the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 11 Dec 2025
Integrating flood-induced population movements into future flood damage estimates in Japan
Abstract. Recent studies have highlighted that flooding can influence population dynamics. However, existing estimates of future flood damage primarily consider population changes driven by births, deaths, and migration unrelated to flooding. As a result, the potential impacts of flood-induced population movements (FIPMs) on future flood damage costs remain largely unexplored. This study evaluated the impacts of FIPMs on future flood damage costs in Japan, a country that faces flood risk and population decline. We develop a methodological framework that uses statistical causal inference to quantify FIPMs, integrates these estimates into future population and land-use projections, and evaluates future flood damage costs under scenarios of climate and land-use change. The results indicate that incorporating FIPMs leads to only modest changes in estimated flood damage costs at the national level (generally below 1 %), and similarly modest impacts at the prefectural level, except for a few prefectures with changes of approximately 2 %. However, greater variability is observed at the municipal level, with approximately 10 % of municipalities experiencing changes exceeding 1 % and some municipalities showing reductions in estimated flood damage costs exceeding 10 %. These findings highlight the importance of accounting for FIPMs in flood risk management frameworks and policy evaluations.
-
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
(1555 KB)
-
Supplement
(435 KB)
-
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(1555 KB) - Metadata XML
-
Supplement
(435 KB) - BibTeX
- EndNote
- Final revised paper
Journal article(s) based on this preprint
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2025-5949', Anonymous Referee #1, 07 Feb 2026
- AC1: 'Reply on RC1', Hayata Yanagihara, 03 May 2026
-
RC2: 'Comment on egusphere-2025-5949', Marijn Ton, 01 Mar 2026
Thank you for the opportunity to review this manuscript. The study aims to estimate the impact of two flood events on population movements in Japan, with the goal of assessing the effectiveness of flood mitigation strategies. Overall, the manuscript is well written and addresses an interesting and relevant research question. The empirical approach is generally sound, but I do have some concerns and suggestions for improvement.
Below I provide several comments that could improve the paper. These points are presented in no particular order.
- My main concern relates to the timing of the analysis. The flood events occurred in 2019 and 2020, while population data are observed from 2005 to 2020 in five-year intervals. As a result, the study can only capture very short-term population movements in response to the floods. Since migration and population adjustments often occur gradually, it would be more informative to examine longer-term effects, for example by including population data for the period after 2020 (e.g., 2020–2025) if such data become available. As it stands, the analysis likely captures only immediate or short-run effects. This limitation also raises some concerns about the usefulness of conducting projections over several decades based on these short-term responses.
- The Introduction is well written and clearly outlines the research problem. However, the empirical strategy, namely the use of a difference-in-differences (DiD) approach, is not sufficiently introduced or motivated. It would be helpful if the authors explained why they chose a DiD framework and whether alternative empirical approaches could have been used. Placing the chosen method more explicitly within the existing literature would improve the paper.
- It would be useful to clarify earlier in the Introduction that the analysis focuses on only two flood events. When reading the Introduction, I initially expected a broader analysis of the effects of floods on population movements across Japan, whereas the study is in fact closer to a case study. In addition, a map showing the locations of the flood events would be helpful. Then we could also see whether the affected areas overlap because I was wondering this when reading the manuscript.
- A large part of the methodology is devoted to explaining the variables, coefficients and fixed effects, which makes the paper somewhat difficult to follow. I would encourage the authors to have another look at the equations and see where they can simplify the notation. For example, \delta_{bf}^{mun} R_{i’, t’, g’}^{bf} is not very readable. The equations might become clearer if the authors use simpler notation for coefficients, such as \beta_{0}, \beta_{1} and \beta_{2}, and call the variables BF, AF and CD, for example. Reducing the number of subscripts and superscripts would likely improve readability. Another suggestion would be to present the variables, coefficients, and fixed effects in a table or in a bullet-point list, so that the reader has a more clear overview.
- In Equations (1) and (2), the authors include a time trend for each cross-sectional unit (grid cells in Equation 1 and municipalities in Equation 2). At the same time, the paper emphasizes the importance of satisfying the parallel trends assumption so that the estimated effect can be attributed to the flood rather than to other factors. The manuscript presents evidence suggesting that the parallel trends assumption holds. Given this, it is unclear why unit-specific time trends are still included in the regression. While the supplementary information cites one study that adopts a similar specification, this alone does not justify the modeling choice. I would like to see a clearer rationale for including these trends, or even better, a robustness check presenting results without the unit-specific time trends, especially since time fixed effects are already included.
