Abstract Large language models (LLMs) are playing a growing role in climate-focused efforts by enabling new opportunities to expand the scale and speed of climate research and communication. However, they also introduce risks that, if left unaddressed, could undermine these benefits, turning the technology into yet another roadblock to much-needed climate action. Drawing on a narrative review, this paper highlights LLMs’ existing applications in support of climate action alongside potential social, environmental, and epistemic implications. Through this critical assessment, the paper underscores the imperative for a responsible approach to LLM development and deployment in this burgeoning domain.

The development of AI models is increasing at a rapid rate. However, when are they ready to be deployed in real-world operational settings? In this paper, we introduce a framework to support such assessments and apply it to Google’s recently released AI-based flood prediction system, which is claimed to achieve “reliability in predicting extreme riverine events” and provide “accurate and timely warnings” that are available “earlier and over larger and more impactful events in ungauged basins”. The system has been integrated into an operational early-warning platform producing open, real-time forecasts in more than 80 countries. While this development promises to usher in a new and exciting age in global flood forecasting, the supporting evidence relies heavily on several subjective choices, the implications of which have not been acknowledged or assessed. Here, we evaluate the consequences of these choices on claims of operational deployment readiness across four dimensions: predictive accuracy, forecast timeliness, the characterization of extreme events, and benchmarking against state-of-the-art models. Our assessment reveals that the system’s actual predictive accuracy is likely to be substantially lower than reported—particularly for extreme events—raising concerns about responsible practices across modelling and publicity in high-stakes applications. The deployment of the Google AI model therefore risks misinforming those who depend on its outputs for evacuation and preparedness decisions, particularly in less-developed countries such as those targeted by the enterprise, given its alarmingly high (>90%) rates of false positives and false negatives. Beyond the immediate operational consequences, if left unaddressed, these outcomes may erode public trust in AI within hydrological sciences. We conclude by calling for greater transparency, accountability, and methodological rigor in the integration of AI into flood forecasting.

Abstract. Within hydrological modelling, a persistent notion exists that a model is a neutral, objective tool. However, this notion has several, potentially harmful, consequences, such as marginalising certain stakeholders. In the critical social sciences, the non-neutrality in methods and research results is an established topic of debate. Thus we propose that in order to deal with it in hydrological modelling, the hydrological modelling network can learn from, and with, critical social sciences. This is a call for responsible modelling – modelling that is accountable, transparent, power-sensitive, situated and reproducible and this responsibility is carried by all actors related to the modelling study. To support our proposition, we have four pillars of arguments, detailing the social aspects in hydrological modelling, insights from the critical social sciences, how to build bridges between sciences, and reflecting on what the hydrological modelling network can learn. We provide several actionable recommendations as a follow-up. The main take-away, from our perspective, is that responsible modelling is a shared responsibility. Therefore, we invite all actors – from the modelling network (from commissioner to modeller to end-user) and society – to take up their share in establishing responsible modelling.

In this Perspective piece, we forward a case for considering geographies of interdependence in understanding and implementing Responsible Innovation. At the core of this framing is an emphasis on how innovation occurring in specific places often depends on, and generates impacts in, places elsewhere. With respect to innovation, this involves taking into account the spaces, relations and networks that such activities dynamically depend on and actively produce. With respect to responsibility, this involves considering the consequences of particular innovation investments and trajectories across specific places, both near and far. Taken together, they highlight how considering interconnected geographical dynamics enlarges and enriches the stakes and commitments of Responsible Innovation. Mining in Australia is analyzed as a case study to demonstrate how tracing geographies of interdependence may contribute to conceptualizing and enacting Responsible Innovation in novel ways.

