Brief communication: How well are Londoners prepared for flooding? An assessment of risk mitigation actions and a pilot comparison of perceived and modelled flood likelihood

Russell, Andrew; Khosa, Prarthna Kaur

doi:10.5194/egusphere-2026-3820

Preprints

https://doi.org/10.5194/egusphere-2026-3820

Preprints

14 Jul 2026

| 14 Jul 2026

Brief communication: How well are Londoners prepared for flooding? An assessment of risk mitigation actions and a pilot comparison of perceived and modelled flood likelihood

Andrew Russell and Prarthna Kaur Khosa

Abstract. Surface water flooding probability in London is high but little is known about how aware Londoners are of their exposure. 183 online survey responses showed large differences between perceived flood likelihood and modelled flood probability for respondents’ homes (where the latter could be confidently determined - full addresses cannot be collected due to ethical constraints, which limited sample size and representativeness): 54.6 % overestimated their flood likelihood, 19.7 % underestimated, and 25.7 % were accurate. Residents in high flood probability locations underestimated their exposure. Self-reported flood probability knowledge increased perception accuracy. Climate change concern was linked to flood probability overestimation and implementation of risk mitigation actions.

Received: 28 Jun 2026 – Discussion started: 14 Jul 2026

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.

Download & links

Preprint (PDF, 527 KB)

Supplement (79 KB)

Download & links

Andrew Russell and Prarthna Kaur Khosa

Status: final response (author comments only)

RC1: 'Comment on egusphere-2026-3820', Milad Basirifard, 14 Jul 2026

The principal percentages appear to be mathematically incorrect. This is the most serious problem. Table 1 states that the comparison sample contains 183 respondents, including 30 who answered “unsure.” Table 2 nevertheless also totals 183 observations while claiming to classify respondents as underestimating, accurately estimating, or overestimating flood probability. The reported counts are 36 underestimated, 47 accurate, and 100 overestimated, which sum to 183. An “unsure” response cannot logically be assigned an ordered difference of one or more probability bands. The numbers strongly suggest that the 30 “unsure” respondents may have been added to the underestimation group: Table 1 implies only approximately six genuine underestimators, while Table 2 reports 36. Moreover, the diagonal percentages in Table 1 appear to produce approximately 45 accurate responses, not 47.
Respondents were asked about the general flood risk level where they live, without specifying the source of flooding. However, the authors compare these answers specifically against surface-water flood probability, selected because it was reportedly the highest modelled probability for the retained addresses.
The government tool requires a full address because flood probability can vary among properties within a postcode. The authors did not collect addresses and instead retained postcodes in which at least 75% of addresses shared a flood category. This does not establish the flood category of a particular respondent’s property. Up to 25% of addresses may still belong to another category, and no evidence is presented showing the expected individual-level classification error. The 75% threshold is arbitrary, and there is no sensitivity analysis using thresholds such as 80%, 90%, or 100%.
The manuscript highlights that all respondents in the medium- or high-probability categories underestimated their exposure. Yet this claim is based on one medium-probability and four high-probability respondents. Five observations cannot support statements about residents in high flood probability locations, as presented in the abstract. The uncertainty is enormous, and the finding could easily be a sampling accident. It should not appear as a principal conclusion. At most, it may be reported as an anecdotal pilot observation requiring confirmation in a risk-stratified sample.

Citation: https://doi.org/10.5194/egusphere-2026-3820-RC1

AC3: 'Reply on RC1', Andrew Russell, 01 Oct 2026

AR: Thank you for the thorough and insightful review - it highlighted some important issues that I am confident we have addressed in these responses, which has significantly improved the quality of the manuscript. Responses are provided in a point-to-point fashion in italics after the relevant comments.

1. The principal percentages appear to be mathematically incorrect. This is the most serious problem. Table 1 states that the comparison sample contains 183 respondents, including 30 who answered “unsure.” Table 2 nevertheless also totals 183 observations while claiming to classify respondents as underestimating, accurately estimating, or overestimating flood probability. The reported counts are 36 underestimated, 47 accurate, and 100 overestimated, which sum to 183. An “unsure” response cannot logically be assigned an ordered difference of one or more probability bands. The numbers strongly suggest that the 30 “unsure” respondents may have been added to the underestimation group: Table 1 implies only approximately six genuine underestimators, while Table 2 reports 36. Moreover, the diagonal percentages in Table 1 appear to produce approximately 45 accurate responses, not 47.