- Equations (3), (4), and (5) are also somewhat difficult to follow. The notation b,c = {1,2,3,4,5} is not immediately clear. It might be simpler and more transparent to refer directly to the corresponding years instead of using index notation.
Citation: https://doi.org/10.5194/egusphere-2025-5949-RC2 -
RC3: 'Reply on RC2', Anonymous Referee #3, 02 Mar 2026
This study quantifies flood-induced population movements (FIPMs) between and within municipalities in response to varying flood magnitudes from heavy precipitation events, using the difference-in-differences (DiD) method, a statistical causal inference technique. The influence of these movements on projected future flood damages from heavy precipitation is then assessed, revealing modest effects at the national level (1–2%) but considerably stronger impacts in some individual municipalities (above 10%). The authors conclude that locally concentrated effects suggest that FIPMs are particularly relevant for flood risk management and policy evaluation at the municipal scale.
I have the following comments:
-
Make explicit throughout that only precipitation-induced flooding (not riverine or coastal) is analysed – currently this is missing in the title, abstract and introduction. At l. 51, consider whether riverine flooding is also relevant to your framing – if it is excluded, state so explicitly; if it is partially covered, clarify the scope.
-
The concluding statement about "importance for flood risk management frameworks" is not convincingly supported by 1–2% national-level effects. Either soften the claim to focus on local relevance (>10% in some municipalities), or strengthen it with additional argumentation.
-
Clarify the distinction between planned relocation (policy-driven resettlement) and spontaneous migration (individual household decisions) early in the introduction.
-
I suggest that Sections 3 onwards should be structured as subsections of the Methods section (Section 2) for consistency and clarity.
-
At l. 84, clarify whether the analysis is based on only two specific extreme events: if so, this is a significant scope limitation that should be stated clearly in both abstract and introduction. Alternatively, clarify whether the two events serve as empirical anchors for estimating movement assumptions used in the broader modelling.
-
l. 333 / Figure 4: Specify the unit clearly – "persons per [what]?" needs to be defined (per km², per municipality, per 1,000 inhabitants, etc.).
-
l. 345: Explain why flood damages are initially reduced under higher climate change scenarios – this is counterintuitive and needs explicit reasoning. Is it due to outmigration from high-risk zones reducing exposed population? Or overall population increase?
-
Figure 5: Add a brief explanatory note in the text clarifying what it means when flood damages increase due to population movement – presumably this reflects in-migration into higher-risk areas, but this should be stated directly.
-
l. 422: The strong general claim is not well-supported given the small national-level numbers and the nuanced discussion preceding it. Consider limiting the claim to the local/municipal scale, where the evidence is stronger.
Citation: https://doi.org/10.5194/egusphere-2025-5949-RC3 - AC3: 'Reply on RC3', Hayata Yanagihara, 03 May 2026
-
- AC2: 'Reply on RC2', Hayata Yanagihara, 03 May 2026
Peer review completion
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2025-5949', Anonymous Referee #1, 07 Feb 2026
The authors present a study titled "Integrating flood-induced population movements into future flood damage estimates in Japan." The manuscript introduces a novel framework for quantifying and predicting population relocation in response to flood events, incorporating climate change effects, to evaluate future population distribution and flood risk in Japan. Overall, I find the study to be well-constructed and the manuscript clearly written. The integration of statistical causal inference with long-term land-use and population projections offers a valuable contribution to socio-hydrological modeling. However, I have identified several critical areas regarding spatial resolution, flood typology, and the modeling of relocation "pull" factors that warrant further attention. I am recommending major revisions to address these methodological limitations and scope concerns. Pending these revisions, I believe the manuscript would be a strong candidate for publication in Natural Hazards and Earth System Sciences (NHESS).
- Methodological Detail (Line 295): The authors cite Yanagihara et al. (2024) for the fluvial flood damage model. However, the current manuscript lacks sufficient detail regarding this model’s internal mechanics. Please include a concise summary of the model’s core equations or assumptions to ensure the reader can follow the assessment without referring exclusively to prior work.
- Flood Typology and Title: The study focuses exclusively on fluvial flooding. Given the significance of pluvial and coastal flooding in Japan, the title should be more specific. In the title, I suggest changing "future flood damage" to "future fluvial flood damage" with corresponding adjustments in the Abstract and throughout the text to avoid over-generalization.
- Spatial Resolution: The use of a 250 m flood grid is relatively coarse for a national-scale study seeking municipal-level precision. This resolution likely misses localized flood risks and micro-topographical features that influence individual relocation decisions. This should be addressed as a limitation in Section 6.3.
