*Editorial* **Flood Risk Governance for More Resilience—Reviewing the Special Issue's Contribution to Existing Insights**

#### **Piotr Matczak 1,\* and Dries L. T. Hegger <sup>2</sup>**


Received: 25 June 2020; Accepted: 23 July 2020; Published: 26 July 2020

**Abstract:** There is lively scholarly and societal debate on the need to diversify flood risk management strategies to contribute to more flood resilience. The latter requires dedicated governance strategies related to which relevant insights are currently emerging. However, more systematic theoretical and empirical insights on how to specify and implement governance strategies are still urgently needed. The Special Issue 'Flood Risk Governance for More Resilience' has brought together nine contributions by renowned flood risk governance scholars that together help to unpack lessons about these governance strategies. This Special Issue's editorial introduces the debate on flood risk governance for more resilience and presents the key findings of the individual contributions to the Special Issue. We show that flood risk governance arrangements in specific regions in the Netherlands, Germany, Switzerland, Bangladesh, France, and Mexico are gradually evolving. A common denominator is that more horizontal forms of governance are under development in which a more diverse array of public and private actors—including citizens, as well as different sectors, is becoming involved. Efforts are underway to establish connectivity between actors, levels, and sectors, both through regional and international exchanges. While lessons on how to do the former successfully are emerging, we notice that these should still be unpacked more fully. Moreover, there is still a need to establish a more open and inclusive societal debate on societal preferences regarding flood risk protection in which all actors with a stake in flood risk governance processes and outcomes can participate.

**Keywords:** flood risk governance; resilience; adaptation; learning; science-policy interactions; interdisciplinarity

#### **1. Introduction: Debate on Resilient Flood Risk Governance**

Flood risks have been increasing worldwide in the last decades. As a result, fundamental and long-lasting assumptions in our understanding of how to deal with floods are being questioned [1–3]. A basic tenet of water managers used to be that physical conditions are relatively stable. However, climate change induces changes in the very nature of these physical conditions [4]. Second, ongoing urbanization exacerbates flood risks in many regions, mostly in deltas [5]. Third, it is increasingly argued that flood risk management cannot be organized solely in a top-down fashion, with national governments, in particular their offices of Public Works, in a leading position [6]. Recent insights show that the long-held belief in our capacity to control natural processes and the wish to adopt managerial reactions have missed the point. Fourth, in accordance with the more general transformation in functioning of public administrations, starting with the participatory revolution in the 1970s, new actors and social groups aspired to have a voice in flood management. On top of all this, since the 1980s, market-based logics have been introduced to the formerly state-controlled water sector. These four

factors challenge the predict-and-control approach that was dominant in flood risk management until recently.

The notion of flood risk governance (FRG) was coined in this respect, depicting a new organizational approach [6], which addresses the turn from flood defense to flood risk management [7,8] and the growing role of previously disregarded stakeholders. Bergsma [3] treats FRG as a distinctly different form of dealing with floods as compared with the welfare state model that forms the backbone of more traditional approaches. First of all, the models differ in terms of the values they rely on. The welfare state model assumes that a centralized actor takes care of the public interest and collects resources to achieve this aim. A basic principle of this model is that flood managers should provide security defined as, ideally, absolute protection. Contrary to the welfare state model, the FRG value system implies that a wide array of public and private actors, including the ones at the local level, becomes involved in FRG. Such diversification of the actor base is expected to allow for cost sharing and more efficient information exchange. Additionally, and more widely debated, the FRG model accepts that not all floods can always be prevented and there will be inevitable losses. The task is to manage these losses by minimizing them in order to make them acceptable. In this respect, Kundzewicz et al. [9] write about the need to move from a 'fail-safe' system to a system that is 'safe-to-fail'.

The involvement of multiple actors, levels, and sectors in FRG implies that multiple types of solutions, but also potentially competing interests, are brought to the table. This makes governance systems more network-like, as opposed to a top-down management setting based on a deterministic view of nature and social life. We witness the effects of this in practice. On the policy level, the discussions on available flood risk management options have started to become more open for stakeholders willing to participate. More often than not, skilled and professionally trained hydrologists have started to be one of the involved groups instead of the ultimate solution provider. As part of this overall shift towards more multi-actor and decentralized governance, as observed in several countries [10], the need for participatory processes has been institutionalized in legal systems. Mandatory stakeholder consultation on implementation, as laid down within the EU Water Framework Directive, is a case in point. Participatory processes in a European context also received a boost through the Aarhus Convention that establishes rights of the public to access environmental information, public participation and justice (https://ec.europa.eu/environment/aarhus/).

Within the flood risk governance debate, new concepts were incorporated. Notably, the notion of resilience has been translated from the field of ecology in which it emerged. In the past years, the resilience notion has been specified for the floods domain, systematically unpacked and translated into governance strategies [5,10,11]. A core tenet of thinking in terms of flood resilience is to search for integrated flood risk management approaches that comprise multiple probability and consequence-reducing strategies. These strategies need to be adjusted to local conditions [5,8,10]. A diversity of strategies is key to advancing flood resilience [11]. The discussion on flood risk governance for more resilience requires input from different perspectives and is therefore inherently multi-disciplinary [8]. Besides hydrological and technical sciences, perspectives of public administration, law, human geography, ecology, amongst others also have the potential to contribute to advancements of more resilient governance settings.

The Special Issue 'Flood Risk Governance for More Resilience' aims to build on the emerging insights referred to above. Apart from the editorial, it includes nine papers from diverse geographical contexts that adopt a diversity of multi-disciplinary perspectives. The current paper provides an overview of the Special Issue and reflects on its contribution to the existing state of the art. In order to do so, section two first provides a rough sketch of current knowledge on flood risk governance for more resilience by discussing some recent well-informed contributions to the debate. This leads to the identification of relevant insights, but also apparent knowledge gaps regarding the aforementioned roles of citizens and stakeholders, FRM policies and measures, and tools. Section three presents the key findings of each of the nine papers in the Special Issue. In section four, we wrap up by reflecting on the

papers' joint contribution to the identified knowledge gap and present a revised research agenda for studies into resilient flood risk governance.

