the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 11 Aug 2025
Assessing spatial heterogeneity of active layer thickness over Arctic-foothills tundra through intensive field sampling and multi-source remote sensing
Abstract. Active layer thickness (ALT) is an essential climate variable for monitoring permafrost degradation, whose deepening can lead to increased greenhouse gas emissions, altered hydrology and ecology, infrastructure damage, and a positive climate feedback. Quantifying ALT spatial heterogeneity remains challenging due to the influence of localized variations in terrain, microclimate, snow/soil properties, vegetation cover, and surface disturbances. It is also unclear how local ALT patterns (e.g., sub-meter to 10 m) and mechanisms scale up to broader landscape footprints (e.g., 10–1000 m) represented from global satellite observations and Earth system models. We assessed ALT spatial heterogeneity in the Arctic-foothills tundra within the Northern Slope of Alaska through intensive field sampling over four 90 m × 90 m plots, combined with multi-source remote sensing and machine learning (ML). Analysis using field observations and ML revealed that vegetation, surface wetness, subsurface rocks, and micro-topography exert strong influence on 5-m ALT variations, whereas terrain controls dominate (~65 % contribution) at coarser 10-m spatial resolution. By leveraging cm-level optical-infrared drone imagery, we further generated 0.1-m ALT maps over a larger 5 km × 5 km region and examined ALT scaling effects. Our analysis showed a quadratic relationship in scale-dependent uncertainties, characterized by a rapid increase in uncertainties at the sub-meter level (e.g., RMSE normalized by the standard deviation of 0.1 m ALT climbed by ~10 %), followed by another 10 % increase from 1 m to 30 m resolution, and more conservative error increase (~5 %) from 30 m to 1,000 m scale. Our study allows for improved interpretation of remote sensing and process-based ALT simulations for the changing Arctic by clarifying scale-dependent uncertainties and underlying mechanisms.
- Preprint
(1225 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 22 Sep 2025)
- RC1: 'Comment on egusphere-2025-3236', Anonymous Referee #1, 25 Aug 2025 reply
-
RC2: 'Comment on egusphere-2025-3236', Anonymous Referee #2, 01 Sep 2025
reply
The manuscript by Du et al. investigates active layer thickness (ALT) in the North Slope of Alaska using a combination of machine learning, intensive field sampling, and multi-source remote sensing data. The authors address ALT variability across multiple spatial scales. The main findings highlight the drivers of ALT at different scales. Additionally, the study quantifies scale-dependent uncertainties in ALT mapping and identifies functional relationships describing how these uncertainties change with spatial resolution. The paper is well-written and addresses an important topic in permafrost research. There are, however, some points that could be further discussed and a number of comments that should be addressed.
The parameter importance at different resolutions is reported and supported by previous studies in the discussion section (L292-302). The discussion could be extended to provide more in-depth interpretation of the underlying mechanisms and implications.
Using the 0.1 m results as a benchmark raises some questions. In Figure 4 (especially panel f), these values appear to differ from both in situ measurements and the 5 m results, which could be further discussed. Since the 0.1 m data are synthetically generated for the scaling effects analysis, it would be helpful if the authors could comment on any potential influence on the results.
Furthermore, the set of predictors chosen for the 5 m and 10 m models are not the same, with many variables omitted from the 10 m model. Could the authors clarify the rationale behind this choice? Since red-edge reflectance has relatively high importance at 5 m, it could have been valuable to investigate the 10 m red-edge from Sentinel-2 by performing super-resolution on the 20 m red-edge band. Additionally, elevation, which appears to have the highest importance for the 10 m model, was not included for the higher-resolution 5 m model. Could the authors comment on these choices and their potential implications?
For the machine learning predictors at 5 m resolution, the DEM derived from RGB imagery was used. Were any differences investigated between this DEM and the ArcticDEM aggregated to the same 5 m resolution? Could the authors discuss any potential consequences of using the RGB-derived DEM compared to ArcticDEM?
The study focuses on a specific region. Could you comment on the transferability of the model to other regions? Would additional in situ training data be required for applying the approach elsewhere? Are similar parameter importances and scaling effects to be expected in other regions, and how representative is the study area for the broader Arctic context?
You are mentioning the importance of not only ALT spatial distributions but also temporal dynamics (L44). Could you maybe comment on the feasibility of resolving ALT temporal dynamics with the current approach? For instance, what is the potential for detecting interannual variations in ALT with the presented approach? In addition, since the Sentinel-2 data originate from multiple summer months (L151), could the authors comment on how intra-seasonal variability might influence the results?
