Exploring the Educational Effects of a Point-Cloud-Derived 3D Urban Model on Residents’ Spatial Understanding and Evacuation Behavioral Intentions for Sustainable Community-Based Tsunami Evacuation Education

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The article presents an innovative approach by developing and evaluating a tsunami evacuation education programme based on a high-fidelity 3D urban model generated through terrestrial laser scanning. In this regard, the integration of real point clouds into an interactive environment, together with the simulation of flooding scenarios, provides an immersive experience that is highly relevant for disaster risk education.

Positively, the work is particularly notable for its ability to transform real geometric data into an experiential learning environment, enabling participants to gain a more intuitive understanding of the spatial distribution of hazards and evacuation conditions.

However, I consider that the study could have been strengthened by incorporating a more advanced spatial analysis component, particularly given the volume of data generated from the point cloud. In particular, it would have been interesting to exploit the point cloud and the reconstructed environment to implement or compare optimal evacuation routes (for example, using pathfinding algorithms or GIS-based analysis), which would have allowed a comparison between human decision-making and computationally derived optimal solutions.

The system primarily focuses on risk perception and participants’ subjective evaluation (pre- and post-experience questionnaires), which places it more within the scope of educational and perceptual assessment rather than spatial analysis or evacuation optimisation. In this sense, the main contribution of the work is formative and cognitive rather than algorithmic.

I also felt that the paper could have benefited from a greater number of visual representations, making fuller use of the point cloud data to provide clearer and more comprehensive illustrations of the reconstructed urban environment and the simulation results. In particular, more detailed or higher-quality visualisations of the inundation scenarios would have improved interpretability. Furthermore, the explicit integration of flood-prone zones as a spatial layer within the 3D model could have strengthened the spatial analysis and enhanced the educational value of the simulation.

For future extensions, it would be beneficial to integrate optimal evacuation route analysis derived from the urban model itself in order to further enrich the evaluation of user behaviour.

Author Response

Dear Editor and Reviewers,
We sincerely thank the reviewers for their careful reading of our manuscript and for their constructive comments. We have revised the manuscript substantially in response to the reviewers’ suggestions. In particular, we added exploratory statistical analyses using the Wilcoxon signed-rank test, calculated effect sizes, examined the internal consistency of the questionnaire constructs using Cronbach’s alpha, clarified that the system is a desktop-based 3D simulation rather than an HMD-based VR system, and revised the wording throughout the manuscript to avoid overstating the generalizability or causal strength of the findings.
We also added further discussion on the limitations of the small sample size, the single-group pre–post design without a control group, the qualitative nature of the tsunami inundation representation, the rationale for using terrestrial laser scanning instead of UAV photogrammetry alone, the usability challenges for middle-aged and older participants, and the potential integration of GIS-based route analysis in future work.
All changes have been incorporated into the revised manuscript.

 

Comment 1
The article presents an innovative approach by developing and evaluating a tsunami evacuation education programme based on a high-fidelity 3D urban model generated through terrestrial laser scanning.

Response
Thank you very much for your positive evaluation of our study. We are pleased that the reviewer recognized the value of integrating real point-cloud data into an interactive tsunami evacuation education environment. In the revised manuscript, we further clarified the contribution of the study as a locally grounded, desktop-based 3D evacuation learning environment using a high-fidelity point-cloud-derived urban model.

Changes made
We revised the Abstract, Introduction, and Conclusion to more clearly describe the contribution of the study while avoiding overgeneralization.

Comment 2
The study could have been strengthened by incorporating a more advanced spatial analysis component, such as pathfinding algorithms or GIS-based analysis.

Response
Thank you for this constructive suggestion. We agree that integrating spatial analysis, GIS-based route analysis, or pathfinding algorithms would significantly strengthen the study. The primary contribution of the present manuscript, however, is the development and exploratory evaluation of an experiential learning environment for tsunami evacuation education, rather than the development of an evacuation-route optimization algorithm.
To address this point, we revised the Discussion to clarify the scope of the present study and added future work on GIS-based network analysis and pathfinding. Specifically, we noted that future studies should compare residents’ selected routes with computationally derived optimal or safer routes using the reconstructed urban model.

Changes made
We added a discussion on the potential integration of spatial analysis and evacuation-route optimization.

Comment 3
The system primarily focuses on risk perception and subjective evaluation, and the main contribution is educational and cognitive rather than algorithmic.

