,"identity":"4aa29d42-d082-4d50-9b1a-ff9492102e97","added_by":"auto","created_at":"2025-08-04 16:51:19","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":6417975,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6174105/v1/d7502446-9080-4a2e-8e3f-94fb59adfa58.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Data-Driven Resistivity Zonation: Integrating Inversion, Kriging, and Clustering to Unravel Subsurface Complexity for Groundwater Exploration","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eGeophysics is a branch of Earth sciences that applies the principles of physics to study the Earth's internal and near-surface structures geophysics. It encompasses various techniques that measure physical properties such as seismic waves, gravitational fields, magnetic anomalies, and electrical conductivity to infer geological formations and subsurface conditions(Lillie \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e1988\u003c/span\u003e; Milsom \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e2003\u003c/span\u003e; Kirsch \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2006\u003c/span\u003e; Pehme \u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e2011\u003c/span\u003e). One of the most widely used geophysical methods is electrical resistivity, which involves measuring the resistance of subsurface materials to the flow of electrical current(Nugraha et al. \u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e2020\u003c/span\u003e, \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e2022a\u003c/span\u003e, \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003eb\u003c/span\u003e). Electrical resistivity methods, commonly referred to as geoelectrical methods, are crucial for exploring subsurface structures, particularly in hydrogeology, engineering geology, and environmental studies(Herman \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2001\u003c/span\u003e; Civita and De Maio \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2004\u003c/span\u003e; Auken et al. \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2009\u003c/span\u003e; Wang et al. \u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e2015\u003c/span\u003e; Gance et al. \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Hasan et al. \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). Among these techniques, Electrical Resistivity Tomography (ERT) has gained significant popularity due to its ability to produce high-resolution two-dimensional and three-dimensional images of subsurface resistivity distributions (Herman \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2001\u003c/span\u003e; Heaney \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2003\u003c/span\u003e). By injecting electrical current into the ground through multiple electrodes and measuring the resulting voltage differences, ERT enables the characterization of geological formations, detection of groundwater reservoirs, and identification of subsurface contaminants (Maz\u0026aacute;č et al. \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e1985\u003c/span\u003e; Auken et al. \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2009\u003c/span\u003e; Jardani et al. \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2013\u003c/span\u003e). The advancement of inversion techniques and data processing methods has further improved the accuracy and applicability of ERT, making it an essential tool for subsurface investigations across various disciplines (Kuras et al. \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2009\u003c/span\u003e; Brunet et al. \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2010\u003c/span\u003e; Tso et al. \u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e2017\u003c/span\u003e; Dietrich et al. \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2018\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eElectrical resistivity tomography (ERT) is one of the most widely used geophysical methods for investigating subsurface structures(Gigu\u0026egrave;re et al. \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2008\u003c/span\u003e; Cl\u0026eacute;ment et al. \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2009\u003c/span\u003e; Coulouma et al. \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2013\u003c/span\u003e; Tsourlos et al. \u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; Bernatek-Jakiel and Kondracka \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Giampaolo et al. \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Martel et al. \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; NuGraha et al. \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Nugraha et al. \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e2022a\u003c/span\u003e). It is based on the measurement of the electrical resistivity of geological formations by injecting electrical current into the ground through a set of electrodes and recording the resulting potential differences(Frohlich and Parke \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e1989\u003c/span\u003e; Milsom and Eriksen \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2011\u003c/span\u003e). The measured resistivity values are influenced by various subsurface parameters, including lithology, porosity, water content, and mineral composition(Telford et al. \u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e1990\u003c/span\u003e, \u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e2012a\u003c/span\u003e, \u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003eb\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eERT has been extensively applied in hydrogeological studies, environmental site assessments, engineering geology, and geotechnical investigations (Gance et al. \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Su et al. \u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e2017\u003c/span\u003e; Benabdelouahab et al. \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2018a\u003c/span\u003e). In groundwater studies, ERT is used to delineate aquifers, identify contamination plumes, and assess groundwater recharge areas (Chandra et al. \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2010\u003c/span\u003e; Mao et al. \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2015\u003c/span\u003e; Benabdelouahab et al. \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2018b\u003c/span\u003e). In environmental studies, it helps in detecting pollutants and mapping subsurface waste deposits(Mao et al. \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2015\u003c/span\u003e). In geotechnical applications, ERT assists in assessing soil stability, detecting fractures, and evaluating foundation conditions (Coulouma et al. \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2013\u003c/span\u003e; Giampaolo et al. \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2016\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eDespite its versatility, one of the main challenges in using ERT is the interpretation of resistivity values, which can be ambiguous(Florsch and Muhlach \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). The same resistivity range can correspond to different geological materials under varying conditions (Florsch and Muhlach \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). For example, dry sand and fractured rock filled with air may have similar resistivity values but vastly different geological implications. Similarly, clay and highly mineralized groundwater can exhibit overlapping resistivity signatures.\u003c/p\u003e \u003cp\u003eTraditionally, ERT data interpretation has relied heavily on qualitative and semi-quantitative methods, where resistivity values are visually compared with known geological references or inferred from experience(NuGraha et al. \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Nugraha et al. \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e2022a\u003c/span\u003e). While effective to some extent, this approach is subjective, lacks reproducibility, and does not fully utilize the spatial variability of resistivity data. The increasing complexity of subsurface investigations demands more advanced and systematic approaches to enhance ERT data interpretation.\u003c/p\u003e \u003cp\u003eTo address these limitations, geostatistical methods and machine learning-based clustering techniques can be integrated into the interpretation process. Geostatistics provides a framework for analyzing the spatial correlation of resistivity values and improving interpolation accuracy(Xu et al. \u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; H\u0026ouml;rning et al. \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Liang et al. \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Meanwhile, clustering algorithms can automatically classify different subsurface zones based on resistivity distributions, reducing subjectivity in interpretation(Audebert et al. \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; Xu et al. \u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e2017\u003c/span\u003e; Shi and Wang \u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eOne of the main challenges in Electrical Resistivity Tomography (ERT) interpretation is the difficulty in defining clear boundaries between subsurface zones(Slater et al. \u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e2002\u003c/span\u003e; Zarroca et al. \u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e2011\u003c/span\u003e; Kim et al. \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; Martorana et al. \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). This issue arises due to several factors, including noise and uncertainties in resistivity measurements, which result from instrumental errors, heterogeneous subsurface properties, and inversion artifacts(Heaney \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2003\u003c/span\u003e; Telford et al. \u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e2012b\u003c/span\u003e, \u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003ea\u003c/span\u003e; Binley \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2015\u003c/span\u003e; Florsch and Muhlach \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Cardarelli and De Donno \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). Additionally, the overlapping resistivity values of different geological materials make it challenging to distinguish between formations with similar electrical properties(Telford et al. \u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e2012a\u003c/span\u003e). Furthermore, there is a lack of standardized quantitative zoning methods, as most existing approaches rely on expert judgment rather than data-driven techniques. Given these challenges, there is a growing need for a more robust and systematic approach that can effectively process ERT data, identify geological patterns, and delineate distinct subsurface zones with higher confidence.\u003c/p\u003e \u003cp\u003eGroundwater is one of the most vital natural resources, providing a crucial supply of drinking water, supporting agricultural irrigation, and sustaining various industrial activities. As surface water sources become increasingly stressed due to climate change and over-extraction, groundwater serves as a critical alternative for human consumption and economic development. Many regions rely heavily on groundwater to meet their water demands, making its sustainable management a top priority. However, the depletion and contamination of groundwater reserves pose significant environmental and socio-economic challenges, highlighting the urgent need for effective exploration and monitoring techniques.\u003c/p\u003e \u003cp\u003eAmong the various methods for groundwater exploration, Electrical Resistivity Tomography (ERT) has proven to be highly effective due to its ability to produce detailed subsurface resistivity distributions. ERT enables the identification of water-bearing formations, aquifer boundaries, and potential contamination zones by measuring electrical resistivity variations in the subsurface. This method is non-invasive, cost-effective, and provides high-resolution imaging, making it particularly useful for hydrogeological studies. By integrating ERT with advanced computational techniques, groundwater exploration can be significantly improved, leading to more accurate assessments of aquifer properties and groundwater potential.\u003c/p\u003e \u003cp\u003eTo optimize groundwater exploration, the integration of resistivity inversion, geostatistical interpolation, and clustering methods offers a powerful approach for subsurface characterization. Inversion techniques refine raw resistivity measurements into a geologically meaningful resistivity model, while geostatistical analysis provides spatial interpolation to enhance the accuracy of subsurface resistivity distribution. Clustering methods further improve subsurface zoning by classifying resistivity values into distinct geological units, reducing ambiguity in groundwater assessments. The combination of these techniques enhances the interpretation of ERT data, minimizing uncertainties and providing a more systematic framework for groundwater exploration and management.