- Multimodal Flooding Impacts: The observational dataset used to quantify FIPMs includes residents affected by various water-related disasters (pluvial, coastal, etc.), yet the predictive model only simulates fluvial events. The authors should discuss the net effect of this discrepancy near Line 375 and in the limitations. Specifically, how does the exclusion of pluvial and coastal flood projections affect the estimated impact of FIPMs the results of the study?
- "Pull" Factors and Relocation Dynamics: The current framework assumes that populations relocate proportionally to existing density. In reality, migration is driven by specific "pull" factors such as employment hubs and family networks—as seen in historical precedents like Hurricane Katrina in the USA, where displacement from New Orleans was highly concentrated in specific regions (e.g., Houston) rather than dispersed uniformly. This simplification should be explicitly noted as a limitation in Section 6.3.
- Temporal Resolution and Validation Discrepancy: Since the model relies on 5-year census data, it inherently captures long-term permanent displacement while overlooking short-term movements. This temporal mismatch may explain the significant discrepancy between the model’s 2020 estimate (~16,000 movements) and the IDMC data (~212,000 movements). This distinction between "forced displacement" and "permanent migration" needs a more thorough treatment in the Discussion.
- Code and Data Availability: The current statements stating "available upon request" do not appear to meet the journal's submission guidelines (https://www.natural-hazards-and-earth-system-sciences.net/submission.html). The authors must deposit the underlying research data and code in a public repository or provide a formal explanation as to why it cannot be made publicly accessible.
Citation: https://doi.org/10.5194/egusphere-2025-5949-RC1 - AC1: 'Reply on RC1', Hayata Yanagihara, 03 May 2026
-
RC2: 'Comment on egusphere-2025-5949', Marijn Ton, 01 Mar 2026
Thank you for the opportunity to review this manuscript. The study aims to estimate the impact of two flood events on population movements in Japan, with the goal of assessing the effectiveness of flood mitigation strategies. Overall, the manuscript is well written and addresses an interesting and relevant research question. The empirical approach is generally sound, but I do have some concerns and suggestions for improvement.
Below I provide several comments that could improve the paper. These points are presented in no particular order.
- My main concern relates to the timing of the analysis. The flood events occurred in 2019 and 2020, while population data are observed from 2005 to 2020 in five-year intervals. As a result, the study can only capture very short-term population movements in response to the floods. Since migration and population adjustments often occur gradually, it would be more informative to examine longer-term effects, for example by including population data for the period after 2020 (e.g., 2020–2025) if such data become available. As it stands, the analysis likely captures only immediate or short-run effects. This limitation also raises some concerns about the usefulness of conducting projections over several decades based on these short-term responses.
- The Introduction is well written and clearly outlines the research problem. However, the empirical strategy, namely the use of a difference-in-differences (DiD) approach, is not sufficiently introduced or motivated. It would be helpful if the authors explained why they chose a DiD framework and whether alternative empirical approaches could have been used. Placing the chosen method more explicitly within the existing literature would improve the paper.
- It would be useful to clarify earlier in the Introduction that the analysis focuses on only two flood events. When reading the Introduction, I initially expected a broader analysis of the effects of floods on population movements across Japan, whereas the study is in fact closer to a case study. In addition, a map showing the locations of the flood events would be helpful. Then we could also see whether the affected areas overlap because I was wondering this when reading the manuscript.
- A large part of the methodology is devoted to explaining the variables, coefficients and fixed effects, which makes the paper somewhat difficult to follow. I would encourage the authors to have another look at the equations and see where they can simplify the notation. For example, \delta_{bf}^{mun} R_{i’, t’, g’}^{bf} is not very readable. The equations might become clearer if the authors use simpler notation for coefficients, such as \beta_{0}, \beta_{1} and \beta_{2}, and call the variables BF, AF and CD, for example. Reducing the number of subscripts and superscripts would likely improve readability. Another suggestion would be to present the variables, coefficients, and fixed effects in a table or in a bullet-point list, so that the reader has a more clear overview.
- In Equations (1) and (2), the authors include a time trend for each cross-sectional unit (grid cells in Equation 1 and municipalities in Equation 2). At the same time, the paper emphasizes the importance of satisfying the parallel trends assumption so that the estimated effect can be attributed to the flood rather than to other factors. The manuscript presents evidence suggesting that the parallel trends assumption holds. Given this, it is unclear why unit-specific time trends are still included in the regression. While the supplementary information cites one study that adopts a similar specification, this alone does not justify the modeling choice. I would like to see a clearer rationale for including these trends, or even better, a robustness check presenting results without the unit-specific time trends, especially since time fixed effects are already included.