Response
Thank you for this accurate interpretation of the study. We agree that the main contribution of the present work is educational and cognitive rather than algorithmic. In the revised manuscript, we clarified this point explicitly in the Discussion.
The study aims to examine whether a locally grounded 3D environment can help residents recognize hazards, reflect on evacuation decisions, and form behavioral intentions. We have therefore revised the manuscript to more clearly position the study within the field of disaster-prevention education and experiential learning, while identifying algorithmic evacuation-route optimization as an important direction for future research.

Changes made
We revised the Discussion to clarify the contribution and scope of the study.

Comment 4
The paper could have benefited from more visual representations and clearer illustrations of the reconstructed urban environment and inundation scenarios.

Response
Thank you for this helpful suggestion. We agree that clearer visual representations would improve the interpretability of the study. In the revised manuscript, we improved the explanation related to the reconstructed urban environment, the inundation scenario, the evacuation start point, the designated shelter, and route-obstruction scenarios. We also added this issue as an important direction for future work, particularly the need to integrate flood-prone zones, road-blockage points, and possible detour routes more explicitly as spatial layers within the 3D model.
These improvements are intended to help readers better understand how the reconstructed urban environment was used for tsunami evacuation education and how inundation and road-blockage scenarios were represented.

Changes made
We revised the explanation to improve the interpretability of the reconstructed urban environment and simulation scenario.

Comment 5
The explicit integration of flood-prone zones as a spatial layer within the 3D model could have strengthened the spatial analysis and enhanced the educational value of the simulation.

Response
Thank you for this valuable suggestion. We agree that explicitly integrating flood-prone zones as spatial layers would strengthen both the spatial interpretation and educational value of the system. In the revised manuscript, we clarified that the inundation scenario was configured with reference to the Kanazawa City tsunami inundation assumption zone and that the visualized inundation was intended to help participants understand the spatial relationship between predicted inundation areas and evacuation routes.
We also added future work noting that flood-prone zones should be integrated more explicitly as spatial layers within the 3D environment and linked to GIS-based evacuation-route analysis.

Changes made
We revised the Methods and Discussion to clarify the use of the tsunami inundation assumption zone and the future integration of flood-prone spatial layers.

Comment 6
For future extensions, it would be beneficial to integrate optimal evacuation route analysis derived from the urban model itself.

Response
Thank you for this suggestion. We fully agree. In the revised manuscript, we added optimal evacuation route analysis as a key direction for future work. We noted that future studies should use the point-cloud-derived urban model and GIS-based network analysis to compare residents’ selected routes with computationally derived safer routes. This would allow future research to evaluate not only participants’ subjective understanding but also the relationship between their route-choice behavior and spatially optimized evacuation routes.

Changes made
We added this point to the Limitations and Future Work section.

Reviewer 2 Report

Comments and Suggestions for Authors

This article studied the use of ground-based lidar scanning to construct a realistic 3D street view model and explored its application in tsunami evacuation education. The topic has good practical significance. The following revisions are suggested:

1. The authors clearly state in the introduction that the purpose of this study is to "quantitatively evaluate" the educational effect. However, in the analysis of the results (pages 546-547), they admit that no inferential statistical tests (such as paired-samples t-tests) were performed, only describing the increase in the mean. For ordinal data collected through Likert scales, comparing only the mean is not rigorous.

2. The study used a custom questionnaire (dimensions D, E, F, G, H, I, J, etc.) to measure participants' spatial understanding and self-efficacy, but the reliability (such as Cronbach's alpha coefficient) and validity analysis of the questionnaire are not mentioned throughout the article.

3. Only 25 residents participated in the experiment, and only 22 provided complete pre- and post-test paired data. For disaster prevention interventions aimed at community dissemination, such a sample size is insufficient to draw strong, generalizable conclusions.

4. The literature review extensively discusses the immersive advantages of VR (virtual reality), but Figure 17 shows that participants actually operate controllers on ordinary desktop monitors (PC monitors). Desktop 3D and head-mounted display (HMD) VR differ fundamentally in spatial perception and motion sickness.

5. The article mentions setting up a "time-varying discharge profile" to simulate the sudden increase in tsunami water volume, but provides no specific physical parameters (such as initial flow velocity, peak flow size, simulation duration, etc.).

6. The team conducted 8 surveys and 652 scanning stations to build the model, which was extremely costly. However, for macroscopic path identification in disaster prevention and evacuation drills, UAV photogrammetry modeling is often more efficient.