\u003c/p\u003e \u003cp\u003eThis study aims to address these challenges by integrating geostatistical interpolation and clustering techniques for automated subsurface zoning using ERT data, with a focus on groundwater exploration.The first objective is to apply geostatistical methods, particularly Kriging, to perform resistivity interpolation and construct a continuous spatial resistivity model. This will provide a smoother and more comprehensive representation of subsurface resistivity distribution. Kriging is a widely used geostatistical interpolation method that provides reliable spatial predictions and has been extensively applied across various Earth science disciplines, including hydrogeology, geophysics, and environmental studies(Javed et al. \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Balacco et al. \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Masoudi et al. \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). In subsurface investigations, Kriging effectively models spatial variability by utilizing statistical relationships between measured data points, making it a powerful tool for mapping geological formations and predicting resource distributions (Singh and Verma \u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Menke \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Singh et al. \u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). The robustness of Kriging in handling spatial uncertainty has led to its adoption in diverse geoscientific applications, such as groundwater potential assessment, mineral exploration, and soil property mapping (Singh and Verma \u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Menke \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Singh et al. \u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e2025\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe second objective is to implement clustering techniques, including K-Means, DBSCAN, and Hierarchical Clustering, to classify subsurface zones based on variations in resistivity values. K-Means is a centroid-based algorithm that partitions data into a predefined number of clusters by minimizing the variance within each cluster (Liang and Yoon \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). DBSCAN (Density-Based Spatial Clustering of Applications with Noise) identifies clusters based on the density of data points, making it particularly effective for detecting irregularly shaped geological formations and filtering out noise (Belyadi and Haghighat \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Ahmadi \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Liu et al. \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). Hierarchical Clustering builds a nested structure of clusters using either an agglomerative (bottom-up) or divisive (top-down) approach, allowing for a more flexible and interpretable clustering process (Seo and Shneiderman \u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e2003\u003c/span\u003e; Theodoridis and Koutroumbas \u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e2009\u003c/span\u003e; Ahmadi \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Dong et al. \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Liu et al. \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). These methods will help in identifying distinct geological formations and detecting patterns that are not easily recognizable through traditional qualitative interpretation.\u003c/p\u003e \u003cp\u003eThe third objective is to evaluate the effectiveness of these clustering approaches in distinguishing geological formations, ensuring that the results align with known geological features. Finally, the study aims to compare the clustering outcomes with geological references to assess their reliability in real-world groundwater exploration applications. By fulfilling these objectives, this research seeks to develop a quantitative and reproducible workflow for improving subsurface characterization using ERT data for groundwater exploration..\u003c/p\u003e \u003cp\u003eThis research holds significant contributions to geophysical and environmental studies, particularly in groundwater exploration, by enhancing the accuracy of ERT interpretation, improving geological zoning, and expanding its applications to various hydrogeological challenges. By integrating geostatistical interpolation with clustering analysis, the proposed approach offers a more precise and systematic classification of subsurface resistivity data, which is crucial for identifying groundwater-bearing formations, delineating aquifers, and assessing water quality. The automatic classification of subsurface zones using clustering techniques reduces reliance on expert judgment, thereby increasing the reproducibility and objectivity of geological interpretations. Moreover, the methodology can be applied to practical groundwater exploration applications such as locating potential aquifers, detecting contamination plumes, and evaluating groundwater recharge zones. Beyond ERT, the framework developed in this study has the potential to be extended to other geophysical datasets, including seismic surveys, gravity measurements, and electromagnetic sounding, enabling a multi-method approach for more comprehensive groundwater resource assessments. This broader applicability underscores the value of integrating advanced data-driven techniques into groundwater exploration and environmental management.\u003c/p\u003e"},{"header":"2. Materials and Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003e2.1 Geology and Hydrogeology\u003c/h2\u003e \u003cp\u003eThe oldest rock unit in the study area is a Tertiary sedimentary sequence of claystone interbedded with sandstone (Yuniardi et al., 2019). This unit is overlain by Pre-Sunda Volcanic Rocks (lava and pyroclastic deposits), followed by the Lower-Upper Pleistocene Sunda Andesite Volcanic unit, Upper Pleistocene Pyroclastic Sunda, early Holocene andesitic Tangkubanparahu, and Late Holocene Pyroclastic deposits. The stratigraphic sequence includes Alluvium (Qa), Colluvium (Qc), Sandy Tuff (Qyd), Breccia and Agglomerate (Qyb), Pumiceous Tuff (Qyt), Undifferentiated Young Volcanic Products (Qyu), Undifferentiated Old Volcanic Products (Qvu), Older Volcanic Products (Qob), Tjitalang Formation (Pt), and Kaliwangu Formation (Pk), composed of various volcanic and sedimentary deposits such as tuff, breccia, lahar, sandstone, conglomerate, marl, and limestone (Yuniardi et al., 2019; Nurfiani \u0026amp; Bouvet de Maisonneuve, 2018; Sabrian et al., 2021).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eTangkubanparahu volcano is part of the Sunda-Tangkubanparahu volcanic system, characterized by seven volcanic phases and four distinct periods of volcanism. Two major caldera-forming eruptions produced extensive pyroclastic deposits, including the Cisarua and Manglayang ignimbrites. The volcanic facies influencing groundwater occurrence include: (1) Volcanic Proximal Facies (3050\u0026ndash;3172 masl) of impermeable andesites, (2) Medial Volcanic Facies (500\u0026ndash;2076 masl) with impermeable pyroclastic deposits and fractured andesitic lava hosting 142 springs (178 l/s), and (3) Distal Volcanic Facies (100\u0026ndash;650 masl) composed of permeable volcanic deposits hosting 53 springs (700 l/s).\u003c/p\u003e \u003cp\u003eGroundwater aquifers in the study area comprise moderately to highly productive units, with variable transmissivity, deep water tables, and high spring discharge (\u0026gt;\u0026thinsp;100 l/s), though well yields remain low (\u0026lt;\u0026thinsp;5 l/s) (Delinom, 2009, 2011). The study area, Cikole-Lembang, West Bandung Regency, lies within the Lembang Groundwater Basin (95.56 km\u0026sup2;), bordered by the Bandung-Soreang, Ciater, and Sumedang groundwater basins. Lembang has a population of approximately 196,690, with tourism as a key economic sector, hosting 36 major destinations. The geological and hydrogeological characteristics of the study area have been comprehensively documented in Nugraha et al. (\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2023\u003c/span\u003e), providing an in-depth analysis of stratigraphy, aquifer properties, and groundwater flow dynamics. Their findings serve as a key reference for understanding groundwater resources in the region.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003e2.2 ERT Data Acquisition and Processing\u003c/h2\u003e \u003cp\u003eA comprehensive explanation of data acquisition techniques, processing, analysis, and interpretation of geoelectrical data, along with its hydrogeological implications, can be found in Nugraha et al. (\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). The present text provides only a summary of their study. This research focuses on an advanced geoelectrical analysis involving kriging and clustering techniques. These analyses utilize the resistivity output from the inversion process conducted by Nugraha et al. (\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2023\u003c/span\u003e).Their study employed a single Electrical Resistivity Tomography (ERT) traverse, which will be further discussed in the subsequent section. The ERT survey utilized a dipole-dipole configuration with a 10-meter electrode spacing over a total length of 550 meters. The dipole-dipole array is defined by its geometric factor (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:K\\)\u003c/span\u003e\u003c/span\u003e), which determines the apparent resistivity (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\rho\\:}_{a}\\)\u003c/span\u003e\u003c/span\u003e) measurement at each point. The geometric factor for a dipole-dipole array is given by (Binley \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2015\u003c/span\u003e; Florsch and Muhlach \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Cardarelli and De Donno \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2019\u003c/span\u003e):\u003cdiv id=\"Equ1\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ1\" name=\"EquationSource\"\u003e\n$$\\:K=\\frac{\\pi\\:an(n+1)(n+2)}{2}$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e1\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eWhere \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:a\\)\u003c/span\u003e\u003c/span\u003e = electrode spacing (10 m) and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:n\\)\u003c/span\u003e\u003c/span\u003e = dipole separation factor (varies from 1 to 6). The apparent resistivity (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\rho\\:}_{a}\\)\u003c/span\u003e\u003c/span\u003e) at each measurement point is calculated using Ohm\u0026rsquo;s Law (Herman \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2001\u003c/span\u003e; Telford et al. \u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e2012a\u003c/span\u003e):\u003cdiv id=\"Equ2\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ2\" name=\"EquationSource\"\u003e\n$$\\:{\\rho\\:}_{a}=K\\cdot\\:V/I$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e2\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eWhere \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:V\\)\u003c/span\u003e\u003c/span\u003e = measured voltage (V) and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:I\\)\u003c/span\u003e\u003c/span\u003e = injected current (A)\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eERT data were acquired using the Advanced Geosciences Inc. (AGI) SuperSting R8, an 8-channel resistivity meter (Nugraha et al. \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Direct current was injected through current electrodes, while voltage responses were recorded by potential electrodes. The survey employed a dipole-dipole array with 56 electrodes at 10-meter spacing, inserted\u0026thinsp;~\u0026thinsp;0.25 meters into the ground, selected for its high horizontal resolution in volcanic sediment aquifers (Neyamadpour et al., 2010). Measurements lasted 1.2 seconds, with two data stacks, or three if the standard deviation exceeded 2%.\u003c/p\u003e \u003cp\u003eData processing and inversion were performed using AGI EarthImager 2D (v2.4.0) (Nugraha et al. \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). A finite-element model with Smooth Model Inversion ensured stability across datasets. Apparent resistivity values were inverted using a least-squares approach to derive the true subsurface resistivity distribution.