Random forest predictions of tundra snow density elevate Arctic soil temperatures in CLM5.0

Rutherford, Johnny; Rutter, Nick; Wake, Leanne; Wooley, Georgina; Boike, Julia; Cannon, Alex J.

doi:10.5194/egusphere-2026-3363

Preprints

https://doi.org/10.5194/egusphere-2026-3363

Preprints

20 Jul 2026

| 20 Jul 2026

Status: this preprint is open for discussion and under review for The Cryosphere (TC).

Random forest predictions of tundra snow density elevate Arctic soil temperatures in CLM5.0

Johnny Rutherford, Nick Rutter, Leanne Wake, Georgina Wooley, Julia Boike, and Alex J. Cannon

Abstract. Arctic snow exerts a critical control on winter soil temperature and carbon exchange, however representation of its properties in Earth System Models (ESMs) remains simplified. In the Community Land Model v5.0 (CLM5.0), recent updates to snow compaction schemes have led to overly dense tundra snow, increasing snow thermal conductivity and conductive heat loss, producing a persistent cold-soil bias. Here we developed a Random Forest (RF) regression model to derive tundra bulk snow density from meteorological variables, trained on six winter seasons (Sep – May) of Arctic SVS2-Crocus (ASC) simulations constrained by in-situ observations from Trail Valley Creek (TVC), Northwest Territories, Canada. RF-derived snow densities capture the lower bulk densities of Arctic tundra snowpacks, which CLM5.0’s compaction scheme overestimates. Application of the RF-derived snow densities to CLM5.0 reduces snow thermal conductivity and enhances snowpack insulation, decreasing soil temperature RMSE by approximately 2–3 °C relative to field measurements (2017–2023) and increasing future winter soil temperature projections (2016–2100) by 4–7 °C under RCP4.5 and 8.5. Such increases in simulated soil temperature highlight the strong sensitivity of CLM5.0’s soil thermal regime to snow physical properties. The RF model reproduces ASC-simulated density evolution with a mean absolute error of 32 kg m⁻³ which falls within typical measurement error for volumetric snow density sampling in Arctic tundra (20–40) kg m⁻³ and matches field measurements more closely than default CLM5.0. Future bulk snow density predictions using the RF model driven by bias-corrected North American-Coordinated Regional Downscaling Experiment (NA-CORDEX) meteorology indicate bulk snow densities 200–450 kg m⁻³ lower than CLM5.0 and more consistent with tundra conditions.

Received: 09 Jun 2026 – Discussion started: 20 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

Johnny Rutherford, Nick Rutter, Leanne Wake, Georgina Wooley, Julia Boike, and Alex J. Cannon

Status: open (until 29 Oct 2026)

