"Quantifying Retrogressive Thaw Slump Mass Wasting and Carbon Mobilisation on the Qinghai-Tibet Plateau Using Multi-Modal Remote Sensing" by Maier et al. is the first study to quantify the impact of RTS on erosion, soil carbon mobilization, and ground ice loss across the entire QTP. Overall, the methods and analyses are well designed and the results represent a meaningful contribution to our current understanding of RTS. However, significant improvement in the clarity of the figures/maps and language is needed before I can recommend publication.

Major comments:

1. In general, I found the organization of the manuscript to need some adjustment, as there was quite a bit of material that seemed to be in the wrong section. For example, the final paragraph of the introduction read more like methods, while the second paragraph of section 2.1 read more like the introduction with numerous citations of other studies. Additionally, paragraphs should be arranged by topic with strong topic sentences. Currently, a lot of paragraphs feel like they are an amalgamation of various topics, many of which are addressed multiple times in multiple paragraphs.

2. All of the figures could use some refinement to improve clarity and visibility, both in the figures themselves and in the associated captions. Details are provided below.

3. I found the distinction between the methods used with the delineations from Xia et al. 2024 and the manual delineations from the DEM difference map difficult to understand quickly. Although (I think) I eventually figured it out after bouncing back and forth between methods, figures, and results text, this point is central to the manuscript and should be clarified.

4. It seems like an ALT of 1 m was assumed in the ground ice loss calculations, despite the fact that the stated average ALT of the QTP is over 2 m. It was unclear to me whether there may be a reason for this discrepancy or if the analysis should be re-run with a different assumed ALT depth.

Section 2.1, paragraph 2: This reads like the introduction section. I would suggest combining and streamlining this information with the information already included about the QTP in the introduction.

Section 2.2: There is already an overview of the methods at the end of the introduction which reads better than this section, so I would highly recommend replacing everything before section 2.2.1 with that section. I found this section too vague as it left me with more questions than answers.

Section 3 (before 3.1): I don't think the commentary on how the results are structured is necessary, as it is only two sections and it follows the structure of the methods.

Section 4.1: The organization of the last two paragraphs in this section could use some work. It is unclear to me which one topic is being discussed in each. Instead it feels like a lot of different topics are mentioned briefly and the comparison between the Arctic and QTP shows up in both paragraphs. Please reorganize and ensure that the topic sentences adequately describe the overarching topic of each paragraph.

Figures/Tables:

Fig. 1.a Although the reader should be able to infer that any pixel with RTS has permafrost, this is not possible to see in this map. Using a hexagonal grid with both color and size for RTS density would allow readers to see both the ground type and RTS density layers across the entire map. For example: https://www.esri.com/arcgis-blog/products/arcgis-pro/mapping/how-to-turn-a-ton-of-overlapping-data-into-a-hexagon-map. I think this would be a great use of this technique and would really make this map pop!

Validation sites show up in the legend but not on the map. Please make sure they are visible. I would recommend including this information here and removing it from Fig. 5, which is really busy.

The West-Central and Central clusters look like one cluster. Are they really different? More information needed in the methods.

1.b. why 2-3 m only? Is this to approximate layers below the active layer?

Fig. 2. "delta A" and "delta V" in caption cannot stand alone. Please spell out the variables rather than using their acronyms.

2a. Not being an expert in radar remote sensing, it is unclear what "geocoding" is and why it shows up twice here.

Fig. 3. I'd also recommend the variable size and color hexagonal grid here. Panels b. and c. are quite small.

Fig. 4.

- "First row displays a box plot of the quantity’s distribution per RTS, the second row the total quantity per subregion, and the third row additional data." This is a bit unclear. I would try something like, "The first row displays a box plot for values associated with individual RTS, the second row displays the total quantity across subregions, and the third row displays additional data that vary between columns."