Explainable Artificial Intelligence (XAI) offers the promise of being able to provide additional insight into complex hydrological problems. As the “new kid on the block”, these methods are embraced enthusiastically and often viewed as offering something radically new and different. However, upon closer inspection, many XAI approaches are very similar to more “traditional” methods of “interrogating” existing models, such as sensitivity or break-even analysis. In fact, the approach of developing data-driven models to obtain a better understanding of hydrological processes to inform the development of more physics-based models is as old as hydrology itself. Consequently, rather than being considered a new approach, XAI should be viewed as part of a long-standing tradition, and XAI methods part of an ever-expanding hydrological modelling toolkit, rather than a silver bullet. Critically, there needs to be shift from focusing on how to best eXplain what AI models have learnt (i.e., the X component of XAI) to developing models that are able to capture relationships that are contained within the data in a robust and reliable fashion (i.e., the AI component of XAI), as there is little value in explaining AI-derived relationships if these do not reflect underlying hydrological processes. However, this is often not the case due to a focus on maximising the predictive ability of AI models “at all costs”, not uncommonly resulting in large models that often have thousands or even millions of parameters that are not well defined. Consequently, these models generally do not capture underlying hydrological processes in a robust and reliable fashion. Finally, there is also a need to stop thinking about XAI as a purely technical approach, but a socio-technical approach that views XAI as a process that can assist with solving problems that are situated within broader social and political contexts.

Whether we, as end-users of technology, are aware of it or not, our societies are becoming increasingly entangled in a complex network of interactions with Artificial Intelligence (AI) systems. This goes beyond what is often called ‘Human-AI collaboration’, involving the broader socio-political systems supporting these technologies <xref ref-type="bibr" rid="ref1" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">[1]</xref> . Faced with these complex interactions, society grapples with the timeless question: where does responsibility lie for the consequences or results produced by AI systems or applications, whether they are successful or not? Is it with the human operator, AI developer, user, or the AI agent itself? In the case of failure, AI cannot be held accountable, as such software systems are not yet recognized as separate legal entities <xref ref-type="bibr" rid="ref2" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">[2]</xref> .

Advancements in large language models (LLMs) provide opportunities to accelerate progress towards the attainment of the Sustainable Development Goals (SDGs). Current research largely overlooks the nuanced benefits and dangers LLMs introduce to sustainability research and communication, as well as broader challenges that need to be addressed in the longer term. This paper overcomes these shortcomings by introducing and discussing a framework that highlights how LLMs can benefit knowledge production, mobilization, and communication in the sustainability sciences, as well as any associated dangers. In addition, it outlines potential long-term challenges that must be acknowledged and addressed to ensure the responsible use of LLMs in advancing sustainability science. A key to the development and use of LLMs for sustainability science is the development of regulatory measures. These measures should be guided by what is needed for expanding sustainability science on the one hand and a holistic view to ensure its responsible use on the other. Failure to reflect and act on this might result in unintended consequences or misuse, making the technology another roadblock to progress towards the SDGs.

The new scientific decade (2023-2032) of the International Association of Hydrological Sciences (IAHS) aims at searching for sustainable solutions to undesired water conditions - whether it be too little, too much or too polluted. Many of the current issues originate from global change, while solutions to problems must embrace local understanding and context. The decade will explore the current water crises by searching for actionable knowledge within three themes: global and local interactions, sustainable solutions and innovative cross-cutting methods. We capitalise on previous IAHS Scientific Decades shaping a trilogy; from Hydrological Predictions (PUB) to Change and Interdisciplinarity (Panta Rhei) to Solutions (HELPING). The vision is to solve fundamental water-related environmental and societal problems by engaging with other disciplines and local stakeholders. The decade endorses mutual learning and co-creation to progress towards UN sustainable development goals. Hence, HELPING is a vehicle for putting science in action, driven by scientists working on local hydrology in coordination with local, regional, and global processes.

