the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Exploring the use of Sentinel-1 to monitor spatial and temporal evolution of permafrost in the Swiss Alps
Abstract. In this work, we assess the performance of Persistent Scatterer Interferometry (PSI) based on Sentinel-1 data in the Swiss Alps, focusing on the analysis of surface displacement in areas located in permafrost. To this end, we exploit a PSI dataset over Canton Wallis and the recently published permafrost and ground ice map of Switzerland (PGIM). First we evaluate the sensitivity to radar detection of the terrain, given a particular slope distribution and given satellite orbital geometries. We find that 92 % of the areas currently labeled as permafrost satisfy the geometric visibility criteria, which is a necessary but not sufficient condition for detection of ground movement by satellite radar. Second, we analyse the PSI surface velocity and displacement time series observed in Canton Wallis in the period 2015 to 2022. We find that the displacements rates appear to be correlated with the occurrence of permafrost and that there is a significant difference in displacement rates between ice-poor and ice-rich permafrost compared to no-permafrost zones. However, the difference between the surface velocities retrieved in ice-poor permafrost and no-permafrost zones are small and additional information would be needed to discriminate ice-poor permafrost zones.
- Preprint
(11955 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2023-2321', Reynald Delaloye, 19 Jan 2024
The reported study is comparing surface motion derived from a single specific product from space-borne radar data (PSI) with on a first hand a modelled spatial distribution of permafrost and related ground-ice content (PGIM), and on the second hand with in-situ geodetic data retrieved for one rock glacier (GNSS). None of these data has been produced by the authors, but are made available by third parties. Whereas PGIM and GNSS are open-access data, the PSI data is not made publicly available by the Canton Valais. PSI and GNSS datasets are measured data, namely observations, whereas PGIM is a model.
The title (“Exploring the use of Sentinel-1 to monitor spatial and temporal evolution of permafrost in the Swiss Alps”) is not reflecting the subject of the paper. There are various possible uses of space-borne radar data (e.g. Strozzi et al. 2020), but only a single one (the persistent scattered interferometry – PSI) is exploited in this paper. Definitely, there is no exploration of the use of space-borne radar data. No aspect of the paper is dedicated to the monitoring of the temporal evolution of the displacement rates. Finally, what is observed are surface changes related to mass movements, but not permafrost.
Starting from these statements, interrogation is rising about the mastering of the intended subject.
The paper comprises major issues, namely:
PSI is a technique limited to the detection of slow movements up to a maximally a few cm (max. about 5 cm) in the LOS direction. There is no word about this limitation in the paper. There is also no word on how the data is produced and what is its accuracy. I suspect that a displacement rate in the order of +/- 1 mm or so is the range (standard deviation) of the uncertainty. Most of the treated data (see fig. 5) lays in this range, removing sense to the statistical analysis as it is performed. But, what is the most important, mass movements faster than 0.05 m/year are NOT detected. And they are frequent, in all regions with or without permafrost (e.g. Barboux et al. 2014, Bertone et al. 2022).
The figure 5 (right) is basically showing us the uncertainty of the PSI data. And whatever the track, there is almost as many movements toward the satellite as away from it. Positive values (movement toward the satellite) are provided by a surface heave or a downward motion on a slope facing the LOS. Where is the sense of treating simultaneously both the positive and the negative values in the statistical analysis ? In addition, the outliers (larger than +/- 10 mm/year) have been removed from the figure… but they are the most interesting in the purpose of this paper (regarding hazards). (On the left figure, I guess that the legend of the vertical axis is wrong (cm/year instead of mm/year)).
The take-home message of the paper is that on ice-rich permafrost the mean motion rate is about 2 mm/year, whereas it is only 1 mm/year on ice-poor permafrost and close to 1 mm/year on non-permafrost area. Regarding my above comments (detection limit, uncertainty and positive values), I don’t see the sense of this even if this is statistically correct.
Observed surface displacement (with PSI) is caused by the downslope movement of the surface relative to processes as diverse as shallow soil creep (e.g. solifluction), permafrost creep (e.g. rock glaciers, push-moraines) or landslides of almost any size (e.g. large deep-seated landslides), among others. Permafrost creep is a process related to the occurrence of frozen ground under specific conditions (slope, thickness, ice content, temperature…), but is not specifically related to the degradation (warming) of permafrost. In addition, vertical movement (mostly subsidence) is expected to occur where either remnants of debris-covered glacier ice (e.g. dead ice) or ground ice exceeding the porosity are melting, without a downslope movement is necessarily taking place. Such a subsidence is mostly occurring during (the second part of) summer and early fall, namely when the PSI observation is undertaken. There is no distinction made between all these different processes by the authors, who tend to attribute most detected motion signals to the occurrence of degrading permafrost and related mass wasting processes (e.g. second case study).
