arxiv: 2604.27126 · v1 · submitted 2026-04-29 · 💻 cs.AI · cs.CE· cs.LG· physics.geo-ph

Recognition: unknown

Unsupervised Electrofacies Classification and Porosity Characterization in the Offshore Keta Basin Using Wireline Logs

Hamdiya Adams , Theophilus Ansah-Narh , Daniel Kwadwo Asiedu , Bruce Kofi Banoeng-Yakubo , Marcellin Atemkeng , Thomas Armah , Richmond Opoku-Sarkodie , Rebecca Davis

show 1 more author

Ezekiel Nii Noye Nortey

Authors on Pith no claims yet

Pith reviewed 2026-05-07 08:57 UTC · model grok-4.3

classification 💻 cs.AI cs.CEcs.LGphysics.geo-ph

keywords electrofacies classificationK-means clusteringwireline logsunsupervised machine learningporosity characterizationKeta Basinoffshore Ghanasubsurface characterisation

0 comments

The pith

K-means clustering on six wireline logs alone identifies four electrofacies tracking shale-to-sandstone transitions and porosity in the Keta Basin.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper demonstrates an unsupervised workflow that groups wireline log data from a single offshore well into electrofacies without any core samples. K-means is run on approximately 11,195 depth points using six standard logs, and the four resulting clusters are checked with inertia and silhouette scores that reach an average of about 0.5. The clusters align with expected geological changes in clay content and rock framework, producing depth-continuous units that range from shale-dominated to cleaner sandstone intervals. A sympathetic reader would see this as evidence that log-only clustering can supply reproducible subsurface characterisation when traditional core data are unavailable in frontier basins.

Core claim

The central claim is that K-means clustering performed in multivariate log space on six wireline logs, validated by quantitative inertia and silhouette diagnostics, yields four clusters that correspond to a geological continuum of electrofacies from shale-rich to sandstone-rich units, thereby establishing a robust and reproducible framework for formation evaluation and porosity characterisation in core-scarce offshore settings.

What carries the argument

K-means clustering applied to six wireline log curves, with cluster quality measured by inertia and average silhouette coefficient.

If this is right

Early-stage formation evaluation becomes feasible in frontier offshore basins where core recovery is limited or absent.
The workflow produces depth-continuous electrofacies that systematically track variations in clay content and porosity.
The method supplies a quantitative, reproducible starting point that can be extended by later integration with other data types.
Moderate but meaningful cluster separation (silhouette ~0.5) is sufficient to reveal geologically interpretable patterns across the logged interval.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The same log-only clustering approach could be applied to other wells or basins that have comparable wireline suites to test whether the number and character of electrofacies remain consistent.
Cross-validation against independent porosity measurements from a different logging tool or a short core interval would strengthen or weaken the link between clusters and reservoir quality.
Extending the workflow to include additional logs or alternative clustering algorithms might tighten the separation between facies and reduce the chance of parameter-driven artifacts.

Load-bearing premise

The four clusters produced by K-means actually represent distinct geological electrofacies rather than statistical groupings driven by the chosen logs or the clustering parameters.

What would settle it

Acquiring core samples or detailed petrographic descriptions from a new well in the Keta Basin and checking whether the predicted electrofacies match the observed rock types and porosity values at the same depths.

read the original abstract

This study presents an unsupervised machine learning workflow for electrofacies analysis in the offshore Keta Basin, Ghana, where core data are scarce. Six standard wireline logs from Well~C were analysed over a depth interval comprising approximately $11{,}195$ samples. K-means clustering was applied in multivariate log space, with the clustering structure evaluated using inertia and silhouette diagnostics. Four clusters were identified, supported by an average silhouette coefficient of approximately $0.50$, indicating moderate but meaningful separation. The resulting electrofacies exhibit systematic, depth-continuous patterns associated with variations in clay content, porosity, and rock framework properties, forming a geological continuum from shale-dominated to cleaner sandstone-dominated units. The results demonstrate that log-only, unsupervised clustering supported by quantitative metrics provides a robust and reproducible framework for subsurface characterisation. The proposed workflow offers a practical tool for early-stage formation evaluation in frontier offshore basins and a foundation for future integrated studies.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