Thank you for this detailed comment that makes more than one important point:

1) Yes, I see that some of the data and/or calculations in Tables 1 and 2 are incorrect. (With 2 authors working on the same document and bringing data across from different spreadsheets, we got some lines mixed up: we apologise and thank the reviewer for their careful review of the numbers.) For Table 1, some of the numbers in the top row were incorrect and these have been corrected. To clarify things, we’ve also added the absolute values to the main part of the table, in addition to the percentages, so that it’s easier to see what the complete situation is. Note that all the numbers in the main part of the table were correct; it was the top row that was wrong.
2) For Table 2, agreed: we made a mistake in how to analyse the “unsure” answers. We coded all the responses numerically to be able to calculate the over- or under-estimation but, in doing this, we brought the “unsure” responses into the calculation when they should have been excluded. Whilst this was clearly a mistake and has been corrected, the act of correcting it re-iterates the main conclusion: that the overestimation of flood risk in this cohort was a bigger problem than underestimation.
The updated tables are included here:

Modelled\Perceived	Unsure (27; 14.8%)	Very low (40; 21.9%)	Low (83; 45.3%)	Medium (29; 15.9%)	High (4; 2.2%)
Very low (167; 91.3%)	25 (15.0%)	38 (22.8%)	˄ 73 (43.7%) ˄	˄ 27 (16.2%) ˄	˄ 4 (2.4%) ˄
Low (11; 6.0%)	2 (18.2%)	˅ 1 (9.1%) ˅	7 (63.6%)	˄ 1 (9.1%) ˄	˄ 0 (0.0%) ˄
Medium (1; 0.5%)	0 (0.0%)	˅ 0 (0.0%) ˅	˅ 1 (100.0%) ˅	0 (0.0%)	˄ 0 (0.0%) ˄
High (4; 2.2%)	0 (0.0%)	˅ 1 (25.0%) ˅	˅ 2 (50.0%) ˅	˅ 1 (25.0%) ˅	0 (0.0%)

Unsure	Underestimated by 2+ probability levels	Underestimated by 1 probability level	Accurate	Overestimated by 1 probability level	Overestimated by 2+ probability levels
27 (14.8%)	3 (1.6%)	3 (1.6%)	45 (24.6%)	74 (40.4%)	31 (16.9%)

2. Respondents were asked about the general flood risk level where they live, without specifying the source of flooding. However, the authors compare these answers specifically against surface-water flood probability, selected because it was reportedly the highest modelled probability for the retained addresses.

That’s correct. Our assumption was that the vast majority of London residents would not distinguish, or understand the difference, between different types of flooding in the context of the survey. Therefore, raising the issue of different flood sources would require a detailed explanation, and potentially branches of questions for probability estimates for different types of flooding, for minimal benefit to the method applied. Instead, given that SWF is the overwhelmingly dominant source of flood risk in London, and that this type of flooding is poorly understood – only 20% of the UK public have a good understanding of the term SWF (https://policycommons.net/artifacts/3330882/surface-water-flooding-social-research/4129734/fragments/) – we maintain that this was a reasonable assumption and the best way of answering the research question. As a point of clarification: we looked at the likelihood of all sources of flooding from the gov.uk tool but SWF always had the highest likelihood for this cohort so that was used in the analysis. We will revise the description of the method to reflect this detail more clearly.

3. The government tool requires a full address because flood probability can vary among properties within a postcode. The authors did not collect addresses and instead retained postcodes in which at least 75% of addresses shared a flood category. This does not establish the flood category of a particular respondent’s property. Up to 25% of addresses may still belong to another category, and no evidence is presented showing the expected individual-level classification error. The 75% threshold is arbitrary, and there is no sensitivity analysis using thresholds such as 80%, 90%, or 100%.