According to RGIK (2023), cited by the authors, active rock glaciers are moving faster than 0.1 m/year. Thus, there are NOT detected with PSI. There are several hundreds, even probably 1’000 active rock glaciers in the Valais Alps. And there are also numerous landslides moving (much) faster than 0.05 cm/year, which are also NOT detected with PSI. Some of the missed landforms are moving more than 10 m/year. Of high concern for the authorities regarding safety issues related to permafrost in debris terrain are mass movements faster than e.g. 1 m/year and in specific locations only. I invite the authors to take a look for instance at the various papers published by Kummert et al. in 2018-2021. One should note also that PSI is not suited to detect/monitor mass movement in dense vegetated (forested) areas, basically in non-permafrost areas in the context of the Valais Alps. Beside that PSI is not suited for detecting this large number of mass movements, and is of limited concern regarding the detection/monitoring of hazardous areas, there is a strong bias in the performed analysis as a large majority of the undetected moving areas are located in the PGIM-modelled “ice-rich” zones.
There are 11 rock glaciers in the Canton Valais, whose surface displacements are monitored by terrestrial ground surveys within the framework of PERMOS. It should be valuable to compare the PSI data with GNSS data on all these sites and not only on the Alpage de Mille rock glacier … which is the slowest one. In fact, all the others are moving >10 cm/year and are not suited for a PSI survey. This should have evidenced to the authors the (important) limitations of PSI for monitoring mass movements in Alpine permafrost areas.
“Ice-rich” and “ice-poor” permafrost is a denomination used in the PGIM model to reflect two different approaches of modelling the distribution of permafrost. There is a lot of uncertainty in the model at the local scale, as it is in any model, and it is not an observation of permafrost occurrence, even less about its ice content. Analyzing the PSI data paying regard to the permafrost model results, and according to the detection limitations, is somehow not relevant. There is in addition something weird in the development of the paper as the authors first use the modelled permafrost zoning to analyze the PSI data, but then suggest to use the PSI data to improve the modelled zoning.
The example close to Stalden is in fact the slope above the “village” of Grächen. The village itself is located on the median-lower part of a very large deep-seated landslide (the village is moving about 0.2 to 0.5 cm/year downslope according to PSI). The area depicted in the figure is the head section of this deep-seated landslide. Almost the entire area is in motion, up to several cm/y. This has nothing to do with the occurrence or not of either ice-poor or ice-rich permafrost ! One could also note that in the lowermost edge of the figure is located the Ritigraben rock glacier, a pretty fast landform causing issues with debris flows occasionally reaching the valley bottom (e.g. Kenner et al. 2017, Lugon and Stoffel 2010), but which is not revealed by PSI data because of velocities (significantly) faster than 0.05 m/year. There is the same PSI issue with the Distelhorn rock glacier just outside the uppermost edge of the figure (Strozzi et al. 2020).
The example close to Zermatt is located on a southern slope, so not really suited for a precise PSI/InSAR analysis.
Finally, one could imagine that if the Canton Valais is contracting a company to provide the PSI data annually, this is because PSI is providing a usefull overall view on slow moving mass wasting over a large region. This is not a finding of this paper.
These are for the main comments. I do not enter further into details. I suggest the rejection of the paper. To improve the consistency of their study, I really encourage the authors to have a closer and thorough look at the geomorphological processes related to the PSI-detected movements and to take care about the limitations of the technique.
References (not already cited in the paper) :
Barboux et al. (2014). Inventorying slope movements in an Alpine environment using DInSAR. Earth Surface Processes and Landforms, 39/15, p2087-2099. DOI: https://doi.org/10.1002/esp.3603
Bertone et al. (2022). Incorporating InSAR kinematics into rock glacier inventories: insights from 11 regions worldwide, The Cryosphere, 16, 2769–2792, https://doi.org/10.5194/tc-16-2769-2022, 2022.
Kenner et al. (2017). Factors Controlling Velocity Variations at Short-Term, Seasonal and Multiyear Time Scales, Ritigraben Rock Glacier, Western Swiss Alps, Permafrost and Periglacial Processes, https://doi.org/10.1002/ppp.1953
Kummert et al. (2018): Regional-scale inventory of periglacial moving landforms connected to the torrential network system, Geogr. Helv., 73, 357–371, https://doi.org/10.5194/gh-73-357-2018, 2018.