\u003c/p\u003e \u003cp\u003eThe inversion process solves for the resistivity model (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\varvec{m}\\)\u003c/span\u003e\u003c/span\u003e) by minimizing the objective function (Herman \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2001\u003c/span\u003e; Telford et al. \u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e2012a\u003c/span\u003e; Binley \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2015\u003c/span\u003e; Florsch and Muhlach \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2018\u003c/span\u003e):\u003cdiv id=\"Equ3\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ3\" name=\"EquationSource\"\u003e\n$$\\:{\\Phi\\:}\\left(\\varvec{m}\\right)=\\left|\\right|{\\varvec{W}}_{\\varvec{d}}({\\varvec{d}}_{obs}-{\\varvec{d}}_{calc})|{|}^{2}+\\lambda\\:|\\left|{\\varvec{W}}_{\\varvec{m}}\\right(\\varvec{m}-{\\varvec{m}}_{0}\\left)\\right|{|}^{2}$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e3\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eWhere \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\varvec{d}}_{obs}\\)\u003c/span\u003e\u003c/span\u003e = observed apparent resistivity data, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\varvec{d}}_{calc}\\)\u003c/span\u003e\u003c/span\u003e = calculated resistivity data from the current model, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\varvec{W}}_{\\varvec{d}}\\)\u003c/span\u003e\u003c/span\u003e = data weighting matrix, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\varvec{m}\\)\u003c/span\u003e\u003c/span\u003e = model parameters (true resistivity values),\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\varvec{m}}_{0}\\)\u003c/span\u003e\u003c/span\u003e = reference model, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\varvec{W}}_{\\varvec{m}}\\)\u003c/span\u003e\u003c/span\u003e = model weighting matrix and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\lambda\\:\\)\u003c/span\u003e\u003c/span\u003e = regularization parameter. The inversion algorithm iteratively updates the resistivity model by solving (Herman \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2001\u003c/span\u003e; Telford et al. \u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e2012a\u003c/span\u003e; Binley \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2015\u003c/span\u003e; Florsch and Muhlach \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2018\u003c/span\u003e) :\u003cdiv id=\"Equ4\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ4\" name=\"EquationSource\"\u003e\n$$\\:{\\varvec{m}}_{k+1}={\\varvec{m}}_{k}+({\\varvec{J}}^{T}{\\varvec{W}}_{\\varvec{d}}^{T}{\\varvec{W}}_{\\varvec{d}}\\varvec{J}+\\lambda\\:{\\varvec{W}}_{\\varvec{m}}^{T}{\\varvec{W}}_{\\varvec{m}}{)}^{-1}{\\varvec{J}}^{T}{\\varvec{W}}_{\\varvec{d}}^{T}({\\varvec{d}}_{obs}-{\\varvec{d}}_{calc})$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e4\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003ewhere \u003cb\u003eJ\u003c/b\u003e is the Jacobian matrix of partial derivatives. The final result is a 2D resistivity model, which provides insights into the geological structures beneath the survey area.\u003c/p\u003e \u003cp\u003eThe default \"surface\" inversion setting was applied for consistency across sites (Vogelgesang et al., 2019). A fine-model inversion was implemented, refining the mesh in near-surface areas while coarsening with depth. Resistivity values below 1 Ohm-m and above 10,000 Ohm-m were excluded as potential erroneous data. The final resistivity model achieved an average root mean square (RMS) error of 2.92% and an average L2 norm ratio of 0.9. Additionally, negative pseudo-resistivity values and non-conforming relative data exceeding 50% were removed (Vogelgesang et al., 2020).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThe computational analysis was performed on a system running Windows 11 Home Single Language (version 23H2) with an Intel(R) Core(TM) i7-12700H \"Alder Lake\" 2.30 GHz processor, 16 GB DDR5 RAM (4800 MHz), NVMe Gen 4x4 SSD, and dual graphics comprising Intel(R) Iris(R) Xe integrated GPU and NVIDIA GeForce RTX 3060 Laptop GPU. The data processing and analysis for Geostatistical interpolation and clustering techniques for subsurface zoning were conducted using Python within the Jupyter Notebook (version 7.0.8) environment, as provided in Anaconda Navigator (version 2.5.2).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003e2.3 Geostatistical Interpolation (Kriging)\u003c/h2\u003e \u003cp\u003eTo refine resistivity estimates and improve spatial resolution, Ordinary Kriging was applied. Kriging is a spatial interpolation technique that predicts unknown values based on known sample points. Kriging requires a variogram model to define spatial correlation. The experimental variogram is calculated as (Singh and Verma \u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Menke \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Balacco et al. \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2023\u003c/span\u003e):\u003cdiv id=\"Equ5\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ5\" name=\"EquationSource\"\u003e\n$$\\:\\gamma\\:\\left(h\\right)=\\frac{1}{2N\\left(h\\right)}\\sum\\:_{i=1}^{N\\left(h\\right)}\\left[Z\\right({x}_{i})-Z({x}_{i}+h){]}^{2}$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e5\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eWhere \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\gamma\\:\\left(h\\right)\\)\u003c/span\u003e\u003c/span\u003e= variogram function for lag distance \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:h\\)\u003c/span\u003e\u003c/span\u003e, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:Z\\left({x}_{i}\\right)\\)\u003c/span\u003e\u003c/span\u003e = resistivity at location \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{x}_{i}\\)\u003c/span\u003e\u003c/span\u003e and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:N\\left(h\\right)\\)\u003c/span\u003e\u003c/span\u003e = number of data pairs at distance \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:h\\)\u003c/span\u003e\u003c/span\u003e. A best-fit theoretical model (e.g., spherical, exponential, Gaussian) was selected based on the variogram shape. The resistivity at an unknown location (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{Z}^{*}\\left(x\\right)\\)\u003c/span\u003e\u003c/span\u003e) is estimated using:\u003cdiv id=\"Equ6\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ6\" name=\"EquationSource\"\u003e\n$$\\:{Z}^{*}\\left(x\\right)=\\sum\\:_{i=1}^{N}{\\lambda\\:}_{i}Z\\left({x}_{i}\\right)$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e6\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003ewhere \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\lambda\\:}_{i}\\)\u003c/span\u003e\u003c/span\u003e are Kriging weights, computed by solving:\u003cdiv id=\"Equ7\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ7\" name=\"EquationSource\"\u003e\n$$\\:\\sum\\:_{j=1}^{N}{\\lambda\\:}_{j}\\gamma\\:({x}_{i},{x}_{j})+\\mu\\:=\\gamma\\:({x}_{i},x)$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e7\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003ewhere \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\mu\\:\\)\u003c/span\u003e\u003c/span\u003e is the Lagrange multiplier ensuring unbiased estimates.\u003c/p\u003e \u003cp\u003eUsing the optimized variogram, Kriging interpolation was performed to generate a smoothed resistivity distribution. This process helps in reducing inversion artifacts and providing a clearer visualization of geological structures.\u003c/p\u003e \u003cp\u003eThis study applies Ordinary Kriging (OK) to interpolate electrical resistivity data collected from geoelectrical surveys. The dataset consists of spatial coordinates (\u003cb\u003eX, Y\u003c/b\u003e) and corresponding resistivity values (\u003cb\u003eZ\u003c/b\u003e), which are used to construct a continuous resistivity model. The primary goal of this approach is to enhance subsurface characterization by generating a high-resolution resistivity distribution while optimizing computational efficiency.\u003c/p\u003e \u003cp\u003eTo achieve this, a structured interpolation grid is defined with \u003cb\u003e30 \u0026times; 30\u003c/b\u003e resolution points, reducing the computational burden compared to a finer mesh. The OrdinaryKriging function from the PyKrige library is utilized, with a spherical variogram model, which is commonly used for modeling subsurface properties with spatial continuity. The interpolation is performed in point mode, where only the 20 closest data points contribute to each prediction, ensuring computational efficiency and preventing overfitting.\u003c/p\u003e \u003cp\u003eThe interpolated resistivity values are then visualized using a contour plot, which provides a clear representation of the spatial resistivity variations. The original data points are overlaid on the contour plot to assess the accuracy of the interpolation, ensuring that key subsurface resistivity patterns are preserved. A viridis colormap is applied to highlight resistivity variations, while a coolwarm colormap is used for the original data points to distinguish measured values from interpolated estimates.\u003c/p\u003e \u003cp\u003eThe application of n_closest_points and the loop-based backend minimizes memory usage and improves interpolation performance, making it suitable for large-scale geophysical studies. The output is saved as a high-resolution image to facilitate further analysis and integration with other geophysical datasets.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003e2.4 Clustering Techniques for Subsurface Zoning\u003c/h2\u003e \u003cp\u003eTo classify different subsurface zones based on resistivity variations, three clustering techniques were applied: K-Means, DBSCAN, and Hierarchical Clustering. These methods allow for a more systematic and data-driven approach to geological zoning.\u003c/p\u003e \u003cp\u003eThe dataset is first loaded from a CSV file containing geophysical survey results with columns X, Depth, and Resistivity. The data is structured into a NumPy array for computational efficiency. The three selected attributes are used as input features for clustering, ensuring that both spatial (X, Depth) and physical (Resistivity) parameters are considered in the classification. This preprocessing step ensures that clustering is performed based on the combined influence of location and resistivity variations.\u003c/p\u003e \u003cdiv id=\"Sec7\" class=\"Section3\"\u003e \u003ch2\u003e2.4.1 K-Means Clustering\u003c/h2\u003e \u003cp\u003eK-Means is a partitioning algorithm that groups data points into K clusters based on their similarity(Menke \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Ahmadi \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Dong et al. \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Liu et al. \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). The algorithm follows these steps: Select K initial centroids randomly, Assign each data point to the nearest centroid based on Euclidean distance, Recalculate cluster centroids iteratively until convergence.\u003c/p\u003e \u003cp\u003eThe optimal number of clusters (K) was determined using the Elbow Method and Silhouette Score.\u003c/p\u003e \u003cp\u003eClusters were determined by minimizing intra-cluster variance:\u003cdiv id=\"Equ8\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ8\" name=\"EquationSource\"\u003e\n$$\\:J=\\sum\\:_{i=1}^{K}\\sum\\:_{j\\in\\:{C}_{i}}||{x}_{j}-{\\mu\\:}_{i}|{|}^{2}$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e8\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003ewhere \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{C}_{i}\\)\u003c/span\u003e\u003c/span\u003e is the set of points in cluster ii, and \u0026micro;i\\mu_i is the centroid.\u003c/p\u003e \u003cp\u003eThis study applies K-Means clustering to classify subsurface zones based on electrical resistivity data collected from geoelectrical surveys. The dataset comprises spatial coordinates (X, Depth) and Resistivity values, which serve as the input parameters for clustering. The primary objective is to identify distinct resistivity zones, which may correspond to different geological formations or hydrogeological conditions.\u003c/p\u003e \u003cp\u003eThe clustering process is conducted using the K-Means algorithm, a widely used unsupervised learning technique for partitioning data into a predefined number of clusters. In this study, the number of clusters is set to three (k\u0026thinsp;=\u0026thinsp;3), representing three distinct resistivity zones. The algorithm iteratively assigns data points to clusters based on their similarity in resistivity values and spatial distribution, ensuring that subsurface zoning is objectively classified.\u003c/p\u003e \u003cp\u003eThe resulting cluster assignments are visualized using a scatter plot, where data points are color-coded according to their respective clusters. A viridis colormap is applied to enhance the visual distinction between zones, and a color bar is included to indicate the cluster labels. The X-axis represents horizontal distance, while the Y-axis corresponds to depth, providing a clear spatial representation of the identified zones.\u003c/p\u003e \u003cp\u003eThe final output, saved as \"Zonation Based on Resistivity.png\", provides valuable insights into the spatial distribution of subsurface resistivity variations. This clustering-based approach reduces subjectivity in geological interpretation and enhances the reproducibility of resistivity-based zoning. The ability to automatically segment resistivity data into meaningful zones improves subsurface characterization for applications such as groundwater exploration, contamination mapping, and geotechnical assessments.