Mapping permafrost distribution and thaw depths in the Inner Tien Shan: A CryoGrid-based modeling approach integrating geophysical data

Mathys, Tamara; Marchenko, Sergey S.; Westermann, Sebastian; Gregor, Luke; Martin, Léo C. P.; Duishonakunov, Murataly; Hauck, Christian; Farzamian, Mohammad; Herring, Teddi; Hilbich, Christin; Kenzhebaev, Ruslan; Esenaman uulu, Mukhammad; Musaev, Timur; Bektursunov, Zhoodarbeshim; Usubaliev, Ryskul; Brining, Isabella; Saks, Tomas; Hoelzle, Martin

doi:10.5194/egusphere-2026-2756

Preprints

https://doi.org/10.5194/egusphere-2026-2756

Preprints

27 Jul 2026

| 27 Jul 2026

Mapping permafrost distribution and thaw depths in the Inner Tien Shan: A CryoGrid-based modeling approach integrating geophysical data

Tamara Mathys, Sergey S. Marchenko, Sebastian Westermann, Luke Gregor, Léo C. P. Martin, Murataly Duishonakunov, Christian Hauck, Mohammad Farzamian, Teddi Herring, Christin Hilbich, Ruslan Kenzhebaev, Mukhammad Esenaman uulu, Timur Musaev, Zhoodarbeshim Bektursunov, Ryskul Usubaliev, Isabella Brining, Tomas Saks, and Martin Hoelzle

Abstract. Permafrost degradation poses increasing risks to infrastructure in high-mountain regions, yet quantitative and spatially distributed assessments remain scarce in data-limited areas such as Central Asia. This study focuses on a 1,563 km² area in Inner Tien Shan, Kyrgyzstan, including the Kumtor gold mine and the Arabel plateau. We applied the numerical permafrost model CryoGrid to simulate permafrost thermal conditions and thaw depth across the study area, informed by geophysically derived stratigraphies and assessed against borehole temperatures and ground-based observations. To represent subsurface heterogeneity, surface-type-specific stratigraphies were developed for six landscape classes derived from petrophysical joint inversion of co-located Electrical Resistivity Tomography (ERT) and Refraction Seismic Tomography (RST) data. Legacy borehole measurements from 1986–1992 were incorporated to provide a multi-decadal validation of ground thermal conditions.

The resulting 30 m resolution maps indicate continuous permafrost across the study area, with mean ground temperatures at 5 m depth ranging from −0.1 °C to −6.5 °C and a spatial mean of −2.9 °C. Mean thaw depth at the end of August 2024 is 2.0 m, with surface type exerting a strong control on spatial patterns. Fine-grained and vegetated sediments on the Arabel plateau show comparatively deep thaw, coinciding with thermokarst features and ground subsidence, whereas coarse-blocky debris and rock glaciers exhibit shallow thaw depths. Comparison with ERT-derived thaw depths from 15 profiles shows agreement generally within about 1 m. Simulated ground temperatures show warming trends of 0.20 °C and 0.18 °C per decade at 5 m and 20 m depth since 1975, corresponding to approximately 1 °C warming since the 1980s. These results demonstrate the value of integrating geophysical data and surface-type-specific stratigraphies into modeling studies to represent spatial variability of mountain permafrost in data-scarce regions.

How to cite. Mathys, T., Marchenko, S. S., Westermann, S., Gregor, L., Martin, L. C. P., Duishonakunov, M., Hauck, C., Farzamian, M., Herring, T., Hilbich, C., Kenzhebaev, R., Esenaman uulu, M., Musaev, T., Bektursunov, Z., Usubaliev, R., Brining, I., Saks, T., and Hoelzle, M.: Mapping permafrost distribution and thaw depths in the Inner Tien Shan: A CryoGrid-based modeling approach integrating geophysical data, EGUsphere [preprint], https://doi.org/10.5194/egusphere-2026-2756, 2026.

Received: 22 May 2026 – Discussion started: 27 Jul 2026

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.

Download & links

Status: final response (author comments only)

RC1:
'Comment on egusphere-2026-2756', Anonymous Referee #1, 03 Aug 2026
The manuscript by Mathys et al. applies a surface-type-specific CryoGrid modeling framework, constrained by geophysical surveys (ERT/RST), borehole and ground temperature observations, and ERA5-based meteorological forcing, to estimate permafrost distribution and thaw depths in the Inner Tien Shan. The study is based on an impressive amount of field data and is one of the few that physically model permafrost distribution at high spatial resolution in a very remote area. The manuscript is well written and addresses important scientific questions, such as how permafrost in arid mountain regions has developed over recent decades, which is crucial for, e.g., freshwater supply in these environments. In my opinion, the paper is suitable for publication after some revisions.
General comments:
The Introduction could be shortened slightly by removing some textbook-like sentences about permafrost, ERT, etc.

Study site: The description is rather brief. I suggest adding some geological information, along with the genesis of the sedimentary cover on the Arabel plateau (Quaternary? Former lake sediments?). This is important for understanding ice content and permafrost dynamics there; 2–3 sentences would be sufficient.

