the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 27 Oct 2025
Signals Without Action: A Value Chain Analysis of Luxembourg’s 2021 Flood Disaster
Abstract. Effective Early Warning Systems are essential for reducing disaster risk, particularly as climate change increases the frequency of extreme events. The July 2021 floods were Luxembourg’s most financially costly disaster to date. Although strong early signals were available and forecast products were accessible, these were not consistently translated into timely warnings or coordinated protective measures. We use a value chain approach to examine how forecast information, institutional responsibilities, and communication processes interacted during the event. Using a structured database questionnaire alongside hydrometeorological data, official documentation, and public communications, the analysis identifies points where early signals did not lead to anticipatory action. The findings show that warning performance was shaped less by technical limitations than by procedural thresholds, institutional fragmentation, and timing mismatches across the chain. A new conceptual model, the Waterdrop Model, is introduced to show how forecast signals can be filtered or delayed within systems not designed to process uncertainty collectively. The results demonstrate that forecasting capacity alone is insufficient. Effective early warning depends on integrated procedures, shared interpretation, and governance arrangements that support timely response under uncertainty.
- Preprint
(1613 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 08 Dec 2025)
-
RC1: 'Comment on egusphere-2025-3892', Michael Szoenyi, 29 Oct 2025
reply
-
AC1: 'Reply on RC1', Jeff Da Costa, 03 Nov 2025
reply
Hello,
Thank you for taking the time to review our manuscript and the really useful feedback provided!
Below is the detailed reply (Author Reply) to each of the points raised in previous RC1 comments (italicised):Generally, this is a very interesting subject well presented and discussed.
The text can be improved by additional consistency of key hazard and risk terms and using them also more consistenty. I am surprised you don't introduce EWS as an acronym since you mention the long term Early Warning System so often.
Author Reply: Thank you for this observation. The decision not to use the acronym EWS was a choice, as the paper aims to remain accessible to both technical and policy audiences. Writing out Early Warning System in full was a choice intended to support readability across disciplines.
Structurally, a weakness in the manuscript can be overcome by consequently introducing every concept and key aspect first before discussing it. You confuse the reader by already discussing warning levels, thresholds and apps and mechanisms before they are properly introduced. This should be done by adding an earlier section on "Early Warning Systems and its governance in Luxembourg" or similar. I would start with the paragraph line 115 as the motivation of your work, then outline the Luxembourg EWS and emergency governance system as suggested above, then continue with current line 110. You currently have the challenge that you describe this in more detail in section 3.1 but this comes too late.
Author Reply: The current sequencing was chosen to introduce the broader conceptual and transboundary context before presenting national governance details. Section 3.1 was placed after the methodological framework to maintain a consistent transition from international to national scales of analysis. We recognise, however, that introducing the structure of Luxembourg’s Early Warning and emergency governance system earlier could improve reader orientation. In a future revision, we plan to add a short bridging paragraph near the start of Section 1 to preview this framework while keeping Section 3.1 as the detailed reference.
Conceptually, I am surprised that in today's focus on "people centered EWS" and the action by the people who need to be safe from the hazard events, neither the established value chain approach nor your work really focuses on that elemental, "first mile" aspect of the people itself but stops with "official decision-makers". Ultimately, the decision-maker is the individual, household head or community that can or cannot keep safe. I would have liked to see this discussed more. See https://www.sciencedirect.com/science/article/pii/S2589004225006145.
Author Reply: Thank you for this comment. We examine how forecasts and warnings were handled within Luxembourg’s national warning system to explain why early signals did not lead to better anticipatory action. We completely agree that the “first mile” is where early warning must succeed. While this paper focuses on institutional processes, these determine whether people receive clear and actionable messages. The findings therefore help to explain why the link between official warnings and public action unfolded in the way it did. Direct community responses were beyond the available data. We share the view expressed by Budimir et al. (2025) that people-centred early warning requires clear connections across all layers of the system. In the revised discussion we will note this explicitly and position Luxembourg’s structural barriers as one factor limiting that connection.
With regards to uncertainty, while I agree with your discussion in 5.2, it requires some further thinking - this is not just about integrating uncertainty into the forecast and how it might trigger higher alert levels, but then also how laypeople both in official functions as well as "residents" need to deal with that uncertainty, too --> otherwise we'll have soon a big discussion about "increasing false alerts" and fatigue of the population. A general improvement on risk education incl. handling natural hazard event uncertainty is required.