We agree that this is significant issue with this type of study and our 75% threshold approach was an imperfect solution to the problem of restrictions on the data that could be collected. Nonetheless, it was a solution that was implementable under the ethical restrictions under which the research could be conducted i.e. participants must not be identifiable and collecting the full address would breach that condition.
We have now conducted a sensitivity analysis, as suggested and these are the results:
- 66.7% of the 183 postcodes had a 100% agreement in flood risk for all addresses; a further 21.3% had 90%-99% agreement; then another 10.9% in the 80%-89% agreement band; and only 1.1% in the 75%-79%.
This is confirmation that we can have high-to-very-high confidence that our comparison of perceived to modelled flood risk is comparing the right data even with the 75% threshold.
Furthermore, the distribution of overestimated, underestimated and accurate responses is almost identical for the 4 thresholds tested:

Cut-off	People kept in the study	% who overestimated	% accurate	% underestimated
75%	100%	58%	26%	2%
80%	98.9%	58%	25%	2%
90%	77.6%	57%	25%	2%
100%	66.7%	55%	27%	2%

However, the broader point that the paper was trying to address – that residents of a major European city have poor understanding of their flood risk – is still addressed despite the issue. Indeed, we are investigating ways in which the research can be undertaken without collecting addresses (or collecting addresses under much more restrictive conditions), which would be a more time consuming and more expensive exercise; these pilot results go a significant way to justifying that time and cost investment.

4. The manuscript highlights that all respondents in the medium- or high-probability categories underestimated their exposure. Yet this claim is based on one medium-probability and four high-probability respondents. Five observations cannot support statements about residents in high flood probability locations, as presented in the abstract. The uncertainty is enormous, and the finding could easily be a sampling accident. It should not appear as a principal conclusion. At most, it may be reported as an anecdotal pilot observation requiring confirmation in a risk-stratified sample.

Agreed, we will remove the finding from the abstract and discuss it as a pilot result that deserves more investigation (which is our intention).

Citation: https://doi.org/10.5194/egusphere-2026-3820-AC3

RC2:
'Comment on egusphere-2026-3820', John Drury, 02 Aug 2026
This paper reports a study comparing Londoners’ perception of flood risk in their own area with estimated flood risk in their area. This is an important topic, as there is previous evidence from a range of other threats that members of the public often underestimate risk likelihood. This has consequences for preparedness and interpretation of threat signals. I am less familiar with the specific literature around perceptions of flood risk, but the authors suggest that this may be mixed, which may be due to changing perceptions of climate crisis. For this reason too, the present study has a strong rationale. The authors’ strategy of comparing participant ratings to existing estimates by postcode seems a reasonable one.
Other than for high-risk areas, the finding of over-estimation (in low risk areas) contrasts with findings on risk in the broader literature on public perceptions of hazard risk, where people often believe ‘it won’t happen’ (or ‘it won’t happen to me’). For example:
Atwood, L. E., & Major, A. M. (2000). Optimism, pessimism, and communication behavior in response to an earthquake prediction. Public Understanding of Science, 9(4), 417–432. https://doi.org/10.1088/0963-6625/9/4/305
And
Kinsey, M. J., Gwynne, S. M. V., Kuligowski, E. D., & Kinateder, M. (2019). Cognitive biases within decision making during fire evacuations. Fire Technology, 55, 465–485. https://doi.org/10.1007/s10694-018-0708-0
As the authors acknowledge, however, the sample is small and not necessarily representative of London (let alone other locations). Therefore it seems correct to treat the study as a pilot.

Specific points.

P 2 ‘Given the particular hazard and vulnerability profile in London’
Explain what that is.

Method

3 Explain why the sample target was 500. Also this seems to be a total recruitment. What was the useable target for analysis, if any?

Report demographics – ages, gender, ethnicity.

I am not sure that a representative sample is a matter of (large) numbers as the authors imply. Approaching representativeness means stratifying certain demographics. If the larger sample doesn’t do this, it’s still not representative.

Technical corrections

The abstract should not include the point about full addresses not being collected – leave this detail to the main body