Post a comment Subscribe to comment alert

RC1: 'Comment on egusphere-2026-3363', Anonymous Referee #1, 14 Sep 2026 reply

This study uses a random forest (RF), trained on six winters of Arctic SVS2-Crocus (ASC) simulations at Trail Valley Creek (TVC), to modify snow thermal conductivity in CLM5.0. The reduction in winter soil-temperature bias is encouraging, and the problem is relevant to The Cryosphere. I am less convinced that the RF improves on existing corrections or that it supports the conclusions drawn from the future runs. I recommend major revisions, particularly to the comparisons and validation described below.
Major comments
1. The authors already show that the Sturm conductivity scheme reduces the cold bias. Why is that configuration absent from the comparisons? Default density with Sturm conductivity should be tested against RF density with Jordan conductivity under the same forcing. A seasonal ASC-density climatology and a simple regression would also be useful baselines. With cumulative snowfall as the dominant predictor (L266), it is difficult to tell how much the RF learns beyond the seasonal cycle. Separating seasonal skill from interannual anomalies would help answer this. This is the kind of improvement we would particularly expect under future warming.
2. I would expect to see an analysis using explainable methods to identify what is actually constrained and improved by the observations. The phrase “constrained by in-situ observations” needs qualification. Observations are used to select ASC seasons and ensemble members; ASC then supplies the training labels. Figure 1 tests against the ensemble mean, so the training and evaluation targets also differ. The late-season RF bias relative to ASC happens to bring predictions closer to the measurements, but it is unclear whether this improvement holds across individual years. Please show comparisons on the campaign dates and distinguish errors against ASC from errors against observations. Agreement near peak SWE cannot establish that the early-winter evolution is correct. More generally, we should have interpretable results to understand what happens when the models are integrated. This would benefit the community beyond simply demonstrating better predictions from black-box machine learning models.
3. The authors acknowledge the problem with random partitioning of daily data, but it remains in the tuning procedure. Use leave-one-season-out validation and report the hyperparameters, feature importance and effect of adding training seasons. The exclusion of wind, and the possible role of shortwave radiation as a seasonal proxy, also deserve explanation. A more fundamental issue is Figure 3: NA-CORDEX winters do not reproduce the weather of the corresponding observed years. The RMSE therefore includes forcing mismatch. The present-day evaluation should use observed TVC meteorology for both models, with errors reported by winter and soil depth. This would give readers a clearer understanding of model performance and avoid overstating predictive accuracy based on the full dataset.
4. In the conductivity-only approach, the RF density used to calculate thermal conductivity differs from that implied by CLM5’s simulated snow mass and layer thickness. Please discuss the physical implications of using these different densities and whether the improved soil-temperature agreement could partly reflect compensating errors. How is the RF bulk density applied to individual snow layers (Eqs. 3–7), and how do the two insertion schemes affect snow depth, SWE and snowpack thermal resistance? Energy diagnostics would also help explain why snow layer scaling produces a warm soil-temperature bias (Line 299-300).
5. I remain doubtful about carrying a relationship learned from six selected historical winters through to 2100 before the earlier concerns have been resolved, clearly explained or evaluated. The authors recognise the RF extrapolation problem, but the Jensen–Shannon analysis only describes a shift in the inputs. It does not tell us how large the density or soil-temperature error might become. Figure 5 combines all projection years into one seasonal curve, which obscures the response to warming. Please separate early-, mid- and late-century density and conductivity, test the RF on withheld warm winters, and compare it with ASC under representative future forcing. The uncertainty analysis should include training-season and ASC-label choices, rather than only climate forcing and CLM parameters. The plateau in Figure 6 needs further support for the proposed latent-heat explanation, including temperature profiles, ice/liquid-water and energy diagnostics. These checks are needed before the 4–7°C offset can be interpreted as a more realistic future projection. A winter temperature approaching 0°C at 2 m alone does not establish when or how much permafrost thaws. In general, future projections are more challenging than present-day simulations and require further investigation and evaluation. I agree that using machine learning as an emulator is a good approach, but the projections are somewhat ambitious for the scope of the current manuscript.
6. For the comparison between the RCP4.5 and RCP8.5 scenarios, how much of the apparent scenario difference comes from ensemble composition or initialisation? A comparison using common GCM–RCM pairs and baseline anomalies would help. The “all members have SWE > 5 mm” criterion may also select different winter windows; their lengths should be reported and a common window tested. Please explain how the carbon-parameter choices affect soil-temperature spread. Without carbon-flux results, the claim that previous carbon-loss projections are a conservative lower bound goes beyond this study. In addition, I suggest that the study, including its title and conclusions, should consistently reflect its single-site scope rather than imply a spatial evaluation.
7. The late-season RF bias relative to ASC brings predictions closer to the observations at TVC, but this may reflect compensation for a site-specific ASC bias. Would the same approach still improve agreement with observations at a site where ASC has a different bias, or could it make the predictions worse? Please discuss this possibility and clarify whether the improvement observed at TVC can reasonably be expected at other sites.
Specific comments
Figure 1: The legend, axis labels and units are difficult to read.
L171–175: Please explain the physical rationale for including SW radiation as a predictor of bulk density (it is largely a proxy for day of year at this latitude) and comment on whether it simply encodes seasonality.
L196, L203: "Kullback-Leiber" should be "Kullback-Leibler".
L199: JSD on a 2-D KDE of only two predictors does not capture the joint distribution of all four predictors.
L264, L367 and L479: Clarify whether the 32 kg m⁻³ value refers to MAE or RMSE and define the evaluation target. Why report different metrics instead of using one consistently?
Section 2.2 and Table 1: Specify temporal aggregation, precipitation-phase partitioning, RF-to-CLM time-step mapping and the handling of snow-free conditions. A more detailed description of the data preprocessing and methods would help readers better evaluate your study.
Figure 3: Please check and clarify the depth or scenario assignments in the caption.
Figure 6: Please define the uncertainty bands more clearly.
L214–215: Supply the missing table number.
Appendix A: Correct “RSME” to “RMSE”.
L232–234: The 2 m soil temperature data are unpublished.
L300–302: The 10–24-day delay in spring warming persists across all configurations and is not fully resolved by the density corrections. Please discuss whether melt-energy or snow-cover-fraction processes contribute to this remaining bias.
L310–312: Mean winter air temperature increases from −13 to −7 °C, but accumulated snowfall stays at 59–62 mm; please state whether precipitation phase is recomputed under warming, since constant snowfall with 6 °C warming seems unlikely at TVC.
L412–414: Dutch et al. (2024) is cited for a factor-of-3 overestimate of Keff. I believe this should be Dutch et al. (2022); please check.
L468–475: The claim that existing permafrost carbon estimates are "a conservative lower bound" is not supported by any carbon result in this paper; please either show fluxes or soften the claim.
L500–503 (Appendix B): Please define the formula for the performance score explicitly and give the scores for all 29 seasons in a supplementary table.