Standard K-means on six logs from one well gives a usable but unvalidated electrofacies map for the Keta Basin.

read the letter

This paper runs K-means on six wireline logs from a single well in the offshore Keta Basin and settles on four clusters with an average silhouette of about 0.5. The clusters are then interpreted as a shale-to-sandstone continuum tied to clay content and porosity. That is the core contribution: a basin-specific case study using a routine unsupervised workflow where core is scarce. The authors show the depth-continuous patterns and report the usual diagnostics (inertia plus silhouette), which is done cleanly enough for an applied paper. The workflow itself is not new, but documenting it for this frontier setting adds a practical data point. The main limitation is the lack of external checks. With only one well and no core or second-well comparison, the geological labels rest on post-hoc pattern matching to the same logs used for clustering. A silhouette of 0.5 is only moderate, and wireline logs are typically correlated, so the partitions could partly reflect feature choice rather than distinct rock types. No sensitivity tests on preprocessing or log selection are described. The claim of a “robust and reproducible framework” therefore overreaches what the evidence shows. This is the kind of paper that belongs in an applied geophysics or petroleum geology journal. Readers working on data-limited basins will find the workflow and the Keta Basin results useful as a starting template, even if they will want to add their own validation steps. It is not a methods advance, but the execution is honest and the target audience is clear. I would send it to peer review with requests to tighten the validation discussion and qualify the strength of the geological interpretation.

Referee Report

2 major / 2 minor

Summary. The manuscript presents an unsupervised K-means clustering workflow applied to six wireline logs (approximately 11,195 samples) from a single well (Well C) in the offshore Keta Basin, Ghana. Four clusters are identified and evaluated using inertia and an average silhouette coefficient of ~0.50; these are interpreted post-hoc as electrofacies forming a shale-to-sandstone geological continuum based on associated variations in clay content, porosity, and rock framework properties. The central claim is that this log-only approach, supported by quantitative internal metrics, provides a robust and reproducible framework for subsurface characterisation in core-scarce frontier basins.

Significance. If the clusters can be shown to correspond to distinct geological electrofacies rather than artifacts of the chosen logs or K=4, the workflow could offer a practical, reproducible tool for early-stage formation evaluation where core data are limited. The moderate silhouette value and single-well scope limit broader impact, but the emphasis on quantitative diagnostics is a positive step over purely qualitative log interpretation.

major comments (2)

[Abstract] Abstract: The headline claim that the four clusters 'exhibit systematic, depth-continuous patterns associated with variations in clay content, porosity, and rock framework properties, forming a geological continuum' rests entirely on post-hoc visual or qualitative association with the same six logs used as input features. No quantitative comparison to independent labels, core descriptions, petrophysical models, or a second well is reported, so the geological interpretation cannot be distinguished from statistical partitioning of correlated log responses.
[Abstract] Abstract and methods (clustering evaluation): An average silhouette coefficient of approximately 0.50 is cited as indicating 'moderate but meaningful separation,' yet no error bars, bootstrap resampling, sensitivity tests on log selection/preprocessing, or comparison to alternative K values or algorithms (e.g., GMM) are provided. In low-dimensional, correlated feature spaces typical of wireline data, such values are compatible with both geologically meaningful and spurious partitions; this directly affects the robustness assertion.

minor comments (2)

[Abstract] Abstract: The notation 'Well~C' appears to be a typesetting artifact; standardise to 'Well C' or define the well name explicitly in the methods.
[Abstract] Abstract: The total sample count (11,195) is given without the corresponding depth interval length or sampling rate; adding this would improve reproducibility.