I am not sure why method limitations are in the method section. I would expect to see this in the discussion, not here.
Citation: https://doi.org/10.5194/egusphere-2026-3820-RC2
- AC2:
  'Reply on RC2', Andrew Russell, 30 Sep 2026
  AR: responses to the (very useful!) review comments have been added in a point-to-point fashion in italics after each comment.
  This paper reports a study comparing Londoners’ perception of flood risk in their own area with estimated flood risk in their area. This is an important topic, as there is previous evidence from a range of other threats that members of the public often underestimate risk likelihood. This has consequences for preparedness and interpretation of threat signals. I am less familiar with the specific literature around perceptions of flood risk, but the authors suggest that this may be mixed, which may be due to changing perceptions of climate crisis. For this reason too, the present study has a strong rationale. The authors’ strategy of comparing participant ratings to existing estimates by postcode seems a reasonable one.
  Other than for high-risk areas, the finding of over-estimation (in low risk areas) contrasts with findings on risk in the broader literature on public perceptions of hazard risk, where people often believe ‘it won’t happen’ (or ‘it won’t happen to me’). For example:
  Atwood, L. E., & Major, A. M. (2000). Optimism, pessimism, and communication behavior in response to an earthquake prediction. Public Understanding of Science, 9(4), 417–432. https://doi.org/10.1088/0963-6625/9/4/305
  And
  Kinsey, M. J., Gwynne, S. M. V., Kuligowski, E. D., & Kinateder, M. (2019). Cognitive biases within decision making during fire evacuations. Fire Technology, 55, 465–485. https://doi.org/10.1007/s10694-018-0708-0
  We will add these references.
  
  As the authors acknowledge, however, the sample is small and not necessarily representative of London (let alone other locations). Therefore it seems correct to treat the study as a pilot.
  
  Specific points.
  P 2 ‘Given the particular hazard and vulnerability profile in London’
  Explain what that is.
  Will revise this to address the comment; it was intended to refer to the previous discussion but this is not clear.
  
  Method
  Explain why the sample target was 500. Also this seems to be a total recruitment. What was the useable target for analysis, if any?
  With the target of 500 we hoped to get 385 responses where the flood risk could be identified to a certain level of confidence (75%) from the government flood risk checker. As this type of had not been done before, we underestimated the number of responses required (this is an interesting result in itself!) 385 was chosen as it could be a representative sample of the overall population of the London postal area (depending on demographics).
  
  Report demographics – ages, gender, ethnicity.
  Will revise this to address the comment, although ethnicity data was not collected. In short:
  18-24 - 7.2%; 25-34 - 35.4%; 35-44 - 27.7%; 45-54 - 16.4%; 55-64 - 10.0%; 65+ - 3.2%
  
  Female - 58.0%; Male - 40.1%; Non-binary - 0.6%; Prefer not to say - 1.3%
  
  I am not sure that a representative sample is a matter of (large) numbers as the authors imply. Approaching representativeness means stratifying certain demographics. If the larger sample doesn’t do this, it’s still not representative.
  Agreed, the sample size quoted is necessary for a representative sample but not sufficient on its own. We will revise the language and also note that this was intended as a pilot study to test the method.
  
  Technical corrections
  The abstract should not include the point about full addresses not being collected – leave this detail to the main body
  Will revise this to address the comment.
  
  I am not sure why method limitations are in the method section. I would expect to see this in the discussion, not here.
  Will revise this to address the comment.
  
  Citation: https://doi.org/10.5194/egusphere-2026-3820-AC2
RC3:
'Comment on egusphere-2026-3820', Anonymous Referee #3, 07 Aug 2026

