the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 30 May 2024
A regional analysis of paraglacial landslide activation in southern coastal Alaska
Abstract. Glaciers worldwide are retreating rapidly due to anthropogenic climate change. One consequence of glacier mass loss is the destabilization of valley walls as the support provided by the glacier changes and eventually vanishes, a process known as ''debuttressing.'' In this work, we examine the evolution of eight large, active instabilities in southern coastal Alaska, a region experiencing some of the fastest glacier retreat worldwide. At half of the sites, the glacier is still in contact with the landslide, while in the other four cases, the terminus retreated past the landslide in recent decades. One site has experienced catastrophic failure; the others have not. We use automatic and manual feature tracking on optical imagery to derive slope movement from the 1980s to present and compare this with glacier terminus retreat and thinning, precipitation, and seismic energy. We find that the majority of sites underwent a pulse of accelerated landslide motion (up to 17 times higher compared to the five years preceding the acceleration) during the studied time period and that the subsequent deformation was independent of the initial activation. In two cases, the acceleration occurred after a particularly rainy month and/or a marked increase (around two times higher than the 1960–2000 average) in glacier thinning. At two further sites, no distinct activation could be detected, though both landslides are known to be moving at velocities below the detection threshold of the methods employed here. In four cases, landslide activation coincided with the rapid retreat (up to 12 times the long term average) of a lake- or marine- terminating glacier past the instability. Our results suggest that landslides adjacent to lake- or marine- terminating glaciers may be especially susceptible to sudden activation, which we hypothesize is due to the faster retreat rates of water-terminating glaciers as well as mechanical and hydrological changes resulting from the replacement of ice with water at the landslide toe. This work shows that glacier retreat can be associated with increasing landslide hazards in various glaciological settings in Alaska, which has implications for the assessment of hazards in a warming world.
- Preprint
(8699 KB) - Metadata XML
-
Supplement
(5179 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2024-1086', Anonymous Referee #1, 09 Jul 2024
General Comment
The main short-coming of this study concerns it’s overarching premise, which is stated as a regional analysis. The concept of regional analysis is not clearly defined in terms of scientific goals or questions being addressed at that scale. The eight landslide objects selected represent 1% of the total landslide inventory (perhaps this is a little higher when only lake/fjord terminating glaciers in southern Alaska are considered), and hence the authors need to consider whether the outcomes of the study have any relevance to the regional-scale understanding of glacier retreat and landslide instability phenomena in the lakes and Fjords of southern Alaska.
In regional landslide studies, the analysis of multiple influencing factors are usually used to estimate susceptibility or plausibility of a landslide occurrence, particularly in terms of identifying patterns and key driving factors, in either a spatial and/or temporal context. The study here lacks this type of broad rationality, but rather focuses on eight landslide case-studies each with their own unique conclusions.
Two options to consider.
- Revise the scope and the purpose of the study to reflect the case-study oriented approach using the eight landslides selected. However, bear in mind that the main outcome(s) lack scientific strength, e.g. it is already well established in the scientific literature that glacier volume reduction (e.g. thinning/retreat) can lead to slope movements, and on rare occasions, catastrophic failure.
- The potential strength of this study lays in the elucidation of the regional distribution of the landslide/glacier interactions, and the broader (spatial) understanding of dominant driving factor. This should include a regional consideration of rainfall and seismic activity (landslide-fault proximity). The current study already goes in that direction but is limited. To highlight a point stated in the conclusions, there is a potentially broader use of ITS-LIVE data on a world-wide scale, but to demonstrate this potential the authors should focus on a truly regional study across the fjords of southern Alaska, for which there is an extensive landslide and glacier inventory.
Other comments
- The manuscript is highly descriptive and long and could do with shortening/condensing, including moving non-essential information to a supplementary section.
- The abstract is vague and requires more context with respect to the conclusions. As an example “..17 times higher compared to five years preceding the acceleration” (see line 10) – what is this referring to? The abstract needs to be more understandable.
- Line 50: The discussion of the Tungnakvislarjökull landslide should be moved to the section dealing with landslides in an Alpine setting.
- Line 50 discussions on landslides in Alpine setting section, the relationship of glacier unloading and landslide movement was established by Kos et al (2016), the statement pertaining to altered groundwater creating critical conditions is a speculative assertion in Glueer et al (2020).
- Line 55, The potential landslide velocity is primarily controlled by litho-structural characteristics, buttressing ice is an external resisting factor with minor influence.
- Line 55, “, and Kos et al (2016) suggested that landslides react rapidly to glacier changes upon crossing a threshold of ice loss”.
- Line 60, Is it correct to write thinning or retreat? Retreat and thinning occurs simultaneously, are there glaciers that retreat without thinning or vice versa?
