We would like to thank the reviewer for their constructive feedback. Please find a point-by-point response below, in Italic. 

General Comments:

Remmers et al. provide an opinion piece on how hydrological modeling could benefit from insights and practices of the critical social sciences. They offer a well-structured and argued discussion on reasons for and possibilities to increase the accountability, transparency and responsibility in hydrological modelling.

I particularly enjoyed the last part of the paper which has a lot of important and well communicated conclusions and action items for the different actors in a hydrological modelling network. I do believe that the introduction and motivational part of the paper can be strengthened by making new terms and ways of thinking more approachable to the reader inexperienced with critical social sciences. I therefore have a few suggestions that I hope can help in this regard.

Generally, I believe this is a very fitting contribution for HESS and a nice piece for the hydrological modelling community to reflect on current and future modelling practices and the impact social aspects might have on our work. Awareness is the first step to change, which is why I recommend publication after minor revisions.

Thank you for your encouraging evaluation of our work, and the constructive feedback provided.

Specific Comments:

[Are models perceived as neutral and objective tools?]

The authors base their motivation on the framing that models are perceived as neutral and objective tools. I would argue that most hydrologic literature (and also the sources cited in line 16) argue that models are hypothesis and therefore not quite as neutral as implied. These (model) hypotheses are generally formed based on a perceptual model which is then translated into the mathematical model that becomes the “tool” we use. As perceptual models are known to be personal and at least in part qualitative, I think we can agree, that by the time the model is formed and ready to be used as a “tool” a lot of social influence has already happened. The authors themselves describe part of this process in their Argument 2. Therefore, I keep on stumbling over the sentence “models are perceived as neutral and objective tools” as something I can’t fully agree with. And I would imagine that this will be the case for most experienced modelers. To engage both groups (the problem aware and less aware modelers and model users) equally well, it might be helpful to simply acknowledge that different groups in the hydrological modelling network are more or less likely to see a hydrological model as a “neutral and objective tool”, but that it is important for everyone to understand what this notion may lead to.

I believe that most of my discomfort comes from the sentence “Within hydrological modelling, a persistent notion exists that a model is a neutral, objective tool” that is used prominently in abstract and introduction. To me it has the disadvantage of veiling and softening the main motivation for this paper (the assumption of a neutral and objective tool is questionable and comes with consequences) and giving an impression of consensus where a spectrum of understanding already exists.

I assume that this comes down to mere nuances of formulation as I realize that “notion” is supposed to imply that “many believe models are objective, but this view is not universally accepted”. I argue, however, that a more direct phrasing of this issue will help the reader to grasp the main point and motivation of this paper more easily and helps to acknowledge that we do not start at zero regarding the awareness of this problem.

I therefore suggest to either change the first sentence of the introduction to be a more direct description of the problem or include a short discussion of the different states of awareness regarding this problem around line 24. I believe this would also make the citations from line 16 more fitting (see minor comments).

We can fully resonate with the feedback from the reviewer on our sentence that “models are generally perceived as neutral and objective tools”. Among many modellers there is wider acknowledgment about the expertise and subjectivity involved in developing the perceptual model, and also the lack of data that leads to methodological underdetermination is generally well recognized – usually in terms of uncertainty. We do feel, however, that this awareness is not always linked to notions of bias, power, or non-neutrality (and that this non-neutrality is often forgotten in decision-support contexts). That being said, rephrasing this sentence is indeed a good suggestion to keep everyone on board in this discussion and to recognize the awareness that is already there. We propose: “Although modellers generally acknowledge that models are hypotheses, determined by experts’ system understanding and subject to uncertainty, this is rarely connected to deeper reflections on how this shapes certain narratives that benefit some while disadvantaging others. The model is recognized as subjective, yet still perceived as neutral, especially in decision-support contexts” . For the Abstract, this sentence would be: “Models are acknowledged as subjective and uncertain, yet are often still perceived as neutral, with little reflection on how they shape narratives that may advantage some while disadvantaging others, especially in decision-support contexts.”

[What is critical social science and how can we benefit from it?]