Thank you for the opportunity to review this manuscript. The manuscript presents a brief assessment of how London residents perceive their flood probability and how these perceptions compare with modelled flood probability. Based on an online survey of 469 respondents, the authors investigate the relationship between perceived and modelled flood likelihood, self-reported knowledge of flood risk, climate change concern, previous flood experience, and household preparedness actions. Then, a subset of 183 respondents is used for the comparison between perceived and modelled flood probability. Overall, I found the manuscript interesting and relevant to the field of flood risk management. I particularly appreciate the attempt to compare perceived flood probability with modelled probability, as this provides a useful perspective on the potential mismatch between general awareness of flooding and individuals' understanding of their own exposure. The manuscript also provides some interesting observations regarding the relationship between perceived risk, climate change concern, and preparedness actions.
I believe the manuscript is suitable for publication after addressing a few comments, mainly concerning the interpretation of some results and the clarification of several concepts used in the survey. Given that this is a brief communication, I do not think that extensive additional analyses are necessary. Rather, I think that some statements could be made more cautious and that some aspects of the interpretation could be clarified.
Overall comment
The manuscript provides a useful observation regarding the mismatch between perceived and modelled flood probability. However, I think the manuscript would benefit from a clearer explanation of the practical implications of this mismatch for flood risk management. For example, could the findings suggest a need for more spatially targeted risk communication, particularly for residents who may underestimate their flood probability? A brief discussion of how the observed mismatch could inform risk communication or preparedness interventions would help clarify the practical relevance of the study.
Specific comments
Page 1, Lines 19–22. The statement that "high property prices have led to many basements being converted into flats (apartments) in London" was not completely clear to me. I initially found it difficult to understand the causal relationship implied by this sentence. Do the authors mean that high property prices have created incentives to convert basement spaces into residential units, thereby increasing the number of people and properties potentially exposed to basement flooding? If so, I think this could be stated more explicitly. A brief clarification would help readers understand why this particular characteristic of London's housing market is relevant to flood vulnerability.
Page 1, Lines 28–31. Since the central focus of the manuscript is the perception of flood probability and its relationship with preparedness, I think it would be useful to briefly introduce here the main factors identified in the literature as influencing flood-risk perception and subsequent mitigation behaviour. The authors already cite Taylor et al. (2014) and Lechowska (2022) in this context. A short sentence summarising the most relevant factors would help establish the conceptual background for the subsequent analysis, particularly Section 3.3 on preparedness actions. This does not need to become an extensive literature review, given the brief communication format, but a slightly clearer link between the existing literature and the variables examined in the survey would strengthen the rationale of the study.
Page 4, Lines 99–104. I appreciate the methodological difficulty of estimating modelled flood probability without collecting full addresses, particularly given the ethical constraints described by the authors. I would nevertheless appreciate a brief justification for the 75% threshold used to retain respondents, whereby at least 75% of addresses within a postcode had to share the same flood-probability category. Why was this threshold considered sufficient to confidently assign the modelled flood probability to a respondent? I understand that this may have been selected as a pragmatic criterion to balance sample size and confidence in the assigned probability, but a brief explanation of the rationale for this threshold would help clarify this methodological choice.
Page 8, Lines 215–220. I found the section concerning self-reported knowledge and flood-probability estimation accuracy particularly interesting, but also one of the weaker parts of the interpretation. The manuscript states that accuracy increased from 16.7% among respondents who felt "not at all informed" to 50.0% among those who felt "very well informed". However, the latter category contains only six respondents. The subsequent statement that "this result implies" that greater outreach could potentially improve the low accuracy observed overall seems too strong given this small sample. I would suggest presenting this as a possible trend rather than as evidence supporting the effectiveness of greater outreach. For example, the authors could state that the result "suggests a possible trend" or that it "should be interpreted cautiously given the small number of respondents in the 'very well informed' category." This would make the interpretation more consistent with the evidence presented.
Page 8, Lines 220–221. The authors report no association between previous flood experience and flood-risk preparedness. I found this result interesting because the Introduction identifies previous flood experience as an important indicator of risk-management actions, based on the literature cited by the authors. However, the full cohort included only four respondents with direct experience of flooding and 20 with indirect experience. Therefore, I think this finding should be interpreted very cautiously. In particular, the absence of an identified association in this sample should not necessarily be interpreted as evidence that previous flood experience is unrelated to preparedness. It may instead reflect the very small number of respondents with previous flood experience. A brief clarification of this point would strengthen the discussion.
Appendix, Section A (Climate change concern, Page 9, Lines 260–266). I found the categories "Very concerned", "Fairly concerned", "Not very concerned", and "Not at all concerned" somewhat difficult to interpret because the manuscript does not indicate whether these categories were associated with any specific definitions or were intentionally based on respondents' subjective assessment. If the latter was the intention, I think this could simply be clarified in the description of the survey. Since climate-change concern is subsequently considered in relation to flood-probability perception and preparedness actions, it would also be useful to briefly acknowledge that these self-reported categories may involve some degree of subjectivity or response heterogeneity. I would not necessarily consider this a major limitation, but a brief clarification would help readers better understand how this variable was interpreted.
Appendix, Section B (Page 10, Lines 270–274). The survey includes the response category "Yes, indirectly (nearby area)" when asking about previous flood experience. I think it would be useful to briefly clarify what the authors intended by "indirectly affected", or indicate whether this was left to the respondent's own interpretation. This would help readers better understand how this category was interpreted when discussing the role of previous flood experience.
I hope these comments are helpful in further improving the clarity and quality of the manuscript. I found the study interesting and relevant, and I appreciate the authors' work on this topic.