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec8\" class=\"Section3\"\u003e \u003ch2\u003e2.4.2 DBSCAN (Density-Based Spatial Clustering of Applications with Noise)\u003c/h2\u003e \u003cp\u003eDBSCAN is a density-based clustering method that identifies high-density regions and separates outliers (Belyadi and Haghighat \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Liang and Yoon \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). It operates as follows: Defines a neighborhood radius (\u003cb\u003eε\u003c/b\u003e) and minimum number of points (MinPts) and Groups points with high local density into clusters also Identifies low-density points as noise (outliers).\u003c/p\u003e \u003cp\u003eDBSCAN is particularly useful for detecting anomalous geological zones, such as faulted or fractured areas. Unlike K-Means, DBSCAN (Density-Based Spatial Clustering of Applications with Noise) does not require a predefined number of clusters. Instead, it relies on two key parameters: eps (ε) \u0026mdash; the neighborhood radius around a point, min_samples \u0026mdash; the minimum number of points required to form a cluster\u003c/p\u003e \u003cp\u003eDBSCAN identifies clusters based on density, making it effective for detecting irregular geological formations and removing noise from the dataset. It assigns core points to dense regions, border points to cluster edges, and noise points as outliers. In this study, eps\u0026thinsp;=\u0026thinsp;10 and min_samples\u0026thinsp;=\u0026thinsp;5 are chosen based on data scale and distribution. The resulting clusters are visualized in a scatter plot, with different colors indicating different density-based clusters. Unlike K-Means, DBSCAN can detect zones of varying shapes, making it useful for identifying complex resistivity structures.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec9\" class=\"Section3\"\u003e \u003ch2\u003e2.4.3 Hierarchical Clustering\u003c/h2\u003e \u003cp\u003eHierarchical clustering builds a tree-like structure (dendrogram) to show relationships between resistivity zones(Seo and Shneiderman \u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e2003\u003c/span\u003e; Theodoridis and Koutroumbas \u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e2009\u003c/span\u003e). The process includes: Computing a distance matrix between all data points, Iteratively merging the closest clusters using linkage criteria (e.g., Ward\u0026rsquo;s method) and Cutting the dendrogram at an optimal level to define geological zones. This method provides a hierarchical view of subsurface formations, offering insights into the relative relationships between different resistivity zones.\u003c/p\u003e \u003cp\u003eHierarchical clustering is performed using the Agglomerative Clustering method with Ward linkage, which minimizes the variance within clusters during merging. Unlike K-Means and DBSCAN, Hierarchical Clustering does not require a predefined number of clusters, and the final grouping is determined through a dendrogram visualization. The steps include:\u003c/p\u003e \u003cp\u003e \u003col\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003eComputing the pairwise distance between all data points.\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003eIteratively merging the closest clusters based on the Ward variance minimization criterion.\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003eConstructing a dendrogram, which shows the hierarchical relationships between clusters.\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003c/ol\u003e \u003c/p\u003e \u003cp\u003eThe dendrogram helps determine the optimal number of clusters by identifying natural splits in the hierarchical tree. In this study, three clusters are selected for visualization. The results are plotted similarly to K-Means and DBSCAN, showing distinct resistivity-based zones.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003e2.5 Evaluation of Clustering Results\u003c/h2\u003e \u003cp\u003eTo assess the performance of clustering techniques, the following evaluation metrics were used:Silhouette Score Measures how well each point fits within its assigned cluster.\u003cdiv id=\"Equ9\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ9\" name=\"EquationSource\"\u003e\n$$\\:S=\\frac{b-a}{max(a,b)}$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e9\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003ewhere a is intra-cluster distance and b is nearest-cluster distance.\u003c/p\u003e \u003c/div\u003e"},{"header":"3. Results and Discussion","content":"\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003e3.1. Results\u003c/h2\u003e \u003cdiv id=\"Sec13\" class=\"Section3\"\u003e \u003ch2\u003e3.1.1 2D Resistivity Inversion Results\u003c/h2\u003e \u003cp\u003eThe following is a summary of the ERT results from previous study conducted by Nugraha et al. (\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). The ERT survey was carried out along a 540-meter east-west oriented line with a dipole-dipole electrode configuration and a 10-meter electrode spacing. The inversion results using EarthImager 2D software indicate a maximum modelable depth of 103 meters, revealing three distinct resistivity zones: low, medium, and high(Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe low resistivity zone (123\u0026ndash;292 Ωm), represented by dark to light blue colors, is observed at depths of 0\u0026ndash;25 meters across the entire section, with greater thickness on the eastern side. The medium resistivity zone (293\u0026ndash;700 Ωm), colored green to yellow, extends from 13 to 77 meters and from 77 to 103 meters, lying beneath the low-resistivity layer. It is thicker in the east and thinner in the west. The high resistivity zone (701\u0026ndash;3875 Ωm) is concentrated in the middle of the cross-section, extending from 27 to 77 meters, with a thickness of about 50 meters. This layer is thicker in the west and thins towards the east.\u003c/p\u003e \u003cp\u003eLithologically, the low and upper-medium resistivity zones correspond to the Tangkubanparahu Pyroclastic Fall 2 (Tjp2), composed of scoria, pumice, and lithic fragments of andesite and basalt, typically loose to consolidated, well-stratified, and weathered brownish-yellow. The high rainfall in the study area, particularly during the rainy season when measurements were conducted, likely contributes to the low resistivity values due to increased water infiltration. The high-resistivity layer is associated with Tangkubanparahu lava (Tl), which consists of basaltic lava flows containing plagioclase, pyroxene, and minor olivine. These lava flows originate from Tangkubanparahu\u0026rsquo;s main crater and are widely exposed in areas such as Cisarua, Sagalaherang, and northern Bandung.\u003c/p\u003e \u003cp\u003eA medium-resistivity layer in the ERT cross-section is interpreted as Tangkubanparahu Pyroclastic Fall 1 (Tjp1), predominantly composed of pumice (1\u0026ndash;5 cm), scoria, and basaltic-andesitic lithic fragments, with occasional volcanic bombs. This unit, characterized by thin layers up to 1 meter thick, has a resistivity of 300\u0026ndash;700 Ωm and a thickness of 71 meters in the west, thinning towards the east. The Sunda Pyroclastic Flow Deposits (Sap) and Sunda Lava (S1) were not detected, likely due to the inversion depth limit of 103 meters, whereas these deposits are estimated to be deeper than 150 meters.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section3\"\u003e \u003ch2\u003e3.1.2 Comparative Analysis of Clustering Methods\u003c/h2\u003e \u003cp\u003eTo further analyze the spatial distribution of resistivity values, three different clustering methods were applied: K-Means Clustering, DBSCAN (Density-Based Spatial Clustering of Applications with Noise) and Hierarchical Clustering (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e)\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cdiv id=\"Sec15\" class=\"Section4\"\u003e \u003ch2\u003e\u003cem\u003e3.\u003c/em\u003e1.\u003cem\u003e2.1 K-Means Clustering\u003c/em\u003e\u003c/h2\u003e \u003cp\u003eThe K-Means clustering method was employed to classify subsurface resistivity distributions into distinct zones. The resulting zonation is depicted in Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003e, where different colors represent different clusters. The K-Means algorithm effectively segmented the resistivity data into three distinct clusters. This clustering technique partitions the dataset based on similarity, ensuring that points within a cluster exhibit comparable resistivity values. The spatial contiguity of the resulting clusters suggests that regions with similar resistivity distributions are systematically grouped, thereby enhancing geological interpretation by delineating subsurface zones with consistent electrical properties.\u003c/p\u003e \u003cp\u003eThe clustering analysis categorized the resistivity data into three distinct groups, designated as Cluster 0, Cluster 1, and Cluster 2. These clusters correspond to low, moderate, and high-resistivity zones, as previously identified through resistivity inversion.\u003c/p\u003e \u003cp\u003e \u003cul\u003e \u003cli\u003e \u003cp\u003e \u003cb\u003eCluster 2 (yellow)\u003c/b\u003e predominantly occupies the shallow subsurface, extending from the surface to approximately 20 meters in depth. The low resistivity values observed in this cluster suggest the presence of water-saturated formations, which may indicate aquifers or weathered volcanic deposits. The extensive distribution of Cluster 2 in the uppermost layers underscores its significance for groundwater exploration, as these formations likely comprise porous and permeable materials conducive to groundwater storage and transmission.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003e \u003cb\u003eCluster 1 (teal)\u003c/b\u003e is primarily observed between 20 and 40 meters depth, suggesting a transitional zone between the water-bearing formations above and the more compact lithologies below. This transition zone may correspond to increasing compaction with depth or a shift from unconsolidated to semi-consolidated deposits.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003e \u003cb\u003eCluster 0 (dark purple)\u003c/b\u003e becomes more prominent at depths of approximately 30 meters and beyond. These clusters are associated with higher resistivity values, indicative of compacted or less permeable rock formations. The presence of Cluster 0 at greater depths aligns with high-resistivity zones, likely representing volcanic rock formations. These formations may function as barriers to groundwater flow due to their lower porosity and permeability.\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e \u003c/p\u003e \u003cp\u003eThe clustering results provide a refined segmentation of the subsurface, facilitating enhanced geological interpretation. The classification of resistivity data into meaningful clusters contributes to a more comprehensive understanding of lithological variations at different depths.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec16\" class=\"Section4\"\u003e \u003ch2\u003e\u003cem\u003e3.\u003c/em\u003e1.\u003cem\u003e2.2. DBSCAN Clustering\u003c/em\u003e\u003c/h2\u003e \u003cp\u003eThe DBSCAN (Density-Based Spatial Clustering of Applications with Noise) algorithm was applied to the resistivity dataset but resulted in a single dominant cluster (Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003e). This suggests that the majority of data points were classified into the same category, limiting its ability to differentiate geological units. Unlike hierarchical clustering, which provided a structured classification, DBSCAN failed to identify distinct subsurface zones in this case.\u003c/p\u003e \u003cp\u003eDBSCAN is particularly useful for detecting clusters of varying density and identifying noise points. However, in this analysis, no significant noise points were detected, indicating that the resistivity dataset is relatively homogeneous. The effectiveness of DBSCAN largely depends on parameter selection, particularly the neighborhood radius (epsilon) and the minimum number of points required to form a cluster. If these parameters are not optimally set, the algorithm may fail to detect meaningful sub-clusters.\u003c/p\u003e \u003cp\u003eThe results indicate that the chosen parameter settings may have been too restrictive, preventing the formation of distinct clusters. Adjusting these parameters could potentially improve DBSCAN\u0026rsquo;s ability to identify subsurface zones with different resistivity characteristics.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec17\" class=\"Section4\"\u003e \u003ch2\u003e\u003cem\u003e3.\u003c/em\u003e1.\u003cem\u003e2.3. Hierarchical Clustering\u003c/em\u003e\u003c/h2\u003e \u003cp\u003eHierarchical clustering provided a structured representation of subsurface resistivity distribution, successfully identifying three distinct clusters (Fig.\u0026nbsp;\u003cspan refid=\"Fig9\" class=\"InternalRef\"\u003e9\u003c/span\u003e). Unlike the K-Means method, hierarchical clustering does not require a predefined number of clusters, allowing for a more flexible and adaptive interpretation. The clustering results were further validated using a dendrogram, as shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig10\" class=\"InternalRef\"\u003e10\u003c/span\u003e. This visualization confirms the presence of natural resistivity groupings, supporting the notion that hierarchical clustering provides a geologically meaningful classification.