Reply

Citation: https://doi.org/10.5194/egusphere-2026-3363-RC1
RC2: 'Comment on egusphere-2026-3363', Mickaël Lalande, 02 Oct 2026 reply

This manuscript addresses an important issue, namely the poor representation of tundra snow properties in CLM5.0 and the resulting winter soil-temperature biases. The attempt to move beyond a simple adjustment of snow thermal conductivity by modifying snow density is potentially useful. However, I do not think the manuscript currently demonstrates that the RF approach provides a more physically realistic snowpack. The main result is that prescribing a lower snow density reduces thermal conductivity and warms the soil, thereby improving agreement with observed soil temperatures, but the snowpack itself is not evaluated sufficiently to establish whether this improvement occurs for the correct physical reasons. Snow depth, SWE, density, thermal conductivity or vertically integrated thermal resistance, snow/surface temperature, and soil temperature should be evaluated together. A further concern is the physical interpretation and implementation of the RF-derived density. The RF seems to predict a single bulk density for the whole snowpack, whereas CLM represents a multilayer snowpack. However, applying a single bulk density to derive each layer's thermal conductivity cannot reproduce the effective thermal behaviour of a strongly stratified Arctic snowpack, because the density–conductivity relationship is nonlinear and thermal resistances combine in series. This is particularly important for tundra snowpacks composed of low-density depth hoar beneath much denser wind slabs. Snow-pit and vertical-profile observations should therefore be used, together with snow depth and SWE, to assess whether the RF implementation actually improves snow stratigraphy and thermal resistance rather than only reproducing a reasonable bulk density. Finally, before comparing the RF against default CLM5.0 and extending the method to future projections, the highly unusual seasonal evolution of default CLM5.0 snow density should first be physically diagnosed. The model starts with extremely high early-winter densities and then exhibits a marked decrease through the season. Comparing the RF against this apparently non-physical baseline without first identifying and resolving the origin of this behaviour in the baseline snow model makes it difficult to determine whether the RF provides any added value. Addressing these issues would require substantial additional model diagnosis, evaluation, and development; I therefore recommend rejection in its present form, while encouraging resubmission of a fundamentally revised study.
Specific comments
Introduction
L42–44: The literature cited here could be expanded to better acknowledge previous efforts to improve Arctic snow representation in land-surface and snow models (e.g., Barrere et al., 2017; Gouttevin et al., 2018; Domine et al., 2019; Royer et al., 2021; Lackner et al., 2022; Woolley et al., 2024; Lalande et al., 2026). Since several of these studies concern land-surface or standalone snow schemes rather than complete ESMs, I also suggest slightly broadening/rephrasing the sentence, e.g., “land-surface snow schemes used within ESMs remain limited in their representation of these processes.”
L44–45: Consider adding Mavrovic et al. (2023, 2025).
L46–49: The statement that deficiencies in tundra snow representation produce systematic winter soil-temperature biases appears too strong. Snow representation is clearly an important contributor, but soil thermal properties, soil moisture/phase change, vegetation and organic layers, and permafrost/ground thermal processes may also contribute to the simulated bias. I suggest softening the wording to indicate that snow deficiencies contribute to rather than solely produce these biases, and, if possible, discussing previous attempts to improve those other processes in CLM and their possible contributions to soil temperature biases.
L57–60: Consider citing Domine et al. (2019) regarding known limitations of detailed snow models in reproducing Arctic snowpack properties, and Domine et al. (2018) alongside Bouvet et al. (2023) for the roles of soil moisture, temperature gradients and wind in depth-hoar formation.
L62–64: This paragraph appears to oversimplify the relationship between depth-hoar formation, bulk snow density and snow insulation. Low-density basal depth hoar can contribute strongly to snowpack thermal resistance, but tundra snowpacks are not necessarily characterised by low bulk density because dense wind slabs may constitute a substantial fraction of the snowpack. For example, Royer et al. (2021) report mean bulk densities of approximately 275, 315, and 409 kg/m3 for boreal forest, tundra, and polar-desert snow, respectively. I therefore suggest avoiding the statement that sustained depth hoar gives tundra snow a “characteristically low bulk density”. It would be more accurate to state that the coexistence of low-conductivity basal depth hoar and dense wind slabs produces a strongly stratified snowpack whose bulk thermal resistance is not necessarily captured by bulk density alone.
L67–69: As noted above, the statement that bulk snow density is generally overestimated in ESMs is too broad. Moreover, the sign of the model bias is clearly model-dependent. In CLASSIC, Lalande et al. (2026) found the opposite behaviour, with default bulk snow density underestimated by more than 50 kg/m3 at several Arctic sites because the original compaction formulation imposed maximum dry-snow densities that were too low for thin, wind-exposed snowpacks. It is therefore model- and parameterisation-dependent.