- Line 70, I have doubts whether this study is a detailed regional
- Line 75, “…and discuss in the context of the possible physical mechanisms behind the slope instabilities (sect 5)”. What of the key factors geology and rock mass characteristics determining physical mechanisms? These are not featured nor discussed in the manuscript.
- Line 95, why are the criteria selected important of relevant? There could certainly be more criteria to consider geological susceptibility, permafrost thermal state etc. These are very important factors (spatially) to consider in a regional study.
- Section 2.1-2.8, these descriptions could be moved to a supplementary section and/or tabulated so they are more easily read.
- Figure 1a, legend for the geological base map should simply indicate the rock types (formation names are not particularly useful for those readers who don’t know the local geology)
- Figure 1e is hard to read, perhaps make it into a separate larger figure. It would be useful to see the distribution of the landslide inventory more clearly
- Arrange the figure 4 sub components in the same order as figures 2 and 3.
- Line 400, How would a precipitation (long-term) trend be an important factor for landslide activation/failure? What is the relationship? E.g. these questions need to be placed in the context of the enormous rainfall that parts of southern Alaska receive.
- Figure 6, This figure is difficult to read, they could be much larger in size, and maybe show only one Landsat image as a reference, with glacier/landslide outlines for each year where there are changes observed.
- Figure 7. What is the significance of showing cumulative monthly precipitation plots in terms of landslide activation/failure when the dark blue lines correspond to other years (light blue lines) where no activation is recorded? Why isn’t rainfall a significant conditioning/triggering factor in southern Alaska?
- Line 415, an untested hypothesis is merely speculation. Is there precedence in the literature for this assertion? Either demonstrates the plausibility of a snow loaded slope triggering movement or leave it out.
- Section 4.4.2, triggering of a landslide would be associated temporally with the earthquake occurrence, but this relationship unfortunately cannot be shown. Lack of evidence in this case doesn’t mean that seismic activity is not important, it may actually be more important than glacial retreat for landslide activation/ongoing failure – this should be expanded in the discussion.
- Line 495, How reliable is it to compare ice thinning thresholds between Alpine landslide/glaciers and landslide/Fjord glaciers? How relevant is 100m of thinning at Moosfluh to the cases in southern Alaska. On this point, please always refer to the primary literature where phenomena/relationships are first reported and then the later studies that find confirmation – the Glueer et al article falls in the latter.
- Lines 560, The several factors that the authors did not consider are central to the discussion, and therefore need to be included in a regional study.
***********
A regional analysis of paraglacial landslide activation in southern coastal Alaska deals with eight landslides, selected from a landslide inventory, each of which is characterized by its proximity to an already retreated or retreating lake/fjord terminating glacier.
The main premise of the study is stated by the authors is as follows:
- “To understand how glacier thinning and retreat control landslide activation and mobility” (line 70).
- “Whether the (re)-activation of paraglacial landslides can be explained by glacier retreat” (line 70).
To answer the above questions, the main methodology and data used in the analysis consisted of the following:
- Velocity analysis of the selected glaciers and landslides utilizing ITS-LIVE annual mosaics. The data for ITS-LIVE is derived from Landsat 4-8 image pairs, has a resolution of 120m and the relevant parameters include velocity magnitude, horizontal components of movement, and uncertainty estimates (section 3.1)
- Surface elevation change analysis, using a total of seven DEMs. The existing DEMs were derived from topographic maps (composite years in at least one product), satellite stereo images and ASTER radiometric data. DEM resolution ranged between 30-48 metres (section 3.2)
- Mapping of the glacier terminus position from Landsat images for the period (1985-2022) (section 3.3)
Supporting methods and data used in the analysis included:
- Determination of ice thickness from global datasets of Farinotti et al. (2019) and Milan et al, (2022), which included estimates of the glacier bed depth (section 3.4.1).
- Temperature and precipitation time series data derived from ERA5 reanalysis for the period 1979-2022 (section 3.4.2).
- Extraction of seismic events from the USGS Earthquake Catalogue for the period 1980-2023 (section 3.4.3).
The main conclusions of the study consist of the following:
- “Six of the eight landslides underwent significant slope acceleration at some stage during the studied time period” (line 620).
- “At four [of the six] sites,…acceleration occurred as the glacier terminus retreated past the landslide area” (line 620).
- “For another two sites, landslide acceleration either coincided with a particularly rainy month or a significant increase in glacier thinning” (line 620).
- “The remaining two sites …showed either slow, constant movement without a specific period of acceleration, or no detectable movement at all” (line 620).