Further comments:
The capitalization in the section titles is not consistent.
L22: North Slope not Northern Slope
L62-63: You mention ALT derivation from LiDAR and from InSAR deformation signals driven by soil freeze–thaw. This connection needs further explanation as ground deformations do not directly translate to active layer thickness estimates without additional assumptions.
L64ff: You state that low-frequency microwave measurements show strong potential for mapping ALT. However, the connection to active layer thickness is not entirely clear, since ALT cannot be directly measured with these observations. Could the authors elaborate on the underlying mechanism or clarify how these measurements can be translated into ALT estimates?
L75: The manuscript mentions “direct microwave sensing of soil profiles” for ALT. However, this is somewhat misleading, as ALT is not directly measurable from microwave observations.
L92: Could you please provide information on the method on which the regional records are based?
L94: The numbering of the study plots starts at Plot 3 rather than Plot 1. Could you clarify why Plots 1 and 2 are not included or why the numbering begins at 3?
L100: Is there a specific reason why the Canada Albers Equal Area Conic projection was chosen for Alaska? This projection results in scenes that are not north-oriented, with north arrows appearing consistently tilted. Could the authors clarify the rationale behind this choice?
Figure 1: It is unclear whether the black rectangles represent the actual footprints of the RGB images, as the displayed RGB images appear to be rotated relative to these rectangles.
L146: The method chosen to aggregate the data is not specified and should be clarified.
L163: re-constucting or reconstructing
L163-165: I find this sentence and Figure 3 somewhat misleading and suggest revisions to avoid confusion. In Figure 3, the arrow from the ‘0.1 m predictors’ to the model could incorrectly suggest that the model was trained at 0.1 m, whereas it was actually applied at that resolution. This should be represented in a way that clearly distinguishes training from application/prediction to avoid any confusion. Additionally, the arrows from ‘Aggregated to 10 m ALT’ and ‘Field ALT sampling’ connect to the predictors, which does not seem correct and should rather lead into the models. Additionally, ‚Scaling analysis‘ should be capitalized for consistency.
Table 1: aspect (on the right side) should be capitalized for consistency
Figure 4: The coordinate labels are difficult to read and should be enlarged for better readability.
L247-249: This statement may be misleading, since NDVI (19.85%) is more important than aspect (17.85%) or slope (17.16%).
Figure 5: It would be helpful to also include elevation, since it is reported as the most important factor.
L296-298: Only the change in R is reported for radar-based observations; it would be helpful to also show their importance relative to other predictors.
Citation: https://doi.org/10.5194/egusphere-2025-3236-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 1,482 | 27 | 7 | 1,516 | 52 | 52 |
- HTML: 1,482
- PDF: 27
- XML: 7
- Total: 1,516
- BibTeX: 52
- EndNote: 52
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Du et al. have submitted an interesting paper that investigates the spatial heterogeneity of the active layer thickness for a region on the North Slope of Alaska. The study combines field sampling, remotely sensed imagery and modelling. The analysis considers the relative influence of various climate and environmental factors on active layer conditions at different spatial resolutions. The study has potential to contribute to improved prediction of active layer thickness in a warming climate. However, I do have some concerns and comments that should be addressed for the manuscript to be acceptable for publication.
Several previous studies have considered the relative influence of environmental factors on the ground thermal regime through analysis of field data collected across environmental gradients and some of the conclusions in the manuscript are not new. Recent papers, including those by some of the coauthors of the submitted manuscript have also considered quantification of active layer at multiple scales in Alaska (e.g. Brodylo et al. 2024). A better description of the novelty and advancement in knowledge of the submitted manuscript compared to earlier studies would be beneficial.
The manuscript would benefit from a better description of the study plots and the broader region considered, such as information on surficial materials, vegetation and topography. Since the focus of the paper is spatial heterogeneity, it would be useful for the reader to have information on the spatial heterogeneity of the factors that are considered in the analysis within the study plots and in the broader area considered.
An improved presentation of results would be beneficial especially for comparison of ALT from field measurements and the modelled results. It is difficult for example, for the reader to compare the measured values to the ML outputs in figure 4 as the study plot area isn’t clear. The text refers to ALT variability near water bodies, but the results presented (i.e. maps) do not allow the reader to see this. Consideration of the accuracy of the models in terms of the entire area covered is fine but it is useful to consider which areas of the study area and under what conditions the accuracy is better. (see below for additional comments)
In the discussion and conclusions, general statements are made but it is unclear how the results and analysis support these statements. There needs to be a clearer link between results and conclusions.