Kummert et al. (2018). Mapping and quantifying sediment transfer between the front of rapidly moving rock glaciers and torrential gullies. Geomorphology 309, 60-76, https://doi.org/10.1016/j.geomorph.2018.02.021
Kummert et al. (2021). Pluri-decadal evolution of rock glaciers surface velocity and its impact on sediment export rates towards high alpine torrents. Earth Surface Processes and Landforms, 46(15), 3213-3227, https://doi.org/10.1002/esp.5231
Lugon et al. (2010). Rock-glacier dynamics and magnitude–frequency relations of debris flows in a high-elevation watershed: Ritigraben, Swiss Alps. Global and Planetary Change, 202-210, https://doi.org/10.1016/j.gloplacha.2010.06.004
Citation: https://doi.org/10.5194/egusphere-2023-2321-RC1 -
AC1: 'Reply on RC1', Kristina Juliana Reinders, 05 Apr 2024
We thank Prof Delaloye for reviewing our manuscript. Despite the intention of the comments provided is evidently not constructive, and in some case the arguments are incorrect, some points that can be indeed implemented in a revised version of our manuscript to better clarify the main goal of our work, as well as to highlight the importance of our results. Here below we provide punctual replies to the concerns raised by Prof. Delaloye. We hope that our replies will convince the editor that the information and analyses provided are very relevant and it is worth to provide a revised version of the manuscript.
Our response is written in bold below each point of Prof Delaloye.
The reported study is comparing surface motion derived from a single specific product from space-borne radar data (PSI) with on a first hand a modelled spatial distribution of permafrost and related ground-ice content (PGIM), and on the second hand with in-situ geodetic data retrieved for one rock glacier (GNSS). None of these data has been produced by the authors, but are made available by third parties. Whereas PGIM and GNSS are open-access data, the PSI data is not made publicly available by the Canton Valais. PSI and GNSS datasets are measured data, namely observations, whereas PGIM is a model.
We interpret this comment as a tentative from Prof. Delaloye to undervalue the analysis performed. We believe that the fact we did not produce the PSI dataset cannot be evaluated in a negative perspective. Many people are approaching more and more PSI datasets provided at regional scales and in many cases, they don’t have the experience or expertise to evaluate the quality and the limitations. In addition, the initial part of our paper indeed presents a new dataset that we have produced, i.e., the Sentinel-1 visibility map at the country scale, including specific considerations on the expected visibility of permafrost regions. This dataset will be useful for the community approaching satellite SAR observations in this specific context and provides important hints on the capability of Sentinel-1 to be used in monitoring applications in Switzerland. In the revised version we will better highlight the importance of our analyses on this dataset and include additional considerations on the potential use of similar datasets to monitor deformation in alpine areas.
The title (“Exploring the use of Sentinel-1 to monitor spatial and temporal evolution of permafrost in the Swiss Alps”) is not reflecting the subject of the paper. There are various possible uses of space-borne radar data (e.g. Strozzi et al. 2020), but only a single one (the persistent scattered interferometry – PSI) is exploited in this paper. Definitely, there is no exploration of the use of space-borne radar data. No aspect of the paper is dedicated to the monitoring of the temporal evolution of the displacement rates. Finally, what is observed are surface changes related to mass movements, but not permafrost. Starting from these statements, interrogation is rising about the mastering of the intended subject.
We thank prof Delaloye for this comment. He is right pointing out that the current title of the manuscript can be misleading; however, it also true that the first sentence of the abstract already clarifies the topic and the goals “In this work, we assess the performance of PSI interferometry based on Sentinel-1 data in the Swiss Alps, focusing on the analysis of surface displacement in areas located in permafrost.”. We will revise the title as follows: “Exploring the potential of Sentinel-1 PSI monitoring in permafrost regions of the Swiss Alps”. Indeed, our focus is to provide additional information on the surface displacements “IN” areas where permafrost has been recognized by scientists mastering this research topic (as for example in the PGIM map), and not “OF” the permafrost process itself. To this end, we want also to clarify that our focus is not on rock glaciers (we will reply to this point in the specific comments below). As a matter of fact, surface velocities implicitly provide also temporal information; however, we agree that this should be removed from the title to avoid misunderstandings. We apologize for the confusion our title might have caused, and we are happy to receive this comment and change the title accordingly.
The paper comprises major issues, namely:
PSI is a technique limited to the detection of slow movements up to a maximally a few cm (max. about 5 cm) in the LOS direction. There is no word about this limitation in the paper. There is also no word on how the data is produced and what is its accuracy. I suspect that a displacement rate in the order of +/- 1 mm or so is the range (standard deviation) of the uncertainty. Most of the treated data (see fig. 5) lays in this range, removing sense to the statistical analysis as it is performed. But, what is the most important, mass movements faster than 0.05 m/year are NOT detected. And they are frequent, in all regions with or without permafrost (e.g. Barboux et al. 2014, Bertone et al. 2022).
