Assessing uncertainties in landslide susceptibility predictions in a changing environment (Styrian Basin, Austria)

Knevels, Raphael; Petschko, Helene; Proske, Herwig; Leopold, Philip; Mishra, Aditya N.; Maraun, Douglas; Brenning, Alexander

doi:https://doi.org/10.5194/nhess-2022-154

Preprints

https://doi.org/10.5194/nhess-2022-154

Preprints

03 Jun 2022

Assessing uncertainties in landslide susceptibility predictions in a changing environment (Styrian Basin, Austria)

Raphael Knevels¹, Helene Petschko¹, Herwig Proske², Philip Leopold³, Aditya N. Mishra⁴, Douglas Maraun⁴, and Alexander Brenning¹ Raphael Knevels et al.

Show author details

¹Department of Geography, Friedrich Schiller University Jena, Jena, 07743, Germany
²Remote Sensing and Geoinformation Department, JOANNEUM RESEARCH Forschungsgesellschaft mbH, Graz, 8010, Austria
³Centre for Low-Emission Transport, AIT Austrian Institute of Technology GmbH, Vienna, 1210, Austria
⁴Wegener Centre for Climate and Global Change, Regional Climate Research Group, Karl-Franzens-University Graz, Graz, 8010, Austria

Received: 27 May 2022 – Discussion started: 03 Jun 2022

Abstract. The assessment of uncertainties in landslide susceptibility modelling in a changing environment is an important, yet often neglected task. In an Austrian case study, we investigated the uncertainty cascade in storylines of landslide susceptibility emerging from climate change and parametric landslide model uncertainty. In June 2009, extreme events of heavy thunderstorms occurred in the Styrian basin, triggering thousands of landslides. Using a storyline approach, we discovered a generally lower landslide susceptibility for pre-industrial climate, while for future climate (2071–2100) a potential increase of 35 % in highly susceptible areas (storyline of much heavier rain) may be compensated by much drier soils (-45 % areas highly susceptible to landsliding). However, the estimated uncertainties in predictions were generally high. While uncertainties related to within-event internal climate model variability were substantially lower than parametric uncertainties of the landslide susceptibility model (ratio of around 0.25), parametric uncertainties were of the same order as the climate scenario uncertainty for the higher warming levels (+3 K and +4 K). We suggest that in future uncertainty assessments, an improved availability of event-based landslide inventories and high-resolution soil and precipitation data will help to reduce parametric uncertainties of landslide susceptibility models used to assess the impacts of climate change on landslide hazard and risk.

Download & links

Preprint (PDF, 12884 KB)

Notice on discussion status
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
Preprint (12884 KB)

Supplement (11884 KB)

Download & links

The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.

Preprint (12884 KB)
Supplement (11884 KB)
BibTeX
EndNote

Journal article(s) based on this preprint

Raphael Knevels et al.

Interactive discussion

Status: closed

RC1:
'Reviewer Comment on nhess-2022-154', Anonymous Referee #1, 30 Jun 2022

The manuscript is a good piece of science addressing a relevant issue, i.e. the evaluation of the uncertainties in landslide susceptibility modeling, taking into account also climate and environmental changes. The manuscript is clear, written in fluent English, and well-organized. The figures and tables are useful for presenting and discussing the results. The introduction is complete and useful for focusing on the topic. The method is well presented and the discussion is clear. However, I've found some issues that need to be addressed and explained before the manuscript can be reconsidered for publication.

Although earlier work by some of the authors is recalled in several places, the whole manuscript is rather long, so I would suggest trying to shorten it by at least 10% of the current length.

In the following, I list the main methodological issues I found in the manuscript. Overall, I think that minor revisions are needed.

---

You use the term "storyline approach". I can't grab what you mean with "storyline" and "storyline approach". It seems to me that this term is not common in landslide analyses. I would suggest adding some explanation

You defined the events that occurred in June 2009 and September 2014 as "extreme". How can you classify such events as "extreme"? Was a statistical analysis carried out?