- I assume the color fill is based on region, not some underlying numerical value associated with those regions? If so, I would suggest retaining the different colors across columns to reflect the variable being displayed, but removing the different shades associated with the regions. Using a similar shading scheme to Fig. 3 made me think at first that it might reflect different average values across regions.

- I assume the dashed lines reflect mean (or median?) values across all the regions? Please add this information to the caption.

Fig. 5. I would recommend removing panel a from this figure by moving the table to its own table and only including the validation sites in Fig. 1. It might also make sense to change the layout to allow panel b to be larger. I imagine panel b on the left with a row of panels c-f on the right would improve the visibility considerably. Captions c-e provide commentary on the meaning of the data without clearly describing what the figures show.

5.b. 2018-2020 delineations are not easy to see at this size. Different colors or linetypes may be necessary, although increasing the size may be sufficient. Personally, I would also flip the order of the years in the legend so that they are ascending. Red delineations on top of green imagery is not great for people with color blindness. I have found yellow to be quite visible on top of imagery, although I'm not sure it would work on the DEM layer.

5.c. I cannot tell what the difference is between the bars and the red-dotted line from the caption, including why there is a single dotted line across all years, but three different bars across the years.

5.d. It is unclear what the mean, minimum and maximum values represent.

5.e. Same as c and d

5.f. "We compare average headwall height of 6 RTS in the Beiluhe River Basin (Central QTP) from VHR DEM (2020) to the maximum elevation loss δhmax of the TanDEM-X DEM and the 2019 optical RTS delineation." The way the caption is phrased it makes it sound like the VHR DEM values are being compared to both the TanDEM-X DEM and the optical delineations rather than being compared to a value derived from both of those products. This could be fixed with "We compare average headwall height of 6 RTS in the Beiluhe River Basin (Central QTP) derived from VHR DEM (2020) to those derived from the maximum elevation loss δhmax of the TanDEM-X DEM and the 2019 optical RTS delineation."

Fig. 6.a. Try a less transparent background on the legend to make it more readable. Make sure the lat and lon graticules are drawn underneath the points to improve visibility of the points. Solid points may also be a better choice for visibility. What is the background raster layer and is it necessary? It also contributes to the difficulty in seeing the points. What do the letters represent? Are they different study sites within the listed publications?

Fig. 6.b. The caption provides commentary without fully describing the figure. The caption contains a colon, the meaning of which is unclear to me. The caption mentions differences based on continent, but it takes a lot of effort to compare between panels a and b to figure out which letters correspond to which continents. I would suggest adding an annotation to the x-axis or somehow grouping the bars to show the continent. If it's not too busy, it would also be useful to add an annotation to the x-axis showing which sites come from which publications.

Table 1. Both SOC mobilization columns are rates, so I'm not sure the names make sense. Maybe "areal SOC mobilization rate" and "individual" or "feature-level" SOC mobiliation rate?

Table S2. What is the difference between OC and OC flux? I didn't think that any fluxes were calculated in this study, just mobilization.

Table S3. Why isn't error reported for the change in area when it is reported for the change in volume?

Line edits:

L26: "air temperatures rise" would be better as "air temperature increases"

Paragraph 1-2 transition is abrupt. A final sentence in P1 that connects the two would be helpful.

L39 and elsewhere: "QTP" should always be preceded by "the"

L42-44: "The QTP is susceptible to permafrost thaw processes that cast strong impacts from permafrost degradation, including threats to local transport and energy infrastructure and ecosystems, as well as to regional climate and water storage". Unclear, specifically the use of "that cast".

P2-3 transition is also abrupt.

L46-50: The first two sentences of this paragraph feel pretty wordy and a bit repetitive. Maybe try something like "RTS are a form of abrupt permafrost thaw that occur when ground ice is exposed, allowing the rapid thaw and downslope movement of debris."

L57: "manifesting itself with increased numbers, sizes, and faster retreat rates" would be better as "manifesting through increased numbers, sizes, and faster retreat rates"

L58: I would break the paragraph at "However, due to their complex spatiotemporal dynamics..." as it's a very long paragraph and this seems like a natural transition.