\u003c/p\u003e \u003cp\u003eTo classify subsurface resistivity patterns, a hierarchical clustering analysis was conducted, and the resulting dendrogram is presented in Fig.\u0026nbsp;\u003cspan refid=\"Fig10\" class=\"InternalRef\"\u003e10\u003c/span\u003e. The analysis provided a structured classification of resistivity data, delineating distinct subsurface zones based on their electrical properties. In the dendrogram, the y-axis represents the Euclidean distance, which serves as a measure of dissimilarity between clusters. Higher values on the y-axis indicate greater differences between clusters, whereas lower values suggest closely related data points. The x-axis displays individual resistivity data points, arranged according to their clustering structure, providing an intuitive understanding of similarities and differences in resistivity values across regions.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThree primary clusters were identified in the analysis. Each cluster is represented by a distinct color in the dendrogram\u0026mdash;blue, green, and orange\u0026mdash;indicating separate resistivity groupings. The presence of these clusters suggests significant variations in subsurface electrical properties. A notable observation in the dendrogram is a major separation at approximately 30,000 on the y-axis. This significant dissimilarity, or highest linkage, indicates that the dataset comprises two dominant resistivity groups with fundamentally different characteristics. Near the lower part of the dendrogram, smaller branches indicate finer subdivisions within each major cluster. These sub-clusters highlight internal variability within the resistivity zones, suggesting localized changes in porosity, mineral composition, or degrees of weathering.\u003c/p\u003e \u003cp\u003eThe clustering results align closely with the resistivity zonation map presented in Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e, reinforcing the validity of the classification approach. In the resistivity zonation map, distinct regions were identified based on interpolation techniques. These regions exhibit clear boundaries, which correspond to the clustering patterns observed in the dendrogram.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec18\" class=\"Section3\"\u003e \u003ch2\u003e3.1.3 Geostatistical Analysis Using Kriging Interpolation\u003c/h2\u003e \u003cp\u003eThe Kriging interpolation method was applied to estimate the spatial distribution of resistivity values based on the measured data points. The result is shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig11\" class=\"InternalRef\"\u003e11\u003c/span\u003e, where the resistivity distribution is interpolated over the entire survey area.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec19\" class=\"Section3\"\u003e \u003ch2\u003e3.7.1 Interpretation of Kriging Results\u003c/h2\u003e \u003cp\u003eKriging interpolation provides a detailed representation of resistivity variations across the study area, offering insights into subsurface geological structures. The original resistivity measurement points, represented by white and blue dots, indicate sampled values at specific depths.\u003c/p\u003e \u003cp\u003eThe color gradient in the Kriging interpolation map reflects resistivity distribution. Low resistivity values appear in dark blue to purple, whereas high resistivity zones are depicted in yellow to red. This gradient helps distinguish different subsurface formations. A notable high-resistivity anomaly exceeding 2000 Ω.m is observed between X\u0026thinsp;=\u0026thinsp;250\u0026ndash;400 meters at depths of 30\u0026ndash;50 meters.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec20\" class=\"Section2\"\u003e \u003ch2\u003e3.2. Discussion\u003c/h2\u003e \u003cdiv id=\"Sec21\" class=\"Section3\"\u003e \u003ch2\u003e3.2.1. Clustering Methods\u003c/h2\u003e \u003cdiv id=\"Sec22\" class=\"Section4\"\u003e \u003ch2\u003e\u003cem\u003e3.\u003c/em\u003e2.1.\u003cem\u003e1 K-Means Clustering\u003c/em\u003e\u003c/h2\u003e \u003cp\u003eA key advantage of the K-Means algorithm is its computational efficiency, which makes it a widely adopted method for clustering large datasets. Its capability to produce clear segmentation facilitates a straightforward analysis of subsurface resistivity variations. However, K-Means operates under the assumption that clusters are spherical and exhibit equal variance. This assumption may introduce inaccuracies when applied to geological formations characterized by irregular geometries and complex boundaries.\u003c/p\u003e \u003cp\u003eGeological structures often exhibit gradual resistivity transitions rather than discrete boundaries. Consequently, the assumption of spherical clusters may not adequately capture the inherent variability of subsurface formations, potentially affecting the reliability of interpretations based on K-Means clustering.\u003c/p\u003e \u003cp\u003eThe predominance of low-resistivity zones in the shallow subsurface supports the hypothesis that these layers contain saturated sediments or weathered volcanic deposits with high porosity, favorable for groundwater accumulation. At intermediate depths, Cluster 1 (teal) represents a transitional zone with moderate resistivity values, indicative of variable porosity and permeability, suggesting the presence of a semi-confined aquifer.\u003c/p\u003e \u003cp\u003eThe transition from Cluster 2 to Cluster 1 at depths of 20\u0026ndash;40 meters marks a change in subsurface conditions, potentially corresponding to increasing compaction, reduced permeability, and altered groundwater flow dynamics. The deepest layers, classified as Cluster 0 (dark purple), are associated with high-resistivity values, likely representing dense, compacted volcanic formations or bedrock, which function as an aquitard, restricting vertical water movement.\u003c/p\u003e \u003cp\u003eWhile K-Means clustering provides a useful classification of resistivity data, its reliance on predefined cluster numbers and its sensitivity to geometric assumptions suggest that integrating alternative clustering methods, such as hierarchical clustering, may enhance the robustness of geological interpretations.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec23\" class=\"Section4\"\u003e \u003ch2\u003e\u003cem\u003e3.\u003c/em\u003e2.1.\u003cem\u003e2. DBSCAN Clustering\u003c/em\u003e\u003c/h2\u003e \u003cp\u003eOne of DBSCAN\u0026rsquo;s strengths is its ability to identify clusters of arbitrary shapes, making it suitable for datasets with complex geological structures. However, its sensitivity to parameter tuning presents a challenge in applications where optimal values are not easily determined. The failure to detect multiple clusters in this study suggests that DBSCAN may not be the most suitable method for classifying resistivity data in the given geological setting.\u003c/p\u003e \u003cp\u003eIn contrast to hierarchical clustering, which effectively captured gradual resistivity variations, DBSCAN\u0026rsquo;s reliance on density thresholds may not be well-suited for datasets where resistivity transitions occur progressively rather than in distinct, high-density clusters. Future studies should explore parameter optimization techniques, such as automated epsilon selection, to enhance DBSCAN\u0026rsquo;s performance in geological applications. Additionally, integrating DBSCAN with other clustering methods could provide a more comprehensive analysis of resistivity data.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec24\" class=\"Section4\"\u003e \u003ch2\u003e\u003cem\u003e3.\u003c/em\u003e2.1.\u003cem\u003e3. Hierarchical Clustering\u003c/em\u003e\u003c/h2\u003e \u003cp\u003eA key advantage of hierarchical clustering is its ability to represent the natural progression of resistivity variations. This characteristic is especially beneficial for geological studies, where subsurface changes typically occur gradually rather than abruptly. The hierarchical structure enables researchers to examine clustering results at different levels of granularity, making it useful for both broad-scale and detailed subsurface analyses.\u003c/p\u003e \u003cp\u003eHierarchical clustering has demonstrated its capability to reveal resistivity transitions, suggesting that it may be more suitable for geological applications compared to K-Means, which tends to impose rigid cluster boundaries. The clustering approach enhances the interpretation of resistivity data by revealing patterns that may not be immediately apparent in conventional interpolation maps. By grouping similar resistivity values, the analysis provides a more structured representation of subsurface properties.\u003c/p\u003e \u003cp\u003eHigh-resistivity zones form a distinct and separate cluster in the analysis, suggesting that these areas are geologically unique and potentially represent dense volcanic rocks or compact lithological formations. High-resistivity zones are often associated with materials such as fresh basaltic lava flows or consolidated rock formations with minimal water content. Their classification as an independent cluster indicates their distinct physical properties. Conversely, the low-resistivity regions form a separate cluster, likely associated with water-saturated layers or clay-rich formations. These materials typically exhibit high conductivity, resulting in lower resistivity values. The presence of a well-defined low-resistivity cluster suggests a clear distinction between water-bearing units and surrounding geological formations, which is crucial for hydrogeological assessments and groundwater exploration.\u003c/p\u003e \u003cp\u003eA significant implication of hierarchical clustering in geoelectrical studies is the clear separation between clusters, suggesting the presence of distinct lithological units within the study area, corresponding to different geological formations. The clustering results support previous interpretations derived from resistivity inversion techniques. The agreement between the method reinforces confidence in the identified resistivity zones.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec25\" class=\"Section4\"\u003e \u003ch2\u003e\u003cem\u003e3.\u003c/em\u003e2.1.\u003cem\u003e4. Implications for Subsurface Resistivity Zonation\u003c/em\u003e\u003c/h2\u003e \u003cp\u003eThe comparison of these clustering methods highlights their respective advantages and limitations in subsurface resistivity analysis. Each technique offers unique insights, but their applicability depends on the characteristics of the dataset.\u003c/p\u003e \u003cp\u003eK-Means provides clear segmentation of resistivity zones, making it a suitable choice for cases where well-defined clusters are expected. However, its assumption of spherical clusters limits its effectiveness in complex geological formations.\u003c/p\u003e \u003cp\u003eDBSCAN, while effective for detecting clusters of varying density, struggled to differentiate multiple clusters in this study. The sensitivity of its parameters suggests that further optimization is necessary for meaningful classification.\u003c/p\u003e \u003cp\u003eHierarchical clustering emerged as the most geologically meaningful approach, as it successfully captured gradual resistivity transitions rather than enforcing rigid boundaries. This aligns better with the nature of geological formations.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec26\" class=\"Section3\"\u003e \u003ch2\u003e3.2.2. Geostatistical Analysis\u003c/h2\u003e \u003cp\u003eThis feature (Fig.\u0026nbsp;\u003cspan refid=\"Fig11\" class=\"InternalRef\"\u003e11\u003c/span\u003e) is highlighted in yellow and suggests the presence of a compact volcanic formation or a dry, resistive layer. Conversely, low-resistivity zones (\u0026lt;\u0026thinsp;500 Ω.m) dominate the upper layers. These regions, shown in dark blue, likely indicate water-bearing formations or weathered volcanic deposits, which are commonly associated with high moisture content or clay materials. The spatial distribution of resistivity values suggests a stratified subsurface structure, where resistive and conductive layers alternate at different depths. The observed transitions provide valuable information for geological and hydrogeological interpretation.\u003c/p\u003e \u003cdiv id=\"Sec27\" class=\"Section4\"\u003e \u003ch2\u003e3.2.2.1. Correlation with Resistivity Inversion and Clustering\u003c/h2\u003e \u003cp\u003eThe high-resistivity anomaly identified through Kriging interpolation aligns with the resistivity inversion results presented in Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e. This consistency confirms the presence of a compact and resistive subsurface layer. Additionally, clustering analysis (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e) highlights this region as a distinct class. The agreement between clustering and Kriging results reinforces the uniqueness of this geological feature. A smooth transition is observed between resistive and conductive zones, which follows a gradual gradient. This pattern suggests that Kriging effectively captures spatial variations in resistivity, reducing abrupt discontinuities in the interpolation model.By integrating resistivity inversion, Kriging interpolation, and clustering, a more comprehensive understanding of subsurface characteristics is achieved. These combined methods validate the presence of geologically distinct units and improve interpretation accuracy.