These statements from prof. Delaloye are wrong. PSI technique has known limitations, and we report them in the discussion section. The velocities that can be measured depend on the sensor considered and on the revisit time. In the specific case of Sentinel-1, the C-Band sensor with revisit times of 6-days allows to detect movements up to 84 cm/year (Wasowsky and Bovenga, 2014; Manconi, 2021). The 1 mm/year is not the range of the standard deviation of PSI. This value is reported in numerous papers over the last 20-25 years to be considered as the accuracy of the technique (among several, Lanari et al., 2007; Crosetto et al., 2016). Such values have been considered the accuracy of PSI even in periods when the quality of the satellite SAR acquisitions was much lower in terms of revisit time (for example ERS-1/2 and ENVISAT ASAR). This means that trends (velocities) in the order of mm/year reported and analyzed in our manuscript can be safely considered to interpret the surface displacement behavior.
We already provided information on the dataset in lines 106-117. In the revised manuscript, we can add more specific explanation about the production of the dataset and include important references to better explain basic PSI concepts, and thus avoid that readers of TC not expert in the field of satellite SAR are driven into erroneous considerations, such as the ones done by prof. Delaloye in this comment.
The figure 5 (right) is basically showing us the uncertainty of the PSI data. And whatever the track, there is almost as many movements toward the satellite as away from it. Positive values (movement toward the satellite) are provided by a surface heave or a downward motion on a slope facing the LOS. Where is the sense of treating simultaneously both the positive and the negative values in the statistical analysis ? In addition, the outliers (larger than +/- 10 mm/year) have been removed from the figure… but they are the most interesting in the purpose of this paper (regarding hazards). (On the left figure, I guess that the legend of the vertical axis is wrong (cm/year instead of mm/year)). The take-home message of the paper is that on ice-rich permafrost the mean motion rate is about 2 mm/year, whereas it is only 1 mm/year on ice-poor permafrost and close to 1 mm/year on non-permafrost area. Regarding my above comments (detection limit, uncertainty and positive values), I don’t see the sense of this even if this is statistically correct.
Regarding the accuracy of the PSI results, we already replied in the comment above. We confirm that the label of the Figure 5 is correct, we show velocities in the mm/year range and not uncertainties!
Regarding the positive/negative LOS direction in the dataset and the statistical analysis performed, we outline here our reasoning. In the ANOVA analysis we applied, one of the conditions is that the observations must be normally distributed (Howell, D. C. (1992). Statistical methods for psychology. PWS-Kent Publishing Co., Page 321). This is the reason why we used the PSI data with their sign.
However, we understand the concerns of prof. Delaloye (this is also a point raised from the reviewer #2) and thus investigated alternative statistical approaches to compare the surface velocities without considering their sign, which intuitively will show similar if not higher differences between the different zones. We already tested the Kruskal-Wallis test on the absolute values of the PSI and the results show that the difference in surface velocities between different permafrost zones are still present and statistically significant. In the revised version, we can modify the statistical analysis and provide additional information on the constraints and limitations.
Observed surface displacement (with PSI) is caused by the downslope movement of the surface relative to processes as diverse as shallow soil creep (e.g. solifluction), permafrost creep (e.g. rock glaciers, push-moraines) or landslides of almost any size (e.g. large deep-seated landslides), among others. Permafrost creep is a process related to the occurrence of frozen ground under specific conditions (slope, thickness, ice content, temperature…), but is not specifically related to the degradation (warming) of permafrost. In addition, vertical movement (mostly subsidence) is expected to occur where either remnants of debris-covered glacier ice (e.g. dead ice) or ground ice exceeding the porosity are melting, without a downslope movement is necessarily taking place. Such a subsidence is mostly occurring during (the second part of) summer and early fall, namely when the PSI observation is undertaken. There is no distinction made between all these different processes by the authors, who tend to attribute most detected motion signals to the occurrence of degrading permafrost and related mass wasting processes (e.g. second case study).
We think this is another misunderstanding by prof. Delaloye, possibly caused by the inaccurate title and some missing explanations in the introduction of our paper, as explained before. The goal of the paper is not to discuss the physical processes causing the surface displacements reported. The main goal is to show that areas mapped in PGIM have different ranges of velocities. To the best of our knowledge, the observation we provide at the regional scale is new and is worth of further investigations. Most of the papers using standard InSAR in the Swiss Alps in the past have looked at rock glaciers, which are not our focus here. Understanding the causes of this difference is beyond the scope of this work and would indeed require additional information that is not available at the spatial and temporal scale needed.
According to RGIK (2023), cited by the authors, active rock glaciers are moving faster than 0.1 m/year. Thus, there are NOT detected with PSI. There are several hundreds, even probably 1’000 active rock glaciers in the Valais Alps. And there are also numerous landslides moving (much) faster than 0.05 cm/year, which are also NOT detected with PSI. Some of the missed landforms are moving more than 10 m/year. Of high concern for the authorities regarding safety issues related to permafrost in debris terrain are mass movements faster than e.g. 1 m/year and in specific locations only. I invite the authors to take a look for instance at the various papers published by Kummert et al. in 2018-2021.
