Rapid landslide risk zoning toward multi-slope units of the Neikuihui tribe for preliminary disaster management

Chung, Chih-Chung; Li, Zih-Yi

doi:https://doi.org/10.5194/nhess-22-1777-2022

Articles | Volume 22, issue 6

https://doi.org/10.5194/nhess-22-1777-2022

© Author(s) 2022. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/nhess-22-1777-2022

© Author(s) 2022. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 22, issue 6

Research article

|

02 Jun 2022

Research article |

| 02 Jun 2022

Rapid landslide risk zoning toward multi-slope units of the Neikuihui tribe for preliminary disaster management

Chih-Chung Chung and Zih-Yi Li

Download

Final revised paper (published on 02 Jun 2022)
Preprint (discussion started on 21 Jan 2022)

Interactive discussion

Status: closed

RC1:
'Comment on nhess-2021-409', Anonymous Referee #1, 07 Mar 2022
General comments

The manuscript describes a rapid landslide risk zoning approach targeted at supporting preliminary disaster management in Taiwan. Overall, this is clearly an application-oriented work, and scientific novelty is somewhat limited. However, since the geological disasters do occur quite frequently in this area, entailing that also case-study-type studies aiming at effective disaster management is of importance, I think that this topic is generally suitable to be published in NHESS.

Please find some specific and technical comments below.

Specific comments

In general, I would term this indicator-based approach as a semi-quantitative method rather than a fully quantitative one throughout the manuscript. Indicators and weights are not set on an fully quantitative basis, but rather by using expert knowledge. Ssusceptibility models based on statistical learning or vulnerability curves based on exact measurements of damage and process intensity would be fully quantitative. Many concepts used here imply equal weights or nearly linear relationships on ordinal scales (e.g. "low - medium - high"). This is not a criticism of the method per se,

Abstract: The information about the "No.11 slope unit" is too specific to be mentioned in the abstract without further context. This particular slope unit will be unknown to a vast majority of readers.

Figure 2: It would be helpful to include the variable displayed in each facet, e.g. in the legend for quicker understanding.

l. 120: "Among the processes of delimited slope units, grid-cells units and slope units are commonly adapted (Reichenbach et al., 2018)." Personally, I find the beginning of this section to be a bit sudden. The concept of slope units should be briefly described here, including methods for delineating them. This would aid the reader, especially those who are not familiar with the concept.

Table 1 mixes two types of information. Suggest to separate the lower part of the table to keep the datasets tidy; otherwise it might be confusing why "Classification" would correspond to 10~14; "Occurrence index" would correspond to 7~9, and "Grades" would correspond to "4~6". This is obviously not meant here.

The same applies to Table 2, Table 4 and Table 5.

3.2.1 Susceptibility analysis: The exact method used is unclear to me. l. 160 describes that "Susceptibility of the landslide was evaluated by weighting factors of the slope degree and distance of river channel, lithology, and dip slope", which might indicate some sort of (logistic?) regression model. However, this assumption somewhat contradicted by Table 1, which looks more like an indicator-based approach (which would be more in line with the rest of the described approach)? Please clarify.

l. 191: "Among the elements of Risk zoning, an exposed object is significant." Please clarify/rephrase this sentence.

3.4 Vulnerability analysis: "Vulnerability analysis in this study initially represents the degree of damage of the exposed object by considering the relative position from the landslide, runout, and deposition area. The closer the distance, the greater the damage and the higher vulnerability." I think that the assumption that vulnerability (denoted as damage in this context) dependes on the distance to the landslide is too simplistic. Many factors do determine physical vulnerability (c.f. https://doi.org/10.1016/j.jhydrol.2022.127501), and even deposition height might not be a fully adequate indicator of physical vulnerability, let alone distance. While I understand why this was done due to practical reasons (lack of better information), I think this should be openly discussed as a limitation.

Figure 12: I suggest to use a "prettier" functional relationship between the two y-axis, e.g. y₂ = 10*y₁ (i.e. scale for total accumulated rainfall is [0, 500]). Otherwise, the horizontal lines are off, and the labels for "Total accumulated rainfall" are floating around without corresponding horizontal grid line.