As a reader I am very interested in what critical social science is and how we can benefit from it. But from the introduction alone I feel I do not yet see what critical social science has to offer that hydrology can learn from. I feel that might mainly be the case because the introduction could often benefit from some specific examples that guide and convince the reader of the storyline instead of making statements that are justified with citations from a field less familiar to the average hydrology reader. I would prefer to be convinced through examples from the literature rather than expected to read all the cited papers myself to reach a similar conclusion. I would appreciate if the authors could include more specific examples from the papers they cite when building their argument in the introduction. More details and general contemplations are then provided in the following chapters. I will provide specifics in the minor comments.

We thank the reviewer for this suggestion, for it is of course essential to keep readers on board and therefore to be clear on what we mean with the critical social sciences. We propose to include the following two examples in the paragraph of line 34, p3, to show examples of what political ecology and STS have revealed.

An example from Science and Technology Studies (STS), that might also be relevant to modelling,  is the research on infrastructures. A well-known case is Star and Strauss (1999), who studied the daily work of nurses in hospitals. Through ethnographic observation, they showed how standardized forms and technologies often failed to reflect the complex daily realities of nursing work. Nurses developed informal workarounds to fit complex reality into the official documentation, as official documentation tended to reflect doctors’ perspectives while marginalizing the experiential knowledge of nurses. As such, it demonstrates that standards and technologies represent certain values and perspectives better than others.

The field of political ecology studies the role of power (aka, politics) in environmental issues. As such, it sheds light on the broader context in which processes take place. An applicable example in hydrology is the famous study on soil erosion in Nepal (Blaikie, 1985). Soil erosion was often framed as a result of poor farming practices by local farmers. Blaikie demonstrates how this is also the result of power structures: Since the majority of land was held by a small elite, small local farmers relied on tenant land farming. Because these leases could be terminated at any time, tenant farmers had little incentive to invest in long-term, sustainable practices like erosion control. This suggests that solutions should focus not merely on training farmers in improved practices, but on enhancing their livelihood security.

Both fields, therefore, offer broader perspectives on the use of technology and  environmental issues. Such a broader perspective, accounting for whose perspectives were involved and evaluating broader political contexts, can enrich our understanding of hydrological models and their place in society.

----

Star SL, Strauss A. Layers of silence, arenas of voice: The ecology of visible and invisible work. Computer Supported Cooperative Work. 1999;8(1-2):9-30. doi: https://doi.org/10.1023/A:1008651105359.

Blaikie, P. (1985). The Political Economy of Soil Erosion in Developing Countries.

Minor/Technical Comments:

• Abstract – “marginalizing certain stakeholders”: is this the most relatable problem to mention at this point? I initially fail to imagine an example of what this might mean and would like to read an “OR” with a more relatable example (maybe overconfidence in model results etc.) or a more specific example of the marginalized stakeholder consequence.

Yes, for us marginalized stakeholders is key – “overconfidence in model results”, as suggested by the reviewer, can also lead to certain stakeholders being marginalized; some people or entities pay the price for this overconfidence. Staying close to the modelling, marginalized stakeholders are those voices that are not heard or represented in the modelling. In a decision-support context, this effect is clear: if models only evaluate the effect on discharge and not on fish population, decisions focused on discharge might negatively affect not only the fish population but also communities depending on these fish. In a scientific context the effect is less direct, but model results can shape discourses, foreclosing alternative frames. One of these discourses, one could argue, is the focus on discharge in hydrological modelling, while for many questions perhaps other fluxes or states are more relevant.



Given that this is the abstract we have added a short example: “.. potentially harmful consequences. One is marginalizing stakeholders, not accounting for or recognizing other perspectives on the issue at hand, which might have called for a different (modelling) approach.”

• Abstract – “The main take-away, from our perspective, is that responsible modelling is a shared responsibility” – This sentence might diminish the contribution of the article a little. I would suggest rephrasing in a way that lists the different contributions of the article. E.g.: We highlight that responsible modelling is a shared task between all actors of a modelling network and provide several actionable recommendations for individual actors to increase their share in facilitating responsible modelling. Or something similar.

We understand and propose:

“The main take-away, from our perspective, is that responsible modelling is a collective responsibility, shared by all actors in a modelling network. We provide several actionable recommendations for individual actors to increase their share in facilitating responsible modelling.”

• L25-26 – I believe these citations see models not as a neutral tool but as a hypothesis that needs testing. I therefore find the referencing questionable with the current phrasing. Especially, since the same citations are used in line 27 when stating that “models are simplifications where we need to make choices on what to represent or not to represent”. Please refine citation usage for these two sections of the paper.