Author Response

Dear Editor and Reviewers,
We sincerely thank the reviewers for their careful reading of our manuscript and for their constructive comments. We have revised the manuscript substantially in response to the reviewers’ suggestions. In particular, we added exploratory statistical analyses using the Wilcoxon signed-rank test, calculated effect sizes, examined the internal consistency of the questionnaire constructs using Cronbach’s alpha, clarified that the system is a desktop-based 3D simulation rather than an HMD-based VR system, and revised the wording throughout the manuscript to avoid overstating the generalizability or causal strength of the findings.
We also added further discussion on the limitations of the small sample size, the single-group pre–post design without a control group, the qualitative nature of the tsunami inundation representation, the rationale for using terrestrial laser scanning instead of UAV photogrammetry alone, the usability challenges for middle-aged and older participants, and the potential integration of GIS-based route analysis in future work.
All changes have been incorporated into the revised manuscript.

 

Comment 1
The manuscript originally stated that the purpose was to quantitatively evaluate the educational effect, but the analysis relied mainly on mean comparisons without inferential statistical tests. For Likert-scale ordinal data, comparing only means is not rigorous.

Response
Thank you for this important comment. We agree that the original manuscript overstated the quantitative nature of the evaluation and that comparing only mean values was insufficient, especially because the questionnaire items were measured using five-point Likert scales.
In the revised manuscript, we addressed this issue in two ways. First, we revised the wording in the Abstract and Introduction to clarify that this study is an exploratory community-based evaluation rather than a confirmatory statistical evaluation. Specifically, we changed expressions such as “quantitatively evaluate” to “explore the preliminary educational effects.”
Second, we added an exploratory pre–post statistical analysis using the Wilcoxon signed-rank test, which is more appropriate for ordinal Likert-scale data and a small paired sample. We also calculated effect sizes. The analysis was conducted using the complete paired responses. The results showed significant improvements in all main corresponding item pairs: understanding of hazardous locations, explanation of evacuation-route choices, consideration of detours, clarity of evacuation timing, and communication with family members and neighbors.
The largest improvement was observed for the D1–F1 pair, which assessed understanding of hazardous locations, with a mean increase of 1.591 points, p = 0.0008, and an effect size of r = 0.718.

Changes made
We revised the Abstract and Introduction and added a new statistical analysis section in the Methods. We also replaced the former descriptive comparison section with a new subsection entitled “Exploratory Pre–Post Comparison of Corresponding Items.”

Comment 2
The study used a custom questionnaire, but the reliability and validity of the questionnaire were not sufficiently discussed.

Response
Thank you for pointing this out. We agree that the original manuscript did not provide sufficient explanation of the questionnaire design, content validity, or reliability.
In the revised manuscript, we added a new subsection explaining the questionnaire design and statistical analysis. We clarified that the questionnaire items were developed based on the learning objectives of the program and prior studies on tsunami evacuation education, disaster-prevention learning, and ICT-based educational environments. We also added that the questionnaire was reviewed by the authors, including researchers with expertise in disaster prevention, urban planning, and 3D visualization, to ensure content validity.
In addition, we calculated Cronbach’s alpha for each multi-item construct. The coefficients ranged from 0.703 to 0.941, indicating acceptable to high internal consistency. For example, the evaluation of the 3D model showed high reliability, α = 0.897, and the post-experience learning-effect items also showed high reliability, α = 0.941. Because the sample size was limited, we interpreted these coefficients as supplementary indicators rather than definitive evidence of scale reliability.

Changes made
We added a new subsection titled “Questionnaire Design and Statistical Analysis” and added a new table presenting Cronbach’s alpha values for each construct.

Comment 3
Only 25 residents participated in the experiment, and only 22 provided complete pre- and post-test paired data. This sample size is insufficient to draw strong, generalizable conclusions.

Response
Thank you for this important observation. We agree that the sample size was small and that the findings should not be generalized to all coastal communities. In the revised manuscript, we clarified that this study should be interpreted as an exploratory community-based case study rather than a confirmatory evaluation.
We also revised the Abstract, Introduction, Discussion, and Conclusion to avoid overstating the strength or generalizability of the results. In particular, we added a limitation statement explaining that the study was based on 25 participants and 22 complete pre–post paired responses.

Changes made
We added a limitation statement in the Discussion and revised the Conclusion.