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec28\" class=\"Section4\"\u003e \u003ch2\u003e3.2.2.2. Geostatistical Model and Variogram Analysis\u003c/h2\u003e \u003cp\u003eThe Kriging interpolation was conducted using a Gaussian variogram model, which effectively describes the spatial continuity of resistivity data. This model ensures a geologically consistent interpolation result.\u003c/p\u003e \u003cp\u003eThe general form of the variogram equation used in the analysis is:\u003cdiv id=\"Equ10\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ10\" name=\"EquationSource\"\u003e\n$$\\:\\gamma\\:\\left(h\\right)={C}_{0}+C(1-{e}^{-h/a})$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e10\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003ewhere \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\gamma\\:\\left(h\\right)\\)\u003c/span\u003e\u003c/span\u003e represents the semi-variance at distance \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:h\\)\u003c/span\u003e\u003c/span\u003e, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{C}_{0}\\)\u003c/span\u003e\u003c/span\u003e is the nugget effect accounting for measurement errors and small-scale variability, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:C\\)\u003c/span\u003e\u003c/span\u003e is the sill, representing the maximum variance, and aa is the range, indicating the distance at which spatial correlation diminishes. By optimizing these variogram parameters, Kriging generates a smooth and geologically meaningful resistivity interpolation. The careful selection of parameters ensures that the interpolated resistivity values accurately reflect underlying geological structures. The nugget effect (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{C}_{0}\\)\u003c/span\u003e\u003c/span\u003e) accounts for random noise and microscale variations that cannot be captured by the main interpolation model. A low nugget value suggests high data quality and minimal measurement errors. The sill (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:C\\)\u003c/span\u003e\u003c/span\u003e) represents the maximum variance observed in the dataset. In this case, the sill value helps determine the overall contrast between high- and low-resistivity zones, providing insights into geological heterogeneity. The range (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:a\\)\u003c/span\u003e\u003c/span\u003e) is particularly important as it defines the spatial correlation limit. Beyond this distance, resistivity values become statistically independent, indicating distinct geological formations. The Gaussian variogram model was selected due to its smooth transition properties. This model is particularly suitable for subsurface resistivity data, where gradual geological changes are expected.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec29\" class=\"Section4\"\u003e \u003ch2\u003e3.2.2.3. Key Insights from Kriging Interpolation\u003c/h2\u003e \u003cp\u003eThe identification of a high-resistivity anomaly across multiple analytical methods suggests a geologically significant feature. This consistency strengthens the interpretation of subsurface structures. Kriging interpolation provides a continuous representation of resistivity variations, reducing uncertainties commonly associated with discrete measurement points. The method enhances spatial understanding by filling data gaps with statistically optimized values. The integration of Kriging results with clustering-based zonation confirms the existence of distinct hydrogeological units. This cross-validation improves confidence in geological interpretations. The combination of geostatistical and geophysical techniques allows for a more robust subsurface characterization. By leveraging multiple methods, the study provides a comprehensive assessment of resistivity variations.Overall, Kriging serves as an essential tool for geoelectrical interpretation, offering a refined approach to resistivity mapping and enhancing the reliability of subsurface models.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec30\" class=\"Section4\"\u003e \u003ch2\u003e3.2.2.4. Comparison with Geostatistical Analysis\u003c/h2\u003e \u003cp\u003eThe clustering results show a strong correlation with Kriging interpolation, further validating the spatial distribution of resistivity values. The areas identified as low-resistivity zones through Kriging were consistently classified within Cluster 2. Similarly, the resistivity inversion model reinforces the clustering-based zonation, demonstrating the robustness of this classification approach. The consistency between multiple analytical methods enhances confidence in the geological interpretation. The integration of geostatistical analysis and clustering techniques provides a comprehensive framework for subsurface characterization. By combining both approaches, uncertainties in resistivity interpretation can be minimized.The ability of clustering to segment resistivity data into meaningful groups complements the continuous nature of Kriging interpolation. While Kriging provides a smooth representation of resistivity variation, clustering introduces discrete classification, aiding geological zonation. This agreement between methods suggests that the clustering-based classification effectively captures key hydrogeological features, reinforcing the validity of resistivity-based subsurface modeling.\u003c/p\u003e \u003cp\u003eIntegrating clustering techniques with geostatistical interpolation offers another avenue for improving subsurface modeling. Kriging interpolation provides a continuous representation of resistivity variations, while clustering introduces discrete classification that highlights key geological boundaries.\u003c/p\u003e \u003cp\u003eBy combining these approaches, researchers can develop a more comprehensive resistivity model. This hybrid methodology would allow for a better understanding of subsurface structures, aiding in groundwater exploration and geological mapping.\u003c/p\u003e \u003cp\u003eThe integration of clustering and interpolation can also help validate resistivity zonation results. If clustering-based classifications align with Kriging-generated resistivity distributions, confidence in the geological interpretation increases. Such cross-validation techniques ensure that subsurface models remain consistent across different analytical frameworks.\u003c/p\u003e \u003cp\u003eFor practical applications, an optimized clustering-interpolation framework could assist in hydrogeological assessments, mineral exploration, and environmental studies. The ability to delineate distinct resistivity zones with greater precision would be particularly beneficial for groundwater resource management.\u003c/p\u003e \u003cp\u003eFuture studies should focus on refining these methodologies, testing their applicability across diverse geological settings. By systematically evaluating different clustering strategies and integrating them with interpolation techniques, researchers can enhance the accuracy and reliability of geoelectrical subsurface models.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec31\" class=\"Section3\"\u003e \u003ch2\u003e3.2.3 Integrated Interpretation of Inversion, Geostatistics, and Clustering for Groundwater Characterization\u003c/h2\u003e \u003cp\u003eThe integration of geoelectrical resistivity inversion, geostatistical analysis, and clustering methods provides a comprehensive framework for characterizing subsurface hydrogeological conditions. The combined approach enhances the interpretation of resistivity variations, delineates groundwater-bearing formations, and improves the reliability of subsurface zonation. This multi-method approach ensures that geological complexity is adequately captured, reducing uncertainties in groundwater exploration and resource management.\u003c/p\u003e \u003cdiv id=\"Sec32\" class=\"Section4\"\u003e \u003ch2\u003e3.2.3.1 Insights from Resistivity Inversion\u003c/h2\u003e \u003cp\u003eResistivity inversion results reveal a stratified subsurface structure consisting of three primary zones: low, medium, and high-resistivity regions. Low-resistivity zones (\u0026lt;\u0026thinsp;30 Ωm) dominate the shallow subsurface, extending to depths of approximately 25 meters, and are interpreted as water-bearing formations. These zones correspond to unconsolidated pyroclastic deposits with high porosity and permeability, suggesting their potential as groundwater reservoirs. The presence of these low-resistivity regions is significant as they likely represent recharge areas where surface water infiltrates into deeper aquifers.\u003c/p\u003e \u003cp\u003eMedium-resistivity zones (30\u0026ndash;100 Ωm) appear as transitional layers, indicative of increasing compaction and reduced permeability with depth. These layers may represent partially weathered volcanic deposits or semi-consolidated sediments that still allow some degree of water movement. The identification of such transition zones is crucial for hydrogeological assessments, as they influence groundwater storage and transmission capabilities.\u003c/p\u003e \u003cp\u003eHigh-resistivity zones (\u0026gt;\u0026thinsp;100 Ωm) represent compact volcanic formations, likely acting as aquitards that restrict groundwater flow. These formations are interpreted as dense basaltic lava flows or compacted pyroclastic deposits that have undergone lithification. The presence of these high-resistivity zones at greater depths suggests the existence of impermeable layers that confine deeper aquifers, potentially leading to artesian conditions.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec33\" class=\"Section4\"\u003e \u003ch2\u003e3.2.3.2 Geostatistical Analysis and Spatial Interpolation\u003c/h2\u003e \u003cp\u003eKriging interpolation effectively captures the spatial variability of resistivity values, providing a smoothed resistivity distribution that reduces inversion artifacts. The interpolated results align well with the inversion model, reinforcing the presence of distinct hydrogeological units. The variogram analysis reveals a moderate spatial correlation, indicating the heterogeneous nature of the subsurface. The application of Kriging interpolation allows for improved visualization of resistivity patterns, aiding in the identification of potential groundwater reservoirs and confining layers.\u003c/p\u003e \u003cp\u003eA notable feature observed in the Kriging model is the presence of lateral variations in resistivity values, which may correspond to geological discontinuities such as faults or fractures. These structural features can play a crucial role in groundwater movement, as faults may act as either conduits or barriers depending on their nature. Identifying such features through geostatistical analysis enhances the ability to predict groundwater flow paths and assess the potential for groundwater extraction.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec34\" class=\"Section4\"\u003e \u003ch2\u003e3.2.3.3 Clustering-Based Subsurface Zonation\u003c/h2\u003e \u003cp\u003eThe clustering analysis further refines resistivity-based zonation by grouping similar resistivity values into distinct classes. K-Means clustering successfully segments the resistivity data into three main clusters, corresponding to low, moderate, and high-resistivity zones. However, its assumption of spherical clusters introduces limitations in capturing gradual geological transitions. The abrupt boundaries imposed by K-Means may not fully represent the continuous nature of subsurface resistivity variations, leading to potential misclassification of transitional zones.\u003c/p\u003e \u003cp\u003eDBSCAN, while effective in detecting density-based clusters, struggled to differentiate multiple zones due to the relatively uniform resistivity distribution. The sensitivity of DBSCAN to parameter selection suggests that further optimization is needed to improve its performance in geological applications. Adjusting the neighborhood radius (epsilon) and the minimum points required to form a cluster could enhance its ability to detect meaningful resistivity groupings.\u003c/p\u003e \u003cp\u003eHierarchical clustering provided the most geologically meaningful classification by preserving gradual resistivity transitions and highlighting natural boundaries between hydrogeological units. The hierarchical dendrogram allows for multi-scale analysis, enabling researchers to examine clustering results at different levels of granularity. This flexibility is particularly useful for identifying both broad-scale lithological trends and localized variations within aquifer systems.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec35\" class=\"Section4\"\u003e \u003ch2\u003e3.2.3.4 Hydrogeological Implications\u003c/h2\u003e \u003cp\u003eThe integrated analysis confirms that the low-resistivity cluster corresponds to water-saturated formations, making it a key target for groundwater exploration. These formations are likely to host unconfined aquifers that contribute to regional groundwater recharge. The extent and thickness of these low-resistivity zones provide valuable information for designing well placement strategies and estimating sustainable yield.