the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 29 Jan 2024
Inundation and evacuation of shoreline populations during landslide-triggered tsunami: An integrated numerical and statistical hazard assessment
Abstract. The volcanic island of Stromboli (Southern Tyrrhenian sea, Italy) is renowned for its persistent, periodic, low-intensity explosive activity, whose spectacular manifestations attract tens of thousands of tourists every year. However, sporadic more intense major explosive and effusive eruptions, and paroxysms, pose serious threats to the island. In addition to direct hazards, granular slides of volcanic debris and pyroclastic avalanches, which can rapidly reach the sea potentially generating tsunamis, are often associated with such unpredictable eruptive activity. Due to the very fast propagation of the tsunami around the island, and the consequent short tsunami warning time (ranging from less than a minute to only a few minutes) mitigation efforts and evacuation from the Strombolian coast must be carefully planned. In this paper, we describe a new, GIS-assisted procedure that allows us to combine the outputs of an ensemble of 156 pre-computed landslide-generated tsunami hazard scenarios (with variable landslide volume, position, and density), statistical exposure data (i.e., the number of inhabitants and tourists) and digital geographic information, to obtain a quantitative (scenario-based) risk analysis. By means of the analysis of the road network and coastal morphology, we develop a model with routes and times to reach a safe area from every pixel in the inundated area, and appraisal for the time needed to escape versus the wave arrival time. This allows us to evaluate and quantify the effectiveness of potential risk mitigation by means of evacuation. The creation of an impact score linking the predicted inundation extent and the tsunami warning signals is intended, in the long term, to predict the intensity of future tsunamis, and to adapt evacuation plans accordingly. The model, here applied to Stromboli, is general, and can be applied to other volcanic islands. Evacuating an island hosting several thousand tourists every summer with very little warning time supports the absolute necessity for such mitigation efforts, aimed at informing hazard planners and managers, and all other stakeholders.
-
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
(4893 KB)
-
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(4893 KB) - Metadata XML
- BibTeX
- EndNote
- Final revised paper
Journal article(s) based on this preprint
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2024-221', Anonymous Referee #1, 19 Apr 2024
MS No.: egusphere-2024-221
MS type: Research article
Please find below my evaluation for the manuscript “Inundation and evacuation of shoreline populations during landslide-triggered tsunami: An integrated numerical and statistical hazard assessment” by Emmie M. Bonilauri et al. for consideration in NHESS.
The Authors present a novel study which combines landslide-triggered tsunami modelling at Stromboli volcano (Italy) with evacuation procedures and their potential effectiveness. The combination of the two aspects allows the authors to test escaping routes, timing needed for evacuation, efficacy of evacuation based on the different time of the day and of different seasons. The manuscript is well written and well designed, accompanied by clear figures. A pleasure to read.
I have a very few concerns regarding the structure or the way topics are investigated, which, if addressed, hopefully may result in higher clarity of the text.
- The first point is related to the choice of the case (i.e. wave height) for which results are actually presented. Whereas in methods, five different volumes (5, 8, 14, 21, and 30 × 106 m3) are indicated to be used with the modelling, the case used just for data presentation is considered a medium case in terms of wave amplitude (which varies 0.22 to 48.1 m as actual total range). By means of demonstration, the mid-range example that falls between these two extremes is scenario P6V30CD0.4, a simulation of a submarine landslide (P6) involving 30 × 106 m3 of volcanic material. So, a 30 × 106 m3 is the maximum used volume for the simulations, and the average example in terms of wave amplitude is obtained when this detaches from P6. I thus assume that the maximum waves with a double amplitude (c.ca 40 meters) result from similar volumes but linked to different conditions (height of collapse, maybe P3). Given that this volume is comparable to the 2002 event, is there a correlation with observations? How is this number compared to the Fornaciai et al. (2019) in the case of a submarine slide? Apparently their run-up is smaller (see their figure 4 for run up 20 million cubic m) or figure 3 (for wave amplitude at the beacons).
- The correlation among amplitude at the beacons and potential inundation area is the weakest point to me. First, while a correlation (with the LASSO approach) is clear, I did not get the point of how the different waveforms were treated (with the submarine landslides I would have expected first negative waves). Second, while a relation is visible, it is speculative in a way that all the synthetic data is not tested with at least a single real case, even for what concerns at least the first part of the process (i.e. relation among landslide volume and wave amplitude). Though beyond the scope of the work, this would reduce a variable and, to my knowledge, wave amplitudes (from the beacons) and landslides volumes (from GIS data) are available for the 2019 and 2021 observed PDCs events (see Ripepe and Lacanna 2024 for the beacons, Calvari et al. 2022 for the PDCs volume).