Discussion: I think that the pros and cons of using an indicator-based approach like this one could be discussed in more detail, including the setup of the single elements (tables).

Discussion: Validation is performed basically with one event on Slope No. 11 during a Typhoon event. This is ok, but n=1 is more anecdotal evidence rather than a convincing sample size for accurate validation of the approach. Some sort of goodness-of-fit metric would be good. Since this can probably not be achieved within the scope of this study, accuracy could be discussed based on plausibility and local expert knowledge, and validation procedures could be outlined.

Data availability: Please consider depositing the data to a more persistent repository, e.g. the PANGAEA data repository or zenodo.

Technical corrections:

Figure 2 / Figure 8: "Slpoe" (probably incorrect layer name)
Citation: https://doi.org/10.5194/nhess-2021-409-RC1
- AC1:
  'Reply on RC1', Chih-Chung Chung, 14 Mar 2022
  Authors appreciate your valuable comments. Following please find the responses in detail:
  Thank you for suggestion, and authors agree the term of “indicator-based”. As mentioned in the manuscript, the evaluation indicators in Table 1 are based on the previous logistic result of 2523 collapses in the southern Taiwan, leading to the five major influence factors: lithology, slope, elevation, structural distance, and channel distance. The authors then modified the results by using expert knowledge to establish a susceptibility grades fitted to the rapid landslide risk zoning toward a tribe area.
  
  Thank you for suggestion. The Abstract content is modified as: “The rapid risk zoning analysis of multi-slope units shows that there is a sloping unit around the Neikuihui tribe with a high level of landslide risk, and this sloping unit did suffer from landslide disasters in the 2016 typhoon event.”
  
  Thank you for suggestion, legends for c and d in Fig.2 have been added for quicker understanding.
  
  Thank you for suggestion. The content is modified with new citations as: “Quantitative geomorphological and environmental analysis requires the adoption of well–defined spatial domains as basic mapping units. The spatial domains provide local boundaries to aggregate environmental and morphometric variables and perform calculations. Thus, they identify the spatial scale of the analysis (Alvioli et al.,2020). Among the spatial domain processes of delimited slope units, grid-cells and slope units are commonly adapted (Reichenbach et al., 2018). Grid cells, typically aligned with a digital elevation model, are the standard mapping unit choice (Alvioli et al.,2020). Despite their popularity and operational advantages, grid-cells have apparent drawbacks for susceptibility modelling (Guzzetti et al., 1999). First, there is no physical relationship between a grid-cell, or a group of grid-cells, and landslides. Landslides result from slope processes acting at different spatial and temporal scales that result in geomorphological forms of very different shapes and sizes (Malamud et al., 2004; Guzzetti et al., 2012) that are difficult to capture by grid-cells accurately. An alternative to grid-cells and unique condition units are slope units, which are hydrological terrain units bounded by drainage and divide lines (Carrara, 1983; Carrara et al., 1991, Carrara et al., 1995; Guzzetti et al., 1999), and corresponds to what a geomorphologist would recognize as a slope. The size of the slope units can be tailored to the type and size of the landslides, allowing for using the geo-environmental information best suited for the specific landslide type. (Carrara et al., 1991; Alvioli et al., 2016).”
  
  Thank you for suggestion, we have modified Table 1 accordingly.
  
  Thank you for suggestion, we have modified Tables accordingly.
  
  Thank you for suggestion. Please refer to the response of the Comment 1.
  
  Thank you for suggestion. The sentence is modified as “It is important to calculate how many household, traffic and public Utilities are exposed in risk zoning.”
  
  Thank you for suggestion. Authors agree that vulnerability depending on the distance to the landslide is simplistic. However, the primary purpose of this study is to establish a raply risk assessment framework for quickly interpreting the collapse of multi-slope units in a tribe area. We have added a description of study limitations to the discussion as follows: “In order to quickly assess the slope collapse risk of an area, especially in the vulnerability analysis, it may be simplistic but efficient to judge the vulnerability score by considering the possible impact area of the slope and the distance from the household / the main traffic road. The main reason is that geological data is limited to comprehensively analyzing the slope collapse risk. However, there are still households in this area, and the economic conditions are disadvantaged. When the survey resources are limited, the administration can easily and quickly remind people in higher-risk areas to relocate to a safe place according to the developed methodology in this study.”
  