\u003c/p\u003e \u003cp\u003eThe transition from low to medium-resistivity zones suggests a gradual compaction of volcanic deposits, influencing groundwater flow and storage. This transition is significant because it marks the boundary between shallow, unconfined aquifers and deeper, semi-confined units. Understanding this boundary is essential for groundwater management, as it determines the connectivity between different aquifer layers and affects groundwater extraction potential.\u003c/p\u003e \u003cp\u003eThe high-resistivity zones function as impermeable layers, restricting vertical water movement and forming potential confined aquifers. These layers may serve as protective barriers that prevent contamination from surface pollutants, thereby enhancing the quality of deeper groundwater reserves. Identifying the extent of these confining units is crucial for assessing the vulnerability of groundwater systems to external influences.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec36\" class=\"Section4\"\u003e \u003ch2\u003e3.2.3.5 Methodological Contributions\u003c/h2\u003e \u003cp\u003eBy integrating resistivity inversion, geostatistical analysis, and clustering techniques, this study offers a robust framework for improving subsurface characterization. The combination of these approaches reduces uncertainty in hydrogeological assessments and enhances the delineation of groundwater reservoirs. The use of multiple methods allows for cross-validation, ensuring that the identified subsurface features are consistently represented across different analytical techniques.\u003c/p\u003e \u003cp\u003eFuture research should explore the optimization of clustering parameters and incorporate additional geophysical methods, such as seismic and electromagnetic surveys, to further refine subsurface models. Integrating borehole data and groundwater chemistry analysis could provide additional validation, strengthening the hydrogeological interpretations derived from resistivity-based studies.\u003c/p\u003e \u003cp\u003eThis multi-method approach demonstrates the effectiveness of combining traditional geophysical techniques with data-driven clustering and spatial analysis, providing a more reliable and systematic classification of groundwater-bearing formations in complex geological settings. The findings of this study contribute to the advancement of groundwater exploration methodologies, offering insights that can be applied to other volcanic terrains and similar hydrogeological environments.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003c/div\u003e"},{"header":"4. Conclusion","content":"\u003cp\u003eThis study integrates geoelectrical resistivity inversion, geostatistical interpolation, and clustering-based zonation to improve subsurface characterization for groundwater exploration. The results confirm that low-resistivity zones (\u0026lt;\u0026thinsp;30 Ωm) correspond to water-bearing formations, highlighting their significance in groundwater recharge and storage. Geostatistical interpolation using Kriging effectively refines resistivity distributions, reducing noise and enhancing spatial interpretation. Among the clustering techniques, Hierarchical Clustering provides the most geologically meaningful classification, as it captures gradual resistivity transitions and natural zonation.\u003c/p\u003e \u003cp\u003eThe findings demonstrate that integrating clustering results with geostatistical interpolation enhances the accuracy of subsurface interpretation. The combined approach strengthens confidence in hydrogeological assessments, particularly in volcanic terrains where complex lithological variations impact groundwater distribution.\u003c/p\u003e \u003cp\u003eFuture studies should focus on refining clustering methodologies, particularly optimizing DBSCAN parameters to improve its performance in resistivity classification. Additionally, incorporating multiple geophysical techniques such as seismic refraction and magnetotellurics could further improve subsurface characterization. Borehole data and groundwater chemistry analysis would also provide valuable validation for resistivity-based interpretations.\u003c/p\u003e \u003cp\u003eOverall, this study highlights the effectiveness of combining geophysical inversion, geostatistical modeling, and machine learning-based clustering for subsurface resistivity analysis. The methodological framework developed here contributes to the advancement of groundwater exploration and geological mapping in complex subsurface environments, offering a robust, data-driven approach for hydrogeological assessments.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAuthor contribution\u003c/strong\u003e: All authors contributed to the study conceptualization. Methodology: G.U.N; Formal analysis and investigation:A.M, H.B and R.F.L; Writing - original draft preparation: G.U.N; Writing - review and editing: A.M, R.F.L, H.B and Y.S; Resources: R.F.L, H.B and Y.S; Supervision: R.F.L , H.B, Y.U and Y.S\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e: No Funding are available.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData availability\u003c/strong\u003e Available under request\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics approval\u0026nbsp;\u003c/strong\u003eNot applicable\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent to participate\u0026nbsp;\u003c/strong\u003eNot applicable\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent to publication\u0026nbsp;\u003c/strong\u003eNot applicable\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interest\u003c/strong\u003e The authors declare no competing interests.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eAhmadi M (2024) Chapter 5 - Clustering. In: Ahmadi M (ed) Artificial Intelligence for a More Sustainable Oil and Gas Industry and the Energy Transition. Elsevier, pp 183\u0026ndash;239\u003c/li\u003e\n\u003cli\u003eAudebert M, Cl\u0026eacute;ment R, Touze-Foltz N, et al (2014) Time-lapse ERT interpretation methodology for leachate injection monitoring based on multiple inversions and a clustering strategy (MICS). J Appl Geophy 111:320\u0026ndash;333. https://doi.org/https://doi.org/10.1016/j.jappgeo.2014.09.024\u003c/li\u003e\n\u003cli\u003eAuken E, Gu\u0026eacute;rin R, de Marsily G, Sailhac P (2009) Hydrogeophysics. Comptes Rendus - Geoscience 341:795\u0026ndash;799. https://doi.org/10.1016/j.crte.2009.09.003\u003c/li\u003e\n\u003cli\u003eBalacco G, Fiorese GD, Alfio MR (2023) Assessment of groundwater nitrate pollution using the Indicator Kriging approach. Groundw Sustain Dev 21:100920. https://doi.org/https://doi.org/10.1016/j.gsd.2023.100920\u003c/li\u003e\n\u003cli\u003eBelyadi H, Haghighat A (2021) Chapter 4 - Unsupervised machine learning: clustering algorithms. In: Belyadi H, Haghighat A (eds) Machine Learning Guide for Oil and Gas Using Python. Gulf Professional Publishing, pp 125\u0026ndash;168\u003c/li\u003e\n\u003cli\u003eBenabdelouahab S, Salhi A, Himi M, et al (2018a) Using resistivity methods to characterize the geometry and assess groundwater vulnerability of a Moroccan coastal aquifer. Groundw Sustain Dev 7:293\u0026ndash;304. https://doi.org/10.1016/j.gsd.2018.07.004\u003c/li\u003e\n\u003cli\u003eBenabdelouahab S, Salhi A, Himi M, et al (2018b) Using resistivity methods to characterize the geometry and assess groundwater vulnerability of a Moroccan coastal aquifer. Groundw Sustain Dev 7:293\u0026ndash;304. https://doi.org/10.1016/j.gsd.2018.07.004\u003c/li\u003e\n\u003cli\u003eBernatek-Jakiel A, Kondracka M (2016) Combining geomorphological mapping and near surface geophysics (GPR and ERT) to study piping systems. Geomorphology 274:193\u0026ndash;209. https://doi.org/10.1016/j.geomorph.2016.09.018\u003c/li\u003e\n\u003cli\u003eBinley A (2015) Tools and Techniques: Electrical Methods. Elsevier B.V.\u003c/li\u003e\n\u003cli\u003eBrunet P, Cl\u0026eacute;ment R, Bouvier C (2010) Monitoring soil water content and deficit using Electrical Resistivity Tomography (ERT) - A case study in the Cevennes area, France. J Hydrol (Amst) 380:146\u0026ndash;153. https://doi.org/10.1016/j.jhydrol.2009.10.032\u003c/li\u003e\n\u003cli\u003eCardarelli E, De Donno G (2019) Advances in electric resistivity tomography: Theory and case studies. Elsevier Inc.\u003c/li\u003e\n\u003cli\u003eChandra S, Dewandel B, Dutta S, Ahmed S (2010) Geophysical model of geological discontinuities in a granitic aquifer: Analyzing small scale variability of electrical resistivity for groundwater occurrences. J Appl Geophy 71:137\u0026ndash;148. https://doi.org/10.1016/j.jappgeo.2010.06.003\u003c/li\u003e\n\u003cli\u003eCivita M, De Maio M (2004) Assessing and mapping groundwater vulnerability to contamination: The Italian \u0026ldquo;combined\u0026rdquo; approach. Geofisica Internacional\u003c/li\u003e\n\u003cli\u003eCl\u0026eacute;ment R, Descloitres M, G\u0026uuml;nther T, et al (2009) Influence of shallow infiltration on time-lapse ERT: Experience of advanced interpretation. Comptes Rendus - Geoscience 341:886\u0026ndash;898. https://doi.org/10.1016/j.crte.2009.07.005\u003c/li\u003e\n\u003cli\u003eCoulouma G, Lagacherie P, Samyn K, Grandjean G (2013) Comparisons of dry ERT, diachronic ERT and the spectral analysis of surface waves for estimating bedrock depth in various Mediterranean landscapes. Geoderma 199:128\u0026ndash;134. https://doi.org/10.1016/j.geoderma.2012.07.026\u003c/li\u003e\n\u003cli\u003eDietrich S, Carrera J, Weinzettel P, Sierra L (2018) Estimation of Specific Yield and its Variability by Electrical Resistivity Tomography. Water Resour Res 54:8653\u0026ndash;8673. https://doi.org/https://doi.org/10.1029/2018WR022938\u003c/li\u003e\n\u003cli\u003eDong Q, Chen X, Huang B (2024) Chapter 15 - Cluster analysis. In: Dong Q, Chen X, Huang B (eds) Data Analysis in Pavement Engineering. Elsevier, pp 289\u0026ndash;301\u003c/li\u003e\n\u003cli\u003eFlorsch N, Muhlach F (2018) Direct Current Electrical Methods. Everyday Applied Geophysics 1 27\u0026ndash;103. https://doi.org/10.1016/b978-1-78548-199-4.50002-7\u003c/li\u003e\n\u003cli\u003eFrohlich RK, Parke CD (1989) The Electrical Resistivity of the Vadose Zone \u0026mdash; Field Survey. Groundwater 27:524\u0026ndash;530. https://doi.org/https://doi.org/10.1111/j.1745-6584.1989.tb01973.x\u003c/li\u003e\n\u003cli\u003eGance J, Malet JP, Supper R, et al (2016) Permanent electrical resistivity measurements for monitoring water circulation in clayey landslides. J Appl Geophy 126:98\u0026ndash;115. https://doi.org/10.1016/j.jappgeo.2016.01.011\u003c/li\u003e\n\u003cli\u003eGiampaolo V, Capozzoli L, Grimaldi S, Rizzo E (2016) Sinkhole risk assessment by ERT: The case study of Sirino Lake (Basilicata, Italy). Geomorphology 253:1\u0026ndash;9. https://doi.org/10.1016/j.geomorph.2015.09.028\u003c/li\u003e\n\u003cli\u003eGigu\u0026egrave;re R, Fradette L, Mignon D, Tanguy PA (2008) ERT algorithms for quantitative concentration measurement of multiphase flows. Chemical Engineering Journal 141:305\u0026ndash;317. https://doi.org/10.1016/j.cej.2008.01.011\u003c/li\u003e\n\u003cli\u003eHasan M, Shang Y, Akhter G, Jin W (2019) Application of VES and ERT for delineation of fresh-saline interface in alluvial aquifers of Lower Bari Doab, Pakistan. J Appl Geophy 164:200\u0026ndash;213. https://doi.org/https://doi.org/10.1016/j.jappgeo.2019.03.013\u003c/li\u003e\n\u003cli\u003eHeaney M (2003) Electrical Conductivity and Resistivity\u003c/li\u003e\n\u003cli\u003eHerman R (2001) An introduction to electrical resistivity in geophysics. Am J Phys 69:943\u0026ndash;952. https://doi.org/10.1119/1.1378013\u003c/li\u003e\n\u003cli\u003eH\u0026ouml;rning S, Gross L, B\u0026aacute;rdossy A (2020) Geostatistical electrical resistivity tomography using random mixing. J Appl Geophy 176:104015. https://doi.org/https://doi.org/10.1016/j.jappgeo.2020.104015\u003c/li\u003e\n\u003cli\u003eJardani A, Revil A, Dupont JP (2013) Stochastic joint inversion of hydrogeophysical data for salt tracer test monitoring and hydraulic conductivity imaging. Adv Water Resour 52:62\u0026ndash;77. https://doi.org/10.1016/j.advwatres.2012.08.005\u003c/li\u003e\n\u003cli\u003eJaved Z, Chand S, Ahmad M (2021) Spatial statistical analysis of groundwater quality based on inverse distance weighting and ordinary kriging in District Sheikhupura, Pakistan. Desalination Water Treat 235:324\u0026ndash;332. https://doi.org/https://doi.org/10.5004/dwt.2021.27653\u003c/li\u003e\n\u003cli\u003eKim JH, Tsourlos P, Yi MJ, Karmis P (2014) Inversion of ERT data with a priori information using variable weighting factors. J Appl Geophy 105:1\u0026ndash;9. https://doi.org/10.1016/j.jappgeo.2014.03.003\u003c/li\u003e\n\u003cli\u003eKirsch R (2006) Groundwater geophysics: a tool for hydrogeology\u003c/li\u003e\n\u003cli\u003eKuras O, Pritchard JD, Meldrum PI, et al (2009) Monitoring hydraulic processes with automated time-lapse electrical resistivity tomography (ALERT). Comptes Rendus - Geoscience 341:868\u0026ndash;885. https://doi.org/10.1016/j.crte.2009.07.010\u003c/li\u003e\n\u003cli\u003eLiang Y, Xia R, Yeh T-CJ, et al (2024) Characterizing preferential infiltration of loess using geostatistical electrical resistivity tomography. Eng Geol 340:107692. https://doi.org/https://doi.org/10.1016/j.enggeo.2024.107692\u003c/li\u003e\n\u003cli\u003eLiang Y, Yoon J-Y (2024) 2 - Fundamentals of machine learning. In: Yoon J-Y, Yu C (eds) Machine Learning and Artificial Intelligence in Chemical and Biological Sensing. Elsevier Science, pp 23\u0026ndash;70\u003c/li\u003e\n\u003cli\u003eLillie RJ (1988) Whole Earth Geophysics. An Introductory Textbook for Geologists and Geophysicists. 