The manuscript “Human displacements from tropical cyclone Idai attributable to climate change” presents a modeling study that assesses the impacts of observed climate change on population displacement risk of tropical cyclone (TC) Idai in Mozambique, using a storyline approach. The work presents an interesting and innovative approach to assessing and attributing TC-related risks, using displacements as the response variable. However, the manuscript in its current form has a number of major limitations which inhibit drawing clear take-home messages and conclusions from it. The following issues need to be addressed to qualify for publication:

Abstract

• The abstract is rather short, focusing mainly on the methods used in the study rather than on its results. To make it more concrete, more detail on the results of the study would be useful.

Introduction

• The introduction section effectively describes the relevance of the work, but uses vague language such as “immense” (line 29); “extensive human suffering” (line 35); “intense storm surge” (line 58). Citing numbers from the relevant literature would considerably improve the introduction.
• In total, the literature review provided in this section is rather limited and could e.g. include the work of Bloemendaal et al. 2020; 2021; 2022.
• It is unclear to me whether “inland (freshwater) flooding” (line 73) refers to pluvial or fluvial flooding (or both?).
• The last sentence of the introduction states that the issues described in that paragraph (lines 79-89) will be picked up later again; however, I could not find the respective section.
• The end of the introduction would profit from a clear statement of the study aims and how they are achieved in the remainder of the paper. Clarification of the fact that this is a modeling study that uses observed displacements as reported in the GIDD as its only empirical evidence would be useful as well.

Methods

• The methods, particularly 2.1 and 2.2, provide rather short and technical descriptions of the modeling work. I would suggest including more detail and reasoning behind the modeling choices and assumptions made, supported by additional literature. Some of these details can be found in the results section (section 3.1), which I suggest moving into the methods section as methodological choices are explained here. More specifically, I was wondering about the following points (not exhaustive):
  • Lines 108-109: Why 10 % decrease?
  • Lines 122-124: Why these increases in sea levels?
  • Line 131: Why 9 arc sec resolution? What are the limitations when merging these datasets (which are known to have large offsets)?
  • Lines 137-139: How is it possible that agreement is highest under the maximum tide assumption, followed by the no tidal adjustment assumption? I would assume that agreement would be second best with the mean tidal assumption.
  • Lines 154-156: FloodScan: What is the resolution? What does “small flood sensitivity” mean?
  • Lines 159-162: The description of the RICorDE algorithm is unclear to me.
• Section 2.1: Would it be possible to validate the coastal flood modeling with observed floodplains derived from satellite imagery, similar to the work of Tellman et al. 2021 (who unfortunately do not cover TC Idai)?
• I am wondering why certain flood depth thresholds are used to calculate the displacement vulnerability ratios instead of determining this threshold from the data, by summing up the population per flood depth increment (starting from the highest depth) until the affected population equals the observed displaced population. In my understanding the minimum flood depth at which people start being displaced could then be derived. Subsequently, this flood depth could be applied to the counterfactual scenarios as well.
• I am generally concerned about conducting a local-scale study with global-scale datasets. Would it be possible to enrich the data and assumptions with local data (e.g. on population, elevation) and literature (e.g. of reported land subsidence, flood extents and depths) as well as with analysis of satellite imagery (e.g. to derive flood extents for validation)? While I am well aware of the data limitations in the study region, location-specific information may be available or, if lacking, should be discussed in more detail. If the use of global data is intentional (e.g. to ensure reproducibility in other regions), this should be discussed as well.
• The entire section has a shift in tenses. Especially from section 2.2 to 2.3, past tense shifts to present tense in most instances.

Results

• As pointed out above, section 3.1 does not fit in the results section as it provides further detail on the modeling assumptions. I therefore suggest moving this section to the methods.
• Line 298: what does p = 0.06 stand for?
• Section 3.2:
  • The first paragraph largely repeats what has already been stated in the methods. I therefore suggest (re)moving it (line 307-314).
  • I suggest extending lines 321-324, describing the results shown in Figure 3 in more detail, which – in my opinion – should make up the largest part of this section.
  • Figure 3: It would be useful to have a box in 3a that delineates the zoomed in areas in 3b-d. The color code in c-d is slightly counterintuitive as more intense colors reflect lower flood depths.
• Section 3.3:
  • The results description is difficult to follow. Might it be better to first describe the factual results, followed by the counterfactual results? Maybe a table would be useful, presenting the results for each SLR assumption, tidal assumption, impact threshold, and hazard (i.e. SLR, wind, both)?
  • It would also be insightful to be walked through the results that are presented in Figure 4. Also, it is unclear to me which change is presented in the bottom panel of Figure 4.
  • Lines 364-367: Is this a hypothetical statement?