#### **2. State of the Art: Knowledge Gaps and Proposed Future Directions in FRG Literature**

#### *2.1. Knowledge Gaps in FRG Literature*

A growing literature has explored the application and specification of the resilience concept to the floods domain. In 2018, Morrison et al. [12] assessed the last decade of scholarship in this field and from this distilled five key avenues for further inquiry:


While Morrison et al. focused on scholarship pertaining to governance for more resilience, McClymont et al. [8] centered more on the dependent variable of what resilience entails and what it is that should be made more resilient. They differentiate between three conceptualizations of flood resilience employed by scholars: engineering resilience, systems resilience, and complex adaptive systems resilience. The authors observe that seldom are all three frameworks taken into account in the analyses, with only 15% of papers covering all three frameworks. In addition, the authors found that studies are restricted in terms of the precise topics they include as elements of resilience. For example, they observe that at larger spatial scales, resilience is often used interchangeably with resistance, a concept with a different normative starting point. Another case in point is that in terms of flood risk management strategies, studies often focus on flood recovery.

According to McClymont et al. [8], a key knowledge gap is the actual and necessary division of responsibilities between actors. They notice a shift in responsibilities from state protection to individual responsibility [3]. They argue, though, that this shift is highly geographically and institutionally situated. Processes leading to or hampering resilience can be strikingly different in the well-developed countries and in the Global South. In addition, across contexts, top-down and bottom-up approaches in building resilience need to be combined in a multi-level governance framework. The authors note that the aforementioned issues are being experimented with in practice, but we still lack systematic insights on how to allocate responsibilities.

The contributions by Morrison et al. [12] and McClymont et al. [8] provide an overview of important knowledge gaps and challenges for research. Morrison et al. emphasize that more specific insights are needed to enhance connectivity in several respects in order to address (institutional) fragmentation: (1) They point at the need to better connect natural and social science-based contributions. Now they are often separated, amongst other reasons because they are fueled by different flows of research funding. (2) They observe that knowledge-action networks need to be strengthened to improve the

science-policy interface in flood risk governance. (3) They point out the need to establish connectivity between professionals and stakeholders and see a lack of tools to address this social dimension of flooding. (4) They point out the need to integrate insights arising from governance research and governance models external to the floods domain into flood risk management literature.

Morrison et al. [12] acknowledge that some progress has been made in relation to the aforementioned knowledge gaps. They show that studies on modeling tools and on stakeholder engagement and public participation have been getting increasing attention. In addition, notable advances have been made regarding the translation of governance-related knowledge to the floods domain. Nevertheless, more progress is still to be made and, in particular, the operationalization of flood resilience requires more research, as also argued by McClymont et al. [8].

#### *2.2. Proposed Future Directions in FRG Policy and Practice*

While engaging with the aforementioned debate on flood risk governance for more resilience, Driessen et al. [5] proposed six governance strategies for improving flood resilience in the face of climate change. These have been identified based on comparative empirical research in a European context; hence, these lessons have been empirically validated to some extent, albeit mostly for Europe.


These strategies have some degree of generalizability as they have been validated in different contexts in Europe while their validity in other contexts is plausible. To the best of our knowledge, these strategies constitute the most encompassing overview of necessary improvements in flood risk governance literature and practice. Therefore, the current paper uses these strategies to organize the key findings of the Special Issue.

#### **3. Logic and Outline of the Special Issue**

In this Special Issue, nine papers deal with various aspects of 'Flood Risk Governance for More Resilience'. Eight papers cover a variety of cases in different countries while one paper is a review paper. Section 3.1 first discusses the key findings of each of the nine papers in the Special Issue in turn. Next, Section 3.2 links these key findings to the six aforementioned governance strategies for improving flood resilience and discusses the contribution of each paper to these strategies.

#### *3.1. A Summary of Papers of the Special Issue*

#### 3.1.1. Flood Risk and Resilience in the Netherlands: In Search of an Adaptive Governance Approach

Astrid Molenveld and Arwin van Buuren [19] have analyzed the discursive shift towards more resilience-based approaches in the Netherlands that took place at the end of the 2000s as well as its implications for flood risk governance practice. At that time, the multilayered safety concept (MLS) was coined and started to be discussed. MLS is a policy concept designed for adaptive flood risk management in the Netherlands. The authors took an innovative conceptual approach by using Elinor Ostrom's Institutional Analysis and Design (IAD) framework to analyze this shift. Relying on her polycentric and adaptive governance framework, they were able to point out a concrete timing of the shift, as they compared the FRG approaches before and after the year 2008. The multi-layered safety approach was introduced as a part of the Dutch Delta Program. It aimed to produce more adaptive FRG as it involved a more polycentric and loosely structured institutional regime. This new regime was to be more inclusive and it was to stem from collaboration of wide set of stakeholders. The 'multilayered safety' policy was intended to be loose but in practice it was found to be fairly tight, i.e., a system where rules strictly determine adaptation action. Authors attribute this relative failure to the fact that dealing with floods has deliberately been presented as a 'tame' problem for decades, which means that existing solutions turned into an institutionalized and hard to change routine. As the authors conclude, 'adopting a more adaptive and polycentric approach necessitates 'untaming' the issue of flood safety'. Issues that have been taken for granted need to be put on the table for discussion in an open debate.

#### 3.1.2. Social Learning in Multilevel Flood Risk Governance: Lessons from the Dutch 'Room for the River Program'

Learning and accumulation of knowledge among actors is widely acknowledged [20] as crucial for water and flood risk governance in particular. Jacomien den Boer, Carel Dieperink and Farhad Mukharov [21] looked at factors influencing learning processes. The authors analyze the example of the Dutch 'Room for the River Program' to identify enabling conditions for social learning in multi-level flood risk governance arrangements. They integrated concepts present in FRG: adaptive co-management, sustainable land and water management, and integrated flood risk management, concluding that all these concepts assume that social learning is a multi-level and multi-stakeholder governance challenge. Thus, cooperation occurs between different sectors, including those of water management, spatial planning, and disaster management. Cross-sectoral cooperation is arguably challenging in itself. It appears even more difficult though, if multilevel cooperation is involved. The study differentiates between four types of factors influencing a cooperation and learning process: attributes of engaged individuals; collaborative arena factors (e.g., mutual trust, communication); organizational factors (e.g., cooperation structures, knowledge sources); and external factors (e.g., crisis events, administrative procedures). Examination of the Dutch Room for the River Program showed that a strong personal commitment to learning and mutual inter-personal trust in working groups were key conditions for successful social learning. External factors played a less significant role.