  The data interval 0-500 for total accumulated rainfall in Fig.12 has been modified as suggested.
  
  Based on the above suggestions, we have supplemented the discussion.
  
  Thank you for suggestion. Since the purpose of this study is to quickly analyze the slope collapse risk in a region through indicators, it is suggested that after the follow-up assessment of the slope collapse risk in a certain region, the analysis results should be verified by interviewing local residents, experts and scholars.
  
  Thank you for your suggestion, we will look for a database suitable for storage.
  
  Thank you for suggestion, the wrong word has been edited.
  
  Citation: https://doi.org/10.5194/nhess-2021-409-AC1
RC2:
'Comment on nhess-2021-409', Anonymous Referee #2, 28 Mar 2022
General comments:

This article presents a rapid risk zoning method together with an example in a small village in Taiwan. This qualitative approach to risk analysis can be a relevant approach to work with scarce data and this study could therefore be an interesting contribution to NHESS. However, the method is not sufficiently well described and does not seem to be appropriate. In addition, the text is not easy to read because of the language quality and the quality of the illustrations is insufficient. In my opinion, an extensive work is needed before this article could be published.

My main concern is the evaluation of the hazard. The hazard is evaluated here by combining a susceptibility and two activity criteria. The susceptibility analysis is based a method developed for deep-seated landslides (unfortunately only published in Chinese) defined as having a surface of at least 10 hectares but is applied here on a region that is around 20 hectares according to the text and about 10 hectares according to my measurements on Figure 1. In addition, the only example of landslide that is given is a landslide with a depth of 30-60 cm, so I doubt that the method developed for deep-seated landslides can be used here. It would be necessary to better describe the type of landslide that is considered to make sure that the method is appropriate. When it comes to the activity criteria, I do not really understand:

why it is divided in two when there is a lot of redundancy between the two grades

why the relation between the bedding and the slope is considered in the activity in addition to dip-slope being a criterion in the susceptibility analysis

what cliff activity means. I assume that it refers to the scarp, but I am a bit confuse here about the landslide type that is considered. I assume those criteria were developed to be applied to evaluate an acceleration of slow-moving landslides.

I would like to see some examples of how those criteria are evaluated on the study area, especially since this analysis is done on aerial photos, while only low-quality images are provided in the article.

Finally, the method is presented as “initially validated” by showing that a landslide occurred where the risk has been identified as the highest. The reason for the risk to be the highest in that slope unit is mainly that it is one of the few where there are elements at risk, although the hazard is relatively high as well. The susceptibility is medium and the activity high, but does the activity include the landslide that is described? Anyway, it is difficult to validate such a method and another aspect that I would like to be discussed is the applicability of the method on a more regional scale. It is thought to be fast, but I suspect that it is relatively time-consuming, and I would therefore like to read more about the context in which it could be used and how it would perform compared to a more regional analysis.

Specific comments:

I am not a native speaker, but I wonder if the word “Tribe” is really what you mean?

l 33-67: there is a long list of articles, but I cannot really understand why they are listed here, how they relate to your work and what knowledge gap your study intends to fill

l 89-93: the description of the area is quite confusing. For example, the last sentence states that “most residents have moved north…” but where were they coming from? And does Neikuihui belong to the Kuihui village?

l 102 (and figure 2): you present two different formation, but the rock type is mentioned only for the Tatongshan/Tatungshan formation. Further details on the Aoti formation would be useful

Figure 1: the map to the left lacks a scale and the north arrow does not corresponds to the image on the right since its top points to the south. Also, the quality of the image on the right is not sufficient to understand what we are looking at. A map with buildings and roads would help the reader. I am also wondering if it is a perspective view.

Figure 2: I assume this one is a perspective view. I am disappointed by the quality since you apparently have a 1m-DEM. Could you improve the quality and provide a hillshade? Also I don’t think that the CS-map is very helpful

Figure 4: The sizes of the slope units are very different. I wonder how it impacts the analysis.