Author Reply: Thank you for this observation. Section 5.2 focuses on how uncertainty affected institutional interpretation and decision-making within Luxembourg’s warning chain. We agree that uncertainty also influences how both officials and the wider public perceive and respond to warnings. While public reactions and alert fatigue were somewhat beyond the available evidence, we recognise that growing exposure to probabilistic forecasts and repeated warnings raises important communication challenges. In the revised discussion we will note that institutional treatment of uncertainty shapes public trust, and that improvements in risk education and communication are important to support both professional and citizen understanding of uncertainty in natural hazard events.
Line 1: If you intend to improve the reach and impact of your publication and would like to engage in a policy dialogue, it might be sensible to slightly change the title of the article. Currently you might be seen as implying that no action at all based on early warning was taken, whereas the reality in the "Bernd" event was more nuanced. Consider changing to "not enough protective" or "no impactful" action or the like.
Author Reply: : Thank you for this helpful comment. The title “Signals Without Action” was chosen to capture the disconnect between available forecasts and the anticipatory measures that followed. Some action did occur, but not early or protective enough to prevent major impacts. The title refers to missed or delayed use of forecast information rather than a total absence of action. We appreciate the suggestion and will consider whether a minor adjustment or clarification in the abstract could better convey this nuance while preserving the central focus on the limited conversion of warning signals into timely protective measures.
Line 2 (and section title 6.2): Was it a disaster? Was there an official "disaster" declaration, or what makes you use the term disaster? Extreme event, yes, but disaster (also compared to Germany)?
Author Reply: The term disaster is used intentionally. The July 2021 floods were formally declared a disaster by the Luxembourg government and met the UNDRR definition as a severe disruption exceeding local coping capacity. While no fatalities occurred and the scale was smaller than in neighbouring Germany for example, the impacts were unprecedented in Luxembourg’s context, with extensive damage, evacuations, and emergency mobilisation across most of the country. Referring to it as an “event” would understate its national significance and overlook how disasters are defined relative to local capacity rather than absolute impact. We will clarify this distinction in the text to acknowledge both the difference in scale and the appropriateness of the term disaster for the Luxembourg case.
Line 26: Imprecise use of "hazard" - hazard is a general concept whereas what Early Warning should achieve is taking action towards an imminent event materializing from that hazard, so correct the first occurrence of hazard and replace it with "(hazard) event" the 2nd time in that line
Author Reply: Thank you for pointing this out. We agree and accept the distinction and will adjust the sentence to specify hazard event in the second occurrence.
Line 27: Name it - it is the "EW4ALL" Initiative
Author Reply: We agree and will include “EW4ALL” .
Line 33: Imprecise description of components for EWS - you might want to distinguish between EWS for climate-related hazards or hydromet hazards, since there are EWS for non-hydromet hazards as well.
Author Reply: he description of Early Warning System components in this section refers specifically to hydrometeorological hazards, as this is the scope of the study. We agree that other types of EWS exist and will clarify this by specifying hydrometeorological where relevant.
Line 36: A key point - you're missing the recipient of the message and the desired action they should take. The value chain should not just be observation-transmission-decision-making (seemingly by the authorities or Emergency Services) but start with the "first mile", the desired action taken by individuals, communities and society at large. EWS inherently should be a system comprising the human element and not just stay within a professional science-tech-public authority bubble. And I say this very consiciously with a first-world, developed-context in mind - you are discussing "Bernd" in Luxembourg but the same thing just happened in Valencia, Spain as well so it is very much also a first world problem so we need to work on that last connection between officials taking decisions and the individual who needs to take action. Consider revising this point throughout the manuscript.
Author Reply: We agree that this connection should be made more explicit. In the revised manuscript we will clarify in Section 2 that the purpose of the value-chain approach is to trace how forecast information moves through the system to its end users, and that the analysis of Luxembourg’s 2021 floods shows how institutional design and communication processes shaped whether information could reach and encourage action at the first mile. We will also note this more clearly in the discussion by highlighting that these structural conditions limited the translation of available warnings into better protective action among the population.