- Equations (3), (4), and (5) are also somewhat difficult to follow. The notation b,c = {1,2,3,4,5} is not immediately clear. It might be simpler and more transparent to refer directly to the corresponding years instead of using index notation.
Citation: https://doi.org/10.5194/egusphere-2025-5949-RC2 -
RC3: 'Reply on RC2', Anonymous Referee #3, 02 Mar 2026
This study quantifies flood-induced population movements (FIPMs) between and within municipalities in response to varying flood magnitudes from heavy precipitation events, using the difference-in-differences (DiD) method, a statistical causal inference technique. The influence of these movements on projected future flood damages from heavy precipitation is then assessed, revealing modest effects at the national level (1–2%) but considerably stronger impacts in some individual municipalities (above 10%). The authors conclude that locally concentrated effects suggest that FIPMs are particularly relevant for flood risk management and policy evaluation at the municipal scale.
I have the following comments:
-
Make explicit throughout that only precipitation-induced flooding (not riverine or coastal) is analysed – currently this is missing in the title, abstract and introduction. At l. 51, consider whether riverine flooding is also relevant to your framing – if it is excluded, state so explicitly; if it is partially covered, clarify the scope.
-
The concluding statement about "importance for flood risk management frameworks" is not convincingly supported by 1–2% national-level effects. Either soften the claim to focus on local relevance (>10% in some municipalities), or strengthen it with additional argumentation.
-
Clarify the distinction between planned relocation (policy-driven resettlement) and spontaneous migration (individual household decisions) early in the introduction.
-
I suggest that Sections 3 onwards should be structured as subsections of the Methods section (Section 2) for consistency and clarity.
-
At l. 84, clarify whether the analysis is based on only two specific extreme events: if so, this is a significant scope limitation that should be stated clearly in both abstract and introduction. Alternatively, clarify whether the two events serve as empirical anchors for estimating movement assumptions used in the broader modelling.
-
l. 333 / Figure 4: Specify the unit clearly – "persons per [what]?" needs to be defined (per km², per municipality, per 1,000 inhabitants, etc.).
-
l. 345: Explain why flood damages are initially reduced under higher climate change scenarios – this is counterintuitive and needs explicit reasoning. Is it due to outmigration from high-risk zones reducing exposed population? Or overall population increase?
-
Figure 5: Add a brief explanatory note in the text clarifying what it means when flood damages increase due to population movement – presumably this reflects in-migration into higher-risk areas, but this should be stated directly.
-
l. 422: The strong general claim is not well-supported given the small national-level numbers and the nuanced discussion preceding it. Consider limiting the claim to the local/municipal scale, where the evidence is stronger.
Citation: https://doi.org/10.5194/egusphere-2025-5949-RC3 - AC3: 'Reply on RC3', Hayata Yanagihara, 03 May 2026
-
- AC2: 'Reply on RC2', Hayata Yanagihara, 03 May 2026
Peer review completion
Journal article(s) based on this preprint
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 1,464 | 772 | 185 | 2,421 | 313 | 134 | 177 |
- HTML: 1,464
- PDF: 772
- XML: 185
- Total: 2,421
- Supplement: 313
- BibTeX: 134
- EndNote: 177
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Hayata Yanagihara
So Kazama
Kei Gomi
Yusuke Hiraga
Atsuya Ikemoto
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(1555 KB) - Metadata XML
-
Supplement
(435 KB) - BibTeX
- EndNote
- Final revised paper
Flooding can influence population movements. However, most studies of future flood damage costs do not consider these movements. We examined how such movements may change future flood damage costs in Japan. National and prefectural effects were small, but some municipalities showed reductions of more than 10 % in these costs. These results show that considering population movements can improve future flood risk planning.
Flooding can influence population movements. However, most studies of future flood damage costs...
The authors present a study titled "Integrating flood-induced population movements into future flood damage estimates in Japan." The manuscript introduces a novel framework for quantifying and predicting population relocation in response to flood events, incorporating climate change effects, to evaluate future population distribution and flood risk in Japan. Overall, I find the study to be well-constructed and the manuscript clearly written. The integration of statistical causal inference with long-term land-use and population projections offers a valuable contribution to socio-hydrological modeling. However, I have identified several critical areas regarding spatial resolution, flood typology, and the modeling of relocation "pull" factors that warrant further attention. I am recommending major revisions to address these methodological limitations and scope concerns. Pending these revisions, I believe the manuscript would be a strong candidate for publication in Natural Hazards and Earth System Sciences (NHESS).