L59: "assessing its impact on regional carbon cycling" should be "assessing their impact on regional carbon cycling"

L65-66: "Nevertheless, to quantify RTS-induced mass wasting and evaluate the potential implications on permafrost carbon emissions, additional datasets are still required - particularly those capturing lateral and vertical change and soil properties." This sounds to me like the data required for estimates of permafrost carbon mobilization and not carbon emissions, which would be much more difficult to estimate. As this manuscript does not focus on carbon emissions but rather carbon mobilization, I would change the wording to reflect that.

L77-94: The final paragraph in the introduction is quite long and provides a more detailed explanation of the methods than is required. I'd recommend keeping only a few sentences about the methods here and replacing most of the overview text in 2.2 with this, as this section reads more smoothly clearly than what is currently in 2.2. Of course, there are a few pieces of information in 2.2 which are not included here, that should be kept.

L100: "holding a total water volume of 3330 km3 in the top 10 m" This seems like unnecessary information to me.

L116: "focussing" is typically spelled "focusing", although I just learned that both are apparently acceptable

L124-125: "we established several validation sites that were spatially distributed throughout the QTP" More detail is needed about how these sites were chosen. Were they a stratified random sample or manually selected or something else?

L136: "The German Synthetic Aperture Radar (SAR) satellite mission TanDEM-X allows us to generate temporally resolved digital elevation models based on bistatic SAR interferometry (InSAR)" I think you should be more direct. "The German Synthetic Aperture Radar (SAR) satellite mission TanDEM-X was used to generate temporally resolved digital elevation models using bistatic SAR interferometry (InSAR)."

L139-152: A lot of this reads like introduction and is provided in more detail than is probably required. I would suggest reframing it to focus on what you did with concise explanations of why you did some things differently than previous studies. For example, "We used year round TanDEM-X observations to achieve full spatial coverage, although previous studies in the Arctic have used only winter data to avoid errors caused by dense and wet tundra vegetation during the growing season (sources). The errors introduced by vegetation charateristics are likely to be negligible in this study due to the low canopy heights in the alpine meadows, steppes, arid desert, and bare ground of the QTP." One more sentence could describe that the snow conditions required for winter data to be used were met.

L167: "The resulting elevation change maps are normally distributed around zero with a SD representing the achievable vertical accuracy of the DEM pair." I think this is only true if you assume there is not widespread subsidence occuring. I'm not aware of any studies that have looked into this off the top of my head, but I wouldn't assume there is no subsidence occuring more broadly as the ALT is deepening and temperatures are rising in the region (I'm sure it would be smaller than the detection threshold for the data, in any case).

L176-177: "we assigned the optical RTS shape that matched the last observation year to ensure a conservative estimation of the eroded volume". This sounds backwards to me. Using the latest observation year should ensure the largest RTS area, which would provide the largest estimate of the eroded volume, unless I'm missing something important.

L198-199: "Therefore, we integrate existing datasets for QTP that define (1) ALT, (2) GI content between 2 and 10 m, and (3) SOC stocks between 0 and 3 m depth." I would suggest including the citations here, even though they are described in the following sentences.

Eq. 3: Are the SOC (below ALT) and SOC (above ALT) switched here?

L270: "corresponding to a relative error in elevation change of ∼ 35%" the phrasing here confused me - I'd either remove it since the magnitude of the error is already listed (and it's pretty clear that it's about 1/3 of the total without complicated head math) or change the wording. Also, please be consistent about including or excluding a space between the number and "%".

L270-271: "The median elevation loss that roughly indicates the RTS headwall height was 1.2 ± 0.4 m." This is a fragment and I'm uncertain how the elevation loss is being related to the headwall height. Is this coming from the allometric scaling? If so, what are the assumptions about RTS shape that are going into the relationship between the headwall height and the volume?