Line 42, 46, 132 etc. etc: See general comment. If you introduce EWS as an acronym you can stay consistent. Why "Warning System" here, is this different from "early" warning system? Later, "early warning" but not system anymore.
Author Reply: We agree that terminology consistency will improve readability. In the revision we will standardise wording to Early Warning System throughout, except where early warning clearly refers to the broader process rather than the system.
Line 79: Word missing? Activating emergency - what? Protocols? Services?
Author Reply: Correct. The intended phrase was activating emergency plans and procedures. We will clarify this wording in the next revision.
Line 100: Between July 12 and July 15?
Author Reply: Correct. The period referenced is between 12 and 15 July 2021, we will make the necessary changes.
Line 104: Imprecise use of "frequent exposure" - don't understand the high time variability element of exposure. Exposure may be low or high and it may change over time, but it rarely fluctuates with a high frequency to make the term "frequent exposure" adequate?
Author Reply: Agreed. The phrase frequent exposure was imprecise. The intended meaning was recurrent or regular exposure over time. We will adjust the wording accordingly in the next revision.
Line 110: Improve sentence. I am unclear what is meant by "higher levels" - warning levels based on thresholds? Or warnings did not reach higher levels of government in terms of the messages reaching further? If the former, you will need to introduce what the warning levels/thresholds in Luxembourg look like first before discussing what happened or did not with warning levels.
Author Reply: The sentence refers to the official colour-coded alert levels rather than levels of government. We will clarify this by rephrasing the sentence to read: “Although forecasts were available, official warnings did not reach higher alert levels until shortly before impacts began to unfold.”
Line 145: See general comment - I perceive the ultimate actor to be the individual, family or community to take action, not the government decision-maker to "tell them what to do" since often this involves still a technical message that does not lead to the desired action since it remains unclear what the desired action should be.
Author Reply: We agree that the effectiveness of early warning also depends on individual and community action. The paper focuses on the institutional level because this is where decisions about warning content, format, and timing are made, which in turn determine how clearly the public can understand and act on them. We will clarify in the discussion that while the analysis centres on institutional processes, these directly influence whether people receive guidance that enables protective action.
Section 3.1: See general comment
Line 180: ", which" unnecessary.
Author Reply: OK
Line 189: In the interest of being more multi-hazard and clear on what's incl. and what's not: Can you specificy whether CGDIS is only responsible for such hazards as mentioned (severe weather, flooding) or wheter it would also respond in non-meteorological natural hazards situations?
Author Reply: CGDIS is responsible for all types of emergencies and hazards in Luxembourg, including meteorological, hydrological and other natural hazard situations. We will clarify this in the text to specify that its mandate covers the full spectrum of civil protection hazards, not only severe weather and flooding.
Line 193: Specify when it had not been implemented (during the July 2021 floods I assume) - or at the time of writing?
Author Reply: The reference concerns the period of the July 2021 floods, when the measure in question had not yet been implemented.
Line 201: You contradict yourself - MeteoLux cannot issue alerts... only alerts from MeteoLux... ? I assume "only forecasts and warnings issued by MeteoLux are considered ..."?
Author Reply: Meteolux publishes both weather warnings and flood warnings (the latter on behalf of the Water Management Administration), but its mandate is limited to the publication of these warnings based on defined physical thresholds. It does not have authority to activate emergency planning or response measures, which fall under the responsibility of the Haut-Commissariat à la Protection Nationale (HCPN). The intended meaning of the sentence was that, under national law, only forecasts and warnings issued by MeteoLux are recognised as the official meteorological source for decision-making. We will rephrase this section to make the distinction between issuing warnings and activating emergency plans clearer.
Table 2 - are these exact translations? I am surprised that for weather the terminology "danger" is used and for flood "risk". Especially the latter I find confusing since we are talking immediate, imminent river conditions leading to danger to lives and livelihoods, rather than the conceptual "risk".