Results: Present the ERT results before you introduce the calculation and validation of the active-layer (AL) depths. At present, you refer to ERT results before ERT is introduced.

I think the comparison to one of the global permafrost models does not add much new information. Your high-resolution permafrost model should be used to evaluate the global models, not the other way around. Figure 17 is confusing. You could perhaps show one of the global models (or all of them) in the Introduction, to illustrate how poorly these models perform in heterogeneous landscapes and to motivate your study, with one or two sentences. I suggest removing Section 4.7.

Discussion: Break up Section 5.1 into subparagraphs with subheadings or similar. This will make the discussion much easier to follow. Also check for unnecessary redundancies.

The heading of Section 5.2 should simply be “Uncertainties”, with subheadings such as “Model setting”, “Validation”, “Clustering”. I personally prefer to discuss uncertainties first and then the scientific implications, although this is a matter of taste.

Conclusions: Conclusions should not be a summary of the manuscript. I therefore recommend short bullet points that clearly highlight the main conclusions of your study. Your current conclusions are lengthy and can be shortened to the key statements.

More detailed comments:
l. 24: There are quite a number of papers, boreholes/thermal observations, maps, and studies from Mongolia, which I think is comparable to your site. However, most are not high-resolution, except for some studies in Hövsgöl. The same applies to l. 49–50.

l. 71 ff.: The whole paragraph is lengthy and too detailed (it contains information more appropriate for the Methods section). Consider shortening it.

Figure 1: Add a key map showing the global or regional location of the study area and remove the outline of Kyrgyzstan, which is confusing. A scale bar for the inset map is missing. Consider showing permafrost distribution based on one of the global models here (see above).

l. 94: “approximately -6.84….” is overly precise. Either write “approximately -7” or remove “approximately”.

l. 129 ff.: This paragraph reads like discussion/uncertainty; here you should simply state which methods you use.

l. 158/159: Remove this paragraph; it appears unnecessary.

l. 169: What does “(?)” after “Richards equation” refer to? Please clarify or remove.

l. 181: Remove the last sentence.

Table 2, caption: What does “FC” stand for? Field capacity? Please explain in the caption. Also clarify what is meant by “soil type”. Is “sand” used for rock glaciers?

Figure 6: Aspect is hard to interpret. You show it as a continuous variable, with colors indicating blue (“cold”) for southerly aspects, which is counterintuitive. I suggest using a classified map, similar to the surface-type map. A map of the clustering would also be useful for the reader.

Figure 7: The text inside the plots is very hard to read. In the caption, spell out SON, etc., in full.

l. 286 ff.: Is the snow factor applied uniformly across the entire study area? Make sure you discuss this choice thoroughly in the Discussion; I miss this part.

Figure 9: ALT development in the borehole shows a distinct knick point at 1 m, after which the ALT deepens almost vertically, at least in the first year. Why is this? Is it realistic? Could it be related to water infiltration or sensor issues?

l. 297 ff.: See the earlier comment—present the ERT results before the ALT validation.

Figure 10: Since you calculated temperatures down to 20 m depth (Figure 12), it would be interesting to see maps for shallower and deeper depths as well, perhaps in the Supplement.

l. 341: This sentence should be moved to the figure caption. I generally prefer that figure explanations are contained in the captions rather than the main text.

Figure 12: Why not use the same y-axis scale in all panels? Then you could delete the last sentence of the figure caption.

Figures 11 and 14 and l. 382 ff.: The boxplots are informative, but I miss a quantitative assessment of whether the different boxplots differ significantly, at least where this is not visually obvious.

l. 412: The entire following paragraph would fit better in the uncertainty discussion rather than in the Results section.

Figure 16: In the upper plot, is the full color ERT and the lighter color the model? This is not obvious from the explanation, where you refer to black and gray as differences. Please clarify.

Section 4.7: As noted above, I recommend removing this section. It adds little new insight, and it is obvious that modeling with 30 m resolution yields more heterogeneous results than with 1000 m resolution.

l. 517–522: Remove this paragraph. For central Asia at >3000 m elevation, you actually have a substantial amount of data.

l. 523 ff.: Check whether this paragraph is redundant.

l. 535: Again, there is substantial literature from Mongolia; please consult papers by Ishikawa, Jambaljav, etc.

l. 768: “Hoelzle 2025” – which journal? Please provide full reference details.

Figures A1, A2, captions: Add a reference to a figure that shows the location of the ERT profiles. As a rule, figure captions should allow the reader to understand the figure without having to read the entire manuscript.