192\u003c/li\u003e\n\u003cli\u003eLiu Z, Gu Z, Liu P (2025) Chapter 9 - Clustering analysis. In: Liu Z, Gu Z, Liu P (eds) Transportation Big Data. Elsevier, pp 287\u0026ndash;315\u003c/li\u003e\n\u003cli\u003eMao D, Revil A, Hort RD, et al (2015) Resistivity and self-potential tomography applied to groundwater remediation and contaminant plumes: Sandbox and field experiments. J Hydrol (Amst) 530:1\u0026ndash;14. https://doi.org/10.1016/j.jhydrol.2015.09.031\u003c/li\u003e\n\u003cli\u003eMartel R, Castellazzi P, Gloaguen E, et al (2018) ERT, GPR, InSAR, and tracer tests to characterize karst aquifer systems under urban areas: The case of Quebec City. Geomorphology 310:45\u0026ndash;56. https://doi.org/10.1016/j.geomorph.2018.03.003\u003c/li\u003e\n\u003cli\u003eMartorana R, Capizzi P, D\u0026rsquo;Alessandro A, Luzio D (2017) Comparison of different sets of array configurations for multichannel 2D ERT acquisition. J Appl Geophy 137:34\u0026ndash;48. https://doi.org/10.1016/j.jappgeo.2016.12.012\u003c/li\u003e\n\u003cli\u003eMasoudi MJ, Ashrafzadeh A, Khaledian M, Janatrostami S (2024) Assessment of groundwater quality for agricultural purposes in Qazvin Province, northwestern Iran: A fuzzy inference and indicator Kriging approach. Environmental and Sustainability Indicators 24:100528. https://doi.org/https://doi.org/10.1016/j.indic.2024.100528\u003c/li\u003e\n\u003cli\u003eMaz\u0026aacute;č O, Kelly WE, Landa I (1985) A hydrogeophysical model for relations between electrical and hydraulic properties of aquifers. J Hydrol (Amst) 79:1\u0026ndash;19. https://doi.org/https://doi.org/10.1016/0022-1694(85)90178-7\u003c/li\u003e\n\u003cli\u003eMenke W (2022) Chapter 10 - Interpolation, Gaussian process regression, and kriging. In: Menke W (ed) Environmental Data Analysis with MatLab\u0026reg; or Python (Third Edition). Academic Press, pp 319\u0026ndash;348\u003c/li\u003e\n\u003cli\u003eMilsom J (2003) Field Geophysics, 3rd ed.\u003c/li\u003e\n\u003cli\u003eMilsom J, Eriksen A (2011) Field Geophysics\u003c/li\u003e\n\u003cli\u003eNuGraha GU, Alam BYCSSS, Nur AA, et al (2021) Vertical Electrical Sounding Exploration of Groundwater in Kertajati, Majalengka, West Java, Indonesia. Indonesian Journal on Geoscience 8:. https://doi.org/10.17014/ijog.8.3.359-369\u003c/li\u003e\n\u003cli\u003eNugraha GU, Bakti H, Lubis RF, et al (2022a) Aquifer vulnerability in the Coastal Northern Part of Lombok Island Indonesia. Environ Dev Sustain 24:1390\u0026ndash;1410. https://doi.org/10.1007/s10668-021-01459-0\u003c/li\u003e\n\u003cli\u003eNugraha GU, Gaol KL, Hartanto P, Bakti H (2020) Aquifer Vulnerability: Its Protection and Management\u0026mdash;A Case Study in Pangkalpinang City, Indonesia. International Journal of Geophysics 2020:8887914. https://doi.org/10.1155/2020/8887914\u003c/li\u003e\n\u003cli\u003eNugraha GU, Nur AA, Pranantya PA, et al (2022b) Analysis of groundwater potential zones using Dar-Zarrouk parameters in Pangkalpinang city, Indonesia. Environ Dev Sustain. https://doi.org/10.1007/s10668-021-02103-7\u003c/li\u003e\n\u003cli\u003eNugraha GU, Nur AA, Sudrajat Y, et al (2023) Sub-surface configuration in the northern part of Lembang groundwater basin recharge area. Appl Water Sci 13:. https://doi.org/10.1007/s13201-023-02004-2\u003c/li\u003e\n\u003cli\u003ePehme PE (2011) Groundwater Geophysics: A Tool for Hydrogeology. Environmental and Engineering Geoscience. https://doi.org/10.2113/gseegeosci.17.1.96\u003c/li\u003e\n\u003cli\u003eSeo J, Shneiderman B (2003) Interactively Exploring Hierarchical Clustering Results. In: BEDERSON BB, SHNEIDERMAN BEN (eds) The Craft of Information Visualization. Morgan Kaufmann, San Francisco, pp 334\u0026ndash;340\u003c/li\u003e\n\u003cli\u003eShi Z, Wang C (2024) Near-surface imaging by joint inversion of ERT and seismic traveltime data with guided FCM clustering. J Appl Geophy 222:105332. https://doi.org/https://doi.org/10.1016/j.jappgeo.2024.105332\u003c/li\u003e\n\u003cli\u003eSingh P, Singh PN, Srivastava S, et al (2025) Chapter 11 - Appraisal of spatial interpolation techniques in predicting soil organic carbon using earth observation datasets. In: Petropoulos GP, Da Silva Fuzzo DF, Triantakonstantis D, et al. (eds) Earth Observation for Monitoring and Modeling Land Use. Elsevier, pp 265\u0026ndash;283\u003c/li\u003e\n\u003cli\u003eSingh P, Verma P (2019) Chapter 5 - A Comparative Study of Spatial Interpolation Technique (IDW and Kriging) for Determining Groundwater Quality. In: Venkatramanan S, Prasanna MV, Chung SY (eds) GIS and Geostatistical Techniques for Groundwater Science. Elsevier, pp 43\u0026ndash;56\u003c/li\u003e\n\u003cli\u003eSlater L, Binley A, Versteeg R, et al (2002) A 3D ERT study of solute transport in a large experimental tank. J Appl Geophy 49:211\u0026ndash;229. https://doi.org/10.1016/S0926-9851(02)00124-6\u003c/li\u003e\n\u003cli\u003eSu L jun, Xu X qian, Geng X yu, Liang S qing (2017) An integrated geophysical approach for investigating hydro-geological characteristics of a debris landslide in the Wenchuan earthquake area. Eng Geol 219:52\u0026ndash;63. https://doi.org/10.1016/j.enggeo.2016.11.020\u003c/li\u003e\n\u003cli\u003eTelford WM, Geldart LP, Sheriff RE (1990) Applied Geophysics - Solid Earth Geophysics\u003c/li\u003e\n\u003cli\u003eTelford WM, Geldart LP, Sheriff RE (2012a) Electrical Properties of Rocks and Minerals. In: Applied Geophysics\u003c/li\u003e\n\u003cli\u003eTelford WM, Geldart LP, Sheriff RE (2012b) Methods Employing Natural Electrical Sources. In: Applied Geophysics\u003c/li\u003e\n\u003cli\u003eTheodoridis S, Koutroumbas K (2009) Chapter 13 - Clustering Algorithms II: Hierarchical Algorithms. In: Theodoridis S, Koutroumbas K (eds) Pattern Recognition (Fourth Edition). Academic Press, Boston, pp 653\u0026ndash;700\u003c/li\u003e\n\u003cli\u003eTso CHM, Kuras O, Wilkinson PB, et al (2017) Improved characterisation and modelling of measurement errors in electrical resistivity tomography (ERT) surveys. J Appl Geophy 146:103\u0026ndash;119. https://doi.org/10.1016/j.jappgeo.2017.09.009\u003c/li\u003e\n\u003cli\u003eTsourlos P, Papadopoulos N, Papazachos C, et al (2014) Efficient 2D inversion of long ERT sections. J Appl Geophy 105:213\u0026ndash;224. https://doi.org/10.1016/j.jappgeo.2014.03.022\u003c/li\u003e\n\u003cli\u003eWang TP, Chen CC, Tong LT, et al (2015) Applying FDEM, ERT and GPR at a site with soil contamination: A case study. J Appl Geophy 121:21\u0026ndash;30. https://doi.org/10.1016/j.jappgeo.2015.07.005\u003c/li\u003e\n\u003cli\u003eXu S, Sirieix C, Marache A, et al (2016) 3D geostatistical modeling of Lascaux hill from ERT data. Eng Geol 213:169\u0026ndash;178. https://doi.org/https://doi.org/10.1016/j.enggeo.2016.09.009\u003c/li\u003e\n\u003cli\u003eXu S, Sirieix C, Riss J, Malaurent P (2017) A clustering approach applied to time-lapse ERT interpretation \u0026mdash; Case study of Lascaux cave. J Appl Geophy 144:115\u0026ndash;124. https://doi.org/https://doi.org/10.1016/j.jappgeo.2017.07.006\u003c/li\u003e\n\u003cli\u003eZarroca M, Bach J, Linares R, Pellicer XM (2011) Electrical methods (VES and ERT) for identifying, mapping and monitoring different saline domains in a coastal plain region (Alt Empord\u0026agrave;, Northern Spain). J Hydrol (Amst) 409:407\u0026ndash;422. https://doi.org/10.1016/j.jhydrol.2011.08.052\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"discover-water","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"diwa","sideBox":"Learn more about [Discover Water](https://www.springer.com/43832)","snPcode":"","submissionUrl":"","title":"Discover Water","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Discover Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Resistivity Inversion, Geostatistical Interpolation, Clustering Analysis, Groundwater Exploration, Subsurface Zonation","lastPublishedDoi":"10.21203/rs.3.rs-6174105/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6174105/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThis study integrates geoelectrical resistivity inversion, geostatistical analysis, and clustering techniques to improve subsurface characterization in Cikole, Lembang, West Bandung Regency. The primary objective is to enhance resistivity-based zonation for groundwater exploration by employing a dipole-dipole array with 10-meter electrode spacing, followed by 2D inversion to visualize resistivity distributions. Geostatistical interpolation using Kriging effectively captures spatial variability, while clustering methods (K-Means, DBSCAN, and Hierarchical Clustering) classify resistivity-based subsurface zones. The results reveal that low-resistivity zones (\u0026lt;\u0026thinsp;30 Ωm) correspond to water-bearing formations, whereas high-resistivity regions (\u0026gt;\u0026thinsp;100 Ωm) are associated with compacted volcanic deposits. Among the clustering methods, Hierarchical Clustering provides the most geologically meaningful classification by preserving gradual resistivity transitions. The integration of clustering with geostatistical interpolation enhances subsurface interpretation, reducing uncertainty in hydrogeological assessments. However, DBSCAN's sensitivity to parameter selection limits its effectiveness in identifying multiple resistivity clusters. This study confirms that the combination of geoelectrical inversion, geostatistics, and clustering improves the accuracy of subsurface mapping, offering a data-driven approach applicable to complex geological environments. Future research should focus on optimizing DBSCAN parameters and incorporating additional geophysical methods to refine groundwater characterization.\u003c/p\u003e","manuscriptTitle":"Data-Driven Resistivity Zonation: Integrating Inversion, Kriging, and Clustering to Unravel Subsurface Complexity for Groundwater Exploration","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-04-24 05:06:52","doi":"10.21203/rs.3.rs-6174105/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-04-22T14:54:45+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-04-22T08:01:04+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"13946762722928929038232969667580247383","date":"2025-04-22T03:25:56+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-04-19T07:14:52+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"144443518637915450601332246395098652842","date":"2025-04-19T01:45:16+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"43075095098383174308414021481367239899","date":"2025-04-18T08:12:57+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"220703891317924352073347164434771736133","date":"2025-04-17T02:53:56+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"152337439502463669977090652613691053977","date":"2025-04-16T11:09:31+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"326326963817591356604845630262953328819","date":"2025-04-16T10:19:58+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"147426711492760480008610129672613502723","date":"2025-04-16T09:40:03+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"232262537274644573427335652020894862891","date":"2025-04-05T10:58:11+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"99344976625251084598476672154737662584","date":"2025-04-05T08:52:35+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"109328381396311712051313202403688273611","date":"2025-04-04T20:41:14+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"20423603118877966993211868593430483530","date":"2025-04-04T15:01:25+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-04-02T14:23:44+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-03-29T03:12:16+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-03-29T03:11:21+00:00","index":"","fulltext":""},{"type":"submitted","content":"Discover Water","date":"2025-03-07T01:47:46+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"discover-water","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"diwa","sideBox":"Learn more about [Discover Water](https://www.springer.com/43832)","snPcode":"","submissionUrl":"","title":"Discover Water","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Discover Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"0ded58af-5e3c-4cf7-a24e-82cdc088bf70","owner":[],"postedDate":"April 24th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2025-08-04T16:45:43+00:00","versionOfRecord":{"articleIdentity":"rs-6174105","link":"https://doi.org/10.1007/s43832-025-00261-7","journal":{"identity":"discover-water","isVorOnly":false,"title":"Discover Water"},"publishedOn":"2025-08-01 16:21:40","publishedOnDateReadable":"August 1st, 2025"},"versionCreatedAt":"2025-04-24 05:06:52","video":"","vorDoi":"10.1007/s43832-025-00261-7","vorDoiUrl":"https://doi.org/10.1007/s43832-025-00261-7","workflowStages":[]},"version":"v1","identity":"rs-6174105","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6174105","identity":"rs-6174105","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}