The organization of some sections of the manuscript could be improved including Section 2 and 4.1 – see further comments below. Tables could also be considered for summarizing ALT conditions for study plots etc.
The authors appear to confuse scale and resolution. Scale and resolution are not the same thing. The authors refer to scale (e.g. finer, coarser scales) in the manuscript, but this is incorrect, and references should be made to resolution.
Editorial revisions have been suggested to improve clarity. Additional comments for the author’s consideration are provided below.
Additional Comments
Title – It might be sufficient to have a shorter title: “…..over Arctic-foothills tundra, North Slope Alaska” and delete the last part. I think you should mention the location as the study is specific to a region.
L16 – Revision suggested for first sentence” Changes in active layer thickness are used as an indicator of permafrost degradation” (I don’t think the ECV part is necessary).
L17-18 – A thickness doesn’t deepen, revise to: “Increases in ALT can….” Infrastructure damage results from ground instability so maybe that should be mentioned as changes in ALT do not directly cause infrastructure damage.
L33 – It should be clear that you are referring to atmospheric warming here rather than permafrost warming.
L34-37 – Consider revising the sentence. Some of these things are a result of permafrost degradation while other things mentioned may promote it. Deepening of a layer doesn’t sound right so refer to thicker active layer.
L40 – Biskaborn et al. (2019) was not about GHG emissions, so I suggest you delete it. You could also consider citing the review paper of Miner et al. (2022).
L41-42 – The accepted definition of the active layer comes from the IPA glossary (van Everdingen et al. 1998) so that should probably be cited. Establishment of active layer thickness as an essential climate variable is described in Smith and Brown (2009) so this is probably better reference than what you use.
L44 – You could just refer to spatial and temporal variability.
L47-49 – You are essentially saying that local microclimate is important.
L55 – Are you referring to process models that determine ALT here? – Obu et al. (2019) model simulates TTOP not ALT.
L61-66 – ALT is inferred from information acquired using these techniques. In the case of geophysical techniques, there can be other factors that result in similar signals. Most geophysical techniques are used to determine frozen/unfrozen interfaces, but this also requires knowledge of geology etc. to make the interpretations. Techniques like InSAR are used to determine changes in surface elevation but freezing and thawing are not the only reason movements of the ground occur. Other remote sensing techniques provide information that is used with thermal models to simulate ALT etc. It is difficult to say that any of these techniques are direct measures of ALT, and you should probably say that they are used to infer or provide data for models to estimate ALT.
L75-77 - Inferred through modelling?
L85 – Do you mean “characterize” or “assess” rather than clarify? I think other studies have determined the various controls. Are you investigating the relevant importance of these?
L87-115 – Study Region section. Normally this would include a general description of the regional setting – climate, geology, vegetation etc. and then details of the study plots would be provided. This section could benefit from better organization.
L88-89 – Normally a more general description of regional climate would be presented first, and this sentence seems out of place (see previous comment). Provide reference period for statements such as this and be clear that it is air temperature (rather than permafrost temperature) that is rising by the amount indicated.
L100 – information on map projections should be provided in the figure caption.
L105 – Figure 1 – If the images include the plot area, then it should be clear what area of them is covered by the plot. The orientation of the plots and images differ so it is difficult for the reader to see the characteristics of the plots. It is also unclear if the scale of map and the images is the same.
L125-126 – It would be better to indicate that mechanical probing to the depth of refusal was used to determine the depth of the frost table. Essentially that is what is done when probing is conducted.
L129-131 – Revision suggested: “Additional observations were made including vegetation type and distribution, occurrence of standing and running water….” Observation of subsurface rocks is mentioned but were there also descriptions of organic layer thickness or surficial materials which are relevant for interpretation of results.
L169-170 – The way this is written it sounds like ALT is directly determined through remote sensing. Don’t you mean that the RF method is used with parameters determined through remote sensing to estimate ALT. Doesn’t the ALT used to develop the model come from the field observations?
L224-255 Section 4.1.1 – It would be better to present results for each plot first and then compare them before giving overall statistics. Clearly there are differences between the plots, and they should be described first. A better presentation of the results of the field sampling should be provided before presenting results of the RF models and the comparison to observed ALT.