You added in the susceptibility analysis the rainfall data on the landslide failure day, i.e. the triggering precipitation conditions. I think this is questionable and in contrast with the theoretical definition of susceptibility (see e.g. Reichenbach et al. (2018) [https://doi.org/10.1016/j.earscirev.2018.03.001]; van Westen et al. (2008) [https://doi.org/10.1016/j.enggeo.2008.03.010]). Landslide susceptibility is "the likelihood of a landslide occurring in an area on the basis of the local terrain and environmental conditions", therefore the triggering rainfall conditions should be removed from this analysis.

You also wrote "For the landslide susceptibility analysis, we linked predisposing and triggering factors to landslide occurrences.". I think this can be considered a methodological issue.

Regarding the environmental change simulation, you wrote (line 153) that "Adopting active forest management in the developed future LULC scenario, coniferous forest was replaced by climate resilient mixed forest".

If I have understood well, all LULC changes were defined to have a potentially positive effect on slope stability. If it is so, why not considering also some "negative" changes?

Furthermore, you wrote (line 161) "Specifically, for each grid cell we determined the maximum three-hour rainfall intensity, and we took the maximum five-day rainfall."

In my opinion, also this is questionable, given that it is not always the most severe rainfall condition during a meteorological event that can trigger landslides. An explanation is needed.

Finally, I suggest using round brackets for units of measurement. Please check all over the text.

Citation: https://doi.org/10.5194/nhess-2022-154-RC1
- AC1: 'Reply on RC1', Raphael Knevels, 16 Aug 2022
  
  The comment was uploaded in the form of a supplement: https://nhess.copernicus.org/preprints/nhess-2022-154/nhess-2022-154-AC1-supplement.pdf
  
  Citation: https://doi.org/10.5194/nhess-2022-154-AC1
RC2:
'Comment on nhess-2022-154', Anonymous Referee #2, 23 Jul 2022
This paper investigated the uncertainty cascade in storylines of landslide susceptibility emerging from climate change and parametric landslide model uncertainty. In general, this paper is interesting and rich in content.  However, there are many mistakes in the basic concept of landslide susceptibility, hazard and risk assessment.  In terms of landslide susceptibility prediction modeling process, the writing of this paper is rather rough.  In terms of organizational structure, the thesis is difficult to understand.  Therefore, it is recommended to reject the paper.

Landslide susceptibility refers to the spatial probability of landslide occurrence affected by landslides them-selves conditioning factors, without considering triggering factors such as heavy rainfall, earthquake, et al. Landslide hazard refers to the spatial and time probability of landslide occurrence under condition factors and trigger factors. Hence, this paper focus on landslide susceptibility affected by land cover change and heavy rainfall. I believe this paper has problems with the basic concepts of landslide susceptibility.

The writing ideas of this paper are very confused, and it is difficult for people to understand the specific steps and methods of the research.  Especially in the introduction and methods section.

Where are the input and output variables of landslide susceptibility prediction modeling described in this paper?  How is the uncertainty problem concerned in this paper quantified?  Where are the environmental factor maps of landslide susceptibility and landslide susceptibility outcome maps?  These problems are not well reflected.

The figures are not clear enough.

The references are not new enough and are not representative enough.

The uncertainty characteristics are assessed by which indexes? These description are not clear.

There is insufficient analysis of feasible solutions to the problems in this paper.
Citation: https://doi.org/10.5194/nhess-2022-154-RC2
- AC2: 'Reply on RC2', Raphael Knevels, 16 Aug 2022
  
  The comment was uploaded in the form of a supplement: https://nhess.copernicus.org/preprints/nhess-2022-154/nhess-2022-154-AC2-supplement.pdf
  