We are aware that fast active rock glaciers are not identified in the Sentinel-1 PSI analysis. Despite, the permafrost community should be interested to know that the PGIM Zone 2 presents larger surface velocities compared to other PGIM areas, EVEN when the most active rock glaciers are not considered! Most of the InSAR investigations in the alps have focused on standard radar interferometry to map and monitor rock glaciers. In many cases (as in the references mentioned by prof. Delaloye) the surface velocities derived for rock glaciers are based on few interferograms, thus covering a very short period. For example, if a rock glacier shows one fringe (2.8 cm displacement) in a C-Band interferogram between 6-days (Sentinel-1 revisit time), this would result in a velocity of more than 1.5 m/year; however, such high velocities might be constrained only on a very short period of time and not really representative of the yearly rate of displacements of the rock glacier. We will better specify in our review that rock glaciers are not the focus of our work and provide solid reasoning of the importance of our observations.
One should note also that PSI is not suited to detect/monitor mass movement in dense vegetated (forested) areas, basically in non-permafrost areas in the context of the Valais Alps.
Prof. Delaloye has ignored or overlooked the section of our manuscript (lines 180-189 and Table 6) where we have analyzed only the differences in velocity between PSI located in different PGIM zones at elevations higher than 2500m, thus above the tree line in the Canton Wallis. We have shown that the differences in velocities between PGIM zones is statistically significant even when considering only PS measurements at elevations above the tree line.
Beside that PSI is not suited for detecting this large number of mass movements, and is of limited concern regarding the detection/monitoring of hazardous areas, there is a strong bias in the performed analysis as a large majority of the undetected moving areas are located in the PGIM-modelled “ice-rich” zones. There are 11 rock glaciers in the Canton Valais, whose surface displacements are monitored by terrestrial ground surveys within the framework of PERMOS. It should be valuable to compare the PSI data with GNSS data on all these sites and not only on the Alpage de Mille rock glacier … which is the slowest one. In fact, all the others are moving >10 cm/year and are not suited for a PSI survey. This should have evidenced to the authors the (important) limitations of PSI for monitoring mass movements in Alpine permafrost areas.
We have made a comparison of the PSI results vs. the PERMOS measurements (section 3.3 and Figure 5). This is the only location where PSI results are available for such comparison. Again, we are aware of the limitations (as also highlighted in the Discussion section). Our argument is that despite not considering the most active rock glaciers, surface velocities in the PGIM zone 2 are higher than the other regions.
“Ice-rich” and “ice-poor” permafrost is a denomination used in the PGIM model to reflect two different approaches of modelling the distribution of permafrost. There is a lot of uncertainty in the model at the local scale, as it is in any model, and it is not an observation of permafrost occurrence, even less about its ice content. Analyzing the PSI data paying regard to the permafrost model results, and according to the detection limitations, is somehow not relevant. There is in addition something weird in the development of the paper as the authors first use the modelled permafrost zoning to analyze the PSI data, but then suggest to use the PSI data to improve the modelled zoning.
This comment is unclear. The limitations of the PGIM are clearly stated in the paper and the readers are referred to the Kenner et al., 2019 paper for more specific information, where the method to generate this map is presented. This is currently the only way we can analyze the surface displacements at regional scale. We agree that the section where we mention that the PSI data could be used to validate and improve the permafrost zoning is highly speculative should be removed (this is also a comment of the reviewer #2). In that section, we wanted to highlight that areas that have different classification in the PGIM behave similarly in terms of surface velocities. This is a point that would need additional investigation to understand the processes behind such behavior.
The example close to Stalden is in fact the slope above the “village” of Grächen. The village itself is located on the median-lower part of a very large deep-seated landslide (the village is moving about 0.2 to 0.5 cm/year downslope according to PSI). The area depicted in the figure is the head section of this deep-seated landslide. Almost the entire area is in motion, up to several cm/y. This has nothing to do with the occurrence or not of either ice-poor or ice-rich permafrost !
This example has been chosen because the area is classified as ice-rich permafrost (zone 2) in the PGIM. In the revised version we can select another example.
One could also note that in the lowermost edge of the figure is located the Ritigraben rock glacier, a pretty fast landform causing issues with debris flows occasionally reaching the valley bottom (e.g. Kenner et al. 2017, Lugon and Stoffel 2010), but which is not revealed by PSI data because of velocities (significantly) faster than 0.05 m/year. There is the same PSI issue with the Distelhorn rock glacier just outside the uppermost edge of the figure (Strozzi et al. 2020).