#### 3.1.3. City-To-City Learning for Urban Resilience: The Case of Water Squares in Rotterdam and Mexico City

Similarly to den Boer et al. [21], Silvana Ilgen, Frans Sengers and Arjan Wardekker [22] have studied city-to-city learning for urban resilience. They assessed the on the ground implementation of water squares in Rotterdam and Mexico City and therewith provided important insights about the functioning of knowledge-action networks (in Morrison et al.'s terms [12]). A key finding of the paper is that city-to-city learning took place within identifiable phases: exploration and marketing (phase 1), building pipelines (phase 2), translation and adoption (phase 3), and internalization and reflection (phase 4). The authors point out that it was critical in a first phase to analyze one's own systems, strengths, and weaknesses, rather than performing an outward-looking search for knowledge or mentees. Next, cities reframed their own narratives to match those of their counterparts as a way to create a mutual understanding of each other's developments. A process of policy and knowledge exchange could take place because of that. However, strong leadership turned out to be necessary to make sure that the acquired knowledge was implemented and retained. Fourth and finally, the authors stress that 'by internalizing such lessons, cities might strengthen not only their own resilience, but also enhance future exchanges with other cities'.

#### 3.1.4. Pluvial Flooding in Utrecht: On Its Way to a Flood-Proof City

Citizen engagement is recognized as an important issue in order to make progress in flood risk governance and hence a recognized study area in flood risk governance analyses. Romy Brockhoff, Steven Koop, and Karin Snel [23] have assessed the topic in the case of pluvial flooding in Utrecht. They assessed to what extent the necessary governance capacities to make the city prepared for pluvial flooding have been developed within the city. The authors depart from the viewpoint that pluvial flooding can be addressed by a single actor that is in the lead, such as the municipality, and indicate that addressing pluvial flooding requires the involvement and engagement of a diverse set of actors, including citizens. Applying the governance capacity framework to the city of Utrecht, the authors found that most governance capacities needed to address pluvial flooding have been relatively well-developed: 'collaboration between public authorities is advanced, sufficient financial resources are available, and smart monitoring that enables high levels of evaluation and learning.' Citizen awareness and engagement is, however, in need of further development. The authors recommend developing financial incentives that invite citizens to take measures to address pluvial flooding on their own properties and advise to further develop arrangements for active citizen engagement. The authors argue that these recommendations are valid for other urbanized areas that will face increased problems with pluvial flooding.

3.1.5. Exploring Science-Policy Interactions in a Technical Policy Field: Climate Change and Flood Risk Management in Austria, Southern Germany, and Switzerland

Science is an important factor in FRG. Scientific evidence and expertise deliver solutions and justification for FRG policies and management. Ralf Nordbeck, Lukas Loeschner, Melani Pelaez Jara and Michael Pregernig [24] in their paper on science-policy interactions in the field of flood risk governance analyzed three Alpine regions, in Switzerland, South Germany, and Austria. They assess science-policy interactions from three perspectives: (i) dynamics of knowledge creation; (ii) institutionalization of the science-policy interface; and (iii) pathways of influence of expertise on policy development. The authors found increasing influence of climate change on flood risk governance in the selected regions. Policies to address climate change were supported by evidence-based arguments. The influence of experts was significant; however, it was mediated by national factors. This social embeddedness of expertise was heavily mediated by the 'political climate'. Scientists had to adjust their knowledge to have their expertise recognized. Notably, in South Germany, the high political profile of the climate change issue inclined scientists to recommend climate change as a significant factor despite a very uncertain scientific basis.

#### 3.1.6. The Costs of Living with Floods in the Jamuna Floodplain in Bangladesh

Ruknul Ferdous, Anna Wesselink, Luigia Brandimarte, Kymo Slager, Margreet Zvarteveen, and Giuliano di Baldassarre [25] assessed how residents of the Jamuna floodplain in Bangladesh responded to flood events. Flood resilience literature often stresses that external shocks can induce learning and therewith facilitate adaptation and transformation, ultimately leading to a more resilient state [26]. A study by Ferdous et al. [25] challenges this assumption. They found that Bangladeshi people do develop strategies to cope with floods, such as relocation, temporary evacuation, change in cropping patterns, and supplementing their income from migrating household members. While these

strategies facilitate short-term coping and reduce the negative impact of floods on their livelihoods, they do not prevent impoverishment. The authors conclude that the inhabitants of the Jamuna floodplain do not achieve successful adaptation and that gradually their situation worsens.

#### 3.1.7. Adaptive Capacities for Diversified Flood Risk Management Strategies: Learning from Pilot Projects

The development of diversified flood risk governance is a process that requires both a vision and capacities. Flavia Simona Cosoveanu, Jean-Marie Buijs, Marloes Bakker, and Teun Terpstra [27] focus on adaptive capacities observed in the implementation of two pilot projects: 'Alblasserwaard-Vijfheerenlanden' (The Netherlands) and the 'Wesermarsch' (Germany). The projects aimed at enhancement of the integration of mitigation and preparedness measures. The authors looked for capacities that were missing, employed, and developed throughout the implementation, via the lens of the Adaptive Capacity Wheel that identifies 22 adaptation capacities instrumental in organizational adaptation, and the Triple Loop Learning approach, that measures the depth of learning in an organization. The study found three capacities particularly important to diversify the current form of Flood Risk Management: the capacity to develop a greater variety of solutions, continuous access to information about diversified FRMS, and collaborative leadership. The authors furthermore found that, in both cases, mostly shallow as opposed to deep learning took place. The study suggests that changing FRG to make it more diversified faces organizational hurdles, as in any organizational change. Thus, any process of updating FRG requires taking into account and anticipating a certain level of organizational resistance.

3.1.8. Reducing Hydro-Meteorological Risk by Nature-Based Solutions: What Do We Know about People's Perceptions?

The recently developed concept of Nature Based Solutions (NBS) points at the possibility of developing approaches 'to sustainably reduce hydro-meteorological risks, providing co-benefits for both ecosystems and affected people'. This way, such approaches are supported by nature, and are cost-effective and conducive to resilience. Sungju Han and Christian Kuhlicke [28] reviewed factors shaping people's perceptions of NBS as a means to reduce hydro-meteorological risks, including floods. The authors identified the following six core topics within which perceptions of NBS were discussed: (1) valuation of the co-benefits; (2) evaluation of risk reduction efficacy; (3) stakeholder participation; (4) socio-economic and location-specific conditions; (5) environmental attitude, and (6) uncertainty. The authors noted ambiguous and even contradictory results of the studies and propose a conceptual model for future research. The model comprises socio-economic-demographic conditions, operational knowledge, trust, threat appraisal, environmental attitudes, and direct interaction with NBS.