Citation: https://doi.org/10.5194/egusphere-2026-3820-RC3
- AC1:
  'Reply on RC3', Andrew Russell, 30 Sep 2026
  AR: I have responded to each point below in italics. This is another excellent review that makes some pertinent points - thanks!
  Thank you for the opportunity to review this manuscript. The manuscript presents a brief assessment of how London residents perceive their flood probability and how these perceptions compare with modelled flood probability. Based on an online survey of 469 respondents, the authors investigate the relationship between perceived and modelled flood likelihood, self-reported knowledge of flood risk, climate change concern, previous flood experience, and household preparedness actions. Then, a subset of 183 respondents is used for the comparison between perceived and modelled flood probability. Overall, I found the manuscript interesting and relevant to the field of flood risk management. I particularly appreciate the attempt to compare perceived flood probability with modelled probability, as this provides a useful perspective on the potential mismatch between general awareness of flooding and individuals' understanding of their own exposure. The manuscript also provides some interesting observations regarding the relationship between perceived risk, climate change concern, and preparedness actions.
  I believe the manuscript is suitable for publication after addressing a few comments, mainly concerning the interpretation of some results and the clarification of several concepts used in the survey. Given that this is a brief communication, I do not think that extensive additional analyses are necessary. Rather, I think that some statements could be made more cautious and that some aspects of the interpretation could be clarified.
  Overall comment
  The manuscript provides a useful observation regarding the mismatch between perceived and modelled flood probability. However, I think the manuscript would benefit from a clearer explanation of the practical implications of this mismatch for flood risk management. For example, could the findings suggest a need for more spatially targeted risk communication, particularly for residents who may underestimate their flood probability? A brief discussion of how the observed mismatch could inform risk communication or preparedness interventions would help clarify the practical relevance of the study.
  Agreed that this would strengthen the argument. This will be added in a revised manuscript.
  
  Specific comments
  Page 1, Lines 19–22. The statement that "high property prices have led to many basements being converted into flats (apartments) in London" was not completely clear to me. I initially found it difficult to understand the causal relationship implied by this sentence. Do the authors mean that high property prices have created incentives to convert basement spaces into residential units, thereby increasing the number of people and properties potentially exposed to basement flooding? If so, I think this could be stated more explicitly. A brief clarification would help readers understand why this particular characteristic of London's housing market is relevant to flood vulnerability.
  Yes, the high potential return on basement conversions incentivises this behaviour. I will clarify this and add a reference – Russell and Jones (2026) – where this is discussed.
  
  Page 1, Lines 28–31. Since the central focus of the manuscript is the perception of flood probability and its relationship with preparedness, I think it would be useful to briefly introduce here the main factors identified in the literature as influencing flood-risk perception and subsequent mitigation behaviour. The authors already cite Taylor et al. (2014) and Lechowska (2022) in this context. A short sentence summarising the most relevant factors would help establish the conceptual background for the subsequent analysis, particularly Section 3.3 on preparedness actions. This does not need to become an extensive literature review, given the brief communication format, but a slightly clearer link between the existing literature and the variables examined in the survey would strengthen the rationale of the study.
  This is an excellent point and we will add a new sentence summarising the link between perception and mitigation response.
  
  Page 4, Lines 99–104. I appreciate the methodological difficulty of estimating modelled flood probability without collecting full addresses, particularly given the ethical constraints described by the authors. I would nevertheless appreciate a brief justification for the 75% threshold used to retain respondents, whereby at least 75% of addresses within a postcode had to share the same flood-probability category. Why was this threshold considered sufficient to confidently assign the modelled flood probability to a respondent? I understand that this may have been selected as a pragmatic criterion to balance sample size and confidence in the assigned probability, but a brief explanation of the rationale for this threshold would help clarify this methodological choice.
  Please see the response to reviewer 1 where this is responded to in detail.
  
  Page 8, Lines 215–220. I found the section concerning self-reported knowledge and flood-probability estimation accuracy particularly interesting, but also one of the weaker parts of the interpretation. The manuscript states that accuracy increased from 16.7% among respondents who felt "not at all informed" to 50.0% among those who felt "very well informed". However, the latter category contains only six respondents. The subsequent statement that "this result implies" that greater outreach could potentially improve the low accuracy observed overall seems too strong given this small sample. I would suggest presenting this as a possible trend rather than as evidence supporting the effectiveness of greater outreach. For example, the authors could state that the result "suggests a possible trend" or that it "should be interpreted cautiously given the small number of respondents in the 'very well informed' category." This would make the interpretation more consistent with the evidence presented.
  I will revise this statement in line with the comment here to more accurately reflect the small sample size.
  