Again, we know about the PSI limitations, and they have been clearly stated in the discussion section. In the revised version, we can better specify the points related to the detection of rock glaciers. We reiterate here that we think it is important to notice that areas classified in the PGIM as zone 2 have surface velocities higher than the other areas.
The example close to Zermatt is located on a southern slope, so not really suited for a precise PSI/InSAR analysis.
The orientation of the slope is south-west and thus visible and suitable for PSI-InSAR.
Finally, one could imagine that if the Canton Valais is contracting a company to provide the PSI data annually, this is because PSI is providing a usefull overall view on slow moving mass wasting over a large region. This is not a finding of this paper.
We honestly do not understand this comment. We did not state in our paper that we discovered that PSI analyses are useful to monitor slow moving processes. The main finding of our paper is that the PSI data show (statistically significant) differences in the surface velocities behavior of different areas.
These are for the main comments. I do not enter further into details. I suggest the rejection of the paper. To improve the consistency of their study, I really encourage the authors to have a closer and thorough look at the geomorphological processes related to the PSI-detected movements and to take care about the limitations of the technique.
References (not already cited in the paper) :
Barboux et al. (2014). Inventorying slope movements in an Alpine environment using DInSAR. Earth Surface Processes and Landforms, 39/15, p2087-2099. DOI: https://doi.org/10.1002/esp.3603
Bertone et al. (2022). Incorporating InSAR kinematics into rock glacier inventories: insights from 11 regions worldwide, The Cryosphere, 16, 2769–2792, https://doi.org/10.5194/tc-16-2769-2022, 2022.
Kenner et al. (2017). Factors Controlling Velocity Variations at Short-Term, Seasonal and Multiyear Time Scales, Ritigraben Rock Glacier, Western Swiss Alps, Permafrost and Periglacial Processes, https://doi.org/10.1002/ppp.1953
Kummert et al. (2018): Regional-scale inventory of periglacial moving landforms connected to the torrential network system, Geogr. Helv., 73, 357–371, https://doi.org/10.5194/gh-73-357-2018, 2018.
Kummert et al. (2018). Mapping and quantifying sediment transfer between the front of rapidly moving rock glaciers and torrential gullies. Geomorphology 309, 60-76, https://doi.org/10.1016/j.geomorph.2018.02.021
Kummert et al. (2021). Pluri-decadal evolution of rock glaciers surface velocity and its impact on sediment export rates towards high alpine torrents. Earth Surface Processes and Landforms, 46(15), 3213-3227, https://doi.org/10.1002/esp.5231
Lugon et al. (2010). Rock-glacier dynamics and magnitude–frequency relations of debris flows in a high-elevation watershed: Ritigraben, Swiss Alps. Global and Planetary Change, 202-210, https://doi.org/10.1016/j.gloplacha.2010.06.004
Thanks for the additional references. The are mostly related to rock glaciers and we can mention them in the revised manuscript in the section where we will clarify that this is not the main target of our investigation.
Citation: https://doi.org/10.5194/egusphere-2023-2321-AC1
-
AC1: 'Reply on RC1', Kristina Juliana Reinders, 05 Apr 2024
-
RC2: 'Comment on egusphere-2023-2321', Anonymous Referee #2, 28 Feb 2024
This paper uses satellite InSAR PSI data to evaluate, PGIM, a permafrost and ground ice map in the Swiss Alps. They reported that 92% of the permafrost areas are visible from Sentinel-1 according to the radar and slope geometries. The authors compared the PSI deformation velocities with the PGIM in a local area in southern Switzerland and reported higher mean deformation velocities in ice-rich permafrost zones than in ice-poor permafrost and no permafrost zones.
I identified two major issues in the work presented in the current format.
(1) The linkage between the ground surface deformation and permafrost zonation is weak and complex, yet poorly described in this paper. In mountain regions such as the Swiss Alps, surface deformation can be caused by numerous processes, many of which, such as landslides and rock slope failure, are not directly related to permafrost. High velocities (>10 mm/yr as adopted by the authors) do not necessarily indicate ice-rich permafrost. On the other hand, intact but inactive rock glaciers may exhibit little deformation but still contain ground ice. Some of these complexities are briefly mentioned in Section 4 (L237). But without building a reliable relationship between surface deformation and alpine permafrost conations, the use of PSI deformation velocities in mapping permafrost zonation is hardly convincing.