  Citation: https://doi.org/10.5194/nhess-2022-154-AC2

Peer review completion

AR: Author's response | RR: Referee report | ED: Editor decision

ED: Publish subject to minor revisions (review by editor) (25 Oct 2022) by Filippo Catani

dear authors,

first of all, let me apologize for the time that this review has taken. This, as you may have supposed, has had a lot to do with two quite contrasting reviews that arrived from the selected referees.
The indecisiveness of the reports forced us to ask for a third review, which unfortunately arrived after the review acceptance time. So, I was left with the only solution of revising the paper myself.
After all this, and despite some issues that you will find below, I am glad to inform you that the paper can be accepted should you be willing to apply some minor modifications.
Please note that, here, "minor" should be intended as something below "major" but still important, in my opinion. So, no other review round will be needed but I will revise your modifications personally, to expedite the acceptance process.
Now, going to the point, the main issues of the manuscript are those highlighted in the review reports, which you will find attached, plus the following, partially overlapping, points:

1. classical susceptibility does not usually include time - it is intended to produce a statistical inference on the relative probability in space. As a consequence, susceptibility is usually based on historical inventories and tries to predict the occurrence of landslides over a time span of the same order of magnitude as the time span of the inventory itself (uncertainties and missing information considered). Instead, you propose here event-based landslide maps to train a statistical model, which seems to be valid only for similar triggering events. In this way, to the extent that your climate-change forecasts are correct, you are predicting the relative response of the landscape to this event type, only. This poses two main concerns:
1A. How can then your samples (being based only on a few rare, documented events) be statistically representative of all the future expected extreme rainfall events?
1B. How can you support (statistically) the inference that the effect produced by the triggering event X in position Y is able to produce the same effect in position Z (where local conditions are different from site Y)?
I am not saying that you model is flawed, as I do not have sufficient information to say so, or the opposite. But maybe you could try to better frame the limits and conceptual context in which you are working. The approach may well be new, but at the moment we can’t say if it is truly appropriate for the usual purposes of susceptibility mapping, that is scenario building, land planning, or risk reduction.

2. The topic in general is far from new. There are several papers published in the last 10 years that propose multi-temporal susceptibility across different climate change scenarios. It would be good if you listed at least some of them, to better set the background as well as to be able to better explain why your approach is novel. There are several papers, as well, that propose model sensitivity in statistical analysis of landslide hazard, including the influence of the different mapping parameters, scales, resolutions, sampling methods and so on. Please also enrich the list of references as related to this aspect. I do not want to add specific suggestions, as it is quite easy to find them. This would also help in framing your own research and the gaps you meant to fill.
3. As a side note, more extreme weather events of the same type are available in the study region, as far as I know. Did you try to include them in the analysis? Or did you try to use them as external validation of your model? (see also references cited below on the Austrian Alps).

4. Please consider that in the Eastern Alps, recent extreme events such as the Vaia storm of 2018 have highlighted that rainfall may start triggering landslides not just during the main storm, which only creates some preconditions, so to speak. But, instead, after several weeks. You use 5-day and 3-hour accumulated rain in your study. How can this influence your results? How can your statistics account for this delayed-time triggering?

Some minor points:
5. In lines 14-15 you report UNISDR and CRED data: Economic Losses, Poverty & Disasters (1998 - 2017) that cover a variety of disasters around the globe; I would suggest including one more study to the introduction, which was solely based on landslides, such as e.g.:
i) Froude, M. J. and Petley, D. N.: Global fatal landslide occurrence from 2004 to 2016, Nat. Hazards Earth Syst. Sci., 18, 2161–2181, https://doi.org/10.5194/nhess-18-2161-2018, 2018.