- Although modelling is not the main core of the paper, a more thorough discussion on the physics of the phenomenon would help in better understanding why position 3 is able to generate the largest amplitudes, even larger than higher positions (0, 1, 2). An explanation on the modification is given in the discussion (flow modification during the flow), but to me this is a major point that should be expanded a bit. I recall again the crater collapse events of 2021, and I wonder if these have occurred with the same volumes close to the shoreline, one may have expected higher waves. Is it a matter of initial state of the material (i.e. solid block vs loose material)?
- A minor point is related to the number of population in accommodation obtained from maximum capacity given in the web (Booking.com, Airbnb, Tripadvisor etc.). Maybe it is a stupid question, but I am wondering if the availability given in booking.com is only a part of the total availability for each single structure (they common allow online reservation only for a part of the rooms). Can this results in an underestimation of total holiday touristic capability? I also suspect that the Scari pier population distribution should be higher (in the range of 100-250) given that during summer days the pier is full of persons waiting for the boats during some periods of the day.
- Another curiosity is related to the waves arrival times and the discussion on submarine and subaerial lava flows. While the effect of subaerial lava flows is discussed, I am wondering if the effect of submarine lava flows can play a role and if this has been tested in some way.
- Another point is related to the 11 explanatory variables of the LASSO approach that are only mentioned in the main text. At least they should be briefly summarized (not only in the supplementary material, where however I could roughly get an idea of which kind of variables were tested).
- A final point is related to the crowd effect. Despite this is not an agent-based evacuation approach, I am wondering how the fastest pedestrian evacuation paths from a danger point to a safe point may be affected by a huge number of persons in the small routes at the same time. Can we say that this is the best-case scenario, i.e. single person escape time and that this could largely increase when a lot of people are together in the small routes?
FIGURES
All figures all well designed. I guess there is no need to specify “Created by INGV-Pisa”. I would add some toponyms (i.e. Spiaggia Lunga) used in the text for clarity.
Citation: https://doi.org/10.5194/egusphere-2024-221-RC1 -
AC1: 'Reply on RC1', Emmie Bonilauri, 22 Jul 2024
Dear RC1,
Thank you very much for your feedback and constructive comments!
Please find attached our reply with your feedback in black and our responses in red. The manuscript has been revised with your comments in mind.
Have a great summer!
Sincerely yours,
Emmie Bonilauri (on behalf of all the authors)
-
RC2: 'Comment on egusphere-2024-221', Anonymous Referee #2, 07 May 2024
Reviewer report for ‘Inundation and evacuation of shoreline populations during landslide-triggered tsunami: An integrated numerical and statistical hazard assessment’ by Emmie M. Bonilauri et al.
General Comments
This paper studies a very challenging problem, that of whether and how residents and visitors to Stromboli can be evacuated to safety in time after a landslide on the flanks of the volcano causes a tsunami. On the whole this is a good-quality paper, presenting modelling of landslide-caused tsunamis at the volcano, methods for their detection and quantification, and estimates of which of the areas at risk can potentially be evacuated in time.
The biggest problem I found in reading this paper was with understanding Section 3.4 on Evacuation Capacity, where it was difficult for me to interpret as some of the key metrics were not clearly enough defined, which made it difficult to understand some of the figures and to confirm some of the stated implications. I will put more specific comments below, but I would very much like to see this section improved as it is important for the paper as a whole.
I also feel that some of the assumptions that have gone in to the evacuation modelling are rather optimistic, for example the assumption of no reaction time. The discussion section makes it clear that the authors are aware of many of these issues, but some further elaboration may be useful.
Specific Comments
Landslide modelling
A challenge with landslide modelling is the wide range of possible initial parameters. The paper considers variations in three landslide parameters: the position upslope, the volume, and the density. The authors mention that off-axis lanslide position and rheology are parameters for future study.
However lines 333-334 could perhaps be revised to be more clear that the volume and position were found to be the most important parameters of those considered in this study. As it is not clear from this study alone that all other parameters are ‘of second order’.
The result that it is landslides from the lowest subaerial position (position 3) caused the highest impact is interesting and a bit suprising. The authors assert that this is related to the landslide not having time to deform before reaching the sea surface. This is a plausible explanation to me, although it would be nice to see this demonstrated with a figure or reference. Since the rate of deformation is related to the rheology, it also makes me question whether the effect of rheology is truly ‘second order’ (see previous paragraph).