Citation: https://doi.org/10.5194/egusphere-2024-1086-RC1 - AC1: 'Reply on RC1', Jane Walden, 15 Nov 2024
-
RC2: 'Comment on egusphere-2024-1086', Stuart Dunning, 22 Jul 2024
This is a really interesting study that adds to our understanding of how large slopes behave in very transient parts of the landscape, that, despite their activity remain poorly quantified. Please take the following as constructive, and, as a means to stimulate discussion which can only improve the final uptake of this good piece of work and drive future work to test your ideas. I do believe the abstract could be far tighter and would benefit from a rewrite, especially when talking about rates – try to keep this to the take home messages.
Although the case-study landslides are from a ‘region’, I do not think that can be considered a ‘regional’ analyses. Eight is not enough to unpick a regional pattern/differences or the behaviour of lake, marine or still ice contact slopes. I would like some further justification for the selection of the 8, how are they representative of the 780, rather than some (rare?) end member? For example, at Ellsworth you note a number of landslides, so, what is the rationale to only characterise one, rather than a suite of them to see if the behaviour (and lags to ice front change) are comparable or if you are picking earlier phases/time zero of instability further up ice? The same goes for the two at Portage, what is the rationale for only looking at the largest one? Can you say why a minimum of 10 million volume was used? Landslides far smaller are able to cause damaging landslide-tsunami unless you can say otherwise? Something that needs to be dealt without throughout is the behaviour of large rockslopes, and, the science of creep that may or may not transition to an eventual catastrophic failure – this would give more context to the role of speeds ups, ie, are they significant, or, not suggestive of any move towards (or away from) some eventual failure that can cause the cascading risks you note.
The authors use ‘debuttressing’ very rapidly in the introduction without critique. It is not universally accepted as a causal mechanism of slope failure (particularly in bedrock) but at various points you seem to be ascribing the word as a process – rather than saying a slope became ice-free for example in line 141. You return to this later, but, I’m not convinced ‘debuttressing’ can be applied, rather than the more correct statement later that ice-loss / thinning has been associated with data and modelling to instability and changes in landslide motion.
Similarly, ‘paraglacial’ landslide formation is used in line 70 with no setup as to what you mean by the use of this term or original references to the term. Are you implying that these landslides did not exist prior to the onset of (this) deglaciation?
I like the reuse of ITS-LIVE for a new purpose, but, are yearly data too crude to characterise the questions – can you provide a little justification for the use of annual data, and the limitations of doing so? You say in Line 204 that using a larger timestep may show more (perhaps part of the rationale for annual data as the finest step), but I disagree automatic feature tracking would not cope with this, is it a matter of the image temporal separation that you feed into automatic optical feature tracking. I am also surprised that the ice velocity product is not used as a supplementary dataset, if might prove to have little use and be ruled out, but, I’d expected to see velocity changes of the ‘buttressing’ ice alongside other plots – can you say why these data are not considered useful? It is clear from some work that ice-velocity can responds to landslide displacement into the glacier, and, it may (debate!) be that velocity of the ice may feedback to the velocity of the landslide(s). It would also be good to see data on Fig 3, is there a glacier response in the velocity field as the landslide deforms into/under the ice?
In the results, text and Figure 4 I can’t reconcile the red bars with the velocity of the landslides, most are clearly moving before the red bar, sometimes for years, how have you decided that this red bar is the onset?
There seems to be considerable disparity in the ITS-LIVE and manual data, for example, Grand Plateau in Fig 4, which one you use fundamentally changes the interpretation. E..g for G Plateau it is either decreasing in velocity as the terminus retreats to the landslide, and then becomes quite constant or, it is near zero then starts to accelerate as the terminus reaches the landslide. There are similar things to resolve with many of the 8 (Tyndall is another where the data are showing two very different patterns, increasing to 2000 with the manual, or, decreasing to ~2005 from a peak.
The precipitation analyses seems to be setting up a model that the landslide must be responding to the rainfall in that year? Why not be plotting some form of precipitation anomaly or metric on a figure like 4 that shows the two variables as a time series? I would want to look at landslide responses to rainfall over time, is there any, and, is it consistent or does any relationship break down after some ‘key’ event? I think the same is true of the seismic data presented, it is an interpreted derivative of velocity you are showing – the accelerated portion, rather than presenting these data to show the full time series links (or not links looking at it). If there is very little link, does it warrant being in the main text, or, in the SI with a few lines saying the seismic intensity time series showed little? For both precip and seismic I do wonder if the annual time series approach to deformation may mask any links.
Discussion: Really interesting ideas deserving of further investigation. You talk of landslide activation, but, as a generalisation here, you are dealing with landslides that were already ‘features’ at the start of the analyses, rather than capturing the true activation/initiation and associated conditions? I wonder (and you should cove this) how representative these forms of landslide are compared to those that show no precursory creep of note (or do they and it has been missed) and fail catastrophically? I wonder around lines 525 onwards where you think about differing controls you should be posing a conceptual diagram to bring these ideas together and allow you to pose questions/ideas around the future of these instabilities.