Comment 4
The literature review discusses immersive VR, but the system used in the experiment was operated on ordinary desktop monitors. Desktop 3D and HMD-based VR differ in spatial perception and motion sickness.

Response
Thank you for this valuable comment. We agree that the original manuscript did not sufficiently distinguish between immersive HMD-based VR and the desktop-based 3D simulation used in this study.
In the revised manuscript, we clarified that the system was implemented as a desktop-based interactive 3D simulation rather than an HMD-based VR system. We also revised related terminology throughout the manuscript. For example, we changed expressions such as “VR-induced sickness” to “discomfort associated with the desktop-based 3D experience.” We also added a sentence in the related studies section to clarify that, although previous studies have often focused on immersive VR, the present study uses a monitor-based interactive evacuation learning environment.

Changes made
We revised the Related Studies, Methods, Results, and Discussion sections to clarify the nature of the system.

Comment 5
The manuscript mentioned a “time-varying discharge profile” for tsunami water volume, but did not provide physical parameters such as flow velocity, peak flow size, or simulation duration.

Response
Thank you for this important comment. We agree that the original description could be misunderstood as a physically calibrated hydrodynamic simulation. In the revised manuscript, we clarified that the tsunami inundation representation was not designed as an engineering-accurate tsunami simulation. Rather, it was developed as an educational visualization to help participants intuitively recognize the approximate spatial extent and direction of inundation in relation to the published tsunami inundation assumption zone.
We added a clear statement that parameters such as flow velocity, discharge volume, arrival time, and water depth should not be interpreted as engineering predictions. We also added a clarification explaining the purpose and scope of the tsunami inundation representation.

Changes made
We revised Section 3.4 to explain the purpose and scope of the tsunami inundation representation.

Comment 6
The team conducted eight surveys and 652 scan stations, which was costly. UAV photogrammetry may be more efficient for macroscopic evacuation route identification.

Response
Thank you for this insightful comment. We agree that UAV photogrammetry is more efficient for wide-area modeling and macroscopic route identification. In the revised manuscript, we added a discussion clarifying why terrestrial laser scanning was selected in this study.
The main objective of this study was not wide-area route mapping alone, but the development of a street-level, highly realistic evacuation learning environment for residents. Therefore, terrestrial laser scanning was used because it can reproduce pedestrian-level spatial conditions, including road width, visibility at corners, the sense of enclosure created by buildings, steps, and possible blockage points. These features are difficult to reproduce sufficiently using aerial data alone but are important for helping residents understand evacuation conditions from a walking perspective.
We also added a statement that future implementations should consider a hybrid workflow combining UAV photogrammetry for wide-area coverage and TLS for critical street-level segments.

Changes made
We added a new discussion on the rationale for using TLS and the potential of a hybrid TLS–UAV workflow.

 

 

Reviewer 3 Report

Comments and Suggestions for Authors

This manuscript explores the effectiveness of tsunami evacuation education based on point cloud 3D models. The topic is in line with the trend of digitalization in disaster prevention education, and the technical applications (BLK360 scanning, Unity fluid and building collapse simulation) have certain practical innovation. However, there are still some shortcomings in this manuscript at present.
1. The internal validity of the research design is weak. The author only used a single group pre-test design (N=25, effective pairing n=22) and did not set up any control group (such as the traditional paper map group or the general CG model group), which cannot exclude confounding variables such as Hawthorne effect, historical effect, or maturity effect. In educational intervention research, the lack of a controlled design makes it difficult to demonstrate the incremental benefits of point cloud models compared to other media. This is an important issue that directly affects the feasibility of the research approach of the entire paper. The author is requested to carefully review and make corresponding modifications.
2. The data analysis remains at the level of descriptive statistics, and the author explicitly avoids inferential testing. In the case of extremely small sample sizes, reporting only mean differences without providing effect sizes, confidence intervals, or significance tests seriously weakens the statistical basis of conclusions. In addition, the questionnaire tool did not report reliability (such as Cronbach's alpha) and validity evidence, and the psychometric quality of the scale cannot be confirmed.
3. Participants had difficulty operating (with an H1 mean of only 3.67) and 60% had no 3D/VR experience, indicating that the usability bottleneck of this technology for the middle-aged and elderly population has not been resolved, which is precisely the core target population of disaster prevention education. The author should delve into this paradox in depth, rather than just listing it as a direction for future improvement.