Table 4: I don’t understand the “raw grades” and “adjusted grades” and why for example the class with “less than one household” (isn’t that 0?) gets a raw grade of 3 and an adjusted grade of 9 when it should be 0 if there is no household. “Households 3 to 4” and “Households 1 to 2” should be renamed “3 to 4 households” and “1 to 2 households” respectively. “More than five households” should be renamed “5 or more households”. The summation og grades gives a “low” level from 1 to 11, but if “less than one household” gives 9 points, then it can’t be below 9. I think the exposure level should be 0 if there is nothing otherwise the method gives an “extremely low risk” when there is actually no risk

Vulnerability analysis: the vulnerability analysis is based on the distance to the landslide, but how are the landslide source, run-out and deposition areas defined?

l 217: The citation refers to an article describing a vulnerability index that is combining several criteria including the physical properties of the buildings. I do not see the similitude with the method you are using

Table 7: Slope unit 4 is considered a dip-slope, but only a tiny portion of it is inside the dip-slope polygon. I wonder if the polygon of dip-slope has been drawn at an appropriate scale for this analysis. Otherwise, is the slope unit well-defined? Or is the rest of the slope unit not a dip-slope because of a geological folding?

Table 12: There are 9 households in unit 5, but 1-2 according to table 11… which one is wrong?

l 341: what is the residential house No. 8. Do you mean a house in slope unit No. 8?

l 395: JSPJ is mentioned in the author contributions, but is not a co-author
Citation: https://doi.org/10.5194/nhess-2021-409-RC2
- AC2:
  'Reply on RC2', Chih-Chung Chung, 05 Apr 2022
  The authors appreciate the advice. The manuscript has been revised according to suggestions. The responses to general comments are detailed as follows.:
  The authors appreciate the suggestion. Susceptibility is typically applied for the landslide risk assessment of large-scale geological conditions accompanied by common environmental factors, such as slope degree and lithology. To rapidly assess the landslide risk of a tribe region, this study refers to the susceptibility findings of deep-seated landslide inventory and carefully includes Activity analysis, especially for a small-scale slope unit. The revised Activity is based on the Activity Area Ratio (AAR) principle and the categories of dip sliding and colluvium indexes from the previous experience. Hence, the proposed Activity aims to examine the evolutions of slope units through DEM and aerial photos of different periods. Regarding this, Activity 1 is modified to measure the activity level of the dip sliding along a slope unit, while Table 2 examines the activity level of the colluvium layer on the surface of a slope unit. The relation between the bedding and the slope considered in the Activity 1 is further applied to comprehensively appraise the rock's exposed condition through the high-resolution DEM and aerial photo interpretation.
  
  Thank you for the correction, the "cliff" should be "scarp", which has been corrected, and the method is suitable for slow-moving slopes.
  
  Thank you for the suggestion. The revised manuscript will add the activity analysis process of slope unit No.11 as an example in the text to explain in detail.
  
  Thank you for the suggestion.
  
  The landslide event in 2016 is not included in the risk analysis.
  
  The proposed method intends to provide a rapid analysis according to the scoring indicators in the tables after obtaining the initial geological, DEM, and aerial photo data. Indeed, identifying scarp in the activity analysis will take a little time because it is necessary to compare the evolution of possible features at different times for the exact location.
  
  According to the official observations in the recent decade, the Neikuihui tribe is the only one with a landslide for validation. Although only one regional analysis is revealed, the authors believe that the proposed rapid risk zoning process is ready to be applied in the next phase at other hillside tribes.
  
  The responses to the specific comments:
  Thank you for the suggestion. The word “Tribe” is what we mean.
  
  Thank you for the suggestion. The paragraph collects significant risk analysis findings and the corresponding applications to reveal the basics of the risk analysis research method. For example, Varnes et al. provided the risk assessment principle in 1984 as Risk = Hazard × Exposure × Vulnerability; Corominas and Mavrouli (2011) stated a completed deep-seated landslide risk assessment must include Susceptibility, Hazard, and Vulnerability. The authors believe these are important and relevant documents closely related to this study.
  