#### 3.1.9. What Can We Learn from Planning Instruments in Flood Prevention? Comparative Illustration to Highlight the Challenges of Governance in Europe

Planning is crucial for flood governance. Mathilde Gralepois, in her paper [29], examines flood prevention planning instruments in three European countries: England, France, and the Netherlands. Maps were shown to be a vital and powerful tool in flood governance, in accordance with the Flood Directive requirements. Maps were not only geographical representations of territories, but—more importantly—a part of legal zoning, excluding certain types of land use. However, as socio-technical objects, flood maps are negotiated by actors engaged in their preparation and application. Gralepois claims that this process laid bare tensions that hamper a potentially beneficial role of planning in flood risk governance. Firstly, local spatial planning often conflicts with flood prevention policies implemented by national authorities. Local bodies often have more precise data and expertise and challenge the strict approach of the central administrations. In England, local planning authorities often disregard the non-mandatory recommendations of the central authorities, in order to continue local development. The situation in France is a manifestation of local authorities' pursuit to extend

their autonomy. In the Netherlands, the situation is more balanced. The Water Impact Assessment is a mandatory procedural instrument to ensure that local plans fit the national criteria, but it is non-binding. Secondly, there is tension in professional culture between planning and prevention, which results in different preferences concerning instruments. The flood management administration prefers instruments referring to hydraulic models. Within the planning domain, allocation of land for different purposes and local development is the main concern. As a result, establishing a platform allowing debate and reconciliation of interests and expertise could contribute to balancing flood prevention and development needs. In France the Flood Risk Prevention Plan, and in the Netherlands, the Water Impact Assessment can play such a role.

#### *3.2. Contribution of the Special Issue to the Discussion about Governance Strategies for Improving Flood Resilience*

#### 3.2.1. Context-Sensitive Diversification of Strategies

Several papers in the Special Issue contributed insights to the governance strategy of achieving a context-sensitive diversification of strategies. Molenveld and Van Buuren [19] illustrated the challenges of discussing and implementing such a diversification in the Netherlands, a country that had relied on predict-and-control approaches and flood defense for decades. They showed that several path-dependency mechanisms may have a tendency to reduce the scope of the debate on diversification of flood risk management strategies and may water down its implementation in practice. In the words of the authors, multi-layered safety was introduced at a time in which flood risk governance was a 'tame' issue, while the issue would have needed 'untaming' first. Regarding context-sensitivity, Ilgen et al. [22] identified an interesting mechanism. They showed that city-to-city learning processes can induce processes of international inter-city comparison, in which actors analyze their own systems and deliberately reframe their own narratives. Gralepois [29] showed that the difference in professional culture between planning and prevention hinders flexibility in applying instruments. In addition, in centralized flood risk governance, actors tend to favor the application of standardized instruments. However, the tension between the need for coordination and the necessity to take local interests into account needs to be acknowledged and addressed. In line with this claim, Cosoveanu et al. [27] showed that pilot projects can provide learning experiences for achieving diversified flood risk management. However, it remains challenging to achieve forms of second order learning that invite actors to reconsider dominant approaches and assumptions. There is a tendency to return to business as usual since higher order learning may lead to resistance. The papers together focus mainly on the policy side of FRG. In particular, factors hampering and enhancing the diversification of flood risk governance strategies have been revealed.

#### 3.2.2. Involvement of Different Sectors in Flood Risk Management Strategies

Ilgen et al. [22] concur with the need to involve a wide range of sectors. They point at mechanisms that can enhance city-to-city learning. They show that what is first needed is to enhance connectivity within and knowledge of their own water governance system. This is a prerequisite to engage in city-to-city learning. In addition, den Boer et al. [21] unpack this key governance strategy. They claim that flood risk management strategies should be linked together and aligned, not stay isolated. Involvement of different sectors (water management/spatial planning/disaster management) is needed. Den Boer et al. argue, on the one hand, that social learning requires multi-sector and multi-level cooperation. On the other hand, they presuppose that integration between different sectors requires and to some extent enables social learning. Gralepois [29] underlines difficulties in cooperation between planners and flood prevention specialists. Examination of the Dutch Room for the River Program showed that a strong personal commitment to learning and mutual inter-personal trust in working groups were key conditions for successful social learning. External factors played a less significant role. This lesson also recurs in the paper by Brockhoff et al. [23]. They show that pluvial flooding requires

the engagement of a diverse set of actors from different sectors and reveal the challenges related to their involvement. These papers explore the topic of stakeholder engagement in FRG. Barriers in engagement of sectors have been indicated as well as difficulties related to an engagement process. The need for learning is articulated in particular.

#### 3.2.3. Involvement of Private Actors Including Residents, Businesses, and NGOs

Iligen et al. [22] note the need to involve private actors in building urban resilience. In addition, Molenveld and Van Buuren [19] have seen increased involvement of other types of actors in flood risk governance. They note, though, that it is difficult to involve these private actors in a Dutch context. Often, newly entering actors are still public rather than private ones. Broadening the actor base—referred to by the authors as achieving more polycentricity as well as becoming more adaptive and less static—is not something that can be achieved overnight. The authors show that in the Dutch case this requires a re-politicization or, in the terms of the authors, an un-taming of a tame problem. Brockhoff et al. [23] claim that pluvial flooding requires the engagement of citizens. Amongst all resources, the substantive contribution that citizens can offer to address pluvial flooding is most in need of improvement. Han and Kuhlicke [28] contributed to the theme of actor involvement by studying a specific response to flood risks: nature-based solutions. They highlighted several relevant elements that co-determine the perception of different types of actors of these nature-based solutions as described in the previous sub-section. These papers contribute to the research gap concerning stakeholder engagement and the reshuffle of responsibilities, from the state to private actors and residents. The studies note both pulling and blocking factors in this respect. The studies suggest that engagement of private actors can be a postulate difficult to achieve due to institutional, economic, and social reasons.

#### 3.2.4. Rules and Regulations

Gralepois' paper [29] focuses on instruments applied in flood risk governance in England, France, and the Netherlands. In all three countries, legal instruments dominate. Moreover, the hierarchical legal structure leaves limited space for diversification of policy instruments. Rules and regulations are implicitly discussed in the papers of the Special Issue, although Molenveld and Van Buuren [19] touch upon this issue when they point at a 'loose' vs. a 'tight' implementation of the multi-layered safety policy, which alludes to the fact that existing rules and regulations in some cases may have a tendency to reinforce stability and path dependency. These papers contribute to the research gap concerning flood risk management policies. The dominant role of legal instruments and their stabilizing character are noted.