(2) The PSI-based deformation velocities adopted in this study may not reflect the true ground motion. First, since the reported PSI velocities are the projection of true ground motion onto the line-of-sight directions (L130). Depending on the slope direction (slope angle and aspect angle) and satellite viewing geometry, the LOS velocities in a mountain region can be in either positive or negative. Taking an average of the LOS velocities would be inevitably biased towards zero, which seems to be the case in this study as the reported mean velocities in all the permafrost zones are all very close to zero (e.g., Table 4, Table 6, Figure 5). In other words, the reported mean LOS velocities could be averaged of positive and negative values from slopes of all different orientations. Second, the reported LOS velocities are mostly within +/- 10 mm/year, even for ice-rich permafrost (e.g., Figure 5). These are much lower than the expected velocities of rock glaciers (on the order of a few cm to meters per year). The three local cases presented in Figures 6 to 8 would be regarded as ‘outliers’ in the histograms of Figure 5. Still, the PSI velocities presented in Figure 6 seem to be only about 50% of the ground-measured velocities.
Minor comments:
L8: displacement (singular)
L11: what kind of additional information
L14: the Arctic experienced the fastest warming (Rantanen et al. The Arctic has warmed nearly four times faster than the globe since 1979. Communications Earth & Environment, 2022, 3(1): 1-10.)
L20: may add ‘area’ or ‘extent’ to specify what has declined
L22: *lower* boundary
L23: the active layer is on top of, not part of, permafrost
Figure 1: add distance scale bars in the maps
L43: permafrost stability is not the same as permafrost zonation
L55: ‘accesible’ should be accessible
L73 and Table 2 caption: ??
L101: In cases of shadow or layover, the R-index should be zero. Shouldn’t these be the cases when S_h = 0 or L = 0, opposite to the authors’ definition?
L109: subdivision (singular)
L117: deformation is measured “in a remote monitoring network of automated measurement stations”, using what method? And why these independent measurements were not used to validate the PSI results?
L212: specifically
Section 2.3: need to provide more details of the PGIM, for instance: what methods were used; what are the main limitations; why need to be evaluated?
L132: a map
L140: what kind of ground temperature, mean annual ground temperature at a particular depth, or something else?
L146: what are the reasons for large uncertainties?
L159: ??
Table 4: better to state that the numbers of PS are reported to their nearest 1000. It is meaningful to report velocity values in two significant digits after the decimal points in mm/yr, given the accuracy level of Sentinel-1 PSI?
Figure 5: how were the outliers identified?
L268: ‘en’ should be ‘and’
Citation: https://doi.org/10.5194/egusphere-2023-2321-RC2 -
AC2: 'Reply on RC2', Kristina Juliana Reinders, 05 Apr 2024
We thank the second reviewer for reviewing our manuscript.
Our response is written in bold below each point of the reviewer.
This paper uses satellite InSAR PSI data to evaluate, PGIM, a permafrost and ground ice map in the Swiss Alps. They reported that 92% of the permafrost areas are visible from Sentinel-1 according to the radar and slope geometries. The authors compared the PSI deformation velocities with the PGIM in a local area in southern Switzerland and reported higher mean deformation velocities in ice-rich permafrost zones than in ice-poor permafrost and no permafrost zones.
I identified two major issues in the work presented in the current format.
(1) The linkage between the ground surface deformation and permafrost zonation is weak and complex, yet poorly described in this paper. In mountain regions such as the Swiss Alps, surface deformation can be caused by numerous processes, many of which, such as landslides and rock slope failure, are not directly related to permafrost. High velocities (>10 mm/yr as adopted by the authors) do not necessarily indicate ice-rich permafrost. On the other hand, intact but inactive rock glaciers may exhibit little deformation but still contain ground ice. Some of these complexities are briefly mentioned in Section 4 (L237). But without building a reliable relationship between surface deformation and alpine permafrost conations, the use of PSI deformation velocities in mapping permafrost zonation is hardly convincing.
We thank reviewer #2 for this comment. We agree that the speculation currently indicated in the manuscript about the use of PSI in the permafrost zonation should be removed. In the revised version of the manuscript, we can better highlight the main result of our work, i.e., the differences in PSI velocities between areas classified differently in the PGIM. We think that this is a new and important information that can be the base for further investigations of the processes leading to deformation in high alpine environments.
(2) The PSI-based deformation velocities adopted in this study may not reflect the true ground motion. First, since the reported PSI velocities are the projection of true ground motion onto the line-of-sight directions (L130). Depending on the slope direction (slope angle and aspect angle) and satellite viewing geometry, the LOS velocities in a mountain region can be in either positive or negative. Taking an average of the LOS velocities would be inevitably biased towards zero, which seems to be the case in this study as the reported mean velocities in all the permafrost zones are all very close to zero (e.g., Table 4, Table 6, Figure 5). In other words, the reported mean LOS velocities could be averaged of positive and negative values from slopes of all different orientations.
This point has been noted also by the reviewer #1. As explained, the statistical analysis performed (ANOVA) required the condition of normally distributed datasets. We agree that this can bias towards zero the averages; however, we also highlight that using absolute values for the velocities will show similar if not higher differences between the velocities. We will modify the statistical analysis as indicated in the reply to reviewer #1.