The notion of convergent and transdisciplinary integration, which is about braiding together different knowledge systems, is becoming the mantra of numerous initiatives aimed at tackling pressing water challenges. Yet, the transition from rhetoric to actual implementation is impeded by incongruence in semantics, methodologies, and discourse among disciplinary scientists and societal actors. This paper confronts these disciplinary barriers by advocating a synthesis of existing and missing links across the frontiers distinguishing hydrology from engineering, the social sciences and economics, Indigenous and place-based knowledge, and studies of other interconnected natural systems such as the atmosphere, cryosphere, and ecosphere. Specifically, we embrace ‘integrated modeling’, in both quantitative and qualitative senses, as a vital exploratory instrument to advance such integration, providing a means to navigate complexity and manage the uncertainty associated with understanding, diagnosing, predicting, and governing human-water systems. While there are, arguably, no bounds to the pursuit of inclusivity in representing the spectrum of natural and human processes around water resources, we advocate that integrated modeling can provide a focused approach to delineating the scope of integration, through the lens of three fundamental questions: a) What is the modeling ‘purpose’? b) What constitutes a sound ‘boundary judgment’? and c) What are the ‘critical uncertainties’ and how do they propagate through interconnected subsystems? More broadly, we call for investigating what constitutes warranted ‘systems complexity’, as opposed to unjustified ‘computational complexity’ when representing complex natural and human-natural systems, with particular attention to interdependencies and feedbacks, nonlinear dynamics and thresholds, hysteresis, time lags, and legacy effects.

This paper introduces and fleshes out a systemic method designed to develop a holistic understanding of states’ behavior in transboundary water conflict and cooperation. Such an approach leverages causality analysis to capture the deep structural characteristics that shape the hydropolitics dynamics and may lead to the evolution of destructive behaviors with severe consequences. The paper does so by using the concepts of the system archetype. The system archetype analysis offers insight into the underlying structures from which the dynamics of hydropolitics emerge over time—cycles of conflict and cooperation. The approach provides riparian states with a diagnostic tool to recognize patterns of destructive behaviors in the management of shared water resources and warning signs that are usually too long overlooked. Using four case studies from different continents, this paper shows how a systems archetype approach is useful for developing a big-picture understanding of the hydropolitical problem, its dynamics, and potential resolution pathways. The systemic lessons learned from these case studies can be used in other contexts, helping policymakers anticipate the destructive and constructive dynamics leading to conflict and cooperation.

There is a growing debate amongst academics and practitioners on whether interventions made, thus far, towards Responsible AI have been enough to engage with the root causes of AI problems. Failure to effect meaningful changes in this system could see these initiatives not reach their potential and lead to the concept becoming another buzzword for companies to use in their marketing campaigns. Systems thinking is often touted as a methodology to manage and effect change; however, there is little practical advice available for decision-makers to include systems thinking insights to work towards Responsible AI. Using the notion of 'leverage zones' adapted from the systems thinking literature, we suggest a novel approach to plan for and experiment with potential initiatives and interventions. This paper presents a conceptual framework called the Five Ps to help practitioners construct and identify holistic interventions that may work towards Responsible AI, from lower-order interventions such as short-term fixes, tweaking algorithms and updating parameters, through to higher-order interventions such as redefining the system's foundational structures that govern those parameters, or challenging the underlying purpose upon which those structures are built and developed in the first place. Finally, we reflect on the framework as a scaffold for transdisciplinary question-asking to improve outcomes towards Responsible AI.

Lessons from the COVID-19 pandemic inspire a guide to recognizing the politics of modeling.

The COVID-19 pandemic has shown the importance of modeling in guiding decision-making for governments and society, and the significant influence that modelers hold, especially during times of crisis. Water modelers may also encounter similar situations where their models are caught up in political debates, shaping people's everyday lives. This paper discusses the cultural and professional norms around water modeling practice that need to be established or revisited in order to make modeling work more responsible, through a review of models developed for COVID-19. It introduces six areas of study for "responsible water modeling" that can advance future theoretical and practical discussions on the topic: (1) building a common appreciation of the concept of responsibility, (2) interactions between science and policy, (3) the influence of boundary judgments on the model's outcome, (4) the politics of uncertainty, (5) stakeholder involvement, and (6) integration and coordinationThe paper suggests that by focusing on these subjects, the fundamental principles and characteristics of responsible modeling can be established in order to address and respond to water challenges while also serving the public good.