#### 3.2.5. Financial and Non-Financial Resources

Ferdous et al. [25] point out the fact that dealing with flooding in a Bangladeshi context can also lead to a vicious circle where coping strategies cannot prevent negative long-term outcomes such as impoverishment. In terms of strategy number five: the resource base (in particular: human capital, but also livelihood resources) is steadily declining, which undermines communities' resilience.

Nordbeck et al. [24] shed light on specific resources for improving flood resilience in the face of climate change. They point to the importance of science-policy interfaces and their role in making flood risk governance climate-sensitive. They show how climate-related information is used in Switzerland, South-Germany, and Austria. The mechanisms through which this information was used differed in the three countries, but in all cases reliance on experts and on evidence-based information was high. The paper shows, however, that the extent to and ways in which scientists' expertise was recognized depended on the political climate in the different countries. Brockhoff et al. [23] in their paper write, amongst other resources, about the substantive contribution that citizens can offer to address pluvial flooding. This resource is said to be most in need of cultivation. Cosoveanu et al. [27] in their assessment of adaptive capacities for diversified flood risk management in two case studies

found that three types of capacities are particularly important: the capacity to develop a greater variety of solutions, continuous access to information about diversified FRMS, and collaborative leadership. The issue of resources dealt with in the papers refers to the division of responsibility between state and non-state actors and flood risk governance policies in general. The studies reveal that significant resources can be found in the state, and in the private sector, the use of the private sector resources is conditional. Moreover, the resources are vulnerable.

#### 3.2.6. Open and Inclusive Societal Debate

The paper by Molenveld and Van Buuren [19] is the only paper that explicitly addresses the need for an open and inclusive societal debate. The forms of re-politicization or un-taming pleaded for by the authors can be seen as a specific way to implement governance strategy number six of Driessen et al. [5]: to achieve an open and inclusive societal debate that leads to the adoption of certain normative principles as the outcome of a political discussion. While the other papers do not address this last governance strategy explicitly, they all implicitly hint at it. Recurring issues in all contributions are that different types of actors are becoming and have to become involved in flood risk governance; that these may have different viewpoints and often vested interests; that institutional change is necessary, and that there are various mechanisms that make institutions relatively inert to change. This is a setting in which, arguably, the need to put un-debated issues up to deliberation again becomes increasingly important. Moreover, the paper by Ferdous et al. [25] strongly points to the potential detrimental and/or distributive effects that a resilience discourse may have, in addition to several positive effects. This shows all the more how important it is that flood risk governance is debated by diverse societal actors.

#### **4. Concluding Remarks and Suggestions for Future Research**

The nine contributions to the Special Issue 'Flood Risk Governance for More Resilience' together have advanced the state of the art in scholarship on flood risk governance. At the same time, they have also laid bare the limitations that this literature still has. A dominant message that can be derived from the papers is that more horizontal forms of governance are being developed in which a more diverse array of public and private actors, including citizens as well as a diverse array of sectors is becoming involved. Efforts are underway to establish connectivity between actors, levels, and sectors, both through regional and international exchanges. The diversity of lessons provided by the papers in the current Special Issue signals the even larger diversity existing in empirical reality. Despite recent progress, including in this Special Issue, it is safe to say that the empirical knowledge base regarding governance strategies for achieving flood resilience still needs to be significantly expanded, whereby different aspects of the six governance strategies for more resilience should be further unpacked. We also note that there is still a need to expand the geographical scope.

Secondly, although the papers demonstrate engagement of various actors within FRG, analyses of the mechanisms through which participation is taking place and the underlying power relations remain a bit shallow. It seems that actor participation often takes place through mechanisms close to cooptation, but these mechanisms are not yet systematically unpacked. In particular, within and between actor conflicts are hardly explored. It is an open question whether these issues of power in participation are missing because they would add too much complexity to research, or that conflicts are "suppressed" in FRG and do not appear as an issue. In any case, it seems unlikely that FRG is always a non-zero-sum game with winners only. Exploration of this aspect is an area for future studies.

A third key message, one closely related to the second, is that the scope of normative debates on what flood resilience entails and who should be resilient to what is still too limited in scope, both in literature and in practice. The Special Issue reconfirms the finding of Driessen et al. [5] that normative debates are often absent or not accessible to all actors with a stake in flood risk governance processes and outcomes. Further lessons on how to improve this practice need to be drawn.

An element in scholarly literature that is relevant here is that insights from the natural and social sciences are often still separated. While flood perception is addressed in FRG studies, most literature adopts a 'realist' approach to the environment. We endorse the suggestion by Birkholz et al. [30] to add more constructivist studies that try to unpack how flood risks and flood risk management approaches are understood, interpreted, framed, and given meaning and how this influences the terms of the debate, including the substantive outcomes for different types of stakeholders in terms of dynamics in power relations and in terms of actual flood protection. Another aspect of the uneasy relation between the social and the natural sciences that appears in the papers is a weak reference to the concepts worked out in the water management literature. For instance, the concept of Integrated Water Resource Management necessitates looking at the processes using the river basin scale as a unit of analysis. This is hardly done in the studies.

The Special Issue combines the notion of flood risk governance with the notion of resilience. The papers discuss several factors enhancing resilience: necessity of learning [22,27]; organizational capacities [22,27]; the role of changes in narratives [22]; infrastructural individual/communal capacities [25]; the role of citizens' engagement [23]; citizen awareness and engagement [23]; "taming" of new policies [19]; problem solving capacities [23]; knowledge/expertise production and application [24]; personal commitment to learning and mutual interpersonal trust [21]. Whether the analyzed influence of particular factors is idiosyncratic or representative of more general patterns is an issue to be corroborated. Notably, the claim by Ferdous et al. [25] that presence of diverse strategies does not automatically imply resilience is deserving of further research.

The papers of this Special Issue rely mostly on case studies. They offer in-depth insights into the process and organization of flood risk governance. A problem with case study research is its limited external validity. In terms of the need for cumulative research, some sort of meta-language would need to be used by researchers to make findings comparable. The key themes of Section 3.2 may provide a first exploration towards such a meta-language.

We encourage scholars and practitioners from diverse scientific perspectives to contribute further to these debates.