6. In lines 19-20, you might want to consider and include the recent works related to landslide susceptibility in the Austrian alps, such as:
i) Lima, P., Steger, S., & Glade, T. (2021). Counteracting flawed landslide data in statistically based landslide susceptibility modelling for very large areas: a national-scale assessment for Austria. Landslides, 18(11), 3531-3546.
ii) Moharrami, M., Naboureh, A., Gudiyangada Nachappa, T., Ghorbanzadeh, O., Guan, X., & Blaschke, T. (2020). National-scale landslide susceptibility mapping in aaustria using fuzzy best-worst multi-criteria decision-making. ISPRS International Journal of Geo-Information, 9(6), 393.
iii) Gudiyangada Nachappa, T., Kienberger, S., Meena, S. R., Hölbling, D., & Blaschke, T. (2020). Comparison and validation of per-pixel and object-based approaches for landslide susceptibility mapping. Geomatics, Natural Hazards and Risk, 11(1), 572-600

7. In lines 63-64, you could try to address the bias associated with downscaling the climate data by using an approach as the one here:
i) Roberts, DR, Wood, WH, Marshall, SJ. Assessments of downscaled climate data with a high-resolution weather station network reveal consistent but predictable bias. Int J Climatol. 2019; 39: 3091– 3103. https://doi.org/10.1002/joc.6005

8. You provide no maps of susceptibility, which could help in identifying the locations of impending occurrences in a spatial way. Could you please add some of the maps (supplementary material as well)?

My recommendation is that you try to answer all reviewers’ questions on a point-to-point basis in a separate document, and modify the original manuscript accordingly, as well. Please also try to reply to the above questions and suggestions.
After you submit the modified version and the response letter, the revision process will continue directly to the Editor’s consideration, without further review rounds.
Thank you for submitting to NHESS and, again, my apologies
FC

Hide

ED: Reconsider after major revisions (further review by editor and referees) (04 Nov 2022) by Filippo Catani

dear authors,

first of all, let me apologize for the time that this review has taken. This, as you may have supposed, has had a lot to do with two quite contrasting reviews that arrived from the selected referees.
The indecisiveness of the reports forced us to ask for a third review, which unfortunately never arrived. So, we had to carefully consider the two reports in the light of the submitted document and materials.
On the basis of the referees’ reports, and the main points summarized below, the paper could be accepted for publication should you be willing to apply some major modifications.
The topic of the manuscript is of large interest to the NH community, and the approach used is quite novel and potentially providing insights into something that is still debated and unsolved. However, there are a few important issues present in the manuscript as highlighted in the review reports, which you will find attached, plus the following, partially overlapping, points:

1. classical susceptibility does not usually include time - it is intended to produce a statistical inference on the relative probability in space. As a consequence, susceptibility is usually based on historical inventories and tries to predict the occurrence of landslides over a time span of the same order of magnitude as the time span of the inventory itself (uncertainties and missing information considered). Instead, you propose here event-based landslide maps to train a statistical model, which seems to be valid only for similar triggering events. In this way, to the extent that your climate-change forecasts are correct, you are predicting the relative response of the landscape to this event type, only. This poses two main concerns:
1A. How can then your samples (being based only on a few rare, documented events) be statistically representative of all the future expected extreme rainfall events?
1B. How can you support (statistically) the inference that the effect produced by the triggering event X in position Y is able to produce the same effect in position Z (where local conditions are different from site Y)?
I am not saying that you model is flawed, as I do not have sufficient information to say so, or the opposite. But maybe you could try to better frame the limits and conceptual context in which you are working. The approach may well be new, but at the moment we can’t say if it is truly appropriate for the usual purposes of susceptibility mapping, that is scenario building, land planning, or risk reduction.

2. The main topic, despite being of general interest, is far from new. There are several papers published in the last 10 years that propose multi-temporal susceptibility across different climate change scenarios. It would be good if you listed at least some of them, to better set the background as well as to be able to better explain why your approach is novel. There are several papers, as well, that propose model sensitivity in statistical analysis of landslide hazard, including the influence of the different mapping parameters, scales, resolutions, sampling methods and so on. Please also enrich the list of references as related to this aspect. I do not want to add specific suggestions, as it is quite easy to find them. This would also help in framing your own research and the gaps you meant to fill.
3. As a side note, more extreme weather events of the same type are available in the study region, as far as I know. Did you try to include them in the analysis? Or did you try to use them as external validation of your model? (see also references cited below on the Austrian Alps).

