arxiv: 2603.13973 · v2 · submitted 2026-03-14 · ⚛️ physics.geo-ph

Recognition: 2 theorem links

· Lean Theorem

Learning relaxation time distributions from spectral induced polarization data with a complex-valued variational autoencoder

Charles L. B\'erub\'e , S\'ebastien Gagnon , Lahiru M.A. Nagasingha , Jean-Luc Gagnon , E. Rachel Kenko , Reza Ghanati , Fr\'ed\'erique Baron

Authors on Pith no claims yet

Pith reviewed 2026-05-15 11:37 UTC · model grok-4.3

classification ⚛️ physics.geo-ph

keywords spectral induced polarizationrelaxation time distributionvariational autoencoderDebye decompositioncomplex-valued datageophysical inversionunsupervised learning

0 comments

The pith

A conditional variational autoencoder learns continuous relaxation time distributions directly from complex resistivity spectra.

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

Spectral induced polarization measures how rocks and soils respond to alternating currents across frequencies to characterize subsurface materials. Conventional Debye decomposition processes each spectrum independently in real space and offers limited uncertainty estimates, which limits performance on mixed datasets. The paper reformulates the task as unsupervised learning and trains a conditional variational autoencoder that operates natively in complex space to map spectra to stable relaxation time distributions. On 140 laboratory and field samples the model produces low reconstruction error and distributions whose parameters track physical grain properties with high correlation.

Core claim

The authors claim that a conditional variational autoencoder trained on complex resistivity spectra learns a shared mapping to continuous relaxation time distributions, achieving reconstruction errors of 0.45 percent for the imaginary component and 0.24 percent for the phase, with statistically significant gains over conventional methods, while the inferred distributions remain physically consistent and their total chargeability and mean relaxation time correlate with polarizable grain content and grain size at coefficients of determination up to 0.98.

What carries the argument

A conditional variational autoencoder (CVAE) that encodes complex resistivity spectra into a latent representation and decodes them to continuous relaxation time distributions.

If this is right

The inferred relaxation time distributions remain stable and physically consistent across the tested granular mixtures, mineralized rocks, and cementitious materials.
Total chargeability of the distributions correlates with polarizable grain content and mean relaxation time correlates with grain size, reaching coefficients of determination up to 0.98.
The learned latent representation organizes spectra such that unsupervised clustering in two dimensions improves the Davies-Bouldin index by nearly a factor of three relative to conventional parameters.
Reconstruction of the imaginary and phase components shows statistically significant improvement over conventional methods with p-values of 4 times 10 to the minus 6 and 2 times 10 to the minus 3.

Where Pith is reading between the lines

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

If the latent space remains structured on broader field datasets, it could support automated grouping of subsurface materials without requiring labeled examples.
The continuous output distributions may reduce manual parameter choices when processing streaming SIP data from long-term monitoring arrays.
Extending the model to joint inversion with other geophysical measurements could tighten uncertainty bounds on subsurface property estimates.

Load-bearing premise

The unsupervised CVAE learns a generalizable mapping from resistivity spectra to physically consistent relaxation time distributions that holds for data outside the 140-sample training set without overfitting or artifacts.

What would settle it

Applying the trained model to a fresh collection of SIP spectra from materials absent from the training set and finding reconstruction errors no lower than those of standard Debye decomposition would falsify the performance claim.

Figures

Figures reproduced from arXiv: 2603.13973 by Charles L. B\'erub\'e, E. Rachel Kenko, Fr\'ed\'erique Baron, Jean-Luc Gagnon, Lahiru M.A. Nagasingha, Reza Ghanati, S\'ebastien Gagnon.

**Figure 2.** Figure 2: Model selection results showing NMAE between measured and modelled com [PITH_FULL_IMAGE:figures/full_fig_p016_2.png] view at source ↗

**Figure 3.** Figure 3: Reconstruction NMAE of complex resistivity spectra as a function of encoder [PITH_FULL_IMAGE:figures/full_fig_p018_3.png] view at source ↗

**Figure 4.** Figure 4: Learning curves of the selected CVAE showing the evolution of the negative [PITH_FULL_IMAGE:figures/full_fig_p019_4.png] view at source ↗

**Figure 5.** Figure 5: Examples of CVAE reconstructions for four SIP spectra drawn from laboratory [PITH_FULL_IMAGE:figures/full_fig_p020_5.png] view at source ↗