**Author Contributions:** Conceptualization, P.M. and D.L.T.H.; methodology, P.M. and D.L.T.H.; formal analysis, P.M. and D.L.T.H.; writing—original draft preparation, P.M. and D.L.T.H.; writing—review and editing, P.M. and D.L.T.H. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Flood Risk and Resilience in the Netherlands: In Search of an Adaptive Governance Approach**

#### **Astrid Molenveld 1,2,\* and Arwin van Buuren <sup>1</sup>**


Received: 30 August 2019; Accepted: 2 December 2019; Published: 5 December 2019

**Abstract:** In the Netherlands, dealing with the risk of flooding in the face of the current climate change requires a governance approach that is less based upon the long-standing tradition of prevention and protection, and more oriented toward ideas of resilience and adaptivity. Such an approach is assumed to be more resilient compared to static approaches and better equipped to deal with the indeterminate character of a problem like flood risk. This article presents the Dutch attempt to introduce a more polycentric and adaptive governance approach in flood management, called multilayered safety (MLS). We studied this approach via interviews and an extensive document study, and analyzed the institutions governing the issue using the Institutional Analysis and Development (IAD) framework of Elinor Ostrom. For years, the issue was in the hands of a small network of actors, mainly occupied by water experts and governed by a strong lead organization and permanent bodies. While introducing a new, more adaptive policy concept the government encountered both resistance and inability within the existing policy regime. This article shows that the issue of flood safety was successfully 'tamed' for decades. Adopting a more adaptive and polycentric approach necessitates 'untaming' the issue of flood safety.

**Keywords:** IAD framework; adaptive governance; multi-level safety; untaming

#### **1. Introduction**

The risk of flooding grows because of more extreme rainfall, sea-level rises, higher river water discharges, and more intense storms. Scholars are, therefore, heavily discussing the basic principles of the current approach of dealing with flood risk. For a long time, the Dutch approach to flood-risk management was mainly based upon the idea of prevention [1]. Over the last 10 centuries, a whole system of dikes and dams was developed in order to protect the Low Countries from flooding. This system was further refined following the major flooding of 1953, after which the world-famous delta works were realized, and a whole regime of legal flood-risk norms was established. To date, the Dutch system of protecting a highly flood-prone country against flooding is seen as an international hallmark of flood-risk management.

This system can be described as a 'domesticated or tamed' strategy [2]. In such a system, an ill-structured problem, such as water safety, is split up into manageable partial tasks [3] (p. 578), resulting in a lack of reflexivity, resilience, and inclusion of new and local experience and knowledge [2] (p. 11). Problem taming is aimed at reducing and controlling wicked problems, by scoping and delineating the problem in such a way that it fits the existing administrative expertise and policy responsibilities [3]. In other words, the existing organizational structure and body of knowledge define how a problem is dealt with. Roberts [4] adds to this the idea that in the case of taming, authority is transferred into the hands of a few selected actors, based upon their hierarchical position and with the authority to deal with the (strictly defined) issue. Roberts [4] wrote that these 'tame' systems

are characterized by authoritative strategies based on preexisting organizational lines of functional specialization [3] (p. 578). Ostrom [5] and also Hardin [6] considered such strategies too narrow to be sustainable. Even agencies involved in water safety feel that taming strategies are undesirable and unsatisfactory, as they are relatively static, difficult to change, and not adaptive to new circumstances.

In the quest for suitable and adaptive governance approaches, we can learn a great deal from the work of Elinor Ostrom [7–9]. She demonstrated that adaptive and polycentric governance systems are able to cope more effectively with such issues [9]. In addition, Ostrom [9,10] asserted that the theoretical support for the positive influence of polycentric and adaptive governance approaches is large (p. 284). She [11] argued that, if governance is adaptive, it will be able to resolve and grasp the issue in a sustainable way.

The Dutch debate on a new approach to flood safety closely resembles the search for a more polycentric governance system. The Dutch Delta Program introduced a new paradigm, called 'multilayered safety' (MLS), based upon a more inclusive scope and more integrative collaboration around flood-risk management. It differs quite significantly from the existing policy paradigm, focused upon flood protection, and is aimed at reorienting the flood management system toward a risk paradigm, in contrast to the dominant probability paradigm [12]. However, it turned out to be very complicated to implement this new approach because it necessitates 'untaming' of the issue of flood safety. Untaming means acknowledging the multifaceted character of the issue, as well as the complexity and controversy surrounding it, and trying to do justice to its wickedness. This approach is highly controversial because policymakers are insecure about the new strategy and scope of the 'reframed problem', and untaming is also very complicated on an institutional level. In this article, we aim to determine whether and to what extent introducing MLS can be seen as an attempt to develop a more polycentric and thus loosely structured institutional regime, as well as to explain its low level of success. In the following chapters, firstly, we address the way in which Ostrom [13] described governance arrangements: the institutional analysis and development (IAD) framework.

Elinor Ostrom et al. developed the IAD framework to assist researchers in studying action situations [8]. According to Ostrom [8], every policy situation has a few general building blocks that can be studied as an action situation, in which actors choose and alternate amongst strategies that lead to changes in the decision-making process [14]. These building blocks include (1) the participants, (2) their positions, (3) and their actions, as well as (4) their control over and (5) information about (6) the cost–benefit analyses and (7) outcomes ([8], (p. 189); [15], (p. 27)). Zooming in on such a specific action situation—such as the flood-safety issue—obviously means a simplification of the complexity inherent in an action situation. However, systematically approaching this action situation and the accompanying governance rules make the IAD framework fruitful when attempting to track changes over time while simultaneously keeping rich descriptions of the case. Understanding such action situations and the related changes is important in efforts to design new approaches [8,16,17]. We deliberately opted for this framework because it offers an extensive typology of (institutional) rules to describe concrete governance situations, but it also helps to distinguish between loose and tighter appearances of such rules. Other frameworks (e.g. [18,19]) mention governance elements as well; however, the IAD framework specifically focuses on the tightness of the rules and thus allows for comparison.

In the following paragraphs, we will first present the material and methods used to analyze the case. Thereafter, we discuss the Dutch case of MLS and study the institutional rules in the case. We end this article with analysis and conclusion.

#### **2. Materials and Methods**

To call climate change in the highly flood-prone Netherlands a complicated problem is quite an understatement. Not only is climate change a contested phenomenon but the strategy to tackle its consequences is also highly disputed within Dutch politics (cf. [12]). The exact consequences of climate change are uncertain, multifaceted, and related to each other in a nonlinear way [20]. There are thus no optimal solutions (i.e., solutions with definitive and objective answers). Climate change adaptation levies enormous costs and has broad consequences for landscape quality and similar issues (cf. [21,22], p. 9).