4. Please consider that in the Eastern Alps, recent extreme events such as the Vaia storm of 2018 have highlighted that rainfall may start triggering landslides not just during the main storm, which only creates some preconditions, so to speak. But, instead, after several weeks. You use 5-day and 3-hour accumulated rain in your study. How can this influence your results? How can your statistics account for this delayed-time triggering?

Some minor points:
5. In lines 14-15 you report UNISDR and CRED data: Economic Losses, Poverty & Disasters (1998 - 2017) that cover a variety of disasters around the globe; I would suggest including one more study to the introduction, which was solely based on landslides, such as e.g.:
i) Froude, M. J. and Petley, D. N.: Global fatal landslide occurrence from 2004 to 2016, Nat. Hazards Earth Syst. Sci., 18, 2161–2181, https://doi.org/10.5194/nhess-18-2161-2018, 2018.

6. In lines 19-20, you might want to consider and include the recent works related to landslide susceptibility in the Austrian alps, such as:
i) Lima, P., Steger, S., & Glade, T. (2021). Counteracting flawed landslide data in statistically based landslide susceptibility modelling for very large areas: a national-scale assessment for Austria. Landslides, 18(11), 3531-3546.
ii) Moharrami, M., Naboureh, A., Gudiyangada Nachappa, T., Ghorbanzadeh, O., Guan, X., & Blaschke, T. (2020). National-scale landslide susceptibility mapping in aaustria using fuzzy best-worst multi-criteria decision-making. ISPRS International Journal of Geo-Information, 9(6), 393.
iii) Gudiyangada Nachappa, T., Kienberger, S., Meena, S. R., Hölbling, D., & Blaschke, T. (2020). Comparison and validation of per-pixel and object-based approaches for landslide susceptibility mapping. Geomatics, Natural Hazards and Risk, 11(1), 572-600

7. In lines 63-64, you could try to address the bias associated with downscaling the climate data by using an approach as the one here:
i) Roberts, DR, Wood, WH, Marshall, SJ. Assessments of downscaled climate data with a high-resolution weather station network reveal consistent but predictable bias. Int J Climatol. 2019; 39: 3091– 3103. https://doi.org/10.1002/joc.6005

8. You provide no maps of susceptibility, which could help in identifying the locations of impending occurrences in a spatial way. Could you please add some of the maps (supplementary material as well)?

My recommendation is that you try to answer all reviewers’ questions on a point-to-point basis in a separate document, and modify the original manuscript accordingly, as well. Please also try to reply to the above questions and suggestions. Please be aware that direct replies in the online discussion web pages are not considered as a response to the decision process, as the latter requires direct upload of the said documents in the manuscript submission and revision page of the journal.
After you submit the modified version and the response letter, the revision process will continue directly to the next review round.

Thank you for submitting to NHESS and, again, my apologies for the delay.
Best regards
FC

Hide

AR by Raphael Knevels on behalf of the Authors (19 Nov 2022) Author's response Author's tracked changes Manuscript

ED: Publish as is (06 Dec 2022) by Filippo Catani

Interactive discussion

Status: closed

RC1:
'Reviewer Comment on nhess-2022-154', Anonymous Referee #1, 30 Jun 2022

The manuscript is a good piece of science addressing a relevant issue, i.e. the evaluation of the uncertainties in landslide susceptibility modeling, taking into account also climate and environmental changes. The manuscript is clear, written in fluent English, and well-organized. The figures and tables are useful for presenting and discussing the results. The introduction is complete and useful for focusing on the topic. The method is well presented and the discussion is clear. However, I've found some issues that need to be addressed and explained before the manuscript can be reconsidered for publication.

Although earlier work by some of the authors is recalled in several places, the whole manuscript is rather long, so I would suggest trying to shorten it by at least 10% of the current length.

In the following, I list the main methodological issues I found in the manuscript. Overall, I think that minor revisions are needed.