  Thank you for the suggestion. The context has been revised: They are the inhabitants who lived here before, named aborigines in Taiwan. Since the Neikuihui tribe has only one external road and frequent rockfall disasters, most residents have moved north to Kuihui Village, and only about 15 households are left in the Neikuihui tribe.
  
  Thanks for your suggestion. The text has been edited: Figure 2e is a 1:25,000 geological map of the Central Geological Survey (2020). The strata include the Tatongshan layer and the Aoti bottom layer, respectively, of which the Tatongshan layer is composed of black hard shale and siltstone interbedded, often forming steep slopes along the river bed. The bottom layer of Aoti is sandstone with a coal seam.
  
  Thanks for the suggestion. Fig. 1 adds a new scale to the main map of Taiwan and provides a compass and road lines corresponding to the Neikuihui tribe area. Since the perspective view is visually complex, we suggest keeping the 3D maps updated with high quality for better visualization.
  
  Thank you for the suggestion. Fig. 2 will be re-output with the highest quality, and the hillshade map is also provided.
  
  Thank you for the suggestion. When delimiting sloping land units, the authors utilized slope aspect, slope, ridgeline, river valley line, and geology for analysis, leading to the different area sizes of the slope units. The authors spent much time examining the activity of the large slope unit through the aerial photo and DEM, as mentioned in the previous response.
  
  Your statement is indeed correct. The authors divide the total score in analysis into three levels (low, medium, and high) or five levels (extremely low, low, medium, high, and extremely high) according to the standard practices on risk management as proposed by previous kinds of literature.
  
  Thank you for the suggestion. The authors add the vulnerability analysis process of slope unit No.11 as an example in the text to explain in detail.
  
  Thank you for the correction. Previously, we mainly quoted the vulnerability index formula mentioned in the article of Papathoma-Köhle et al. (2019). Based on your suggestion, we replaced it with the literature of Papathoma-Köhle et al. (2017), which is closer to the research content.
  
  Thank you for the suggestion. The range and location of the dip slope at slope unit 4 are from the 1/25,000-scale official distribution maps of geologically sensitive areas provided by the Geological Survey of the Ministry of Economic Affairs of Taiwan. Despite the range and location of the dip slope being rough as mapping to the slope unit 4, this study preliminarily assumes the slope unit 4 as a dip slope.
  
  Thank you for the correction. The authors have corrected the number of households in No.5. in Table 12, and changed the total score and figures accordingly.
  
  Thank you for the correction. We removed the word No.8.
  
  Thank you for the suggestion. This part is wrongly planted and has been modified to CCC.
  
  Citation: https://doi.org/10.5194/nhess-2021-409-AC2

Peer review completion

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

ED: Reconsider after major revisions (further review by editor and referees) (05 Apr 2022) by Yves Bühler

AR by Chih-Chung Chung on behalf of the Authors (19 Apr 2022) Author's response Author's tracked changes Manuscript

ED: Referee Nomination & Report Request started (21 Apr 2022) by Yves Bühler

RR by Anonymous Referee #1 (29 Apr 2022)

Suggestions for revision or reasons for rejection

Thanks to the authors for answering all comments raised in the first round of reviews in an adequate manner. Please note that line numbers refer to the version containing authors* tracked changes.

I do have some additional comments, which need further clarification. Many of them are of a merely technical nature. However, I do advise to revise parts of the discussion section to be more precise and explicit about benefits and limitations of the method.