L68–75: Although density is an important first-order control on snow thermal conductivity, Keff is not uniquely determined by density. Snow microstructure, grain and bond geometry, anisotropy and metamorphic history can lead to substantially different conductivities at the same density. Laboratory studies have explicitly shown that Keff can evolve significantly during temperature-gradient metamorphism even when snow density remains essentially unchanged (e.g., Schneebeli & Sokratov, 2004; Satyawali et al., 2008). This is particularly relevant for Arctic depth hoar, which is strongly anisotropic and undergoes pronounced microstructural evolution (e.g., Domine et al., 2011; Löwe et al., 2013; Calonne et al., 2019; Fourteau et al., 2021). The authors should acknowledge this limitation and be less categorical in stating that density drives thermal conductivity. Using density as a first-order proxy for Keff is nevertheless a reasonable and widely used approximation in current snow models, as in density-based parameterisations such as Calonne et al. (2011), but an improved bulk density does not necessarily imply a more physically realistic thermal conductivity.
L75–80: The interpretation of the Sturm et al. (1997) formulation should be more nuanced. A clear strength of Sturm et al. (1997) is that its dataset includes natural high-latitude snow, including depth-hoar-type snow, whereas Calonne et al. (2011) was not calibrated specifically against representative natural Arctic depth hoar. However, the Sturm relationship is based on heated-needle measurements, and this method can underestimate snow thermal conductivity because of contact-resistance effects and assumptions inherent to the measurement technique, with potentially important errors for fragile and strongly anisotropic depth hoar (Fourteau et al., 2022). Therefore, neither formulation can presently be considered unambiguously more appropriate for natural Arctic depth hoar. Second, the statement that the winter soil-temperature cold bias is caused by excessive snow density and Keff is stronger than what the previous studies demonstrate. Dutch et al. (2022) and Damseaux et al. (2025) clearly show that soil temperature is highly sensitive to the conductivity parameterisation, but they do not comprehensively show good overall model performance across snow depth, SWE, density, stratigraphy, thermal resistance, and the soil/vegetation state. The improvement obtained with Sturm may therefore partly reflect compensation among snowpack biases rather than identification of the unique physical source of the soil-temperature error. This interpretation should be presented as a plausible contributing mechanism rather than an established causal diagnosis.
L80–95: The ability of RF models to represent nonlinear relationships is not in question, but nonlinear empirical parameterisations are also widely used in process-based snow models (e.g. density-dependent thermal conductivity or compaction formulations). Furthermore, the manuscript should distinguish between capturing nonlinear statistical relationships between meteorological predictors and bulk density and explicitly representing the physical processes responsible for tundra snow evolution. The RF used here emulates ASC bulk-density evolution but does not explicitly represent depth-hoar formation, vapour transport, wind-slab formation or snow metamorphism. I therefore suggest avoiding wording such as “captures the physical processes driving tundra snowpack evolution” and referring instead to the nonlinear controls or statistical relationships learned from the ASC simulations.
L117–120: Please provide the absolute errors in addition to the relative 41% reduction (for both depth hoar and wind slab if possible), which would give a more balanced indication of both the improvement and the remaining model errors.
Methods
L135–150 and Sect. 2.6: More information is needed on how TVC is represented in CLM5.0. The site is described as highly heterogeneous in vegetation, topography and snow redistribution, yet the manuscript does not clearly state the prescribed PFT(s), soil properties, soil initialisation, vegetation fractions, or whether hummock/microtopographic effects are represented. These choices are important when attributing soil-temperature biases primarily to snow properties. Please provide the relevant site-specific CLM5.0 configuration rather than relying only on the reference to Rutherford et al. (2025), and discuss how representative the selected model point is of the snow and soil observations used for evaluation. Please also discuss whether the snow-cover-fraction parameterisation is activated or not for point-scale simulation. The role of wind redistribution also deserves more attention. The manuscript itself notes that snow depth at TVC differs strongly between drift and tundra locations, yet the forcing/model framework appears to treat snowfall as a local one-dimensional input. Please clarify whether any snow redistribution, erosion or deposition is represented, and how the locations of meteorological forcing, snow-course measurements and soil-temperature observations relate to one another spatially.