Author Reply: During the July 2021 floods, the meteorological and hydrological alert systems in Luxembourg used distinct terminologies. MeteoLux applied the descriptors Danger (orange) and Extreme Danger (red) for meteorological warnings, while the Water Management Administration used escalation phases for flooding: Pre-Alert Phase (Minor Flood Risk) orange, and Alert Phase (Major Flood Risk) red. The wording in Table 2 reproduces the official terminology in use at that time. However, the brackets around Pre-Alert Phase may have caused confusion. We agree that this should be clarified and will present it exactly as defined in the official framework. Minor and Major flood risk were the official descriptors of the corresponding alert levels. The flood terminology was updated nationally in late 2024 to mirror the meteorological colour-code categories, but these later changes are outside the scope of this study and not relevant to the situation in 2021. The terminology valid in July 2021 can be verified through the archived Infocrise website:
https://web.archive.org/web/20240930004300/https://infocrise.public.lu/en/inondations/phases-alerte.html
For all information about current governmental intervention plans please see www.infocrise.luLine 279: July 17-18
Author Reply: Ok
Line 280: Side comment: I like the use of "insured damages" as according to the UNFCCC this would be the right, precise terminology. The industry would be using "insured losses" which according to UNFCCC definition makes no sense since they are reimbursed for recovery. We might want to influence the industry to adjust wording...
Author Reply: Agreed. We will retain the term insured damages as suggested, consistent with UNFCCC definitions.
Line 302: Correct, and remove track changes
Author Reply: Ok
Line 377: See general comment on people. This is the first time that you mention people ("residents") rather than stay with in the official chain, and it was not yet clarified how communication from EWS should lead to action by those residents. Consider revising with a people-centered re-focus.
Author Reply: We will make this connection clearer by adding a short sentence where residents are first mentioned to explain how official warnings are intended to reach the public through established communication channels. In the discussion we will also note that the institutional processes analysed in this paper determine how effectively information can be translated into guidance that supports protective action by residents.
Line 386: You mention France, but not in the outline of the study that France was one of the countries affected by Bernd.
Author Reply: Agreed. We will note that France was also affected by the July 2021 floods to complete the regional context.
Paragraph line 468: I think you're missing a point here. Although the neighboring countries acted differently (and you recommend Lux. to do the same), it seemingly made no difference to how their EWS performed - those countries, particularly Germany, are the ones where so many lives were lost and people reported not received a (meaningful) warning...
Author Reply: The comparison with neighbouring countries is meant to illustrate differences in institutional procedures, not performance outcomes. We will clarify that while Germany and Belgium experienced far greater impacts, their systems also faced serious challenges. The point is that procedural design influences how forecasts are handled, but this alone does not determine outcomes. We will adjust the phrasing to make this distinction clear.
Line 521: I feel this is a bit jumping to conclusions - the public needs actionable messages, not just an alert. You may want to go back to the discussion on Gouv-Alert earlier and review not only if/when it was activated or not (the tech glitch), but also what it would have done without it - would these have been actionable messages or still unclear what action to take? Compare with your reference to Spain's DANA, where the system technically worked, but was both activated late AND did not incl. clear messages tailored to different audiences.
Author Reply: We will revisit the section on GouvAlert to clarify that the issue was not only its non-activation but also the nature of the messages it delivers. We will also reference the Spain DANA example to underline that technical activation alone does not ensure effective communication or behavioural response.
Line 547 and 550, but also refer to comment on choice of the title - if you want to get policy action you may want to tone done - policy makers in Lux may justifiably say that you suggest Germany did better but actually despite having the advantage in 2021 that you suggest Lux also improves on the result in Germany was not better at all, so this is not the decisive factor...
Author Reply: The intention is not to suggest that Germany performed better, but to highlight differences in system design and mandate structure rather than outcomes. We will adjust the phrasing in this section to ensure that the comparison is presented neutrally and does not imply relative performance, but instead focuses on how procedural design shaped the handling of forecasts in each country.
Line 567: Suggest you revise to the 5 steps of the disaster risk management (DRM) cycle. Also, don't use "mitigation" in this context but (corrective and prospective) risk reduction. Mitigation in today's climate crisis should be reserved as a term for climate mitigation (CO2) rather than use it confusingly as a seeming synonym for risk reduction.
Author Reply: We agree to replace mitigation with risk reduction where appropriate. However, we do not adopt the cyclical DRM model, as the paper conceptualises disasters as systemic outcomes of governance and design rather than sequential phases. This approach aligns with (Bosher et al 2021) DOI: https://doi.org/10.1108/DPM-03-2021-0071
Line 580: I assume "response" here is not meant sensu strictu (responding to crisis) but rather as a response to the trigger -therefore I would use "action" in line with your title etc.