These papers were cited in these contexts because they also make the claim that models are generally perceived as neutral – to subsequently attack this claim. Given that we have rephrased this sentence in response to the first point raised by this reviewer, we feel we could make the newly proposed sentence without any further citation.

• L21-23 – I think this part would benefit from at least one very specific example. I can offer a potential example of first nations in Canada suffering from not being included as stakeholder during dam construction. Maybe the introduction of this paper can be a good starting point to investigate specific examples: https://www.tandfonline.com/doi/full/10.1080/08941920.2018.1451582

Thank you for this suggestion. Dam construction is indeed exactly an example where models are used to justify the construction, while many examples exist where certain stakeholders are not involved in the process and as such, marginalized. We decided to include another dam example, where the model is prominently present. We propose to add the following examples after line 30 (.. by assumed neutrality.):

An example of the politics of modeling can be found in Sanz et al. (2019), where a groundwater model was developed to determine which farmers should receive compensation for halting groundwater pumping to ensure continuous flow in the Júcar River (Spain). In this case, the model acted as a referee to determine who could benefit from the compensation scheme. Packett et al. (2020) highlight numerous examples of the politics of modelling along the lines of gender. For instance, in a case studied by Zwarteveen (1997) in Nepal, men and women worked cooperatively as co-farmers but prioritized different aspects of water flow. Men, responsible for land preparation, focused on water arriving at the start of the irrigation season, while women, who managed weeds, needed consistent water throughout the season. An irrigation distribution model optimized for either water arrival or water sustainment would thus benefit either men or women in their activities. Large-scale infrastructures, such as interbasin water transfers, present more explicit examples of contested cases where stakeholder perspectives were marginalized during model development, leading to the underestimation or disregard of impacts on those stakeholders. For example, Godinez-Madrigal et al. (2019) discuss a case in Mexico where a model was used to justify an interbasin water transfer, yet failed to account for the full range of effects on communities in the donor basin.”

• L28-29 – “This can result in injustices: some groups being overlooked […]”: I find it very difficult to jump between processes that become invisible vs. groups being overlooked etc. These are very different aspects of modelling consequences, and I believe it would be helpful to elaborate a bit on potential path dependencies or describe these different aspects in a bit more context than currently done.

We hope that the examples we have suggested above make the jump somewhat smaller.

• L35 – the comma should be a dash to fit the beginning of the sentence?

Agree, we will adapt this.

• L38 – STS as an abbreviation that is not used again in this paper, so it can be removed

We believe it might be helpful to keep the abbreviation in, because this field is often simply referred to as “STS” and might therefore now be better recognizable to people not yet familiar with this field.

• L38 – “provide insights into how to analyze and deal with non-neutrality”: Would it be helpful to include an example of what is being done in this science so the hydrological reader gets and idea what might be worth implementing? This might provide further support to the next sentence calling for more responsible modelling.

In response to the suggestion above, we propose to include an example from political ecology and STS on how these fields contribute to understanding and dealing with non-neutrality. Together with the hydrological modelling examples, we think the introduction now more convincing and compelling, for which we would like to thank the reviewer.

• L80 ff – “Proske et al.“ and equifinality in cloud microphysics. I would argue we have good examples of equifinal model performance in hydrology. I would suggest using a hydrology example here?

Yes, true. We propose

Khatami, S., Peel, M. C., Peterson, T. J., & Western, A. W. (2019). Equifinality and flux mapping: A new approach to model evaluation and process representation under uncertainty. Water Resources Research, 55, 8922–8941. https://doi.org/10.1029/2018WR023750

• L104 – something seems to be wrong with the citation (?), please check

Thank you, we will fix this. The following citation was missing:



Mark S. Reed, Anil Graves, Norman Dandy, Helena Posthumus, Klaus Hubacek, Joe Morris, Christina Prell, Claire H. Quinn, Lindsay C. Stringer, Who's in and why? A typology of stakeholder analysis methods for natural resource management, Journal of Environmental Management, Volume 90, Issue 5, 2009, https://doi.org/10.1016/j.jenvman.2009.01.001.

• L106 – consider removing the “obviously”

Agree.