L145–150: Precipitation uncertainty should be discussed in greater detail because snowfall rate and cumulative snowfall are direct RF predictors. Arctic weighing-gauge records can remain highly uncertain even after standard filtering and undercatch correction. For example, in ongoing work at Bylot Island, a single-wire Geonor record showed substantial event-scale ambiguities associated with increased instrumental noise, possible snow/ice capping or bridging at the gauge inlet, and delayed release events. Reliable snowfall reconstruction therefore requires multi-source quality control using time-lapse cameras, independent snow-height stakes/sensors, wind observations, and additional ancillary datasets. Similar issues may affect TVC, particularly because snow redistribution is strong there. Please clarify the TVC precipitation instrumentation, shielding, gap filling, undercatch correction and any independent event-level validation used to assess the reliability of snowfall and cumulative snowfall. Without a sufficiently constrained snowfall forcing and satisfactory simulations of snow depth and density, it is difficult to assess subsequent biases in snow thermal conductivity and soil temperature.
L155–165: Please provide the dates and values of the snow-course measurements used to select the ASC training seasons.
L165–180: Show feature importance. The exclusion of wind and dominance of cumulative snowfall deserve discussion, particularly given the manuscript's emphasis on wind slab.
L213–230: Clarify the distinction between layer density and bulk snow density and explain exactly how one RF-derived bulk density is mapped onto the multilayer CLM snowpack. The terminology appears inconsistent: ρ_c,jis a layer-resolved density, described as “snow layer bulk density”, whereas ρ_ML is a whole-snowpack bulk density predicted by the RF. These quantities should not be referred to interchangeably as “bulk snow density”. Is the RF predicting each snow layer density, or only the bulk snow density (i.e., the whole snowpack)? Is the RF able to predict the inverted density profiles expected for Arctic snowpacks? And if not, do you expect the thermal regime within the snowpack to be correct or not?
You also describe m_ice and m_liq as fractions of ice and liquid, while in Appendix C they are described as “Layer ice mass per unit area”, which seems more correct in terms of units.
L270–285 / Figure 2: Before the RF approach can be meaningfully evaluated against CLM5.0, the apparently non-physical evolution of the default CLM5.0 snow density needs to be diagnosed and resolved. CLM5.0 starts the season with extremely high bulk densities (~450–500 kg/m3, occasionally reaching ~650 kg/m3) and then shows a persistent decrease toward ~350 kg/m3 through winter. Although decreases in whole-snowpack bulk density can occur through fresh snowfall, changes in snow-cover fraction, or layer restructuring, the magnitude and persistence of this behaviour require a clear process-level explanation. The authors should first identify whether it originates in the fresh-snow-density formulation, the compaction scheme, melt/refreezing processes, or how the multilayer model diagnoses whole-snowpack bulk density, and evaluate the associated snow depth, SWE, layer densities, and thicknesses. In its current form, the study moves directly from an unresolved and apparently unrealistic CLM5.0 density evolution to replacing it statistically, making it impossible to determine whether the RF corrects a genuine physical deficiency or simply bypasses a model/configuration problem. This is particularly concerning if the RF provides only a single whole-snowpack bulk density that is then applied across the multilayer snowpack: such an approach removes the influence of the physically important vertical density structure from the thermal-conductivity calculation, especially the contrast between low-density depth hoar and dense wind slabs that controls thermal resistance. This baseline physical problem should therefore be resolved before presenting the RF as an improvement or extending the analysis to future projections.
Concluding remarks and recommendation
In my view, these issues cannot be resolved through additional discussion alone. They require a process-level diagnosis and, where necessary, further development and evaluation of the default CLM5.0 snowpack, clarification and potentially revision of the RF-to-CLM implementation, and a substantially broader present-day snowpack assessment before future projections can be interpreted with confidence. I therefore recommend rejection in its present form, while encouraging resubmission once these issues have been addressed. The use of an RF approach is not inherently inappropriate, particularly given the continuing difficulty of physically reproducing the strongly stratified density profiles of Arctic snowpacks. However, the authors must demonstrate its added value rather than assume it. In particular, a revised study should show that the RF is able to reproduce the characteristic vertical structure and seasonal evolution of Arctic snow density, including the contrast between low-density depth hoar and dense wind slabs, and that this leads to a more physically consistent representation of snow depth, SWE, thermal resistance and snow–soil heat transfer.
Technical comments
L55–60: update Damseaux et al. to the final 2025 publication.
L215: missing table number.
Eq. 3: mice and mliq are masses per unit area, not fractions.
L497: “RSME” → “RMSE”.