Discussion & conclusion

• While section 4 provides an interesting discussion of technical aspects and limitations of the model assumptions, it lacks more in-depth discussion of the uncertainties stemming from the modeling approach and assumptions as well as the input data used, which are important to contextualize the results. More systematic discussion of these uncertainties would provide interesting insights.
• Furthermore, the discussion largely lacks reflection on the implications of the work in terms of research and policy-making. In the abstract, the authors state that the study “provides a blueprint for event-based displacement attribution” (line 27) which is not elaborated further. Also, the authors state that the storyline approach “raises risk awareness in a more tangible way” (line 455), which I fully agree with; however, reflections on how to use the results of this work to raise awareness (both in society and in policy-making) are missing. In my opinion, such reflections would add substantial value to the messages conveyed with this research.
• Reflections on needs and/or plans for future work would be useful (which also connects to my previous comment). This could include points such as (not exhaustive): Can/Will this approach be used in other contexts? Which aspects of the work can be improved in future work? How may future climate change affect displacement risk, considering both changes in climate as well as changes in socioeconomic conditions.

References

Bloemendaal et al. 2020: https://doi.org/10.1038/s41597-020-0381-2

Bloemendaal et al. 2021: https://doi.org/10.1088/1748-9326/abd131

Bloemendaal et al. 2022:  https://doi.org/10.1126/sciadv.abm8438

Tellman et al. 2021: https://doi.org/10.1038/s41586-021-03695-w

Overview

This study aims to model and identify the ‘excess’ population displacement triggered by tropical cyclone Idai that can be attributed to climate change. By utilising a ‘storyline approach’ the study compares actual recorded displacement against estimated levels of displacement derived from counterfactual scenarios of mean sea level and maximum wind conditions without contributions of climate change. The conclusion of the study is that the impact of climate change has increased displacement risk by between 3.1 to 3.5%, corresponding to 16,000 – 17,000 displaced people.

The approach and results are interesting and of value in exploring the impact of climate change on environmental shocks and stressors, and the resulting affect this will have on patterns of human mobility. As such, the study is worthy of publication. The paper itself, however, requires major revisions prior to publication, both in terms of writing and structure, as well as locating the research in the wider landscape of work exploring the climate-migration nexus, and how this study contributes to the wider discussions.

The key points for the authors:

1. The paper would be improved from restructuring and a careful revision of language employed. Much of the information that is presented in the ‘results’ section details the ‘methodology’ in constructing the study. The results are not clearly detailed, with confusing language, for example, around ‘increases’ in displacement between differing counterfactual scenarios, despite these describing lower levels of displacement from those reported ‘factually’. More generally, the authors employ ambiguous terms such as ‘massive’ or ‘immense’ which are better avoided.
2. The paper refers to ‘displacement’ continually without definition of the forms of human mobility this refers to. It is understood that this definition drawn from GIDD IDMC data, and therefore captures a wide range of mobility in response to rapid-onset events, but the definition employed by IDMC (or lack thereof) should be acknowledged.
3. Such approaches are necessarily constructed on a number of assumptions, particularly in this case the relationship between the intensity of impact of the tropical cyclone and numbers of people displaced. In reality, the relationship between environmental shocks and stressors and human mobility is not direct but rather influenced by pre-existing socio-political-economic conditions. There are also significant uncertainties associated with the IDMC data on which the study is based, as these figures are drawn from secondary sources. Given the validity of approach, the authors can afford to be clearer in outlining the assumptions on which the study is based.
4. The paper does not detail the further implications of the findings, for example in terms of adaptation planning for exposed communities, nor specific priorities for future research. Again, this study would benefit significantly from being grounded in a wider debate of attributing shifts in human mobility to the impacts of a changing climate.

More detailed comments are given by line number below.

Abstract

16:       ‘massive’ is a strange adjective to use, particularly in the first sentence.

17:       ‘frequency and intensity of such events is affected by anthropogenic climate change’ - statement seems to presuppose the subject of the paper.

21-22:   Terms used to describe methodology of ‘storm surge modelling’ could be detailed more clearly. It is worth defining the actual approach and methodology employed.

25:       Are these percentages compared to number of people displaced, or of exposed populations?

Introduction

• Does not refer to the potential impact of climate change on TC storm track. This isn’t something that is necessary for inclusion in the study itself, but perhaps should be referred to in context. This is highlighted in the Discussions/Conclusions, but can be stated here as well.

29:       Use of term ‘immense’ is ambiguous - is there a clearer alternative?

32-33:   ‘massive damages to housing and infrastructure’ is again ambiguous. Is there a quantifiable range, or example time period?

33-34:   Are there additional comparative sources for the estimates of those displaced? This and the Desai et al. (2021) reference are drawing on IDMC reporting, so would benefit from other source corroborating or providing range. Is there an IPCC figure?

A definition of what forms of human mobility are considered ‘displacement’ in the context of this study is required. It is noted that in section ‘2.4 Displacement’ the GIDD source (IDMC, 2022) is described as lacking granularity. The definition of displacement used by IDMC, however, should be outlined.