Author Reply: Ok, We will adjust the wording to action for consistency with the paper’s framing and title.
Line 592: Not hazard but (hazard) event
Author Reply: Agreed. We will replace hazard with hazard event at this point to reflect the more precise terminology.
End of 6.3, and conclusion in 7: Nothing to challenge the conclusion as stated per se, but I am lacking again the final step from officials being "warned" or informed to reaching the "resident" that you prominently mention elsewhere - you could outline that maybe this was outside of the scope of your work and the suggested model, but that more is needed on designing and reviewing how targeted messages must be developed/improved to really get protective action by those "residents".
Author Reply: Thank you and we agree. In the revision we will clarify at the end of Section 6.3 and in the conclusion that the study focuses on institutional processes up to the point where information is issued to the public. We will note that the next step, designing and evaluating how messages reach and support residents in taking protective action lies beyond the scope of this analysis but represents an essential direction for further research and system improvement.
Citation: https://doi.org/10.5194/egusphere-2025-3892-AC1
-
AC1: 'Reply on RC1', Jeff Da Costa, 03 Nov 2025
reply
-
CC1: 'Comment on egusphere-2025-3892', Alexander Fekete, 09 Nov 2025
reply
This is a very well written paper on an important case study of a small country, Luxembourg, affected by the 2021 floods in Western Europe. The paper highlights important issues of coordination, as well as warning and governance failures that are highly interesting also to an international audience. While the paper does a good job of documentation and going in detail by providing a great overview over planning documents and detailing flood and warning aspects, it also has certain shortcomings which I would like to address to suggest improvements for the paper.
More explanation and maybe even visualization is recommended for the value chain analysis and model. This seems to be the conceptual background and backbone selected for the paper and in some descriptive way it is taken up again here and there. But a consistent usage, and maybe more explicit visualization or string and use of certain components could be useful to even better align the story of the paper. The two sources provided also did not help me to understand the value chain model approach, so please add more sources.
As a novelty and interesting name, the paper suggests a water drop model. It does not become fully clear whether this model has been independently developed by the group of researchers, or whether it is a model already developed, and for whom and for which purposes. Reading the paper, it mainly seems to be represented by an illustrated graphic. The graphic is nice, but appears to be mainly a combination of a system representation with a boundary and input and output features. It also contains part of a subsystem represented as a triangle in which a tree model like an event three leads to a single output that is then connecting to the warning. It is especially this vector connecting to warning that is emphasized as the main failure in the whole impact chain. While this is plausible, one might also argue that the usage of representation of the tree model makes it look like this was an inevitable outcome. This is a schematic representation problem of similar diagrams used in quality management, such as the Ishikawa diagram or tree models or bowtie analysis and others. Critic around such pipe, or tunnel approaches, or “Nürnberger Trichter” might not be known to a larger international audience and I just want to leave this as a hint for the writers of the study. I think no defense is needed here and it does not need to be modified in the paper. But I must confess that I don’t think that the diagram is so very novel as a method and I think it is at least in some parts suggestive, but which is OK. however, to call this a water drop model might be a little bit overstretched in my opinion.
What is more important is to ask which of the contributing factors on the left side were key in the final outcome that the warning failed? And if and how they can be compared to each other? The description shows this, but I’m not sure that the part with the threshold is really so convincing. Is this really a threshold problem or not just also a problem of observation or governance and therefore already covered by the two other main contributing factors that led to the failure?
I think it has more potential because also in other areas in Germany, the misinterpretation of such rain and river gauges, and prediction curves were part of the problem. Maybe it could be categorized into certain domains of problem types to make the three contributing factors a bit easier to differentiate. The threshold problem could be a communication, interpretation, or transfer of knowledge problem maybe?
I would also suggest to make more usage of the very nice and inspiring diagram from Brian Golding here. Which major gaps of knowledge domains and missing bridges between them are the main aspect behind the three contributing factors analyzed? And again, which types of values and which type types of chains or components of a chain are analyzed here and also in your water drop model diagram? I think there’s more in that study and idea than what is already described, which is why I encouraged to think about it once again.Finally, the paper is mainly descriptive like an extended report version. This is also very good and necessary since publications about Luxembourg and the flood damages in 2021 are scarce and the paper therefore is a very important contribution. The writers also made a very good job, but I would encourage to include some more features of scientific assessment such as adding a limitation section to the discussion. It would be very illustrative for other readers to understand how the approach of the value chain, and the application of the water drop model have also shown certain challenges, which might help others who use the approach to better conduct it themselves.