Mickaël Lalande
References
Barrere, M., Domine, F., Decharme, B., Morin, S., Vionnet, V., & Lafaysse, M. (2017). Evaluating the performance of coupled snow–soil models in SURFEXv8 to simulate the permafrost thermal regime at a high Arctic site. Geoscientific Model Development, 10(9), 3461–3479. https://doi.org/10.5194/gmd-10-3461-2017
Calonne, N., Flin, F., Morin, S., Lesaffre, B., du Roscoat, S. R., & Geindreau, C. (2011). Numerical and experimental investigations of the effective thermal conductivity of snow. Geophysical Research Letters, 38(23), L23501. https://doi.org/10.1029/2011GL049234
Calonne, N., Milliancourt, L., Burr, A., Philip, A., Martin, C. L., Flin, F., & Geindreau, C. (2019). Thermal Conductivity of Snow, Firn, and Porous Ice From 3-D Image-Based Computations. Geophysical Research Letters, 46(22), 13079–13089. https://doi.org/10.1029/2019GL085228
Damseaux, A., Matthes, H., Dutch, V. R., Wake, L., & Rutter, N. (2025). Impact of snow thermal conductivity schemes on pan-Arctic permafrost dynamics in the Community Land Model version 5.0. The Cryosphere, 19(4), 1539–1558. https://doi.org/10.5194/tc-19-1539-2025
Domine, F., Belke-Brea, M., Sarrazin, D., Arnaud, L., Barrere, M., & Poirier, M. (2018). Soil moisture, wind speed and depth hoar formation in the Arctic snowpack. Journal of Glaciology, 64(248), 990–1002. https://doi.org/10.1017/jog.2018.89
Domine, F., Bock, J., Morin, S., & Giraud, G. (2011). Linking the effective thermal conductivity of snow to its shear strength and density. Journal of Geophysical Research: Earth Surface, 116(F4). https://doi.org/10.1029/2011JF002000
Domine, F., Picard, G., Morin, S., Barrere, M., Madore, J.-B., & Langlois, A. (2019). Major Issues in Simulating Some Arctic Snowpack Properties Using Current Detailed Snow Physics Models: Consequences for the Thermal Regime and Water Budget of Permafrost. Journal of Advances in Modeling Earth Systems, 11(1), 34–44. https://doi.org/10.1029/2018MS001445
Dutch, V. R., Rutter, N., Wake, L., Sandells, M., Derksen, C., Walker, B., Hould Gosselin, G., Sonnentag, O., Essery, R., Kelly, R., Marsh, P., King, J., & Boike, J. (2022). Impact of measured and simulated tundra snowpack properties on heat transfer. The Cryosphere, 16(10), 4201–4222. https://doi.org/10.5194/tc-16-4201-2022
Fourteau, K., Domine, F., & Hagenmuller, P. (2021). Impact of water vapor diffusion and latent heat on the effective thermal conductivity of snow. The Cryosphere, 15(6), 2739–2755. https://doi.org/10.5194/tc-15-2739-2021
Fourteau, K., Hagenmuller, P., Roulle, J., & Domine, F. (2022). On the use of heated needle probes for measuring snow thermal conductivity. Journal of Glaciology, 68(270), 705–719. https://doi.org/10.1017/jog.2021.127
Gouttevin, I., Langer, M., Löwe, H., Boike, J., Proksch, M., & Schneebeli, M. (2018). Observation and modelling of snow at a polygonal tundra permafrost site: Spatial variability and thermal implications. The Cryosphere, 12(11), 3693–3717. https://doi.org/10.5194/tc-12-3693-2018