42-43: Emmanuel, 1987 seems like quite an ‘old’ reference. Following papers relevant for increasing TC intensity potentially:

• Emanuel, K. (2005). Increasing destructiveness of tropical cyclones over the past 30 years. Nature, 436(7051), 686–688. https://doi.org/10.1038/nature03906
• Emanuel, K. A. (2013). Downscaling CMIP5 climate models shows increased tropical cyclone activity over the 21st century. Proceedings of the National Academy of Sciences, 110(30), 12219– https://doi.org/10.1073/pnas.1301293110

46-48: ‘substantially’ unnecessary if IPCC figures are included in parenthesis - better to just cite the IPCC reference

Methods

Coastal Flood Modelling

108-110: Why does the study focus on an equivalent 10% increase in intensity due to changing climate?

“For the counterfactual scenarios with modified TC intensity, we 107 multiplied all wind speed values along the track by a scalar factor of 0.9 (for a decrease of 108 10% in intensity). The central pressure at each track position is increased by 0.1 times the 109 difference between central pressure and environmental pressure.” (Mester et al., 2023)

The Knutson et al. (2020) paper cited suggests that the “…Fifteen individual scaled global estimates are all positive, with a mean and median increase (range) of about 5% (1%–10%)” (Knutson et al., 2020). Therefore the 10% difference that this study ascribes to climate change is at the very upper end of this boundary.

The explanation of this is outlined in ‘Results’ section, lines 284-301. These sections should perhaps be combined so the explanation / justification of this are presented together. It is noted - as highlighted in the paper - that the 10% change is at the upper estimate of other studies. Should a number of model runs at different degrees of change be presented?

Displacement

195 onwards: Necessarily, the relationship between the impact of the TC and displacement is artificially deterministic. Within the context of this model, however, this approach seems reasonable.

Results

Counterfactuals

242 onwards: This section seems to be more focused on methodology rather than results. Can it be relocated within the paper?

323: “Notably, in a world without climate change, the area inundated by 100 cm or more 323 is dramatically reduced.” Although this is demonstrated in comparison of fig.3 (c) and (d), it seems that the areas where there is most significant difference in inundation (e.g. western bank of main flood valley) are those that are least densely populated according to fig.3 (b). This perhaps should be highlighted?

Can the area described in figs.3 (b-d) be outlined on fig.3 (a) alongside that of wider ‘area of interest’ so it is easier to see the relationship between (a) and (b-d)?

Displacement

337-344: The finding that it is a combination of TC intensity and tide assumptions that have a more significant impact on displacement in comparison to SLR is notable and could be highlighted in the Abstract

340 onwards: The results described are not always clear and therefore can perhaps be better expressed.

The baseline for the study is the ‘factual’ displacement of 478,000 people. The counter factual scenarios derived from the model suggest that fewer people would be displaced without contribution of climate change.

Although the line "The coupled effect of higher wind speeds and higher sea level increases the number of affected people and displacements by up to 43,300 and 16,500 (maximum tide) 350 and 44,300 and 17,100 (monthly mean), respectively.” is accurate in terms of relative difference to other counterfactuals, it is potentially unclear as it refers to an ‘increases in the number of affected people and displacements’.

For consistency, it may be clearer to use ‘relative change’ as the base metric throughout, as described in the ‘bottom’ panel of fig.4. This would clarify the relative differences between the scenarios in relation to the ‘factual’ baseline.

Conclusion

387: The conclusion focuses on ‘about 17,000’ additional displacements. This should be given as a range reflecting the different tidal assumptions.

392 onwards: Points regarding the potential impacts of inland flooding and high wind speeds on displacement are noted and provide useful context for the limitations of the study.

What is lacking, however, is an acknowledgement of the complexity of the relationship between environmental impacts, pre-existing socio-economic conditions and displacement. This is particularly important as what forms of human mobility constitute ‘displacement’ are not defined within the paper. Although the assumptions on which the relationship between the impact of TC and displacement are valid in the context of this study, the assumptions themselves should be acknowledged. For reference:

• Cattaneo, C., Beine, M., Fröhlich, C. J., Kniveton, D., Martinez-Zarzoso, I., Mastrorillo, M., Millock, K., Piguet, E., & Schraven, B. (2019). Human Migration in the Era of Climate Change. Review of Environmental Economics and Policy, 13(2), 189–206. https://doi.org/10.1093/reep/rez008
• Foresight: Migration and Global Environmental Change (Final Project Report, p. 234). (2011). The Government Office for Science. https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/287717/11-1116-migration-and-global-environmental-change.pdf

There are also significant uncertainties and limitation associated with the methodology of deriving and estimating ‘factual’ displacement figures by IDMC and these should also be acknowledged and discussed.

Although in a later section (433-441) socio-economic conditions are referred to in terms of a potential change in vulnerability, this is separate point relating to future changes with respect to exposure indices.