L275-276: "An α value of 1.30 indicates that RTS in QTP during the last decade followed a relationship between linear growth (α = 1.0) and exponential growth (α > 1.5)". Since the exponent does not change across values of x, this is a power relationship not an exponential relationship. The wording here also makes it sound like growth in RTS over time rather than the relationship between area and volume. Please revise the wording.

L280-281: "We calculated that ∼ 64 % of the thawed ground ice was located in the first metre under the active layer (2 - 3 m)". Are you assuming an ALT of 1 m? This doesn't make sense to me, given that the average ALT of QTP cited in the methods section is 2.34 m.

L287-288: "We normalised the total quantities per subregion by RTS count and analysed the distributions to ensure better comparability." This reads like methods and is unnecessary here.

L291-293: "This points to a greater elastic distortion in the degree to which the concavity increases volumetrically with changing area for the mountainous northeast QTP (see Van Der Sluijs et al. (2023) for coefficient interpretation, Fig. S1 and Fig. S3)." This should be in the discussion. The wording is unclear to me.

L295-296: "but the total amount is relatively low for the entire QTP compared to the eroded volume". It is unclear to me what point is trying to be made here, as the ice content is necessarily less than the total volume.

L298-299: "However, the distribution of SOC mobilisation follows the spatial trend present in the SOC stocks from Wang et al. (2021)". This belongs in the discussion.

L302-304: "We manually delineated the ablation zones for 307 RTS scars on elevation change maps (2011 - 2019) at all validation sites and statistically analysed the potential differences from the RTS labels (2018, 2019, 2020) in the optical inventory of Xia et al. (2022)." This does not need to be repeated in the results section. The topic sentence here would better be an overarching statement about the agreement of the two methods.

L309-311: "Figure 5 b shows an example RTS in the Beiluhe River Basin in Central QTP including the delineations of the optical inventory (Xia et al., 2022) of the summers one year before, the same year and one year after the SAR observation used for the DEM generation, as well as the manually delineated ablation zone visible on the elevation change map." This is unnecessary.

L312-316: "A RTS delineation based on multispectral imagery defines the boundary of the thaw feature by the difference in the spectral signal between disturbed and intact vegetation, whereas elevation maps showcase RTS activity by elevation loss (ablation at the headwall) and gain (accumulation at the floor). Typically, a larger planimetric area is present in RTS delineations based on optical images that often comprise most of the recently active slump floor and accumulation zone compared to the DEM-based boundaries." This is methods/discussion and does not belong here.

L316-320: "Comparing the RTS delineation area AXia with the active erosion area visible on the elevation change map, we see that the delineations for 2019 and 2020 include substantially more ablation area than what is distinguishable as RTS-induced erosion on the elevation change maps (Fig. 5 b, c)." It is unclear to me why these numbers are so different if both are derived from the DEM difference map and one is just a hand-delineation of more or less contiguous ablation area while the other is includes all pixels showing ablation. I also had a really hard time keeping track of what the difference between these two methods is, so I think it is worth trying to find a concise label for the two methods that can help remind the reader.

L321-322: what is the difference in observation time? I thought both methods used the same DEM difference map which was fairly well matched to the dates of the optical RTS labels?

L324-325: "The lower threshold for distinguishing an RTS from background noise is substantially higher in the elevation change maps". Unclear language. What is the lower threshold?

L331: "the optical label of 2019 shows the best fit." I don't think this provides much value, since they're not measuring the same thing.

L332-334: "For all regions that did not undergo any erosion in the RTS scar or in the vicinity, we assume that the elevation change is normally distributed around zero with a standard deviation of approximately 2 - 3 m in flat terrain for TanDEM-X-derived DEMs." Couldn't it also be that the average elevation value was different than zero, but by an amount smaller than could be detected?