  Page 8, Lines 220–221. The authors report no association between previous flood experience and flood-risk preparedness. I found this result interesting because the Introduction identifies previous flood experience as an important indicator of risk-management actions, based on the literature cited by the authors. However, the full cohort included only four respondents with direct experience of flooding and 20 with indirect experience. Therefore, I think this finding should be interpreted very cautiously. In particular, the absence of an identified association in this sample should not necessarily be interpreted as evidence that previous flood experience is unrelated to preparedness. It may instead reflect the very small number of respondents with previous flood experience. A brief clarification of this point would strengthen the discussion.
  I will revise this statement in line with the comment here to more accurately reflect the small number of respondents who reported previous experiences with flooding.
  
  Appendix, Section A (Climate change concern, Page 9, Lines 260–266). I found the categories "Very concerned", "Fairly concerned", "Not very concerned", and "Not at all concerned" somewhat difficult to interpret because the manuscript does not indicate whether these categories were associated with any specific definitions or were intentionally based on respondents' subjective assessment. If the latter was the intention, I think this could simply be clarified in the description of the survey. Since climate-change concern is subsequently considered in relation to flood-probability perception and preparedness actions, it would also be useful to briefly acknowledge that these self-reported categories may involve some degree of subjectivity or response heterogeneity. I would not necessarily consider this a major limitation, but a brief clarification would help readers better understand how this variable was interpreted.
  I will add a statement to clarify the subjectivity involved in the scale of concern. The scale was used as it replicates the DESNZ Public Attitudes Tracker, which asks questions about climate change with this same subjective scale and, therefore, would allow us to compare results in a longer analysis e.g. https://www.gov.uk/government/statistics/desnz-public-attitudes-tracker-spring-2026/desnz-public-attitudes-tracker-net-zero-and-climate-change-spring-2026-uk
  
  Appendix, Section B (Page 10, Lines 270–274). The survey includes the response category "Yes, indirectly (nearby area)" when asking about previous flood experience. I think it would be useful to briefly clarify what the authors intended by "indirectly affected", or indicate whether this was left to the respondent's own interpretation. This would help readers better understand how this category was interpreted when discussing the role of previous flood experience.
  If respondents selected "Yes, indirectly (nearby area)" for Q7 then Q8 would be available to them to select what they mean or to write free text with more details. This was selected by 25 of the 469 respondents but we didn't analyses these data in this paper.
  
  I hope these comments are helpful in further improving the clarity and quality of the manuscript. I found the study interesting and relevant, and I appreciate the authors' work on this topic.
  
  Citation: https://doi.org/10.5194/egusphere-2026-3820-AC1

Andrew Russell and Prarthna Kaur Khosa

Supplement

https://doi.org/10.5194/egusphere-2026-3820-supplement

Andrew Russell and Prarthna Kaur Khosa

Viewed

Total article views: 377 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	Supplement	BibTeX	EndNote
247	91	39	377	56	35	30

HTML: 247
PDF: 91
XML: 39
Total: 377
Supplement: 56
BibTeX: 35
EndNote: 30

Views and downloads (calculated since 14 Jul 2026)

Month	HTML	PDF	XML	Total
Jul 2026	126	50	11	187
Aug 2026	41	10	2	53
Sep 2026	40	16	8	64
Oct 2026	40	15	18	73

Cumulative views and downloads (calculated since 14 Jul 2026)

Month	HTML	PDF	XML	Total
Jul 2026	126	50	11	187
Aug 2026	41	10	2	53
Sep 2026	40	16	8	64
Oct 2026	40	15	18	73

Viewed (geographical distribution)

Total article views: 306 (including HTML, PDF, and XML) Thereof 306 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 24 Oct 2026

Short summary

As our climate changes and flooding becomes more likely, particularly in big cities, we need to know how accurately people assess the probability of their homes flooding and what actions they are taking to respond to that. In this research, we surveyed people in London to work out how well their assessments of flood potential align with official assessments and what people are doing about it. We find that most people overestimate their exposure but those at most risk underestimate it.


Total:	0
HTML:	0
PDF:	0
XML:	0