Description Lessons from the COVID-19 pandemic inspire a guide to recognizing the politics of modeling Lessons from the COVID-19 pandemic inspire a guide to recognizing the politics of modeling

There is a growing debate amongst academics and practitioners on whether interventions made, thus far, towards Responsible AI would have been enough to engage with root causes of AI problems. Failure to effect meaningful changes in this system could see these initiatives to not reach their potential and lead to the concept becoming another buzzword for companies to use in their marketing campaigns. We propose that there is an opportunity to improve the extent to which interventions are understood to be effective in their contribution to the change required for Responsible AI. Using the notions of leverage zones adapted from the 'Systems Thinking' literature, we suggest a novel approach to evaluate the effectiveness of interventions, to focus on those that may bring about the real change that is needed. In this paper we argue that insights from using this perspective demonstrate that the majority of current initiatives taken by various actors in the field, focus on low-order interventions, such as short-term fixes, tweaking algorithms and updating parameters, absent from higher-order interventions, such as redefining the system's foundational structures that govern those parameters, or challenging the underlying purpose upon which those structures are built and developed in the first place(high-leverage). This paper presents a conceptual framework called the Five Ps to identify interventions towards Responsible AI and provides a scaffold for transdisciplinary question asking to improve outcomes towards Responsible AI.

Studying the relationship between water expertise and the state's governance is important as it helps to explain the mechanism by which a certain group of experts rise to power, speaking for water—its challenges, and opportunities. This is particularly of concern in the times of crisis when the society does not know where to turn, and who to trust. Some aspects of this relationship have been addressed in the literature through now-familiar notions such as hydraulic bureaucracy and the hydraulic mission, in which the prevailing role of water engineers in problem framing and communicating solutions has been brought into the spotlight. However, the reciprocal nature of this relationship, particularly in difficult times when the society is fraught with fear of an uncertain future, has remained heavily under-researched. To fill this gap, this paper suggests we can productively draw on the concepts of “co-production” and “epistemic community”. Using Iran's looming water crisis, the paper provides an example of how governance and water engineering co-produce one another through an ongoing process of mutual constitution. On one hand, engineering artifacts are integral part of state-making process; while on the other hand, water engineers become the gatekeepers of knowledge-making processes. This creates a hegemonic power for water engineers, their epistemic practices, and institutions of power. This research also illustrates how this co-production reinforces the epistemic injustice in water governance by marginalizing non-engineering communities most particularly indigenous knowledge-holders. This is, of course, a great concern as it can lead to depoliticization of the water crisis, monopolization of water science, and demonization of participation in water governance.

The COVID-19 pandemic brought into focus the crucial role of computational modeling and simulation in informing and/or advising governments and societies, demonstrating the profound influence that modelers can have, particularly in times of crisis. This global experience has (re)shaped political and public demands on science and technology and the ways in which we incorporate scientific findings in our everyday lives. Through a brief review of models developed for COVID-19, this paper discusses the urgent need to move towards ‘responsible modeling’ in which social, ethical, and cultural issues claim the same importance as technical aspects. This includes a new social contract between modeler and society on one hand, and science and policy on the other hand. The characteristics of modeling work under this new paradigm, however, are still not well understood and are a topic of debate. This paper aims to fill this gap by proposing a new research agenda that can facilitate the discussion about ‘responsible modeling’ and the overarching principles that underpin the concept’s aims and objectives.

Water governing systems are twisted with complex interplays among levels and scales which embody their structures. Typically, the mismatch between human-generated and natural systems produces externalities and inefficiencies reflectable in spatial scales. The largely known problem of fit in water governance is investigated to detect the issues of fit between administrative/institutional scales and the hydrological one in a lake basin. To implement the idea, constraining the level of analysis interlinked to the concentrated levels of administration in spatial scales, the fit of the governing system was analyzed by means of statistical mechanics. Modeling the structure of water demand/supply governing system in a given region through the Curie-Weiss Mean Field approximation, the system cost in relation to its structure and fit was appraised and compared with two other conceptual structures in the Urmia Lake Basin in Iran. The methodology articulated an analysis framework for exploring the effectiveness of the formulated water demand/supply governing system and its fit to the relevant hydrological system. The findings of this study may help developing strategies to encourage adaptations, rescaling/reforms for effective watershed management.