L236-240 – These features do not appear to be visible on the maps in Figure 4 so difficult for the reader to see how you arrive at these interpretations. There appears to be substantial difference between Plot 5 (g) observed ALT and modelled (e) – observed values appear to be less than modelled.
Figure 4 – The presentation does not allow the reader to compare the observed to the model outputs as it is unclear how the plot area in first column fits on the maps in the other two columns. The plot area should be clearly shown on the other plots. For plot 6 the rest of the plot area should be shown in (j) with grey shading for example to indicate area that couldn’t be probed.
L245-255 – We would expect warmer conditions and greater ALT on south facing vs north facing slopes – can you say anything about this based on observed results. Note that some of the factors considered are related. For example, vegetation will depend on elevation and aspect. Drainage and therefore surface wetness (affects vegetation) will depend on topography.
L262-264 – Words like “rapid” and “slower” imply that a change over time is being considered but that is not the case here. It would be better to refer to is a smaller increase in uncertainty at coarser (or lower) resolution (note on you map in Figure 4 the x axis appears to be the resolution, not scale – resolution and scale are not the same thing).
L276-278 – “Coarser scale” is incorrect, it should be “coarser resolution”. Air temperature affects surface temperature (as do local environmental factors that affect microclimate) which influence the ground thermal regime (ground temperature) and therefore active layer conditions.
L275-302 – Section 5.1 – There are a lot of general statements from the literature, but very little analysis is presented to show the relative importance of the various factors mentioned. Information on snow cover, soil texture, groundwater flow etc. has not been presented and it is unclear how these things may vary over the study area. You mention that ALT is greater in areas with standing water or adjacent to creeks but not clear from results presented (e.g. maps) that this is the case.
L282-285 – Revise “deeper ALT” to “greater ALT” (a thickness can’t be deeper – same issue with shallow ALT). Latent heat is also an important factor with respect to the effect that wet conditions have on the ground thermal regime.
L287-290 – Note that shrubs can promote snow accumulation and other studies have shown that this leads to warmer winter ground temperatures (e.g. Palmer et al. 2012; Morse et al. 2012; Way and Lapalme 2021; Kropp et al. 2020) – winter conditions will influence ALT and it is not as simple as implied in the text. Way and Lapalme (2021) also showed that the insulating effect of snow outweighs the shading effect of shrubs.
L306 – I think you mean greater uncertainty in ALT at sub-metre resolution.
L322-323 – This is a general statement but not a conclusion of your study – there was no investigation of GHG emission, infrastructure issues etc.
L350 – References – check URL links numbers as some of them do not seem to work. I noticed this with a few ERL publications.
L599 – Biskaborn et al. has many more coauthors so “and coauthors” should be added after the last author given. Same comment for Obu et al. in line 457.
References cited in comments
Kropp, H. et al., 2021. Shallow soils are warmer under trees and tall shrubs across Arctic and Boreal ecosystems. Environmental Research Letters, 16: 015001. doi: 10.1088/1748-9326/abc994
Miner, K.R., Turetsky, M.R., Malina, E., Bartsch, A., Tamminen, J., McGuire, A.D., Fix, A., Sweeney, C., Elder, C.D., and Miller, C.E. 2022. Permafrost carbon emissions in a changing Arctic. Nature Reviews Earth & Environment, 3: 55-67. doi:10.1038/s43017-021-00230-3
Morse, P.D., Burn, C.R., and Kokelj, S.V. 2012. Influence of snow on near-surface ground temperatures in upland and alluvial environments of the outer Mackenzie Delta, Northwest Territories. Canadian Journal Earth Sciences, 49: 895-913. doi:10.1139/E2012-012
Palmer, M.J., Burn, C.R., and Kokelj, S.V. 2012. Factors influencing permafrost temperatures across tree line in the uplands east of the Mackenzie Delta, 2004–2010. Canadian Journal of Earth Sciences, 49: 877-894. doi:10.1139/E2012-002
Smith, S. and Brown, J., 2009. Permafrost: permafrost and seasonally frozen ground, T7. Global Terrestrial Observing System GTOS 62, Food and Agriculture Organization of the United Nations (FAO), Rome.
Way, R.G., and Lapalme, C.M. 2021. Does tall vegetation warm or cool the ground surface? Constraining the ground thermal impacts of upright vegetation in northern environments. Environmental Research Letters, 16: 054077. doi:10.1088/1748-9326/abef31