Lackner, G., Domine, F., Nadeau, D. F., Lafaysse, M., & Dumont, M. (2022). Snow properties at the forest–tundra ecotone: Predominance of water vapor fluxes even in deep, moderately cold snowpacks. The Cryosphere, 16(8), 3357–3373. https://doi.org/10.5194/tc-16-3357-2022
Lalande, M., Roy, A., Wang, L., Verseghy, D., Vionnet, V., Dominé, F., & Kinnard, C. (2026). Improving the CLASSIC (v1.8) Snow Model to Better Simulate Arctic Snowpacks. EGUsphere, 1–57. https://doi.org/10.5194/egusphere-2026-492
Löwe, H., Riche, F., & Schneebeli, M. (2013). A general treatment of snow microstructure exemplified by an improved relation for thermal conductivity. The Cryosphere, 7(5), 1473–1480. https://doi.org/10.5194/tc-7-1473-2013
Mavrovic, A., Sonnentag, O., Lemmetyinen, J., Voigt, C., Rutter, N., Mann, P., Sylvain, J.-D., & Roy, A. (2023). Environmental controls of winter soil carbon dioxide fluxes in boreal and tundra environments. Biogeosciences, 20(24), 5087–5108. https://doi.org/10.5194/bg-20-5087-2023
Mavrovic, A., Sonnentag, O., Voigt, C., Lemmetyinen, J., Aurela, M., & Roy, A. (2025). Winter Methane Fluxes Over Boreal and Arctic Environments. Geophysical Research Letters, 52(21), e2025GL118367. https://doi.org/10.1029/2025GL118367
Royer, A., Domine, F., Roy, A., Langlois, A., Marchand, N., & Davesne, G. (2021). New northern snowpack classification linked to vegetation cover on a latitudinal mega-transect across northeastern Canada. Écoscience, 28(3–4), 225–242. https://doi.org/10.1080/11956860.2021.1898775
Royer, A., Picard, G., Vargel, C., Langlois, A., Gouttevin, I., & Dumont, M. (2021). Improved Simulation of Arctic Circumpolar Land Area Snow Properties and Soil Temperatures. Frontiers in Earth Science, 9(June), 1–19. https://doi.org/10.3389/feart.2021.685140
Rutherford, J., Rutter, N., Wake, L., & Cannon, A. J. (2025). Snow thermal conductivity controls future winter carbon emissions in shrub–tundra. Biogeosciences, 22(18), 5031–5049. https://doi.org/10.5194/bg-22-5031-2025
Satyawali, P. K., Singh, A. K., Dewali, S. K., Kumar, P., & Kumar, V. (2008). Time dependence of snow microstructure and associated effective thermal conductivity. Annals of Glaciology, 49, 43–50. https://doi.org/10.3189/172756408787814753
Schneebeli, M., & Sokratov, S. A. (2004). Tomography of temperature gradient metamorphism of snow and associated changes in heat conductivity. Hydrological Processes, 18(18), 3655–3665. https://doi.org/10.1002/hyp.5800
Sturm, M., Holmgren, J., König, M., & Morris, K. (1997). The thermal conductivity of seasonal snow. Journal of Glaciology, 43(143), 26–41. https://doi.org/10.3189/S0022143000002781
Woolley, G. J., Rutter, N., Wake, L., Vionnet, V., Derksen, C., Essery, R., Marsh, P., Tutton, R., Walker, B., Lafaysse, M., & Pritchard, D. (2024). Multi-physics ensemble modelling of Arctic tundra snowpack properties. The Cryosphere, 18(12), 5685–5711. https://doi.org/10.5194/tc-18-5685-2024