**Figure 6.** Figure 6: Frequency-dependent reconstruction error of the CVAE expressed as the NMAE. [PITH_FULL_IMAGE:figures/full_fig_p021_6.png] view at source ↗

**Figure 7.** Figure 7: Posterior RTD of the 2.5 % graphite–sand mixture. The carpet plot shows 100 [PITH_FULL_IMAGE:figures/full_fig_p022_7.png] view at source ↗

**Figure 8.** Figure 8: Block-normalized heatmap of the posterior RTD across all samples. Each row [PITH_FULL_IMAGE:figures/full_fig_p023_8.png] view at source ↗

**Figure 9.** Figure 9: shows the joint distribution of truncated mean relaxation time τ¯[10−4,101] and total chargeability m[10−4,101] across all material groups. The results reveal separation between material classes, with posterior uncertainty remaining small relative to inter-group variability, which indicates that the learned representations provide stable and discriminative RTD parameters. 10−3 10−2 10−1 Mean relaxation tim… view at source ↗

**Figure 10.** Figure 10: Truncated normalized total chargeability [PITH_FULL_IMAGE:figures/full_fig_p027_10.png] view at source ↗

**Figure 11.** Figure 11: Truncated mean relaxation time τ¯[10−4,101] , in ms, as a function of inclusion diameter (d), in mm, for stainless steel spheres and cylinders in feldspar sand. Error bars represent posterior uncertainty from the probabilistic DD. The data show a systematic increase of τ¯ with inclusion size for both geometries. For spherical inclusions, the fitted scaling exponent is α = 1.00 ± 0.12, with a coefficient o… view at source ↗

**Figure 12.** Figure 12: Relative sensitivity structure of the CVAE parameters. The top panel shows [PITH_FULL_IMAGE:figures/full_fig_p029_12.png] view at source ↗

**Figure 13.** Figure 13: Two-dimensional UMAP embedding of the CVAE latent space. Each marker [PITH_FULL_IMAGE:figures/full_fig_p031_13.png] view at source ↗

**Figure 14.** Figure 14: Examples of synthetic complex resistivity spectra generated by sampling the [PITH_FULL_IMAGE:figures/full_fig_p032_14.png] view at source ↗

read the original abstract

Spectral induced polarization (SIP) is a geophysical method used to characterize subsurface materials. It measures the frequency-dependent complex resistivity of rocks and soils through the application of a small alternating current in the subsurface or in laboratory samples. Debye decomposition (DD) is a standard method for analyzing and interpreting SIP data, as it allows estimation of the relaxation time distribution (RTD) of geomaterials. However, conventional DD approaches treat measurements independently, work in real-valued spaces despite the complex-valued nature of SIP data, and provide limited uncertainty quantification. These limitations reduce the effectiveness of conventional DD on heterogeneous datasets. We reformulate DD as an unsupervised machine learning problem and introduce a conditional variational autoencoder (CVAE) that learns a shared mapping from resistivity spectra to continuous RTDs. The model is validated on a dataset comprising 140 laboratory and field SIP measurements of granular mixtures, mineralized rocks, and cementitious materials. The CVAE operates in complex-valued data space and achieves reconstruction errors of 0.45 % and 0.24 % for the imaginary and phase components of resistivity, respectively, with statistically significant improvements over conventional methods (p-values of 4x10^-6 and 2x10^-3). The inferred RTDs are stable and physically consistent, and their total chargeability and mean relaxation time correlate with polarizable grain content and grain size, respectively, with coefficients of determination up to 0.98. An additional contribution of the proposed method is the learned latent representation, which organizes SIP spectra into a structured space. Unsupervised clustering in a two-dimensional projection of this space improves the Davies--Bouldin index by nearly a factor of three relative to conventional RTD parameters.