Analysis of Signals
The use of LASSO penalised linear regression to improve the prediction of inundation is interesting and could be a useful tool in many circumstances. The explanation of the method in the main text and the appendix was quite brief, and the supplied reference Giraud (2021) also rather hard to follow, so any additional explanation of how the method was applied to this problem would be welcome.
My main concern here is how robust the LASSO regression algorithm would be in scenarios that differ in one way or another from those on which it has been trained. For example: landslides that occur off axis, or have different rheologies to those assumed; or how the algorithm would work if there was not a singular landslide but one quickly followed by another (as in 2002 but with a shorter gap) such that there was a superposition of waves.
Another question is how to determine time t=0 instrumentally from the water-level data, and hence when to extract data to use in the algorithm. There is also a potentially longer wait to collect all of the datapoints which may limit the use in some near-source cases. To study all of these things would require another paper, for now I just suggest the authors consider a bit more discussion.
In Appendix A, I was confused for a while by the way that ‘X’ was used both for the inundation of cells and for the tsunami detector time series. Maybe use a different letter for the inundations and clarify how ‘Y’ is calcuated from it?
Evacuation Capacity
I found Section 3.4 quite hard to follow, mostly because the concept of ‘warning time’ (both ‘real’ and ‘needed’ ) was introduced without really clear definitions. With some effort I could establish what I think these are, but I think it would be better to spell this out more explicitly. Similarly it would be good to be fully clear about what the maximum and minumum of these times were calculated.
Based on my interpretation (which could be wrong) I found figure 11 a bit difficult to understand. As by my understanding, the cells that need the most warning time (hollow bars) are unrelated (or even inversely related) to those which need the most warning time (solid bars), yet because they appear next to each other in the figures I initally thought that they would be the same locations.
Pedestrian Evacuation Model
The evacuation model used is relatively simple, and the most important thing for this paper is to make sure that the simplifications and approximations are well documented and explained. Simplifications I’m aware of (some of which are mentioned in the text) include:
- Assumption of no reaction time
- Assumption of no warning dissemination time
- Assumtion of no variation in walking speed
- Assumption of no congestion
While some of these could be approximated by extending the current model, ultimately this problem is really calling out for a full agent-based modelling approach (though I’m not suggesting that the authors need to do that for this paper).
Summary
Although I have made some critical comments above, in all I find this to be a valuable and well-written multi-disciplinary paper, certainly worthy of publication after some minor adjustments.
Citation: https://doi.org/10.5194/egusphere-2024-221-RC2 -
AC2: 'Reply on RC2', Emmie Bonilauri, 22 Jul 2024
Dear RC2,
Thank you very much for your feedback and constructive comments!
Please find attached our reply with your feedback in black and our responses in red. The manuscript has been revised with your comments in mind.
Have a great summer!
Sincerely yours,
Emmie Bonilauri (on behalf of all the authors)
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2024-221', Anonymous Referee #1, 19 Apr 2024
MS No.: egusphere-2024-221
MS type: Research article
Please find below my evaluation for the manuscript “Inundation and evacuation of shoreline populations during landslide-triggered tsunami: An integrated numerical and statistical hazard assessment” by Emmie M. Bonilauri et al. for consideration in NHESS.
The Authors present a novel study which combines landslide-triggered tsunami modelling at Stromboli volcano (Italy) with evacuation procedures and their potential effectiveness. The combination of the two aspects allows the authors to test escaping routes, timing needed for evacuation, efficacy of evacuation based on the different time of the day and of different seasons. The manuscript is well written and well designed, accompanied by clear figures. A pleasure to read.
I have a very few concerns regarding the structure or the way topics are investigated, which, if addressed, hopefully may result in higher clarity of the text.
- The first point is related to the choice of the case (i.e. wave height) for which results are actually presented. Whereas in methods, five different volumes (5, 8, 14, 21, and 30 × 106 m3) are indicated to be used with the modelling, the case used just for data presentation is considered a medium case in terms of wave amplitude (which varies 0.22 to 48.1 m as actual total range). By means of demonstration, the mid-range example that falls between these two extremes is scenario P6V30CD0.4, a simulation of a submarine landslide (P6) involving 30 × 106 m3 of volcanic material. So, a 30 × 106 m3 is the maximum used volume for the simulations, and the average example in terms of wave amplitude is obtained when this detaches from P6. I thus assume that the maximum waves with a double amplitude (c.ca 40 meters) result from similar volumes but linked to different conditions (height of collapse, maybe P3). Given that this volume is comparable to the 2002 event, is there a correlation with observations? How is this number compared to the Fornaciai et al. (2019) in the case of a submarine slide? Apparently their run-up is smaller (see their figure 4 for run up 20 million cubic m) or figure 3 (for wave amplitude at the beacons).