For a long time, flood safety was considered a technical issue that could be solved by means of technical expertise. There was a normative consensus in Dutch society and strong support for governmental action to safeguard the Netherlands against flooding, especially after the floods of 1953. With the help of a set of clear norms for flood safety, a whole system of dike maintenance, inspection, and enforcement was developed that functioned quite well. These norms were entirely based upon the "probability" of a flood: the dikes have to meet a certain standard that equals the probability of flooding of 1 in 1000, 3000, or 10,000 years (dependent upon population density and infrastructure development).

However, the consequences of climate change (more soil erosion, higher river water discharges, as well as sea-level rises) put this technical system under significant pressure [23]. It becomes increasingly difficult to implement the appropriate measures to meet the norms [23]. Moreover, it is recognized that meeting the norm does not mean that a flood cannot occur. Although the likelihood is low due to the impact of climate change, the risk of a flood increases. Sixty years after defining the legal norms, the consequences of a potential flood (and thus the risk) are much higher because investments in the infrastructure behind dikes (houses, companies, infrastructure, etc.) are enormous, and more people live in flood-prone areas. Thus, a potential flood will cause more casualties and economic losses. Hurricanes Katrina and Harvey were an important triggering event that fueled a discussion about the consequences of such a disaster [24].

At the same time, the normative consensus underlying the traditional 'protection' paradigm also eroded to some extent. Citizens increasingly question dike-enforcement projects. They object to the impact of dikes on their environment and the negative consequences for ecological, landscape, and cultural values. In various cases, (recurring) dike-enforcement operations have met fierce resistance from inhabitants who remained unconvinced of the necessity and asked for alternative solutions [25]. In the scientific community, the concern is growing that structural flood-protection measures have 'created flood disasters' [26]. The focus has gradually shifted from a 'protect-and-react' regime (managing the flood) which results in a 'safe development paradox' [27] by continuing the exposure of communities to the changing residual risk (the remaining chance that a disaster will occur), toward a regime in which the changing nature of the risk is managed proactively (managing the risk and strengthening resilience). Many authors have stressed that adapting to climate change for reasons of flood safety necessitates more resilient socio-ecological systems to deal with 'unexpected' shocks.

In the context of the Dutch Delta Program, several proposals have been developed to increase resilience and adaptiveness in the Dutch flood domain [28]. However, these attempts have only resulted in marginal changes in the dominant policy paradigm. In this article, we analyze these proposals with the help of the IAD framework of Ostrom [5,8,13] to understand the efforts to realize a more adaptive and polycentric governance approach. The second author of this article extensively analyzed the single case study described in this article. He was involved in two important policy-advice trajectories regarding the revision of Dutch flood-risk management [15,29]. The Dutch Ministry of Infrastructure and Environment commissioned these trajectories to explore the possible governance approaches for implementing the new paradigm of MLS. Furthermore, he was the main evaluator of three pilots in which the new paradigm was tested [30]. The pilots were meant to explore both the physical and technical possibilities of MLS and the possibilities for an alternative governance approach to explore and implement such an alternative flood-risk strategy.

Given the research aim of this study, we opted for a qualitative approach that focuses on how involved actors interpret the change in policy paradigm, as well as how the governance arrangement and accompanying rules evolve. The more factual question about what really changed can be answered by analyzing policy documents and procedures to identify which elements of the Dutch flood management regime are described therein. To gain additional insights into how these changes were

perceived, more than 50 key players were interviewed, and 10 focus groups with experts and officials were conducted. In addition, we analyzed contributions to the debate on flood-risk management reform and entries on diverse professional Internet forums. In the interviews, more than 30 experts, civil servants, and responsible authorities at national, regional, and local levels participated, and some respondents were interviewed twice. Furthermore, a survey was conducted, in which preferences regarding the new flood-risk regime were investigated [15]. In addition, the second author was actively engaged in the case as a participatory observer. He observed meetings among experts, practitioners, and officials who were discussing the issue of multilayered safety. Over 10 interactive meetings were attended between 2012 and 2015 to discuss the concept of MLS.

#### *2.1. The Institutional Analysis and Development (IAD) Framework*

Institutional rules consist of procedures and mechanisms that actors agree upon jointly (i.e., rules-in-use in the action situation) and influence how particular positions, actors, information, and actions are defined within the action situation [31]. These rules constitute the capacity 'to overcome dilemmas and create effective governance' (Blomquist, as quoted in [11]). In other words, the outcome of an action situation is based upon the extent of organized (or collective) action between independent organizations, in cooperation, to achieve a goal [8].

#### 2.1.1. Operationalizing the Rules-in-Use in Flood Safety

While the IAD framework helps to sketch the action situation and then to study the institutional rules governing the situation, we still need a systematic way to compare how the rules change over time. In this article, we develop the rules-in-use—attached to specific times in the development of the approach to govern flood safety—of the IAD framework (see Figure 1 [8]) to make them suitable to study the case, as well as to compare the governance approach over time. We do not claim that we have found a way to study all types of governance approaches that exist, and we are well aware that this framework can be operationalized in many different ways. As Ostrom [13] concluded, an overall testable set of rules is not achievable, as there are too many rule configurations possible because governance approaches vary too much.

**Figure 1.** Institutional analysis and development framework [8].

In the end, how loose or tight the approach is, defines how stringently the governance approach presses organizations [32]. Many authors, including Elinor Ostrom herself, have made a distinction between tight, less tight, or lax rules [17,33].

In tight approaches, (self-established) authority is often the way to encourage organizations to follow-up on decisions and objectives [34] (i.e., superiors or lead organizations determine the actions of their subordinates). Tight approaches (or tightly coupled systems) are comparable with what Chisholm [34] called transitivity. A tight approach refers to a hierarchical chain of actors and assumes that, whenever an actor A takes a decision, actor B follows this decision. Translated to the action situation and the rule configuration, an action situation is tightly governed if institutional rules strictly determine how participants should adapt [31]. If rules are tight, participants have limited options, which leads to less uncertainty. Therefore, Roberts [4] called setting up a tight governance approach a 'taming strategy'. In such situations, power is in the hands of a few actors with authority, and it is much easier to predict outcomes compared to a situation in which rules are less strict [4,33]. It also has a downside. Elmore [35] argued that tight approaches could lead to more checks, norms, and formal decision points, which can create delays. Furthermore, many authors stress that such tight approaches can be met with reluctance or resistance by organizations because tight governance can diminish the autonomy of organizations ([8] (p. 284), [31,36]).