The change from landslide-ice to landslide-water, you say that this changes toe saturation. Are these not wet/warm glaciers where the margins and base are likely to be very wet already? What is instantly (on a geological time scale) is the yield stress of what the slope is resting against as you say. Is that more or less likely to induce failure as compared to a land terminating removal of ice at the slide margins?
Lines 583 around tidal/wave influences needs to be supported by your data.
It feels a shame to end the discussion with a following section on the limitations, it is quite a negative finish, it would be nicer to keep this with the methods, as, that is when I had those questions to ask (unless you have to match journal convention).
LINE COMMENTS
Line 39-40: Few cases studies in depth, are you saying there isa far larger population of landslide-tsunami that are unstudied, or, there have been few? I’m also unclear if you are ending the paragraph referring to all large landslides associated with a glacier, or, the ones that enter water bodies specifically.
Line 63: I’m not sure I recognise post-failure instabilities. They are either an instability, or, a relict landslide/mass movement deposit?
Line 63: I presume it is not only retreat, but also thinning that is contributing this mass
Line 89: Jumping between ‘local’ SE Alaska numbers of relevance e.g. uplift, to National (e.g. earthquakes).
Line 178: You can say ‘may’ control, or, set this as a formal hypothesis to test, but as written it looks like you have concluded that thinning/retreat DO control ‘mobility’, a term I’m not sure is what is often used, over deformation, creep, motion etc, as the norm is often that mobility is around the final position of a failed mass?
Line 184: Why are they alternative drivers? We know precip and seismics can have an effect, so, why not here assume that all may play a role, but some might be more significant, and, there is the possibility that some may actually compound the deformation process-response?
Line 186: I presume ITS-LIVE is optimised for the glacier areas rather than the boundaries of scene pairs where edge effects may play a part, do the data extend far beyond the slide boundaries, with equal errors to the rest?
Line 209: But, you might be looking for retrogressive behaviour or evidence of future upslope instability?
Line 211: I’m not clear what you mean by the average of individual features, do you use a statistical treatment of the per-pixel displacements, or, a subset of these that have been quality controlled? Do you use off-landslide pixels as a control measure?
Line 239: Happy to talk to you about break-point/change-point analyses to statistically begin to ID the rapid phases in a quant/reproducible way.
Line 248: Give indication of when in the year you make these measures without the need to look at SI please.
Line 285: There is still debate on the role of shaking: https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2021JF006242
Line 303: Not including the manual measurements?
Lines 581/2- not convinced by pers. Comm. part given they are co-authors, can data snapshot not be down in the SI?
FIGURE COMMENTS:
Fig 1. On (e) having ice as white when the ocean also is does not work. Faults (and the difference in dashed lines) needs to be in the legend. The landslide locations A-D are as capitals, the panel letters are lower case. Move the p.frost prob lower, the legend does not sit well in the figure layout. (a) It might be my PDF version, but the water as grey is a bit odd and counter-intuitive, suggest some coherence with water as suggested in (e) and perhaps some version of blue. The geology legend has unequal line spacing and second line indent. I would also expand b, c, d so that the panels match the width of a. Is there more use of the ice on the panels to be had, perhaps use ice-velocity data? Thre are no coordinate crosses in b-d?
Figure 2: Ice velocity data semi transparent may be helpful? Personal preference perhaps, but I do not like the coordinate crosses at all (same in Fig 1), I would rather an outside grid/ticks and a legend encased in solid white. When are the Planet Labs images from? Summer 2023? I hope it becomes clear in the text how you determine the subglacial extent of instability as it is a really interesting open question.
Figure 3: It seems a wasted opportunity not to plot the glacier velocity product on, is there evidence of ice deformation in response to where the slide mass is?
Figure 4: A very nice figure. I do wonder on other ways to plot some of these data in addition…..perhaps cumulative displacement, do be more in keeping with retreat, and/or, the ice thickness data, again, cumulative rather than a rate – it depends what you think is the important driver, the absolute change in thickness relative to the landslide, or the rate of change? I would like to see ‘major’ events incorporated e.g. crack opening using symbols.
Figure 5: I am not clear on how you decided the lower limit for the landslides? All the subglacial reconstructions suggest a steep trough ie no landslide toe. Is this likely ‘real’ ie they are scoured away, or, a function of the inversion process to yield thickness?
Figure 6: Sorry, this figure I struggle with and I would see moved to SI.
Citation: https://doi.org/10.5194/egusphere-2024-1086-RC2 - AC2: 'Reply on RC2', Jane Walden, 15 Nov 2024
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 352 | 181 | 31 | 564 | 28 | 14 | 13 |
- HTML: 352
- PDF: 181
- XML: 31
- Total: 564
- Supplement: 28
- BibTeX: 14
- EndNote: 13
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1