---

You use the term "storyline approach". I can't grab what you mean with "storyline" and "storyline approach". It seems to me that this term is not common in landslide analyses. I would suggest adding some explanation

You defined the events that occurred in June 2009 and September 2014 as "extreme". How can you classify such events as "extreme"? Was a statistical analysis carried out?

You added in the susceptibility analysis the rainfall data on the landslide failure day, i.e. the triggering precipitation conditions. I think this is questionable and in contrast with the theoretical definition of susceptibility (see e.g. Reichenbach et al. (2018) [https://doi.org/10.1016/j.earscirev.2018.03.001]; van Westen et al. (2008) [https://doi.org/10.1016/j.enggeo.2008.03.010]). Landslide susceptibility is "the likelihood of a landslide occurring in an area on the basis of the local terrain and environmental conditions", therefore the triggering rainfall conditions should be removed from this analysis.

You also wrote "For the landslide susceptibility analysis, we linked predisposing and triggering factors to landslide occurrences.". I think this can be considered a methodological issue.

Regarding the environmental change simulation, you wrote (line 153) that "Adopting active forest management in the developed future LULC scenario, coniferous forest was replaced by climate resilient mixed forest".

If I have understood well, all LULC changes were defined to have a potentially positive effect on slope stability. If it is so, why not considering also some "negative" changes?

Furthermore, you wrote (line 161) "Specifically, for each grid cell we determined the maximum three-hour rainfall intensity, and we took the maximum five-day rainfall."

In my opinion, also this is questionable, given that it is not always the most severe rainfall condition during a meteorological event that can trigger landslides. An explanation is needed.

Finally, I suggest using round brackets for units of measurement. Please check all over the text.

Citation: https://doi.org/10.5194/nhess-2022-154-RC1
- AC1: 'Reply on RC1', Raphael Knevels, 16 Aug 2022
  
  The comment was uploaded in the form of a supplement: https://nhess.copernicus.org/preprints/nhess-2022-154/nhess-2022-154-AC1-supplement.pdf
  
  Citation: https://doi.org/10.5194/nhess-2022-154-AC1
RC2:
'Comment on nhess-2022-154', Anonymous Referee #2, 23 Jul 2022
This paper investigated the uncertainty cascade in storylines of landslide susceptibility emerging from climate change and parametric landslide model uncertainty. In general, this paper is interesting and rich in content.  However, there are many mistakes in the basic concept of landslide susceptibility, hazard and risk assessment.  In terms of landslide susceptibility prediction modeling process, the writing of this paper is rather rough.  In terms of organizational structure, the thesis is difficult to understand.  Therefore, it is recommended to reject the paper.

Landslide susceptibility refers to the spatial probability of landslide occurrence affected by landslides them-selves conditioning factors, without considering triggering factors such as heavy rainfall, earthquake, et al. Landslide hazard refers to the spatial and time probability of landslide occurrence under condition factors and trigger factors. Hence, this paper focus on landslide susceptibility affected by land cover change and heavy rainfall. I believe this paper has problems with the basic concepts of landslide susceptibility.

The writing ideas of this paper are very confused, and it is difficult for people to understand the specific steps and methods of the research.  Especially in the introduction and methods section.

Where are the input and output variables of landslide susceptibility prediction modeling described in this paper?  How is the uncertainty problem concerned in this paper quantified?  Where are the environmental factor maps of landslide susceptibility and landslide susceptibility outcome maps?  These problems are not well reflected.

The figures are not clear enough.

The references are not new enough and are not representative enough.

The uncertainty characteristics are assessed by which indexes? These description are not clear.