• L124 – do the critical social sciences or a specific publication provide some sort of glossary or terminology framework that could be referred to here? If a hydrologist would want to learn about this vocabulary, where could he start?

More in general, we will add a reference to Moon and Blackman (2014). Wesselink et al. (2017) would be a good starting point more applied to hydrology.

Moon K, Blackman D. A guide to understanding social science research for natural scientists. Conserv Biol. 2014 Oct;28(5):1167-77. doi: 10.1111/cobi.12326. Epub 2014 Jun 24. PMID: 24962114.

• L144 – is there any example or guide on how to start if an author would want to write and add a reflexivity statement to their work?

Yes, the current sentinel paper for this is Holmes (2020). Also the terminology overview in Malterud (2001) is well known in this context. We will add a description and reference to these two papers.

Kirsti Malterud, Qualitative research: standards, challenges, and guidelines, The Lancet, Volume 358, Issue 9280, 2001, Pages 483-488,https://doi.org/10.1016/S0140-6736(01)05627-6.

Holmes, Andrew Gary Darwin. “Researcher Positionality - A Consideration of Its Influence and Place in Qualitative Research - A New Researcher Guide.” Shanlax International Journal of Education, vol. 8, no. 4, 2020, pp. 1-10.

• L157-158 – “can have ethical implications in society AND water management”?

Agree.

• L160 – Is there one outcome for the development of ethics of artificial intelligence that could be named as being useful/adaptable to hydrology?

Yes, we agree. The development of ethics in AI and responsible AI is useful and adaptable to hydrology. One recent development in this space is the growing emphasis on how system conceptualization can help with more reflective practices around AI modeling. For example, Nabavi and Browne (2023) propose the Five Ps framework to guide AI researchers and practitioners in situating their modeling work as interventions within competing perspectives on what constitutes a problem and how that framing can drive certain solution(s). The problem-solution dynamic can then be situated in specific zones of intervention - Parameter, Process, Pathway, and Purpose - each offering a leverage point with varying potential for change.

For example, when AI modelers address responsible AI challenges within the Parameter zone, they typically focus on quantifiable improvements through numerical adjustments and parameter tuning. In contrast, challenges identified in the Purpose zone seeks to engage with the fundamental norms, values, and worldviews embedded in modeling practices, which of course invites deeper questions, such as: what broader societal or ecological goals should guide modeling practice? (e.g., equity, resilience). This framework supports hydrological modelers in openly reflecting on their role in problem framing and discussing intervention zone. More importantly, it enables them to assess their work's transformative potential in catalyzing systemic changes, particularly for models designed with specific impacts in mind

We will add this to the section on AI.

Nabavi, E., Browne, C. Leverage zones in Responsible AI: towards a systems thinking conceptualization. Humanit Soc Sci Commun 10, 82 (2023). https://doi.org/10.1057/s41599-023-01579-0

• L165 – and again it would be great to read an example to make these new abstract ideas easier to grasp

Here an example on how ontology shapes hydrology.

“For example, hydrologists often distinguish between epistemic and aleatoric uncertainty. Recognizing aleatoric uncertainty, that is, uncertainty due to inherent randomness in natural processes, presupposes a belief that the world is not entirely deterministic. This illustrates how one’s worldview, or ontology, influences which types of uncertainty are considered meaningful to study. The same applies to epistemology, the theory of knowledge: modeling aligns well with a Newtonian perspective, which assumes that natural laws can be discovered and represented objectively. In contrast, a constructivist would argue that all knowledge is socially constructed, and thus would immediately question the idea of a single 'best' model, highlighting the partial and situated nature of modeling.”

• Title for 4 – just a personal preference, but I would probably write “building bridges between (two) scientific disciplines” – but up to the authors

We agree with the suggestion as it better reflects the section, and will therefore adopt it. Thanks for the suggestion!

• I really like part 4! Do you have suggestions on how teachers should be educated/ can educate themselves on this if they would like to incorporate it in their classes? I asked this before, but can you maybe reference sources that would help the motivated reader to get started on writing a positionality statement?

We will again refer to the positionality reference mentioned above as a starting point. Furthermore, in response to reviewer 2, we will include our own positionality statement. In response to the question on Section 6 below, we added more concrete examples there.

