Balancing Hazard Exposure and Walking Distance in Evacuation Route Planning during Earthquake Disasters

Round 1

Reviewer 1 Report

I am honored to be able to review your manuscript. The purpose of the study and the analysis procedure was very clear and easy to understand.

However, I think that the following points need some modification.

1. The significance of conducting this research in Gyeongju, South Korea is weak. Certainly, the 2016 Gyeongju earthquake caused enormous damage. Also, the buildings in Gyeongju are old. However, I do not understand how much it is for the whole of South Korea. It is necessary to clarify the positioning of Gyeongju. In other words, it will be necessary to theoretically strengthen the reason for selecting the study area. (Because most readers of this journal are not familiar with Korean geography)

2. The strength of this research is that the research results are generalized and can be applied to other areas. However, the discussion section should mention not only generality but also peculiarities. This research is only a simulation result in Gyeongju. The regional characteristics of Gyeongju should be reflected in the simulation. Using Figure 8, you can see a slight discussion of regional characteristics. But it is only mentioned the spatial pattern of population distribution. I'm not familiar with Gyeongju, but I guess it's probably historically famous and has many tourists. In other words, it is possible to interpret the simulation results from the socio-cultural aspects of the target area and discover the peculiarities. Finding generality and peculiarity, and comparing them will make the discussion more meaningful.

3. Although related to 2, the data of the mobile phone is valuable and I think it is quite useful. Can't you use the attributes of that data to discuss behavioral characteristics of evacuees? The weakness of this study is that it does not distinguish between those who need special support during evacuation and those who are healthy. It may be necessary to review studies targeting earthquake-prone countries (such as Japan and Indonesia) and reconsider the originality of this study.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

Although most results are clearly stated in this short paper, I do have some questions and remarks. The most important ones are:

• Some statements are missing references in the introduction and authors should elaborate a bit more on the previous studies in this field. After reading, the general idea of the paper is clear, but not the state of the art.
• My biggest question is whether the proposed method is significantly better than the other ones. We can see that distance is slightly larger than the first two and hazard is slightly smaller. Is this improvement large enough to consider this method? For example, does it require more computational resources than the first two methods, hence, limiting the real-time calculations? Does the little gain in hazard outweigh the rise in total distance (during which people are exposed to other dangers)? Will people be prepared to walk a longer route, when they know a closer shelter? I would like to see those questions at least considered in the paper and a quantification or thorough analysis on why the proposed method is the best. Are results statistically significant?
• In the methods section the datasets and algorithms are explained, but it is not clear how this is calculated, in which programming language/ GIS/… If the algorithm is programmed, more info is needed on which heuristic has been used. Is it greedy? How is decided who gets shelter first?
• Why is the mean distance of the USA so small? We would expect large distances here. Is it because shelters are almost never full or because they are close together?

Smaller remarks, which should be implemented in the text

• 1, L.30 Reference?
• 2, L.44 Shouldn’t this be a new paragraph?
• 2, L.83 A figure of the datasets used might be useful.
• What is the source of the building layer?
• Was the telecom data collected on the exact moment of the earthquake or is it an estimation based on data of another day at the same time?
• What are the facility areas on figure 1?
• 3, L.106 I would like to have some more information on the Building Act. When was this revised? Against what force should a building be protected? Why do you choose the 30 years boundary for the buildings?
• 3, L.111 Reference for the korea disaster management?
• Table 1 is not clear to me. Does the number of floors mean that two floors must follow the regulations or that a building can only have two floors in 2018?
• Figure 2: Are there regions with more risk buildings or is it more or less equally spread throughout the city?
• In Table 2 it would be useful to list the algorithms that use the different variables in an extra column.
• Maybe it is better to use subtitles for the explanation of the methods instead of underlining the method.
• Which method is used now when an earthquake occurs?
• 5, L.150 At step 2 another letter should be used than j because another shelter is being considered. This step can be repeated a number of times, which should be clear in the steps.
• One population node equals one cell in the grid of Fig. 1?
• 5, L.162 It is not clear why you mention the shelter capacity. Was it not needed in the previous method as well? What is the implication on the calculation?
• The minimum hazard method also takes capacity into account, which is not clear from the name as opposed to the previous method.
• 6, L. 197 Step 1 should be more explained, it is not clear to me how this group is calculated.
• 6, L.209 So, for every cell in the grid of Figure 1, 20 paths are calculated? To 20 different shelters? Why 20? When you order the routes from least to most hazard, which route is mostly chosen by the algorithm when distance is been taken into account? Is it mostly the 20^th route (a lot of hazard) or the first route (least hazard)? I would like to see some more statistics on that.
• It would be useful to include the shelters on figure 7.

To summarize, More analysis on the differences between the methods is needed to publish this research and a more extended introduction and methods section.

Author Response

Please see the attachment.

Round 2

Reviewer 1 Report

Thank you very much for considering my suggestions. All the points have been resolved and improved the manuscript.

Author Response

The corresponding author (David Lee) is a native, fluent English speaker, and has extensively proofread this paper for spelling, grammar, and clarity.

Reviewer 2 Report

The authors considered all remarks and have made substantial improvements to the manuscript. Still some minor details remain:

• Table 1: p-values are usually presented as ‘.001’ instead of ‘0.001’ and when they are significant an ‘*’ is placed aster the value. This way, readers can see the actual value instead of ‘<0.001’.
• English remains a bit strange (e.g. L.60 these studies attempted to improve on…), I recommend to proofread the article again.

Author Response

The corresponding author (David Lee) is a native, fluent English speaker, and has extensively proofread this paper for spelling, grammar, and clarity.

Title of the Paper: “Balancing hazard exposure and walking distance in evacuation route planning during earthquake disasters” 

To the Editor and Reviewer:

We deeply appreciate the reviewers’ constructive comments as well as the associate editor’s offer to revise and resubmit. We have carefully revised the paper following the reviewers’ comments and suggestions, and extensively proofread the entire paper for correctness and clarity as well. We again thank the editor and the reviewers and welcome any further feedback on ways to improve the paper.

Response to Comments

Point 1: Table 1: p-values are usually presented as ‘.001’ instead of ‘0.001’ and when they are significant an ‘*’ is placed aster the value. This way, readers can see the actual value instead of ‘<0.001’.

Response 1: We have revised Table 1. We changed p-values to t-values and represented a significance as ‘*’.

Revised table in chapter 3.1:

Table 4. Two sample t-test results of evacuation route analysis

*** p < .001

Author Response File: Author Response.docx