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

CVAE reformulation of Debye decomposition gets low reconstruction errors and strong physical correlations on 140 samples, but the lack of any held-out testing leaves generalization unproven.

read the letter

The paper's main move is to treat Debye decomposition as an unsupervised complex-valued conditional variational autoencoder problem that learns a shared mapping from SIP spectra to continuous relaxation time distributions. On their 140 lab and field samples it reports 0.45 % and 0.24 % reconstruction errors on the imaginary and phase components, statistically better than conventional DD, plus RTD-derived chargeability and mean relaxation time that line up with polarizable grain content and grain size at R² values up to 0.98. The latent space also gives cleaner unsupervised clustering, cutting the Davies-Bouldin index by nearly a factor of three. Those are the concrete numbers the work delivers. The complex-valued formulation and the fact that the model produces stable, physically plausible RTDs without treating each spectrum in isolation are the parts that feel like genuine progress over the independent real-valued baselines cited in the abstract. The latent-space organization is an extra practical bonus for exploratory analysis of heterogeneous datasets. The soft spot is straightforward: all the reported errors, p-values, and correlations are computed on the full 140-sample collection with no train-test split, cross-validation, or external test set described. A conditional VAE has enough capacity to fit dataset-specific patterns, especially when the data mix lab mixtures and field measurements. Without evidence that the reconstruction fidelity and the grain-property correlations survive on spectra the model has not seen, it is hard to know whether the learned mapping is general or just well-tuned to this collection. The physical-consistency claims rest on the same in-sample numbers, so they inherit the same uncertainty. This is the kind of paper that belongs in a geophysics methods journal for readers who work with SIP for resource or environmental characterization. Someone already using Debye decomposition on noisy or mixed datasets could pick up the CVAE idea and test it themselves. I would send it to peer review because the reformulation is new, the quantitative claims are explicit, and the method is reproducible enough that referees can check the architecture and training details. The validation gap is fixable with proper splits and would be the main thing to ask for.

Referee Report

2 major / 0 minor

Summary. The paper reformulates Debye decomposition of spectral induced polarization (SIP) data as an unsupervised learning task and proposes a conditional variational autoencoder (CVAE) that maps complex resistivity spectra to continuous relaxation time distributions (RTDs). Validated on 140 laboratory and field SIP measurements of granular mixtures, mineralized rocks, and cementitious materials, the CVAE achieves reconstruction errors of 0.45% (imaginary) and 0.24% (phase), statistically significant improvements over conventional methods, RTD-derived parameters correlating with grain content and size (R² up to 0.98), and a latent space that improves unsupervised clustering by nearly a factor of three.

Significance. If the central claims hold, the work would introduce a data-driven, complex-valued framework for SIP interpretation that learns shared mappings across heterogeneous datasets, yields stable and physically consistent RTDs, and provides a structured latent representation for clustering, offering a clear advance over independent, real-valued Debye decomposition approaches in geophysical characterization.

major comments (2)

[Abstract] Abstract: the validation reports reconstruction errors, p-values, R² correlations, and clustering improvements exclusively on the full set of 140 samples with no mention of train/test splits, cross-validation, or external test data. This directly undermines the claim of a generalizable mapping from spectra to physically consistent RTDs, as the CVAE could overfit dataset-specific artifacts rather than learn a robust unsupervised representation.
[Abstract] Abstract: the comparison to conventional DD baselines is performed on the identical 140-sample set used for CVAE training and evaluation, so the reported statistically significant improvements (p=4e-6 and 2e-3) do not yet demonstrate out-of-distribution performance or robustness on unseen SIP spectra.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback. We agree that the abstract should more clearly describe the evaluation strategy to support claims of generalizability. We will revise the manuscript to address both major comments.

read point-by-point responses

Referee: [Abstract] Abstract: the validation reports reconstruction errors, p-values, R² correlations, and clustering improvements exclusively on the full set of 140 samples with no mention of train/test splits, cross-validation, or external test data. This directly undermines the claim of a generalizable mapping from spectra to physically consistent RTDs, as the CVAE could overfit dataset-specific artifacts rather than learn a robust unsupervised representation.

Authors: We acknowledge that the abstract does not specify the train/test procedure. The reported metrics reflect performance on the full 140-sample collection to demonstrate learning across heterogeneous SIP data. In the revised version we will update the abstract and add a dedicated evaluation subsection that reports 5-fold cross-validation results (training on 112 samples, testing on 28 held-out samples per fold) with average reconstruction errors, p-values, and clustering metrics computed only on the unseen folds. This will directly address generalizability within the available dataset. revision: yes
Referee: [Abstract] Abstract: the comparison to conventional DD baselines is performed on the identical 140-sample set used for CVAE training and evaluation, so the reported statistically significant improvements (p=4e-6 and 2e-3) do not yet demonstrate out-of-distribution performance or robustness on unseen SIP spectra.