L208 – The sentence about flexible modelling frameworks seems a bit detached. Or at least the context of why it comes up here does not seem to be explained in a convincing way. Maybe the authors can consider rephrasing the sentence and making the connection between diversity of approaches, flexible modelling frameworks and different context a bit clearer.

Agree. Modular modelling frameworks directly address some of the issues brought up (underdetermination, the subsequent required pluriformity), but are themselves of course again part of  a certain path (which processes are included, how, etc). We propose to remove the sentence, and bring up the Modular Modelling Frameworks in the introduction where we introduce underdetermination and models as hypotheses.

• L235 – should there be a period/full stop at the end of the sentence?

Yes, thank, this will be added.

• Section 6 – is there a possibility of providing an example for each point mentioned to make it easier for the reader to find a starting point? E.g. what type of assumptions could a model user ask for that might be relevant. How does he know what to ask for? Is there an example of a positionality statement a modeler could look at? Are there resources for reflexivity practices? Are there resources available each actor could look at to get started? To avoid people taking this as recipe you already have the follow up statement that anyone needs to adapt all this to his own working environment.

We will elaborate each point with a more concrete example. Here already a few suggestions:



If you are a model user (i.e. someone who analyses and uses model results), you can consider asking the modeller for the main assumptions in the model, and the trust that the modeler has in the model results. One way to explore this, together with the modellers, could be a serious game, such as presented here: https://floodskinner.games/projects/adventures-in-model-land/

If you are teaching the next generation of hydrological modellers, you can consider incorporating reflexivity practices and social science basics in your lecture, computer practical, course, or curriculum. This could be an interesting starting point:

https://meetingorganizer.copernicus.org/EGU25/EGU25-11346.html

If you are overseeing a modelling team, you can consider having a discussion on internalised assumptions in your way of working, also known as entrenched workflows (Levine and Wilson, 2013). Situated modelling, as suggested by Klein et al. (2023) could be a good starting point: https://www.tandfonline.com/doi/full/10.1080/26395916.2024.2361706

• Conclusion – it might be helpful to have the definition of what you consider a hydrological modelling network to be a bit earlier then in the conclusions.

Agree. It is first introduced in Line 44. We will add this there.

• References – ter Horst et al. “Making a case for power-sensitive water modelling: a literature review” is still cited as a discussion paper. But the final version of the paper is already available: https://hess.copernicus.org/articles/28/4157/2024/

We will adapt this.

We would like to thank the reviewer for their constructive feedback. Below we provide a point-by-point response in Italic. 

This is an interesting opinion paper providing an overview of the current research on social aspects of modelling, and proposing actionable recommendations for the modelling community. Please find my suggestions for further improvement below.

Figure 1

In my opinion the figure is not very useful in this form. For some arguments it is not obvious where they should be positioned (or it is obvious in which case the figure is arguably not very useful). I am also wondering why ‘Model Problem’ is on the left side (where ‘Society’ is) while the ‘Modeller’ is on the right side (where ‘Modelling Community’ is). It could also be the other way around. However, I do not have a suggestion for improvement so you could also keep it, possibly with minor adjustments. An alternative would be to organize the figure along the lines of the text, i.e. social aspects, insights from social sciences, building bridges between sciences, reflecting (note that this goes, more or less, from defining the problem (could be left in the figure) towards possible solutions (could be on the right side of the figure)).

Thank you for the suggestion to rethink the figure. It is true that “Model problem” is placed in “Society”, but note that the “Modeller”-circle  and “Modelling community” -circle also overlap with society. We will emphasize these overlaps in more detail, because they are the core of our story: modellers, and the modelling community, are part of a society, and address problems that are embedded in society. We will explore different configurations of the figure to see how we can improve clarity. We want to emphasize the overlap, because that is what this figure aimed to convey.

Line 15

‘hydrological modelling’. Please define, what type of models? For instance, is the discussion here about model concepts/equations or (also) about the software implementation? Also, is the discussion about forward simulation models (any form) or also about models relying on statistical learning (including machine learning) that are mostly not run forward in time – note that in both models types, observational data are used and there are currently blends, often referred to as hybrid models. Also, do you refer to the activity of model building? Or also other steps in the model development cycle (e.g. calibration, application).