Simulated Author's Rebuttal

2 responses · 1 unresolved

We thank the referee for their constructive feedback on our manuscript. We provide point-by-point responses to the major comments below.

read point-by-point responses

Referee: [Abstract] Abstract: The headline claim that the four clusters 'exhibit systematic, depth-continuous patterns associated with variations in clay content, porosity, and rock framework properties, forming a geological continuum' rests entirely on post-hoc visual or qualitative association with the same six logs used as input features. No quantitative comparison to independent labels, core descriptions, petrophysical models, or a second well is reported, so the geological interpretation cannot be distinguished from statistical partitioning of correlated log responses.

Authors: We acknowledge that the geological interpretation is post-hoc and based on the correlation between the input logs and known formation properties. Since core data are not available for this well, we cannot provide quantitative validation against core descriptions or independent labels. The systematic depth-continuous patterns are evident in the results and consistent with regional geology. In the revised manuscript, we will tone down the claim in the abstract to describe the clusters as electrofacies inferred from log responses forming a continuum, and add a section on the limitations of the approach in core-scarce settings. We will also reference standard petrophysical models to support the associations. revision: partial
Referee: [Abstract] Abstract and methods (clustering evaluation): An average silhouette coefficient of approximately 0.50 is cited as indicating 'moderate but meaningful separation,' yet no error bars, bootstrap resampling, sensitivity tests on log selection/preprocessing, or comparison to alternative K values or algorithms (e.g., GMM) are provided. In low-dimensional, correlated feature spaces typical of wireline data, such values are compatible with both geologically meaningful and spurious partitions; this directly affects the robustness assertion.

Authors: The reported silhouette value of 0.50 is described as moderate in the manuscript, which is appropriate for wireline log data. To address concerns about robustness, we will include in the revision: sensitivity tests to log selection and preprocessing, bootstrap resampling to provide uncertainty estimates on the silhouette coefficient, and a comparison of results with Gaussian Mixture Models (GMM) as an alternative algorithm. These additions will help demonstrate that the partition is stable and not spurious. revision: yes

standing simulated objections not resolved

We are unable to conduct quantitative comparisons to core data, petrophysical models from core, or data from a second well, because the analysis uses a single well and no core data are available in this frontier basin.

Circularity Check

0 steps flagged

No circularity in the unsupervised clustering workflow

full rationale

The paper applies standard K-means directly to the six wireline logs over the 11,195-sample interval and reports internal diagnostics (inertia, average silhouette ~0.50) to select K=4. Cluster labels are then interpreted post-hoc in terms of clay content and porosity trends visible in the same logs. No derivation, equation, or fitted parameter is presented that reduces to its own inputs by construction; the central claim is simply that the resulting partition is reproducible and geologically interpretable. No self-citations, uniqueness theorems, or ansatzes are invoked to justify the method or the number of clusters. The workflow therefore remains self-contained as an application of off-the-shelf unsupervised ML with quantitative internal validation.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim rests on the assumption that log-space clusters map to geological facies; only the number of clusters is tuned via diagnostics.

free parameters (1)

number_of_clusters = 4
Selected as four after inspecting inertia and silhouette scores on the multivariate log data

axioms (1)

domain assumption Distinct clusters in multivariate wireline log space correspond to geologically meaningful electrofacies
Invoked when interpreting the K-means output as a geological continuum from shale to sandstone

pith-pipeline@v0.9.0 · 5518 in / 1185 out tokens · 40064 ms · 2026-05-07T08:57:56.346886+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

18 extracted references

[1]

Estimation of ash, moisture content and detection of coal lithofacies from well logs using regression and artificial neural network modelling,

S. Ghosh, R. Chatterjee, and P. Shanker, “Estimation of ash, moisture content and detection of coal lithofacies from well logs using regression and artificial neural network modelling,”Fuel, vol. 177, pp. 279–287, 2016

2016
[2]

Application of neural networks to identify lithofacies from well logs,

Y . Zhang, H. A. Salisch, and J. G. McPherson, “Application of neural networks to identify lithofacies from well logs,”Exploration Geophysics, vol. 30, no. 1-2, pp. 45–49, 1999