Authors: We agree that out-of-distribution testing would strengthen the robustness claim. In the revision we will add an explicit out-of-sample experiment: a material-type hold-out (e.g., all cementitious samples) is excluded from training, the CVAE is retrained on the remaining data, and reconstruction error, RTD stability, and parameter correlations are compared against DD on the held-out spectra. These new results will be summarized in the abstract and detailed in the results section. revision: yes

Circularity Check

0 steps flagged

No significant circularity in CVAE-based RTD mapping

full rationale

The paper reformulates Debye decomposition as an unsupervised CVAE learning problem that reconstructs complex resistivity spectra while producing RTDs and a latent space. Reconstruction errors (0.45%/0.24%) are standard autoencoder training metrics on the 140-sample set, but the load-bearing claims rest on independent external correlations (R^2 up to 0.98 with measured grain content/size) and clustering improvements versus conventional DD baselines. No equations reduce by construction to inputs, no fitted parameters are relabeled as predictions, and no self-citation chain or ansatz smuggling supports the central mapping. The derivation remains self-contained against external physical benchmarks.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim rests on standard neural network assumptions and data-driven fitting rather than new physical axioms or invented entities.

free parameters (1)

CVAE architecture and training hyperparameters
Latent dimension, layer sizes, learning rate, and regularization parameters are chosen and fitted to the 140-sample dataset.

axioms (1)

domain assumption A neural network can learn a shared mapping from complex resistivity spectra to continuous relaxation time distributions without explicit physical equations.
Invoked in the reformulation of DD as an unsupervised ML problem.

pith-pipeline@v0.9.0 · 5656 in / 1349 out tokens · 60903 ms · 2026-05-15T11:37:24.649081+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

Cost.FunctionalEquation washburn_uniqueness_aczel unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

We reformulate DD as an unsupervised machine learning problem and introduce a conditional variational autoencoder (CVAE) that learns a shared mapping from resistivity spectra to continuous RTDs... The CVAE operates in complex-valued data space and achieves reconstruction errors of 0.45 % and 0.24 %...
Foundation.RealityFromDistinction reality_from_one_distinction unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

The decoder... maps latent distribution samples z to the parameters of a logarithmic RTD γθ and its associated complex resistivity spectrum ˆρ̃.

What do these tags mean?

matches: The paper's claim is directly supported by a theorem in the formal canon.
supports: The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends: The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses: The paper appears to rely on the theorem as machinery.
contradicts: The paper's claim conflicts with a theorem or certificate in the canon.
unclear: Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

26 extracted references · 26 canonical work pages

[1]

Hydrology and Earth System Sciences 17, 4079–4094

Spectral induced polarization measurements for predicting the hydraulic conductivity in sandy aquifers. Hydrology and Earth System Sciences 17, 4079–4094. doi:10.5194/hess-17-4079-2013. Bairlein, K., Hördt, A., Nordsiek, S.,

work page doi:10.5194/hess-17-4079-2013 2013
[2]

Near Surface Geophysics 12, 667–678

The influence on sample preparation on spectral induced polarization of unconsolidated sediments. Near Surface Geophysics 12, 667–678. doi:https://doi.org/10.3997/ 1873-0604.2014023. Benzetta, R.,

work page arXiv
[3]

Geophysics 88, E79– E90

Bayesian inference of petrophysical properties with generative spectral induced polarization models. Geophysics 88, E79– E90. doi:10.1190/geo2022-0495.1. 38 Bérubé, C.L., Chouteau, M., Shamsipour, P., Enkin, R.J., Olivo, G.R.,

work page doi:10.1190/geo2022-0495.1
[4]

Computers & Geosciences 105, 51–64

Bayesian inference of spectral induced polarization parameters for labora- tory complex resistivity measurements of rocks and soils. Computers & Geosciences 105, 51–64. doi:10.1016/j.cageo.2017.05.001. Bérubé, C.L., Gagnon, S., Kenko, E.R., Gagnon, J.L., Nagasingha, L.M.A., Ghanati, R., Baron, F.,

work page doi:10.1016/j.cageo.2017.05.001 2017
[5]