Good point. We mean: The practice of hydrological modelling, from developing and implementing model code, to applying the model (including for instance calibration) to address a certain question or issue. We generally focus on numerical models, but believe our arguments are valid for data-driven modelling too. Therefore we propose to add the following sentence:

“Here, we consider hydrological modelling the act, or practice from developing and implementing model code, to setting-up and applying the model to address a certain question or issue. We are mainly inspired by numerical modelling, but believe our arguments apply to data-driven modelling as well.”

Line 15-18

It is argued here that models are neutral because they are influenced by society. This is indeed the case. However, in addition, models are influenced by the social network within the modelling community (see e.g. Babel et al, 2019). Thus, social factors within the modelling community as well as influences from outside (society) are important in making models non-neutral. In my opinion both aspects need to be highlighted here.

In addition, it would be good to define ‘neutral’. It seems you consider it as a synonym for ‘objective’ but these may be different concepts. It seems the references provided do not clearly define ‘neutral’.

Thank you for the suggestion. Yes, the modelling community also shapes the way we model, this is our “Argument 2” (The modelling process itself is a social product). We will indeed add this to the introduction where we discuss the neutrality of models.



Considering the difference between objective and neutral; these are indeed, in our perspective, not the same. We believe objectivity stems more from methodological approach, while neutrality refers more to its effects in the world. In response to reviewer 1, we acknowledge that many modellers are aware of the subjectivity involved with modelling, but do not question the subsequent neutrality of models – where we will define neutrality as “not taking sides”.

Line 30-31

‘Simultaneously, ignoring the political side …’ I have a similar comment here as given above (line 15-18). This sentence (line 30-31) seems to imply you consider mainly influence from society (‘political side of models’) on the model (and model community). However also within the model community social factors influence modelling.

Agree, in response to reviewer 1 we will include explicit examples here. In response to this comment, we will also include an example from power effects within the scientific community. “Another example, at the scientific community level, is for instance that some large institutions fund hydrological research with the requirement to use their data. This shows the power position that they have, because it will result in scientific publications in which their data is used and as such, legitimized, even if better alternatives would have been available.”

Section 2 Social aspects in hydrological modelling

You distinguish three ‘arguments’. The description of these need to be improved in my opinion. Please let me explain how I see it and how I recommend describing this. You are free following a different approach but please consider my line of reasoning below.

The first argument (in my opinion) should be about how society affects models and modelling. Society is a stakeholder and as a result, society influences the ‘shape’ of models. The current text however only explains _that_ models are embedded in society. In addition (and more importantly), it needs to describe that society influences models/modelling and how.

We understand the position of the reviewer and agree. Right now we only describe how the problems we address with models are embedded within society, but not that these models are a product of this same society (and argument we made in Melsen et al. 2018). We will add the following sentence at the end of arg 1:

“That being said, it should be recognized that the modelling itself is also the result of the society in which it was shaped. Norms, values, and discourses commonly accepted with a society provide natural boundaries within which the hydrological model can be developed and is accepted. Even more, what is considered a problem is determined by societal standards. For instance, flood risk might be considered differently at different places.”

Similar to the first argument, the second argument needs to describe that social aspects within the modelling community influence models/modelling and how. It does. However, in my opinion it can be improved by also explaining the mechanisms. Equifinality is relevant here (multiple models are ‘possible’) but the mechanisms that lead to these particular different models are equally important and could be described. One of the mechanisms is ‘habits’, as described in Babel et al (2019). You use it as a reference for equifinality but if I am correct, we did not discuss it in our paper (Babel et al (2019)). The paper mainly explains _how_ social factors lead to particular models (what you would call ‘non-neutral’ models).

We understand and agree with the reviewer. The reference to Babel et al. (2019) was not intended to refer to the equifinality but to the social processes in that sentence, but we propose to rephrase to be more explicit. “These social processes include habit (Babel et al, 2019), institutional legacy (Addor and Melsen, 2019), and peer experience (Melsen, 2022). As elaborated in Melsen et al. (2025) for the Nash-Sutcliffe efficiency, modelling standards are not purely technical but socially negotiated, in this example shaped by American engineering societies and an active modelling community. Together, these studies highlight that modelling is not just a technical exercise, but a socially learned and negotiated practice.”