- The correlation among amplitude at the beacons and potential inundation area is the weakest point to me. First, while a correlation (with the LASSO approach) is clear, I did not get the point of how the different waveforms were treated (with the submarine landslides I would have expected first negative waves). Second, while a relation is visible, it is speculative in a way that all the synthetic data is not tested with at least a single real case, even for what concerns at least the first part of the process (i.e. relation among landslide volume and wave amplitude). Though beyond the scope of the work, this would reduce a variable and, to my knowledge, wave amplitudes (from the beacons) and landslides volumes (from GIS data) are available for the 2019 and 2021 observed PDCs events (see Ripepe and Lacanna 2024 for the beacons, Calvari et al. 2022 for the PDCs volume).
- Although modelling is not the main core of the paper, a more thorough discussion on the physics of the phenomenon would help in better understanding why position 3 is able to generate the largest amplitudes, even larger than higher positions (0, 1, 2). An explanation on the modification is given in the discussion (flow modification during the flow), but to me this is a major point that should be expanded a bit. I recall again the crater collapse events of 2021, and I wonder if these have occurred with the same volumes close to the shoreline, one may have expected higher waves. Is it a matter of initial state of the material (i.e. solid block vs loose material)?
- A minor point is related to the number of population in accommodation obtained from maximum capacity given in the web (Booking.com, Airbnb, Tripadvisor etc.). Maybe it is a stupid question, but I am wondering if the availability given in booking.com is only a part of the total availability for each single structure (they common allow online reservation only for a part of the rooms). Can this results in an underestimation of total holiday touristic capability? I also suspect that the Scari pier population distribution should be higher (in the range of 100-250) given that during summer days the pier is full of persons waiting for the boats during some periods of the day.
- Another curiosity is related to the waves arrival times and the discussion on submarine and subaerial lava flows. While the effect of subaerial lava flows is discussed, I am wondering if the effect of submarine lava flows can play a role and if this has been tested in some way.
- Another point is related to the 11 explanatory variables of the LASSO approach that are only mentioned in the main text. At least they should be briefly summarized (not only in the supplementary material, where however I could roughly get an idea of which kind of variables were tested).
- A final point is related to the crowd effect. Despite this is not an agent-based evacuation approach, I am wondering how the fastest pedestrian evacuation paths from a danger point to a safe point may be affected by a huge number of persons in the small routes at the same time. Can we say that this is the best-case scenario, i.e. single person escape time and that this could largely increase when a lot of people are together in the small routes?
FIGURES
All figures all well designed. I guess there is no need to specify “Created by INGV-Pisa”. I would add some toponyms (i.e. Spiaggia Lunga) used in the text for clarity.
Citation: https://doi.org/10.5194/egusphere-2024-221-RC1 -
AC1: 'Reply on RC1', Emmie Bonilauri, 22 Jul 2024
Dear RC1,
Thank you very much for your feedback and constructive comments!
Please find attached our reply with your feedback in black and our responses in red. The manuscript has been revised with your comments in mind.
Have a great summer!
Sincerely yours,
Emmie Bonilauri (on behalf of all the authors)
-
RC2: 'Comment on egusphere-2024-221', Anonymous Referee #2, 07 May 2024
Reviewer report for ‘Inundation and evacuation of shoreline populations during landslide-triggered tsunami: An integrated numerical and statistical hazard assessment’ by Emmie M. Bonilauri et al.
General Comments
This paper studies a very challenging problem, that of whether and how residents and visitors to Stromboli can be evacuated to safety in time after a landslide on the flanks of the volcano causes a tsunami. On the whole this is a good-quality paper, presenting modelling of landslide-caused tsunamis at the volcano, methods for their detection and quantification, and estimates of which of the areas at risk can potentially be evacuated in time.
The biggest problem I found in reading this paper was with understanding Section 3.4 on Evacuation Capacity, where it was difficult for me to interpret as some of the key metrics were not clearly enough defined, which made it difficult to understand some of the figures and to confirm some of the stated implications. I will put more specific comments below, but I would very much like to see this section improved as it is important for the paper as a whole.