Maybe also some overarching research question or guiding thoughts from the beginning of the paper could be taken up more prominently at the end or of a summary of main innovations or insights could be given.
My apologies for extended comments, but I hope it does not prompt for large additional sections. I’m looking forward to the publication of this paper.
Citation: https://doi.org/10.5194/egusphere-2025-3892-CC1 -
AC2: 'Reply on CC1', Jeff Da Costa, 12 Nov 2025
reply
We thank the commenter sincerely for taking the time to read our paper in such detail and for offering thoughtful, constructive feedback. We are very grateful for the encouraging words about the paper’s contribution and for the care with which conceptual aspects were considered. The suggestions are valuable and help us reflect further on how to make our analytical approach clearer and more accessible to readers.
The value-chain framework forms the conceptual backbone of the paper. Following the WMO/WWRP Value Chain Project (WMO, 2024; Ebert et al., 2023; Hoffmann et al., 2023), we understand early warning as a process linking forecast generation, translation, communication, and decision. This framing structures both the empirical reconstruction (Sections 2–4) and the analytical discussion (Sections 5–6). We applied the framework operationally, using its nodes–actors–flows logic and the Database Questionnaire developed under the WMO/WWRP HIWeather project (Hoffmann et al., 2023) to guide data collection and the organisation of results. This questionnaire, designed to document the end-to-end flow of information and decision-making in high-impact weather events, informed how we traced actors, information products, decisions, and outcomes across Luxembourg’s warning system. It enabled us to identify where forecast value was maintained or constrained along the chain. The official WMO Value-Chain Framework and supporting materials are publicly available via the WMO Library (https://library.wmo.int/idurl/4/69103)
We developed the Waterdrop Model for this study. It extends the reconstructed value chain and translates its logic into a diagnostic structure that helps explain why forecast information was not converted into better anticipatory protective action. Its purpose is to visualise how institutional design and procedural thresholds influence the operational use of information. In doing so, the model aligns with calls to move beyond linear or cyclical representations of disaster processes (Neal, 1997; Bosher et al., 2021). The Waterdrop Model conceptualises disaster risk as a multi-layered process shaped by governance and mandate boundaries, not by event sequences alone.
More information here:Bosher, L., Chmutina, K., & van Niekerk, D. (2021). Stop going around in circles: towards a reconceptualisation of disaster risk management phases. Disaster Prevention and Management: An International Journal, 30(4/5), 525-537.
Neal, D. M. (1997). Reconsidering the phases of disaster. International Journal of Mass Emergencies & Disasters, 15(2), 239-264.
We appreciate the insightful remarks regarding the potential for schematic forms to appear deterministic. This observation is well taken. The comparison with Ishikawa and bow-tie diagrams is helpful, as it underlines the importance of clarifying that the model shows interacting processes, not deterministic cause-effect chains. We will refine the caption and accompanying text accordingly to make this interpretive purpose more clear.
We value the question about the relative weight of the contributing factors. The “threshold” aspect is not purely technical but procedural, referring to institutional points at which interpretation and responsibility shift. We will clarify in Section 6 that thresholds intersect with governance and interpretation and that their role is systemic, not isolated.
The suggestion to link the analysis more closely to Golding et al.’s “valleys of death” diagram is appreciated and well taken. The WMO/WWRP Value Chain framework applied in this paper was itself influenced by Golding et al.’s work, which conceptualises the gaps between scientific knowledge, service delivery, and decision-making. The Waterdrop Model builds on this foundation by illustrating how such gaps manifested in Luxembourg’s warning system through institutional mandates and communication boundaries. We will make this conceptual link explicit in the revised text. A detailed mapping of individual knowledge domains, as proposed in the comment, would require additional data and targeted analysis beyond the scope of the present study, but it represents a promising direction for future research.
We emphasise that although the paper contains detailed documentation, its aim is analytical. The reconstruction of the 2021 floods provides the empirical basis to explain how institutional structure and mandate configurations shaped forecast interpretation. The integration of the Value Chain approach and the Waterdrop Model provides a systematic means to analyse these mechanisms. We will include a short paragraph on methodological limitations in the discussion to outline the scope of available evidence and the interpretive boundaries of the Waterdrop Model.