Geophysical Journal International 243, ggaf348

Complex-valued neural networks for spectral induced polarization applications. Geophysical Journal International 243, ggaf348. doi:10.1093/gji/ggaf348. Bérubé, C.L., Olivo, G.R., Chouteau, M., Perrouty, S.,

work page doi:10.1093/gji/ggaf348
[6]

Geophysics 84, B135–B151

Mineralog- ical and textural controls on spectral induced polarization signatures of the Canadian Malartic gold deposit: Applications to mineral exploration. Geophysics 84, B135–B151. doi:10.1190/geo2018-0404.1. Flores Orozco, A., Kemna, A., Oberdörster, C., Zschornack, L., Leven, C., Dietrich, P., Weiss, H.,

work page doi:10.1190/geo2018-0404.1
[7]

Journal of Contaminant Hydrology 136-137, 131–144

Delineation of subsurface hydrocarbon con- tamination at a former hydrogenation plant using spectral induced po- larization imaging. Journal of Contaminant Hydrology 136-137, 131–144. doi:10.1016/j.jconhyd.2012.06.001. Florsch, N., Revil, A., Camerlynck, C.,

work page doi:10.1016/j.jconhyd.2012.06.001 2012
[8]

Journal of Applied Geophysics 109, 119–132

Inversion of generalized relax- ation time distributions with optimized damping parameter. Journal of Applied Geophysics 109, 119–132. doi:10.1016/j.jappgeo.2014.07.013. Ghorbani, A., Camerlynck, C., Florsch, N., Cosenza, P., Revil, A.,

work page doi:10.1016/j.jappgeo.2014.07.013 2014
[9]

Geophysical Prospecting 55, 589–605

Bayesian inference of the Cole–Cole parameters from time- and frequency- domain induced polarization. Geophysical Prospecting 55, 589–605. doi:10.1111/j.1365-2478.2007.00627.x. Grenon, C.,

work page doi:10.1111/j.1365-2478.2007.00627.x 2007
[10]

Journal of Applied Geophysics 98, 44–53

Time domain spectral in- duced polarization of disseminated electronic conductors: Laboratory data analysis through the Debye decomposition approach. Journal of Applied Geophysics 98, 44–53. doi:10.1016/j.jappgeo.2013.07.008. Gurin, G., Tarasov, A., Il’in, Y., Titov, K.,

work page doi:10.1016/j.jappgeo.2013.07.008 2013
[11]

Russian Geology and Geophysics 56, 1757–1771

Application of the De- bye decomposition approach to analysis of induced-polarization profiling 39 data (Julietta gold-silver deposit, Magadan Region). Russian Geology and Geophysics 56, 1757–1771. doi:10.1016/j.rgg.2015.11.008. Hase, J., Gurin, G., Titov, K., Kemna, A.,

work page doi:10.1016/j.rgg.2015.11.008 2015
[12]

Geophysics 77, E159–E170

Markov-chain Monte Carlo estimation of distributed Debye relaxations in spectral induced po- larization. Geophysics 77, E159–E170. doi:10.1190/geo2011-0244.1. Khajehnouri, Y., Chouteau, M., Rivard, P., Bérubé, C.L.,

work page doi:10.1190/geo2011-0244.1
[13]

Construction and Building Materials 210, 1–12

Measuring electrical properties of mortar and concrete samples using the spectral in- duced polarization method: laboratory set-up. Construction and Building Materials 210, 1–12. doi:10.1016/j.conbuildmat.2019.03.160. Khajehnouri, Y., Chouteau, M., Rivard, P., Bérubé, C.L., 2020a. Non- destructive non-invasive assessment of the development of alkali-silic...

work page doi:10.1016/j.conbuildmat.2019.03.160 2019
[14]

(Eds.), 3rd International Conference on Learn- ing Representations, ICLR 2015, San Diego, CA, USA, May 7-9, 2015, Conference Track Proceedings, pp

Adam: A Method for Stochastic Optimization, in: Bengio, Y., LeCun, Y. (Eds.), 3rd International Conference on Learn- ing Representations, ICLR 2015, San Diego, CA, USA, May 7-9, 2015, Conference Track Proceedings, pp. 8024–8035. Kingma, D.P., Welling, M.,

work page 2015
[15]

(Eds.), 2nd International Conference on Learning Representations, ICLR 2014, Banff, AB, Canada, April 14-16, 2014, Con- ference Track Proceedings, pp