Figure A7: I suggest including one of the global models already in Figure 1, see above)
Citation: https://doi.org/10.5194/egusphere-2026-2756-RC1
RC2:
'Comment on egusphere-2026-2756', Anonymous Referee #2, 06 Nov 2026
Dear authors:
The manuscript by Mathys et al. is an extremely well-written and scientifically rigorous application of the CryoGrid Community Model (CCM) to map permafrost conditions in the Inner Tien Shan, Kyrgyzstan, Central Asia. The article advances the use of the CCM by applying petrophysically joint inversion of Electrical Resistivity Tomography (ERT) and Seismic Refraction Tomography (SRT) data to define stratigraphy classes, in particular the mineral and air (if applicable) fractions of subsurface grid cells for each surface type. The authors defined multiple surface types (rock glaciers, debris, fine-grained sediment, bedrock, etc.) and spatially modelled permafrost temperatures and seasonal thaw depths for these different surface types with site-specific topo-climatic factors. To enable computationally efficient use of the CCM, the authors grouped the ground surface grid cells into clusters based on elevation, slope angle, aspect, sky view factor, and the surface type. Ultimately, the authors produced high-resolution (30 m) maps of permafrost temperature and thaw depths (August 2024), providing new insights into controls on the ground thermal regime of mountain permafrost beyond just elevation. By comparing their modelled thaw depths to ERT-derived thaw depths, the authors identified potential shortcomings in the application of ERT to properly delineate the zero-degree isotherm in coarse blocky terrain. Borehole validation enabled a rigorous analysis of modelled permafrost temperatures, pointing to seasonally dependent errors associated with inaccurate snow representation in the CCM. Overall, I classify this paper as "accepted subject to minor revisions". I organize my comments below into individual scientific points and technical (typographical, stylistic) points.
Scientific points:
The PJI results were used to define stratigraphy layers in the CCM. In the near surface, did you consider employing "STRAT_linear" instead of "STRAT_layers" for the mineral and water+ice fractions? It is possible to linearly interpolate the porosity between depths, which may improve modelling results where gradual changes in the subsurface composition are present. For example, the SED example in Figure 5 demonstrates that the CCM stratigraphy overestimates the mineral fraction in the upper 1 metre (relevant for seasonal phase changes and latent heat).

By overestimating the mineral fraction in the CCM, the modelled responses are likely to show temperatures that are too warm during MAM and too cold during DJF. In fact, this is what is apparent in Figure 7 (panel B) for the SED site A1023C. Employing a linearly interpolated mineral fraction in the uppermost unit could improve model accuracy.

The geothermal heat flux is constant in all simulations. Could the bias shown in Figure 8D be due to an inaccurate representation of this parameter at this location? Considering that the results for nearby Borehole 51_2 look very good, I suspect the bias is due to another factor. Could you please comment on this?

The discrepancy between the modelled active layer depths and the ERT-derived active layer depths for rock glaciers and coarse block debris is interesting and worthy of further discussion. I agree with the authors that the maximum vertical resistivity gradient in this setting likely corresponds to the ice-rich permafrost boundary as opposed to the thaw depth at the time of measurement. If the electrode spacing were reduced to provide better vertical resolution, one might see two vertical resistivity contrasts, with the weaker one corresponding to the thaw depth at the time of measurement.

Line 42: Consider adding "low ice content" as a factor for enhanced thaw depth for bedrock.
Figure 1: Please add the seismic lines to this figure.
Line 127: For the maximum vertical resistivity gradient, could you please clarify if you calculated this in linear or log space?
Technical points:
Line 26: The definition of the "active layer" is repeated.
Line 72: The line "our primary goals are" is repeated.
Figure 1: It is a bit difficult to see the lake colour when printed.
Line 90: There is an apostrophe glitch for 3200 and 4900 m a.s.l.
Line 169: There is a reference link glitch after "Richards equation".
Figure 8: The modelling envelopes are nearly impossible to see when printed.
Figure 16: For panel B, are you sure the X-axis is not "ERT minus modelled ALT"?
Citation: https://doi.org/10.5194/egusphere-2026-2756-RC2

Viewed

Total article views: 530 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
322	153	55	530	50	64

HTML: 322
PDF: 153
XML: 55
Total: 530
BibTeX: 50
EndNote: 64

Views and downloads (calculated since 27 Jul 2026)

Month	HTML	PDF	XML	Total
Jul 2026	70	26	5	101
Aug 2026	56	27	6	89
Sep 2026	83	46	19	148
Oct 2026	80	45	19	144
Nov 2026	33	9	6	48

Cumulative views and downloads (calculated since 27 Jul 2026)

Month	HTML	PDF	XML	Total
Jul 2026	70	26	5	101
Aug 2026	56	27	6	89
Sep 2026	83	46	19	148
Oct 2026	80	45	19	144
Nov 2026	33	9	6	48

Viewed (geographical distribution)

Total article views: 498 (including HTML, PDF, and XML) Thereof 498 with geography defined and 0 with unknown origin.

Country	#	Views	%

Cited

Latest update: 08 Nov 2026

Short summary

CryoGrid simulations constrained by geophysical and borehole data were used to assess permafrost thermal conditions and thaw depth in the Inner Tien Shan (Kyrgyzstan) at 30 m resolution. Surface-type-specific stratigraphies derived from geophysical surveys improved representation of subsurface heterogeneity. Results indicate widespread but spatially variable permafrost, with evidence of approximately 1 °C warming since the 1980s and increasing thaw depths.


Total:	0
HTML:	0
PDF:	0
XML:	0

Mapping permafrost distribution and thaw depths in the Inner Tien Shan: A CryoGrid-based modeling approach integrating geophysical data

Viewed

Viewed (geographical distribution)

Cited

1 citations as recorded by crossref.