Author Response

Dear Editor and Reviewers,
We sincerely thank the reviewers for their careful reading of our manuscript and for their constructive comments. We have revised the manuscript substantially in response to the reviewers’ suggestions. In particular, we added exploratory statistical analyses using the Wilcoxon signed-rank test, calculated effect sizes, examined the internal consistency of the questionnaire constructs using Cronbach’s alpha, clarified that the system is a desktop-based 3D simulation rather than an HMD-based VR system, and revised the wording throughout the manuscript to avoid overstating the generalizability or causal strength of the findings.
We also added further discussion on the limitations of the small sample size, the single-group pre–post design without a control group, the qualitative nature of the tsunami inundation representation, the rationale for using terrestrial laser scanning instead of UAV photogrammetry alone, the usability challenges for middle-aged and older participants, and the potential integration of GIS-based route analysis in future work.

 

Comment 1
The internal validity of the research design is weak. The study used only a single-group pre-test design and did not set up any control group, such as a traditional paper map group or a general CG model group. Therefore, confounding variables such as the Hawthorne effect, history effect, or maturation effect cannot be excluded.

Response
Thank you for this important comment. We agree that the single-group pre–post design is a major limitation of the present study. Because this study did not include a control or comparison group, such as a traditional paper-map-based training group or a generic CG-model group, it cannot fully exclude possible confounding factors such as the Hawthorne effect, history effect, or maturation effect.
In response to this comment, we revised the manuscript to clarify that the present study should not be interpreted as providing causal evidence of the superiority of point-cloud-derived 3D models over other educational media. Rather, the study is positioned as an exploratory community-based case study that provides preliminary evidence of perceived educational effects and implementation feasibility.
We also added a limitation statement explaining that future studies should employ controlled or comparative designs to examine the incremental educational benefits of point-cloud-derived 3D models. In particular, comparisons with paper hazard maps, conventional two-dimensional evacuation materials, or generic CG-based 3D models will be necessary to clarify the unique contribution of locally measured point-cloud-derived models.

Changes made
We revised the Discussion and Conclusion to clarify the limitation of the single-group pre–post design and to avoid causal overinterpretation.

Comment 2
The data analysis remains at the level of descriptive statistics, and the manuscript does not provide sufficient inferential testing, effect sizes, confidence intervals, or reliability and validity evidence for the questionnaire.

Response
Thank you for pointing out this important issue. We agree that the original manuscript relied too heavily on descriptive statistics and did not sufficiently report inferential analysis, effect sizes, or evidence regarding the reliability and validity of the questionnaire.
In the revised manuscript, we added an exploratory pre–post statistical analysis using the Wilcoxon signed-rank test, which is appropriate for ordinal Likert-scale data and a small paired sample. We also calculated effect sizes for the corresponding pre–post item pairs. The results showed significant improvements in all main corresponding item pairs, including understanding of hazardous locations, explanation of evacuation-route choices, consideration of detours, clarity of evacuation timing, and communication with family members and neighbors.
In addition, we added an explanation of the questionnaire design and content validity. The questionnaire items were developed based on the learning objectives of the program and prior studies on tsunami evacuation education, disaster-prevention learning, and ICT-based educational environments. The items were reviewed by the authors, including researchers with expertise in disaster prevention, urban planning, and 3D visualization. We also calculated Cronbach’s alpha for each multi-item construct. Because the sample size was limited, these reliability coefficients are interpreted as supplementary indicators rather than definitive evidence of scale reliability.

Changes made
We added a new subsection titled “Questionnaire Design and Statistical Analysis” and added new tables presenting the Wilcoxon signed-rank test results, effect sizes, and Cronbach’s alpha values.

Comment 3
Participants had difficulty operating the system, with an H1 mean of only 3.67, and 60% had no 3D/VR experience. This indicates that the usability bottleneck for middle-aged and older residents has not been resolved.

Response
Thank you for this insightful comment. We agree that the usability issue is not a minor technical limitation but a central challenge for the social implementation of this type of disaster-prevention education. The target users of community-based tsunami evacuation education include many middle-aged and older residents, yet the system partly depends on familiarity with 3D environments and controller-based navigation. This creates an important implementation paradox.
In the revised manuscript, we expanded the Discussion to address this issue more explicitly. Although the participants evaluated the 3D model and its educational usefulness highly, the score for ease of operation was lower than those of other items. In addition, 60% of participants had no prior experience with 3D/VR technologies. These findings suggest that high-fidelity 3D visualization alone is not sufficient for effective community deployment. Interface design, operational support, facilitation, and simplified navigation methods are essential conditions for practical use among middle-aged and older residents.