Reply

Citation: https://doi.org/10.5194/egusphere-2026-3363-RC2

Johnny Rutherford, Nick Rutter, Leanne Wake, Georgina Wooley, Julia Boike, and Alex J. Cannon

Viewed

Total article views: 333 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
195	97	41	333	33	26

HTML: 195
PDF: 97
XML: 41
Total: 333
BibTeX: 33
EndNote: 26

Views and downloads (calculated since 20 Jul 2026)

Month	HTML	PDF	XML	Total
Jul 2026	70	23	8	101
Aug 2026	19	16	2	37
Sep 2026	70	42	16	128
Oct 2026	36	16	15	67

Cumulative views and downloads (calculated since 20 Jul 2026)

Month	HTML	PDF	XML	Total
Jul 2026	70	23	8	101
Aug 2026	19	16	2	37
Sep 2026	70	42	16	128
Oct 2026	36	16	15	67

Viewed (geographical distribution)

Total article views: 316 (including HTML, PDF, and XML) Thereof 316 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 12 Oct 2026

Short summary

The Arctic is warming rapidly, threatening to release vast carbon stores from frozen soils. Climate models consistently misrepresent Arctic snow, causing soils to appear artificially cold. We used a random forest model to better predict winter snow density and fed these improved estimates into a widely used climate model which raised projected soil temperatures by up to 7 °C, suggesting current forecasts of soil temperature and carbon release are likely underestimated.


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