1999
[3]

Hybrid modeling of deep neural networks and unsupervised machine learning algorithms for missing well log prediction based on geological lithofacies similarities,

W. Hussain, M. Luo, M. Ali, I. Sadiq, E. E. Kasala, T. Aziz, and Z. Batool, “Hybrid modeling of deep neural networks and unsupervised machine learning algorithms for missing well log prediction based on geological lithofacies similarities,”Journal of Applied Geophysics, p. 105846, 2025

2025
[4]

Unsuper- vised identification of electrofacies employing machine learning,

I. Emelyanova, M. Pervukhina, M. Clennell, and C. Dyt, “Unsuper- vised identification of electrofacies employing machine learning,” in 79th EAGE Conference and Exhibition 2017-Workshops. European Association of Geoscientists & Engineers, 2017, pp. cp–519

2017
[5]

Geology and total petroleum systems of the gulf of guinea province of west africa,

M. E. Brownfield and R. R. Charpentier, “Geology and total petroleum systems of the gulf of guinea province of west africa,” US Geological Survey, Tech. Rep., 2006

2006
[6]

Evidence for transform margin evolution from the ivory coast–ghana continental margin,

J. Mascle and E. Blarez, “Evidence for transform margin evolution from the ivory coast–ghana continental margin,”Nature, vol. 326, no. 6111, pp. 378–381, 1987

1987
[7]

Liuet al.,Principles and applications of well logging

H. Liuet al.,Principles and applications of well logging. Springer, 2017

2017
[8]

D. V . Ellis and J. M. Singer,Well logging for earth scientists. Springer, 2007, vol. 692

2007
[9]

The geological interpretation of well logs; 2,

M. Rider, “The geological interpretation of well logs; 2,” 1996

1996
[10]

G. B. Asquith, D. Krygowski, and C. R. Gibson,Basic well log analysis. American Association of Petroleum Geologists Tulsa, 2004, vol. 16

2004
[11]

Clustering in geo-data science: Navigating uncertainty to select the most reliable method,

B. Sadeghi, “Clustering in geo-data science: Navigating uncertainty to select the most reliable method,”Ore Geology Reviews, p. 106591, 2025

2025
[12]

k- means clustering as tool for multivariate geophysical data analysis. an application to shallow fault zone imaging,

M. G. Di Giuseppe, A. Troiano, C. Troise, and G. De Natale, “k- means clustering as tool for multivariate geophysical data analysis. an application to shallow fault zone imaging,”Journal of Applied Geophysics, vol. 101, pp. 108–115, 2014

2014
[13]

Data clustering: 50 years beyond k-means,

A. K. Jain, “Data clustering: 50 years beyond k-means,”Pattern recog- nition letters, vol. 31, no. 8, pp. 651–666, 2010

2010
[14]

Silhouettes: a graphical aid to the interpretation and validation of cluster analysis,

P. J. Rousseeuw, “Silhouettes: a graphical aid to the interpretation and validation of cluster analysis,”Journal of computational and applied mathematics, vol. 20, pp. 53–65, 1987

1987
[15]

Tiab and E

D. Tiab and E. C. Donaldson,Petrophysics: theory and practice of measuring reservoir rock and fluid transport properties. Elsevier, 2024

2024
[16]

Methods for identifying complex lithologies from log data based on machine learning,

M. Liu, S. Hu, J. Zhang, and Y . Zou, “Methods for identifying complex lithologies from log data based on machine learning,”Unconventional Resources, vol. 3, pp. 20–29, 2023

2023
[17]

Vaidya,Principles of Petroleum Geoscience

A. Vaidya,Principles of Petroleum Geoscience. Educohack Press, 2025

2025
[18]

Well logging: Principles, applications and uncertain- ties,

N. H. Mondol, “Well logging: Principles, applications and uncertain- ties,” inPetroleum geoscience: From sedimentary environments to rock physics. Springer, 2015, pp. 385–425

2015