L334-337: "For volume estimation based on optical delineations, stable areas containing noise are likely included since most optical RTS boundaries are larger than the actual active erosion area. Although this minimally affects the total volume change estimate due to the low magnitude in negative elevation change, it adds additional random errors, thus contributing to the overall uncertainty budget." This is discussion.

L358: "Normalised by study area". It is unclear what has been normalized here.

L362-363: "Most of these sites are mainly characterised by smaller and shallower RTS located on lake shores in relatively flat terrain". Is "these sites" referring to the remaining sites investigated by Bernhard et al. 2022? Please clarify.

L368: "Based on the fitting of a Ordinary Least Square model". "A" should be "an".

L376: "scaling coefficients below α < 1.3". This is repetitive. Either spell it out or use the symbols.

L385: "shallowed" should be "shallow"; "whereby" doesn't seem like quite the right word here

L394-395: "Therefore, we should only see these results and comparisons as another indicator that RTS development in the QTP is on a level similar to the known hotspots in the Arctic permafrost region." I would not recommend starting a paragraph with "therefore" as it implies a direct relationship with the previous sentence. This statement is also very weak sounding and undercuts the importance of this study. Please rephrase.

L427: "the hinterlands". This isn't the right word. Maybe try "adjacent areas"?

L463-465: "In situ observations from northeast QTP showed a 23 to 37 % loss in surface (> 20 cm) (Mu et al., 2017, 2020; Liu et al., 2018) and slightly less (∼ 20 - 28 %) for SOC stocks in deeper soils (Wu et al., 2018; Yi et al., 2025)." Loss of C? To the atmosphere? In an incubation or field measurement? During what time period? From an RTS? More detail is needed here to understand how this fits in.

L467-469: "Wang et al. (2020) estimated for QTP 0.19 to 0.38 × 101 3 kg C (or 1.9 - 3.8 Pg C) from formerly frozen SOC will be subject to decomposition upon permafrost thaw until 2100 under moderate and high-emission climate scenarios that could switch QTP from a carbon sink to a source." The wording is a bit clunky here. Please rephrase.

L476-477: "However, we found that only 2 % of SOC mobilised by RTS activity contributed to the total loss of SOC stocks at soil depths below 3 m." Unclear, please rephrase.

L479-483: "Parts of mobilised SOC remain on the slump floor and are available for microbial decomposition and release as greenhouse gases (Wang et al., 2024) or deposited and even stabilised (Thomas et al., 2023; Liu et al., 2021, 2018; Mu et al., 2017), while other parts, together with thawed material, are

laterally transported downslope into adjacent river and lake systems and SOC can undergo complex water chemistry processes such as dissolution or sedimentation (Lewkowicz and Way, 2019)." This is pretty long. I'd recommend breaking it up into two sentences.

L498-499: "Some permafrost regions in the Arctic have been focus areas and DEMs of two to three time steps in the last decade exist." This feels unnecessary to me.

L519: "On the QTP, no study, to our best knowledge," would be better as "To our knowledge, no study on the QTP"

L520-522: "These limitations highlight that applying such a scaling law to optical RTS inventories should be done carefully and rather to obtain regional estimates on material erosion volume and mass wasting derivatives such as ground ice loss or SOC mobilisation." Unclear, particularly "and rather". Please rephrase.

L526-528: "We could show that remote sensing data with a spatial resolution of ∼ 10 m misses ∼ 35 % of RTS features that could be found in higher resolution PlanetScope images which, however, only accounted for a difference of < 1 % of the eroded material volume (Fig. 5 c)." It is unclear what "could show" means here? Does your work show this? If so, how?

General comments

This paper presents a high quality analysis of a unique dataset and methodology for the quantification of RTS activities including material erosion volumes and SOC mobilisation.

The methodology presented is robust and the interpretation is thorough.

Technical corrections

Figure 5 caption: “(d) The distribution of the RTS ablation area from 2019 is the closest to the actual mean ablation area.”: Based on the figure, it seems that 2020 is closest to the mean ablation area.