There is insufficient analysis of feasible solutions to the problems in this paper.
Citation: https://doi.org/10.5194/nhess-2022-154-RC2
- AC2: 'Reply on RC2', Raphael Knevels, 16 Aug 2022
  
  The comment was uploaded in the form of a supplement: https://nhess.copernicus.org/preprints/nhess-2022-154/nhess-2022-154-AC2-supplement.pdf
  
  Citation: https://doi.org/10.5194/nhess-2022-154-AC2

Peer review completion

AR: Author's response | RR: Referee report | ED: Editor decision

ED: Publish subject to minor revisions (review by editor) (25 Oct 2022) by Filippo Catani

Hide

ED: Reconsider after major revisions (further review by editor and referees) (04 Nov 2022) by Filippo Catani

dear authors,

first of all, let me apologize for the time that this review has taken. This, as you may have supposed, has had a lot to do with two quite contrasting reviews that arrived from the selected referees.
The indecisiveness of the reports forced us to ask for a third review, which unfortunately never arrived. So, we had to carefully consider the two reports in the light of the submitted document and materials.
On the basis of the referees’ reports, and the main points summarized below, the paper could be accepted for publication should you be willing to apply some major modifications.
The topic of the manuscript is of large interest to the NH community, and the approach used is quite novel and potentially providing insights into something that is still debated and unsolved. However, there are a few important issues present in the manuscript as highlighted in the review reports, which you will find attached, plus the following, partially overlapping, points:

1. classical susceptibility does not usually include time - it is intended to produce a statistical inference on the relative probability in space. As a consequence, susceptibility is usually based on historical inventories and tries to predict the occurrence of landslides over a time span of the same order of magnitude as the time span of the inventory itself (uncertainties and missing information considered). Instead, you propose here event-based landslide maps to train a statistical model, which seems to be valid only for similar triggering events. In this way, to the extent that your climate-change forecasts are correct, you are predicting the relative response of the landscape to this event type, only. This poses two main concerns:
1A. How can then your samples (being based only on a few rare, documented events) be statistically representative of all the future expected extreme rainfall events?
1B. How can you support (statistically) the inference that the effect produced by the triggering event X in position Y is able to produce the same effect in position Z (where local conditions are different from site Y)?
I am not saying that you model is flawed, as I do not have sufficient information to say so, or the opposite. But maybe you could try to better frame the limits and conceptual context in which you are working. The approach may well be new, but at the moment we can’t say if it is truly appropriate for the usual purposes of susceptibility mapping, that is scenario building, land planning, or risk reduction.

2. The main topic, despite being of general interest, is far from new. There are several papers published in the last 10 years that propose multi-temporal susceptibility across different climate change scenarios. It would be good if you listed at least some of them, to better set the background as well as to be able to better explain why your approach is novel. There are several papers, as well, that propose model sensitivity in statistical analysis of landslide hazard, including the influence of the different mapping parameters, scales, resolutions, sampling methods and so on. Please also enrich the list of references as related to this aspect. I do not want to add specific suggestions, as it is quite easy to find them. This would also help in framing your own research and the gaps you meant to fill.
3. As a side note, more extreme weather events of the same type are available in the study region, as far as I know. Did you try to include them in the analysis? Or did you try to use them as external validation of your model? (see also references cited below on the Austrian Alps).

4. Please consider that in the Eastern Alps, recent extreme events such as the Vaia storm of 2018 have highlighted that rainfall may start triggering landslides not just during the main storm, which only creates some preconditions, so to speak. But, instead, after several weeks. You use 5-day and 3-hour accumulated rain in your study. How can this influence your results? How can your statistics account for this delayed-time triggering?

Some minor points:
5. In lines 14-15 you report UNISDR and CRED data: Economic Losses, Poverty & Disasters (1998 - 2017) that cover a variety of disasters around the globe; I would suggest including one more study to the introduction, which was solely based on landslides, such as e.g.:
i) Froude, M. J. and Petley, D. N.: Global fatal landslide occurrence from 2004 to 2016, Nat. Hazards Earth Syst. Sci., 18, 2161–2181, https://doi.org/10.5194/nhess-18-2161-2018, 2018.