Changes made
We revised the Discussion to interpret the usability bottleneck as a core implementation issue rather than merely a future technical improvement.

All changes have been incorporated into the revised manuscript.

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

While the article received some improvements after the previous revision, several issues remain:

1. Although the article used the Wilcoxon signed-rank test to analyze 22 valid paired samples, the sample size is extremely small (N=22). Precisely calculating p-values ​​to four decimal places (e.g., P=0.0008) and drawing overly certain conclusions is easily questionable.

2. Table 7 shows the Cronbach's alpha values ​​calculated based on 22-24 samples. In statistics, the internal consistency coefficient calculated with such a small sample size fluctuates greatly and is not stable enough. Although the authors have mentioned it as a "supplementary indicator," it is recommended to further clarify that this data is for preliminary reference only to avoid misleading readers' judgment of the scale's reliability.

3. Lack of data on younger populations. As a disaster prevention education system that introduces new desktop 3D/VR technologies, the audience's "digital literacy" is a key variable. However, according to the statistics in Table 2, the participants' age range is from over 40 to over 80 years old, with a 0% participation rate among young people/teenagers under 40. Therefore, the paper mentions "difficulty in understanding the operation (H1 score of 3.67)" and "difficulty for elderly people in operating the handle." This is largely due to bias caused by the aging of the sample.

4. In the exploratory before-and-after comparison in Table 6, the authors performed Wilcoxon signed-rank tests on five related questions (D1-F1, D2-F2, D2-F3, D3-F4, D5-F5) and claimed that all showed significant improvements. Performing multiple hypothesis tests on the same set of data increases the probability of "Type I errors (false positives)." If Bonferroni correction is used, the significance level (alpha) should be adjusted to 0.05 / 5 = 0.01. If the threshold is 0.01, then D2-F2 (p=0.0133), D3-F4 (p=0.0357), and D5-F5 (p=0.0132) in Table 6 are actually not statistically significant.

5. The authors honestly acknowledge that the Zibra Liquid fluid simulation has not been precisely calibrated for hydrodynamic and other physical parameters (such as flow velocity, water depth, and arrival time) and cannot be used as an engineering prediction. However, in Table 8, residents rated the "realism of flooding and collapse (G8)" as high as 4.17. This poses a potential danger: residents may mistake inaccurate flow velocities and water depths for "real objective laws" and memorize them. In a real disaster, such erroneous physical intuition could harm them. Residents' blind trust in uncalibrated physical phenomena should be avoided.

Author Response

Comment 1:
Although the article used the Wilcoxon signed-rank test to analyze 22 valid paired samples, the sample size is extremely small (N=22). Precisely calculating p-values to four decimal places (e.g., P=0.0008) and drawing overly certain conclusions is easily questionable.

Response:
Thank you for this important comment. We agree that the small sample size requires cautious interpretation and that overly certain conclusions should be avoided. In response, we revised the Abstract, Results, Discussion, and Conclusions to emphasize that the findings should be interpreted as preliminary and exploratory. We also revised the wording so that the statistical results are not presented as definitive evidence. In particular, we avoided making overly strong causal claims and clarified that the results provide preliminary evidence of possible educational and cognitive effects. Where appropriate, we also revised the presentation of p-values in the main text to avoid giving an impression of excessive precision, while retaining sufficient statistical information for transparency.

Comment 2:
Table 7 shows the Cronbach's alpha values calculated based on 22–24 samples. In statistics, the internal consistency coefficient calculated with such a small sample size fluctuates greatly and is not stable enough. Although the authors have mentioned it as a “supplementary indicator,” it is recommended to further clarify that this data is for preliminary reference only to avoid misleading readers' judgment of the scale's reliability.

Response:
Thank you for this helpful suggestion. We agree that Cronbach’s alpha values based on such a small sample should not be interpreted as stable or definitive evidence of scale reliability. We have therefore revised the explanation of Table 7 and the relevant text in the Methods and Results sections to state more clearly that these coefficients are provided only as preliminary reference values. We also added a cautionary statement that the internal consistency estimates may fluctuate substantially because of the small sample size and should not be regarded as conclusive evidence of scale reliability. This revision was made to prevent readers from overinterpreting the reliability analysis.