We appreciate the suggestion to restate the overarching research question more prominently at the end. In the revised conclusion, we will explicitly return to the central question of how forecast signals were or were not translated into better anticipatory action and summarise that institutional design largely determined the space for action. This synthesis will highlight the broader relevance of combining the Value Chain framework with the Waterdrop Model to analyse governance in early-warning systems.
We thank the commenter once again for the generous engagement and constructive feedback. The comments have been read with care and will directly guide improvements to the paper. We are grateful for the opportunity to clarify these points and for the collegial spirit in which the feedback was offered.
Citation: https://doi.org/10.5194/egusphere-2025-3892-AC2
-
AC2: 'Reply on CC1', Jeff Da Costa, 12 Nov 2025
reply
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 179 | 48 | 15 | 242 | 8 | 7 |
- HTML: 179
- PDF: 48
- XML: 15
- Total: 242
- BibTeX: 8
- EndNote: 7
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Generally, this is a very interesting subject well presented and discussed.
The text can be improved by additional consistency of key hazard and risk terms and using them also more consistenty. I am surprised you don't introduce EWS as an acronym since you mention the long term Early Warning System so often.
Structurally, a weakness in the manuscript can be overcome by consequently introducing every concept and key aspect first before discussing it. You confuse the reader by already discussing warning levels, thresholds and apps and mechanisms before they are properly introduced. This should be done by adding an earlier section on "Early Warning Systems and its governance in Luxembourg" or similar. I would start with the paragraph line 115 as the motivation of your work, then outline the Luxembourg EWS and emergency governance system as suggested above, then continue with current line 110. You currently have the challenge that you describe this in more detail in section 3.1 but this comes too late.
Conceptually, I am surprised that in today's focus on "people centered EWS" and the action by the people who need to be safe from the hazard events, neither the established value chain approach nor your work really focuses on that elemental, "first mile" aspect of the people itself but stops with "official decision-makers". Ultimately, the decision-maker is the individual, household head or community that can or cannot keep safe. I would have liked to see this discussed more. See https://www.sciencedirect.com/science/article/pii/S2589004225006145.
With regards to uncertainty, while I agree with your discussion in 5.2, it requires some further thinking - this is not just about integrating uncertainty into the forecast and how it might trigger higher alert levels, but then also how laypeople both in official functions as well as "residents" need to deal with that uncertainty, too --> otherwise we'll have soon a big discussion about "increasing false alerts" and fatigue of the population. A general improvement on risk education incl. handling natural hazard event uncertainty is required.
Line 1: If you intend to improve the reach and impact of your publication and would like to engage in a policy dialogue, it might be sensible to slightly change the title of the article. Currently you might be seen as implying that no action at all based on early warning was taken, whereas the reality in the "Bernd" event was more nuanced. Consider changing to "not enough protective" or "no impactful" action or the like.
Line 2 (and section title 6.2): Was it a disaster? Was there an official "disaster" declaration, or what makes you use the term disaster? Extreme event, yes, but disaster (also compared to Germany)?
Line 26: Imprecise use of "hazard" - hazard is a general concept whereas what Early Warning should achieve is taking action towards an imminent event materializing from that hazard, so correct the first occurrence of hazard and replace it with "(hazard) event" the 2nd time in that line
Line 27: Name it - it is the "EW4ALL" Initiative
Line 33: Imprecise description of components for EWS - you might want to distinguish between EWS for climate-related hazards or hydromet hazards, since there are EWS for non-hydromet hazards as well.
Line 36: A key point - you're missing the recipient of the message and the desired action they should take. The value chain should not just be observation-transmission-decision-making (seemingly by the authorities or Emergency Services) but start with the "first mile", the desired action taken by individuals, communities and society at large. EWS inherently should be a system comprising the human element and not just stay within a professional science-tech-public authority bubble. And I say this very consiciously with a first-world, developed-context in mind - you are discussing "Bernd" in Luxembourg but the same thing just happened in Valencia, Spain as well so it is very much also a first world problem so we need to work on that last connection between officials taking decisions and the individual who needs to take action. Consider revising this point throughout the manuscript.