The third argument, in my opinion, should be about the fact that models have implications for society (political, ethical) and that it is thus extremely important (also outside academia) to describe and discuss how ‘neutral’ they are as they have impact outside academia. This third argument does not discuss how social factors influence models (like the first and second argument). Instead, it describes the relevance of modelling choices for society. I do not see how ‘the previous arguments come together’ (line 98) here (at least for me it is a confusing statement); if you are convinced this is the case please improve the explanation.

We propose to remove the statement about arguments coming together (we meant that both arguments contribute to the political and ethical implications of models). We propose the following as a start for this paragraph:



“Recognizing that models are shaped by society (Arg 1) and the scientific community (Arg 2) is extremely relevant, because models also shape society themselves: they have political and ethical implications.”

Line 160

Ethics from AI could indeed be used for ethics in numerical modelling (is this manuscript about numerical modelling only?). In my opinion this deserves somewhat more discussion. For instance, try to summarize the ethics field in AI and give suggestions how it could be converted to numerical modelling (or what we could learn from it).

Thank you for this relevant question. As indeed clarified above in response to this review, we mainly draw from numerical modelling, although we believe our arguments are equally valid for AI. We will elaborate on our AI argument.

The development of ethics in AI and responsible AI is useful and adaptable to hydrology. One recent development in this space is the growing emphasis on how system conceptualization can help with more reflective practices around AI modeling. For example, Nabavi and Browne (2023) propose the Five Ps framework to guide AI researchers and practitioners in situating their modeling work as interventions within competing perspectives on what constitutes a problem and how that framing can drive certain solution(s). The problem-solution dynamic can then be situated in specific zones of intervention - Parameter, Process, Pathway, and Purpose - each offering a leverage point with varying potential for change.

For example, when AI modelers address responsible AI challenges within the Parameter zone, they typically focus on quantifiable improvements through numerical adjustments and parameter tuning. In contrast, challenges identified in the Purpose zone seeks to engage with the fundamental norms, values, and worldviews embedded in modeling practices, which of course invites deeper questions, such as: what broader societal or ecological goals should guide modeling practice? (e.g., equity, resilience). This framework supports hydrological modelers in openly reflecting on their role in problem framing and discussing intervention zone. More importantly, it enables them to assess their work's transformative potential in catalyzing systemic changes, particularly for models designed with specific impacts in mind

We will add this to the section on AI.

Nabavi, E., Browne, C. Leverage zones in Responsible AI: towards a systems thinking conceptualization. Humanit Soc Sci Commun 10, 82 (2023). https://doi.org/10.1057/s41599-023-01579-0

Positional statement

Consider including a positional statement (or a short description of the background of the authors).

Ha, good point! That is the least we could do.

Positionality of the authors

The authors consist of a group of scholars, each critically engaging with the act of modelling from different entry points. We have diverse disciplinary backgrounds, including hydrological modelling, climate modelling, water governance, STS, and political ecology. Some approach modeling from direct technical experience, while others critique it through a political and social lens. Our perspectives on modeling range from pragmatic to deeply skeptical. Some of us actively use models in our work, while others grapple with finding ways to use models while also acknowledging their limitations, partiality and inherent injustices. Most of us are affiliated with institutions in the Global North and have a privileged (high-educated) position, which influences how we access, use, and critique modelling tools. We recognize that this academic and geographic positioning may limit our engagement with those most affected by modeling outcomes, although some of us have close relationships with people that have experienced marginalization through models. Several of us are actively involved with teaching hydrological modelling.  Some of us explicitly integrate reflexive practices in teaching, such as discussion on ontology, uncertainty, and situatedness, although we acknowledge that this is still evolving. Differences among the authors, in terms of background and experience, have been a source of productive dialogue, although, despite these differences, there was generally wide felt agreement.

Minor comments

Line 20

Reference(s) seem to be missing (after Packett et al, 2020)

We will fix this.

Line 22

‘might’. Consider ‘may’ or ‘will’

Thank you, we opt for “will”.

Line 64

‘purely technical’. Technical does not need to be neutral (not at all, see e.g. work by Latour). Reword and avoid ‘technical’ here.

Agree, actually this is the whole point we try to make.. We propose to reword to:

“to acknowledge that modelling is not neutral, but actively shapes worlds (Krueger and Albra, 2022).”

Line 110

‘tools and theoretical frameworks’, rewrite ‘theoretical frameworks and tools’ (theory comes first, tools are derived from the theory).

Agree, we will switch the order.