6. In lines 19-20, you might want to consider and include the recent works related to landslide susceptibility in the Austrian alps, such as:
i) Lima, P., Steger, S., & Glade, T. (2021). Counteracting flawed landslide data in statistically based landslide susceptibility modelling for very large areas: a national-scale assessment for Austria. Landslides, 18(11), 3531-3546.
ii) Moharrami, M., Naboureh, A., Gudiyangada Nachappa, T., Ghorbanzadeh, O., Guan, X., & Blaschke, T. (2020). National-scale landslide susceptibility mapping in aaustria using fuzzy best-worst multi-criteria decision-making. ISPRS International Journal of Geo-Information, 9(6), 393.
iii) Gudiyangada Nachappa, T., Kienberger, S., Meena, S. R., Hölbling, D., & Blaschke, T. (2020). Comparison and validation of per-pixel and object-based approaches for landslide susceptibility mapping. Geomatics, Natural Hazards and Risk, 11(1), 572-600

7. In lines 63-64, you could try to address the bias associated with downscaling the climate data by using an approach as the one here:
i) Roberts, DR, Wood, WH, Marshall, SJ. Assessments of downscaled climate data with a high-resolution weather station network reveal consistent but predictable bias. Int J Climatol. 2019; 39: 3091– 3103. https://doi.org/10.1002/joc.6005

8. You provide no maps of susceptibility, which could help in identifying the locations of impending occurrences in a spatial way. Could you please add some of the maps (supplementary material as well)?

My recommendation is that you try to answer all reviewers’ questions on a point-to-point basis in a separate document, and modify the original manuscript accordingly, as well. Please also try to reply to the above questions and suggestions. Please be aware that direct replies in the online discussion web pages are not considered as a response to the decision process, as the latter requires direct upload of the said documents in the manuscript submission and revision page of the journal.
After you submit the modified version and the response letter, the revision process will continue directly to the next review round.

Thank you for submitting to NHESS and, again, my apologies for the delay.
Best regards
FC

Hide

AR by Raphael Knevels on behalf of the Authors (19 Nov 2022) Author's response Author's tracked changes Manuscript

ED: Publish as is (06 Dec 2022) by Filippo Catani

Journal article(s) based on this preprint

Raphael Knevels et al.

Supplement

https://doi.org/10.5194/nhess-2022-154-supplement

Data sets

Event-based landslide susceptibility models (Styrian Basin, Austria) Knevels, Raphael; Petschko, Helene; Proske, Herwig; Leopold, Philip; Maraun, Douglas; Brenning, Alexander https://doi.org/10.5281/zenodo.6365228

Raphael Knevels et al.

Viewed

Total article views: 527 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	Supplement	BibTeX	EndNote
369	143	15	527	31	2	4

HTML: 369
PDF: 143
XML: 15
Total: 527
Supplement: 31
BibTeX: 2
EndNote: 4

Views and downloads (calculated since 03 Jun 2022)

Month	HTML	PDF	XML	Total
Jun 2022	179	54	6	239
Jul 2022	52	21	5	78
Aug 2022	47	25	4	76
Sep 2022	18	8	0	26
Oct 2022	25	18	0	43
Nov 2022	21	6	0	27
Dec 2022	15	5	0	20
Jan 2023	12	6	0	18

Cumulative views and downloads (calculated since 03 Jun 2022)

Month	HTML	PDF	XML	Total
Jun 2022	179	54	6	239
Jul 2022	52	21	5	78
Aug 2022	47	25	4	76
Sep 2022	18	8	0	26
Oct 2022	25	18	0	43
Nov 2022	21	6	0	27
Dec 2022	15	5	0	20
Jan 2023	12	6	0	18

Viewed (geographical distribution)

Total article views: 492 (including HTML, PDF, and XML) Thereof 492 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 18 Jan 2023

Download

The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.

Preprint (12884 KB)
Metadata XML

Short summary

In the summer 2009 & 2014, rainfall events occurred in the Styrian basin (Austria), triggering thousands of landslides. Landslide storylines help to show potential future changes under changing environmental conditions. The often neglected uncertainty quantification was the aim of this study. We found uncertainty arising from the landslide model to be of the same order as climate scenario uncertainty. Understanding the dimensions of uncertainty is crucial for allowing informed decision making.


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