Second, the reported LOS velocities are mostly within +/- 10 mm/year, even for ice-rich permafrost (e.g., Figure 5). These are much lower than the expected velocities of rock glaciers (on the order of a few cm to meters per year). The three local cases presented in Figures 6 to 8 would be regarded as ‘outliers’ in the histograms of Figure 5. Still, the PSI velocities presented in Figure 6 seem to be only about 50% of the ground-measured velocities.
Most of rock glaciers do not appear in the PSI dataset because their surface changes are too high and produce decorrelation of the SAR phase. This can be due to high velocities but also because of sharp changes of the electromagnetic properties at their surface. We will better specify this point in the revised version. Despite, we think that the fact that surface velocities in the PGIM zone 2 are higher than in other zones even without considering rock glaciers is an important finding. In the revised version, we will modify the representation of the dataset by using violin plots, which are better suited compared with box plots to highlight the distribution of the data and the impact of outliers.
Minor comments:
L8: displacement (singular)
The sentence is ‘displacement time series’, singular is correct.
L11: what kind of additional information
We will add the following text, ‘such as geological maps, knowledge of the local situation, on-site field measurements’.
L14: the Arctic experienced the fastest warming (Rantanen et al. The Arctic has warmed nearly four times faster than the globe since 1979. Communications Earth & Environment, 2022, 3(1): 1-10.)
Despite our article is about alpine permafrost, we can add the reference mentioned by the 2nd reviewer is about permafrost in the arctic in introduction to broaden the relevance of the topic investigated.
L20: may add ‘area’ or ‘extent’ to specify what has declined
We will add the word ‘areas’.
L22: *lower* boundary
We will add the word ‘lower’.
L23: the active layer is on top of, not part of, permafrost
We will change the sentence into: “On top of the permafrost, an active layer that freezes and thaws each year, is located”.
Figure 1: add distance scale bars in the maps
We will add the distance scale bar.
L43: permafrost stability is not the same as permafrost zonation
We agree and we will change this sentence accordingly to avoid confusion.
L55: ‘accesible’ should be accessible
Thanks for spotting the typo, we will correct the word.
L73 and Table 2 caption: ??
We will correct this, thanks.
L101: In cases of shadow or layover, the R-index should be zero. Shouldn’t these be the cases when S_h = 0 or L = 0, opposite to the authors’ definition?
We thank the reviewer for this remark. This is a typo in line 101 and will correct it.
L109: subdivision (singular)
We will correct this.
L117: deformation is measured “in a remote monitoring network of automated measurement stations”, using what method? And why these independent measurements were not used to validate the PSI results?
In the Swiss Permafrost Bulletin 2022 (about the Permos data) this is explained in Chapter 5.1. ‘Annual terrestrial geodetic surveys are performed using high precision differential GNSS or total stations’.
We already explained in our reply to reviewer#1 why we only used one location with PERMOS data and not all locations.
L212: specifically
There is no word ‘specifically’ in Line 212 and therefore we do not understand this comment.
Section 2.3: need to provide more details of the PGIM, for instance: what methods were used; what are the main limitations; why need to be evaluated?
As in our reply to reviewer #1: The limitations of the PGIM are clearly stated in the paper and the readers are referred to the Kenner et al., 2019 paper for more specific information, where the method to generate this map is presented.
L132: a map
We will correct this, thanks.
L140: what kind of ground temperature, mean annual ground temperature at a particular depth, or something else?
In the paper of Kenner et al., 2019, to which we refer to, this information is explained in detail Chapter 2.1.
L146: what are the reasons for large uncertainties?
In the paper of Kenner et al., 2019, to which we refer to, this information is explained in detail Chapter 4.1.
L159: ??
We will correct this, thanks.
Table 4: better to state that the numbers of PS are reported to their nearest 1000.
We will add this.
It is meaningful to report velocity values in two significant digits after the decimal points in mm/yr, given the accuracy level of Sentinel-1 PSI?
We can change the amount of digits.
Figure 5: how were the outliers identified?
The outliers are calculated as follow, using the interquartile range (IQR) criterion.
Q1 = first quartile
Q3 = third quartile
IQR = Q3 -Q1
Upper fence = Q3 + (1.5 * IQR)
Lower fence = Q1 – (1.5 * IQR)
However, all outliers were included in the analyses. Please see also our reply to reviewer #1
L268: ‘en’ should be ‘and’
We will correct this, thanks.
Citation: https://doi.org/10.5194/egusphere-2023-2321-AC2
-
AC2: 'Reply on RC2', Kristina Juliana Reinders, 05 Apr 2024
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
531 | 135 | 39 | 705 | 26 | 37 |
- HTML: 531
- PDF: 135
- XML: 39
- Total: 705
- BibTeX: 26
- EndNote: 37
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1