Auto-Encoding Variational Bayes, in: Bengio, Y., LeCun, Y. (Eds.), 2nd International Conference on Learning Representations, ICLR 2014, Banff, AB, Canada, April 14-16, 2014, Con- ference Track Proceedings, pp. 1–14. Martin, T., Weller, A., Behling, L.,

work page 2014
[16]

Geophysical Journal Interna- tional 228, 275–290

Desaturation effects of pyrite–sand mixtures on induced polarization signals. Geophysical Journal Interna- tional 228, 275–290. doi:10.1093/gji/ggab333. 40 McInnes, L., Healy, J., Saul, N., Großberger, L.,

work page doi:10.1093/gji/ggab333
[17]

Morgan, F.D., Lesmes, D.P.,

doi:10.21105/joss.00861. Morgan, F.D., Lesmes, D.P.,

work page doi:10.21105/joss.00861
[18]

The Journal of Chemical Physics 100, 671–681

Inversion for dielectric relaxation spectra. The Journal of Chemical Physics 100, 671–681. doi:10.1063/1.466932. Nordsiek, S., Weller, A.,

work page doi:10.1063/1.466932
[19]

Geophysics 73, F235–F245

Anewapproachtofittinginduced-polarization spectra. Geophysics 73, F235–F245. doi:10.1190/1.2987412. Paszke, A., Gross, S., Massa, F., Lerer, A., Bradbury, J., Chanan, G., Killeen, T., Lin, Z., Gimelshein, N., Antiga, L., Desmaison, A., Kopf, A., Yang, E., DeVito, Z., Raison, M., Tejani, A., Chilamkurthy, S., Steiner, B., Fang, L., Bai, J., Chintala, S.,

work page doi:10.1190/1.2987412
[20]

IEEE Transactions on Signal Processing 44, 2637–2640

Second-order complex random vectors and normal distributions. IEEE Transactions on Signal Processing 44, 2637–2640. doi:10.1109/78.539051. Revil, A., Florsch, N., Camerlynck, C.,

work page doi:10.1109/78.539051
[21]

Near Sur- face Geophysics 10, 503–515

Spectral Induced Polarization (SIP) signatures of clayey soils containing toluene. Near Sur- face Geophysics 10, 503–515. doi:https://doi.org/10.3997/1873-0604. 2012015. Virtue, P., Yu, S.X., Lustig, M.,

work page doi:10.3997/1873-0604
[22]

3953–3957

Better than real: Complex-valued neural nets for MRI fingerprinting, in: 2017 IEEE International Con- ference on Image Processing (ICIP), pp. 3953–3957. doi:10.1109/ICIP. 2017.8297024. iSSN: 2381-8549. Volkmann, J., Klitzsch, N.,

work page doi:10.1109/icip 2017
[23]

Journal of Applied Geophysics 114, 191–201

Wideband impedance spectroscopy from 1 mHz to 10 MHz by combination of four- and two-electrode methods. Journal of Applied Geophysics 114, 191–201. doi:10.1016/j.jappgeo. 2015.01.012. Weigand, M., Kemna, A.,

work page doi:10.1016/j.jappgeo 2015
[24]

Computers & Geosciences 86, 34–45

Debye decomposition of time-lapse spectral induced polarisation data. Computers & Geosciences 86, 34–45. doi:10. 1016/j.cageo.2015.09.021. Weller, A., Nordsiek, S., Debschütz, W., 2010a. Estimating permeability of sandstone samples by nuclear magnetic resonance and spectral-induced polarization. Geophysics 75, E215–E226. doi:10.1190/1.3507304. Weller, A.,...

work page doi:10.1190/1.3507304 2015
[25]

Geophysics 78, D315–D325

On the relationship between in- duced polarization and surface conductivity: Implications for petrophysi- cal interpretation of electrical measurements. Geophysics 78, D315–D325. doi:10.1190/geo2013-0076.1. Weller, A., Slater, L., Nordsiek, S., Ntarlagiannis, D., 2010b. On the estima- tion of specific surface per unit pore volume from induced polarization...

work page doi:10.1190/geo2013-0076.1
[26]

Zorin, N.,

doi:10.1029/2010JB007526. Zorin, N.,

work page doi:10.1029/2010jb007526