Comment 3:
Lack of data on younger populations. As a disaster prevention education system that introduces new desktop 3D/VR technologies, the audience's “digital literacy” is a key variable. However, according to the statistics in Table 2, the participants' age range is from over 40 to over 80 years old, with a 0% participation rate among young people/teenagers under 40. Therefore, the paper mentions “difficulty in understanding the operation (H1 score of 3.67)” and “difficulty for elderly people in operating the handle.” This is largely due to bias caused by the aging of the sample.

Response:
Thank you for pointing this out. We agree that the absence of younger participants is an important limitation, particularly because digital literacy may strongly affect the usability and perceived difficulty of a desktop-based 3D evacuation system. We have revised the Discussion and Limitations sections to explicitly state that the participant group was limited to middle-aged and older residents and that the findings, especially those related to operation difficulty and controller usability, may reflect the age composition and digital literacy of the sample. We also clarified that the results should not be generalized to younger populations or users with different levels of digital experience. In addition, we added a statement that future studies should include younger participants, students, and more age-diverse samples, and should examine the effects of digital literacy on learning outcomes and usability.

Comment 4:
In the exploratory before-and-after comparison in Table 6, the authors performed Wilcoxon signed-rank tests on five related questions (D1–F1, D2–F2, D2–F3, D3–F4, D5–F5) and claimed that all showed significant improvements. Performing multiple hypothesis tests on the same set of data increases the probability of Type I errors. If Bonferroni correction is used, the significance level should be adjusted to 0.05 / 5 = 0.01. If the threshold is 0.01, then D2–F2, D3–F4, and D5–F5 are actually not statistically significant.

Response:
Thank you very much for this important statistical comment. We agree with the reviewer’s point. In the revised manuscript, we applied a Bonferroni-adjusted significance threshold of α = 0.01 for the five corresponding item pairs. We revised the Abstract, Results, Discussion, and Conclusions accordingly. We no longer state that all five item pairs showed statistically significant improvements after correction. Instead, we now distinguish between results that remained below the Bonferroni-adjusted threshold and those that showed positive descriptive changes with uncorrected p-values below 0.05 but did not meet the adjusted threshold. Specifically, the improvement in understanding hazardous locations and the improvement in consideration of detours and alternative routes were interpreted as the most statistically robust findings, whereas the other item pairs were described more cautiously as preliminary descriptive trends. This revision reduces the risk of Type I error and presents the exploratory analysis more appropriately.

Comment 5:
The authors honestly acknowledge that the Zibra Liquid fluid simulation has not been precisely calibrated for hydrodynamic and other physical parameters and cannot be used as an engineering prediction. However, in Table 8, residents rated the “realism of flooding and collapse (G8)” as high as 4.17. This poses a potential danger: residents may mistake inaccurate flow velocities and water depths for “real objective laws” and memorize them. In a real disaster, such erroneous physical intuition could harm them. Residents' blind trust in uncalibrated physical phenomena should be avoided.

Response:
Thank you for this very important comment. We fully agree that uncalibrated visual simulations must not lead residents to misunderstand flow velocity, water depth, arrival time, or structural collapse behavior as physically accurate predictions. In response, we revised the manuscript to clarify that the Zibra Liquid simulation and collapse representation were used only for educational visualization and risk communication, not for engineering prediction. We also revised the explanation of Table 8 to clarify that the G8 item reflects participants’ perceived visual realism or plausibility, rather than the physical accuracy of inundation or collapse behavior.
Furthermore, we added a cautionary explanation in the Methods and Discussion sections stating that participants were informed before the experience that the inundation and collapse representations should not be interpreted as precise predictions of actual tsunami behavior. We also added a limitation that future implementations should incorporate validated hydrodynamic information, official hazard data, or calibrated simulation parameters when the system is used for more detailed risk communication. These revisions were made to avoid encouraging blind trust in uncalibrated physical phenomena and to clearly define the educational scope of the system.

Reviewer 3 Report

Comments and Suggestions for Authors

The author has carefully revised the paper based on the review comments. I don't have any more questions.

Author Response

Comment:
The author has carefully revised the paper based on the review comments. I don't have any more questions.

Response:
We sincerely thank the reviewer for the positive evaluation of our revised manuscript. We are grateful for the reviewer’s careful reading and constructive feedback during the review process. We have further revised the manuscript in response to the remaining comments from Reviewer 2 and have highlighted all changes in the revised version.