- l.16: "assess" or "integrate" instead of "apply the susceptibility, ..."
- l.94: To whom does "They are" refer to in this sentence? This is unclear, please rephrase.
- l.98: I suggest to use "basis" instead of "basement"
- l.100: I'm wondering if it is sufficiently clear to a wider audience, what a "CS map" is. I suggest add a half sentence explaining the actual type of output in this type of topographic map. See also the next point.
- Figure 2 (c) and (d): A continuous scale legend is somewhat unorthodox for discrete features ranging from "Ridge" to "Valley". Using Ridge and Valley is fine, but I suggest to also add the actually plotted variables (i.e. their ranges) there. Also, if Figs. (c) and (d) are included for comparative purposes, wouldn't it be better to use the same color palette?
- l.146: "Martinello et al. (2020) brought the Imera Settentrionale watershed in northern Sicily, Italy, as the research scope ..." is a somewhat strange sentence. Why not just "Based on an analysis of the Imera Settentrionale watershed in northern Sicily, Italy, Martinello et al. (2020) found slope units to be superior for representing landslide susceptibility as, as opposed to (I suppose: grid cells) due to (reason here)".
- l.186: lowercase s in susceptibility
- l.188: I would replace "The statistical one uses numerous landslides to analyze the corresponding unstable slopes." with "Statistical approaches require representative landslide inventories as training data set which are used to characterize the corresponding unstable slopes."
- l. 188f: "Factors such as topographical and geological conditions were marked with weights and rankings statistically, leading to objective results in practice." This sentence is unclear. Is this a general description following up on the description of statistical methods? Than the tense should be present tense. Or was this done in this study? If yes, how?
- l. 191: lowercase s in susceptibility
- l. 191: please state "using logistic regression" explicitly, the term "logistically" is incorrect in this context
- l.192: Does "adjacent conditions" mean that the features "distance to rivers" and "distance to faults" were considered as independent variables in the regression model?
- l. 194: "Afterward, regional statistical results were applied to calculate the occurrence index normalized from 1 to 2." I cannot really follow what was done here. Were the prediction results of the logistic regression (i.e. values in the interval [0,1]) transformed into the range [1,2]? If so, why is this a normalization? Normalization usually refers to a transformation into a standard normal distribution with mean zero and sd = 1.
- Upon reading this previous paragraph (Beginning of Section 3.2.1) a couple of times I think that I am still slightly confused here, since to me there is no clear differentiation between the susceptibility map (based on the logistic regression model) and the indicator-based approach. I suggest to streamline this section a bit to avoid misunderstandings here.
- 205ff: I would write activity with a lowercase a.
- Figure 8: This image seems to include a NIR band rather than being a standard VIS orthophoto?
- Figure 11: Is there a slope unit in light green ("extremely low")? (This is hard to see in the version provided for review). If not, this category can be removed from the legend for clarity.
- 461: what is "semi-experienced statistical prediction"?
- 464: "the quantitative method is introduced" this is unclear. where or to whom is it introduced, and why is the introduction relevant?
- 466: "A qualitative description does not need to quantify each factor, which is described hierarchically". What is described hierarchically? Please be more explicit.
- "Therefore, the initial risk assessment is more suitable with qualitative descriptions." I would rather say it is easier to accomplish, not necessarily more suitable.
- 470: "The Risk zoning framework designed in this study includes the Activity assessment when grading the Hazard, which can make up for the lack of time change in the Susceptibility assessment and represent the actual site activities. However, it requires little time to analyze visually." (1) What do you mean with "ack of time change in the Susceptibility assessment"? (2) What does "it requires little time to analyse visually?" mean? It requires little time would mean that this is very fast, but I am under the impression that this is rather a drawback, and that some effort is required for this analysis?
- l. 479: "the economic conditions are disadvantaged" does not seem to be proper English to me.
- Again, dropbox is not the optimal place to store research data. Please deposit the data on a proper data repository.

Hide

ED: Publish subject to minor revisions (review by editor) (29 Apr 2022) by Yves Bühler

AR by Chih-Chung Chung on behalf of the Authors (04 May 2022) Author's response Author's tracked changes Manuscript

ED: Publish subject to technical corrections (05 May 2022) by Yves Bühler

AR by Chih-Chung Chung on behalf of the Authors (06 May 2022) Author's response Manuscript

Short summary

The Neikuihui tribe in northern Taiwan faces landslides during rainfall events. Since the government needs to respond with disaster management for the most at-risk tribes, this study develops rapid risk zoning, which involves the susceptibility, activity, exposure, and vulnerability of each slope unit of the area. Results reveal that one of the slope units of the Neikuihui tribal area has a higher risk and did suffer a landslide during the typhoon in 2016.