I also feel that some of the assumptions that have gone in to the evacuation modelling are rather optimistic, for example the assumption of no reaction time. The discussion section makes it clear that the authors are aware of many of these issues, but some further elaboration may be useful.
Specific Comments
Landslide modelling
A challenge with landslide modelling is the wide range of possible initial parameters. The paper considers variations in three landslide parameters: the position upslope, the volume, and the density. The authors mention that off-axis lanslide position and rheology are parameters for future study.
However lines 333-334 could perhaps be revised to be more clear that the volume and position were found to be the most important parameters of those considered in this study. As it is not clear from this study alone that all other parameters are ‘of second order’.
The result that it is landslides from the lowest subaerial position (position 3) caused the highest impact is interesting and a bit suprising. The authors assert that this is related to the landslide not having time to deform before reaching the sea surface. This is a plausible explanation to me, although it would be nice to see this demonstrated with a figure or reference. Since the rate of deformation is related to the rheology, it also makes me question whether the effect of rheology is truly ‘second order’ (see previous paragraph).
Analysis of Signals
The use of LASSO penalised linear regression to improve the prediction of inundation is interesting and could be a useful tool in many circumstances. The explanation of the method in the main text and the appendix was quite brief, and the supplied reference Giraud (2021) also rather hard to follow, so any additional explanation of how the method was applied to this problem would be welcome.
My main concern here is how robust the LASSO regression algorithm would be in scenarios that differ in one way or another from those on which it has been trained. For example: landslides that occur off axis, or have different rheologies to those assumed; or how the algorithm would work if there was not a singular landslide but one quickly followed by another (as in 2002 but with a shorter gap) such that there was a superposition of waves.
Another question is how to determine time t=0 instrumentally from the water-level data, and hence when to extract data to use in the algorithm. There is also a potentially longer wait to collect all of the datapoints which may limit the use in some near-source cases. To study all of these things would require another paper, for now I just suggest the authors consider a bit more discussion.
In Appendix A, I was confused for a while by the way that ‘X’ was used both for the inundation of cells and for the tsunami detector time series. Maybe use a different letter for the inundations and clarify how ‘Y’ is calcuated from it?
Evacuation Capacity
I found Section 3.4 quite hard to follow, mostly because the concept of ‘warning time’ (both ‘real’ and ‘needed’ ) was introduced without really clear definitions. With some effort I could establish what I think these are, but I think it would be better to spell this out more explicitly. Similarly it would be good to be fully clear about what the maximum and minumum of these times were calculated.
Based on my interpretation (which could be wrong) I found figure 11 a bit difficult to understand. As by my understanding, the cells that need the most warning time (hollow bars) are unrelated (or even inversely related) to those which need the most warning time (solid bars), yet because they appear next to each other in the figures I initally thought that they would be the same locations.
Pedestrian Evacuation Model
The evacuation model used is relatively simple, and the most important thing for this paper is to make sure that the simplifications and approximations are well documented and explained. Simplifications I’m aware of (some of which are mentioned in the text) include:
- Assumption of no reaction time
- Assumption of no warning dissemination time
- Assumtion of no variation in walking speed
- Assumption of no congestion
While some of these could be approximated by extending the current model, ultimately this problem is really calling out for a full agent-based modelling approach (though I’m not suggesting that the authors need to do that for this paper).
Summary
Although I have made some critical comments above, in all I find this to be a valuable and well-written multi-disciplinary paper, certainly worthy of publication after some minor adjustments.
Citation: https://doi.org/10.5194/egusphere-2024-221-RC2 -
AC2: 'Reply on RC2', Emmie Bonilauri, 22 Jul 2024
Dear RC2,
Thank you very much for your feedback and constructive comments!
Please find attached our reply with your feedback in black and our responses in red. The manuscript has been revised with your comments in mind.
Have a great summer!
Sincerely yours,
Emmie Bonilauri (on behalf of all the authors)
Peer review completion
Journal article(s) based on this preprint
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 339 | 165 | 34 | 538 | 18 | 21 |
- HTML: 339
- PDF: 165
- XML: 34
- Total: 538
- BibTeX: 18
- EndNote: 21
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Emmie M. Bonilauri
Catherine Aaron
Matteo Cerminara
Raphaël Paris
Tomaso Esposti Ongaro
Benedetta Calusi
Domenico Mangione
Andrew J. L. Harris
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(4893 KB) - Metadata XML