Line 42, 46, 132 etc. etc: See general comment. If you introduce EWS as an acronym you can stay consistent. Why "Warning System" here, is this different from "early" warning system? Later, "early warning" but not system anymore.
Line 79: Word missing? Activating emergency - what? Protocols? Services?
Line 100: Between July 12 and July 15?
Line 104: Imprecise use of "frequent exposure" - don't understand the high time variability element of exposure. Exposure may be low or high and it may change over time, but it rarely fluctuates with a high frequency to make the term "frequent exposure" adequate?
Line 110: Improve sentence. I am unclear what is meant by "higher levels" - warning levels based on thresholds? Or warnings did not reach higher levels of government in terms of the messages reaching further? If the former, you will need to introduce what the warning levels/thresholds in Luxembourg look like first before discussing what happened or did not with warning levels.
Line 145: See general comment - I perceive the ultimate actor to be the individual, family or community to take action, not the government decision-maker to "tell them what to do" since often this involves still a technical message that does not lead to the desired action since it remains unclear what the desired action should be.
Section 3.1: See general comment
Line 180: ", which" unnecessary.
Line 189: In the interest of being more multi-hazard and clear on what's incl. and what's not: Can you specificy whether CGDIS is only responsible for such hazards as mentioned (severe weather, flooding) or wheter it would also respond in non-meteorological natural hazards situations?
Line 193: Specify when it had not been implemented (during the July 2021 floods I assume) - or at the time of writing?
Line 201: You contradict yourself - MeteoLux cannot issue alerts... only alerts from MeteoLux... ? I assume "only forecasts and warnings issued by MeteoLux are considered ..."?
Table 2 - are these exact translations? I am surprised that for weather the terminology "danger" is used and for flood "risk". Especially the latter I find confusing since we are talking immediate, imminent river conditions leading to danger to lives and livelihoods, rather than the conceptual "risk".
Line 279: July 17-18
Line 280: Side comment: I like the use of "insured damages" as according to the UNFCCC this would be the right, precise terminology. The industry would be using "insured losses" which according to UNFCCC definition makes no sense since they are reimbursed for recovery. We might want to influence the industry to adjust wording...
Line 302: Correct, and remove track changes
Line 377: See general comment on people. This is the first time that you mention people ("residents") rather than stay with in the official chain, and it was not yet clarified how communication from EWS should lead to action by those residents. Consider revising with a people-centered re-focus.
Line 386: You mention France, but not in the outline of the study that France was one of the countries affected by Bernd.
Paragraph line 468: I think you're missing a point here. Although the neighboring countries acted differently (and you recommend Lux. to do the same), it seemingly made no difference to how their EWS performed - those countries, particularly Germany, are the ones where so many lives were lost and people reported not received a (meaningful) warning...
Line 521: I feel this is a bit jumping to conclusions - the public needs actionable messages, not just an alert. You may want to go back to the discussion on Gouv-Alert earlier and review not only if/when it was activated or not (the tech glitch), but also what it would have done without it - would these have been actionable messages or still unclear what action to take? Compare with your reference to Spain's DANA, where the system technically worked, but was both activated late AND did not incl. clear messages tailored to different audiences.
Line 547 and 550, but also refer to comment on choice of the title - if you want to get policy action you may want to tone done - policy makers in Lux may justifiably say that you suggest Germany did better but actually despite having the advantage in 2021 that you suggest Lux also improves on the result in Germany was not better at all, so this is not the decisive factor...
Line 567: Suggest you revise to the 5 steps of the disaster risk management (DRM) cycle. Also, don't use "mitigation" in this context but (corrective and prospective) risk reduction. Mitigation in today's climate crisis should be reserved as a term for climate mitigation (CO2) rather than use it confusingly as a seeming synonym for risk reduction.
Line 580: I assume "response" here is not meant sensu strictu (responding to crisis) but rather as a response to the trigger -therefore I would use "action" in line with your title etc.
Line 592: Not hazard but (hazard) event
End of 6.3, and conclusion in 7: Nothing to challenge the conclusion as stated per se, but I am lacking again the final step from officials being "warned" or informed to reaching the "resident" that you prominently mention elsewhere - you could outline that maybe this was outside of the scope of your work and the suggested model, but that more is needed on designing and reviewing how targeted messages must be developed/improved to really get protective action by those "residents".