Loose approaches lie on the other end of the scale and are similar to self-organizational networks, in which autonomous—not guided by binding rules—participants discuss and contemplate issues [31]. Which actors are necessary at the table is determined based on the task at hand, not because of the organizational chart [34,36]. Roles and actions are continuously adjusted based on experience, and tasks are generally established by negotiation among the participants. A loose approach is almost like an 'adaptive device' ([35] p. 608). Participants have leeway in such approaches, which they can use to deal with conflicting or complex demands. However, not adapting, adhering, or exhibiting deviant behavior may lead to distrust [37]. In this way, a loose approach is almost like a self-reinforcing mechanism because the rules integrate and bind participants based on the issue, trust, and loyalty [31]. As such systems are flat—there is no one 'center of authority'—they may appear to be disordered ([34], (p. 54), [38]).

Both types have positive sides and challenges [31,39]. Besides the pitfalls per type, Benz [31] (p. 14) also wrote that, if the type of approach does not fit the organizational style, policy implementation and decision-making are both likely to end in an impasse. The tight approach resembles what Rittel and Webber called 'the systems approach of the first generation'. The second generation resembles a loose governance approach, based upon an open dialogue with participants, explorative argumentation, and an iterative search about the characteristics of the problem and a possible solution [21].

When researchers think of Elinor Ostrom and the IAD framework, studying common pool issues is what immediately comes to mind. Although the IAD framework is often used to analyze common-pool issues like forest, irrigation, or fisheries management, the framework is not restricted to studying common-pool resources. Elinor Ostrom [11] (p. 646) shows that the IAD framework is compatible with *public goods* as well [40]. Furthermore, the framework has been used by many other researchers focusing on flood risk e.g., [41,42]. The following paragraphs describe the elements of the IAD framework and the way we operationalize each element by using the four-point scale, leading to a rich and fine-grained method to score governance approaches on all their rules ranging from (1) tight rules to (4) loose rules.

#### 2.1.2. Boundary Rules

The participants constitute the first element of the action situation [8]. In this article, organizations are considered to be the participants. Boundary rules describe how exclusive or open the collaboration is to participants, in the beginning, or along the way. These rules also indicate who can say something about 'the water issue': who is obliged to cooperate, which actors are excluded, and how participants can exit the action situation. Jordan and Schubert [43] described the boundary rules with three dimensions, including the number of participants, whether they are sectoral or trans-sectoral, and the stability.

To operationalize this rule, two aspects were combined: the level of stability (based on [43,44]) and the diversity of participants (based on [45]). How open or closed the boundaries are can be described by the level of stability, which ranges from stable to unstable [43]. For the diversity of the participants, a distinction was made between vertical and horizontal linkages, in line with Bouckaert et al. [45]. One could think of vertical linkages between different hierarchical layers of organizations, as well as horizontal linkages between organizations of different policy sectors, portfolios, or different parts of the triple helix (universities, companies, and governments).

A tight participant constellation is highly institutionalized, formal, obliged, and restricted in terms of access to new participants. An open, ad hoc participant constellation, based on voluntary participation, is a loose arena. In loose approaches, the participant constellation is highly diverse, and the boundaries are open, fluent, and accessible to new participants. In tight approaches (see the left column in Table 1), the participant constellation is closed, and clusters' participants from one vertical hierarchy. The four institutionalization forms in Table 1 were borrowed from Van Waarden [44] and [43] combined with the vertical/horizontal linkages of Bouckaert et al. [45]. This dimension does not concern the centralization of power or the positions in the governance approach, it merely concerns the access and boundaries of the constellation.

#### 2.1.3. Positions Rules

Secondly, next to the participant constellation, participants are, because of rules, agreements, and mandates, situated in different positions [8]. The actor's position in an action situation determines the extent of influence, as well as the bargaining and veto power, of a certain actor [46]. The operationalization of this dimension builds on the work of Provan and Kenis [47].

Positions may be, at one extreme, collectively structured (i.e., shared) by a separately created organization that deals with the administration, communication, and coordination tasks [47]. An action situation governed by a separately created administrative organization is the most highly structured and tightest form of governance. At the other extreme, positions may not be structured at all and thus participant governed, which is a very loose type of governance. In such a situation, members themselves govern their actions. Another example of loose governance occurs when the network (or collective body) acts as the basic entity where activities are governed, such as when a network functions as the collective structure in which consensus can be reached. Column 2 of this dimension describes a situation in which a single participant takes on the role of a lead organization [47].

#### 2.1.4. Choice Rules

Choice rules determine the leeway a participant has, as described in the mission of the action arena, in terms of when the participant may, must (not), or should take action. Although Ostrom [8] employed the 'attributes, deontic, aim, conditions, or else' (ADICO) framework [48] as a systematic way to describe the rules-in-form [49], we used it to operationalize the leeway on the organization has. Ostrom distinguished between a rule, a norm, and a strategy (see also [49]). The ADICO grammar can help us determine if the objective is governed by a leading rule, a norm, or a strategy.


An objective containing each of the ADICO components is considered to be a rule. An objective containing the first four components (ADIC, i.e., those without sanction) is characterized as a norm. Objectives that only contain an attribute, aim, and condition (AIC)—and no sanction or prescription—are considered to be strategies [50]. If there is a tight and guiding rule, there is a strong degree of certainty [51] about what is expected of the participants. The loosest type of rule is a statement, which consists of an attribute (A) and an aim (I).

#### 2.1.5. Information Rules

The fourth element of the IAD framework (see Table 1) is the information a participant has about how to reach the strategy and about the tasks of the other participants. This dimension does not stress the frequency or the amount of information, rather, it stresses the *kind* of information [8]. A situation in which a participant has a unique part of the information can lead to protection and 'bargaining', and it is the actor with the most information who has a dominant position [52]. In the situation of incomplete information and deficits, the strategies of the participants are 'messy' and uncoordinated [53]. Hood [54] distinguished between information that tightly and loosely presses organizations to adapt. Information is tight when responsibilities are ex-ante thoroughly described, involving the separation of 'coordination' and 'implementation' activities ([51], (p. 12); [52], (p. 34)). Another tight type also concerns an ex-ante described strategy and tasks, without this separation in 'coordination' and 'implementation' activities.

Information about the strategy and the tasks set during the process in the core group is looser. The group can agree on the strategy and tasks as a norm, but this can also be left to the participants themselves. In such situations, there is a rich exchange of information, with a lot of leeway for learning or adaptation [55]. Information, here, is seen as a collective asset. Participants of such loose types communicate on a personal, rather than formal, level with each other, which can result in more 'noise' and different conceptions of the task at hand [56]. Furthermore, participants who do not participate in the coordination process lack the information of those who do.

#### 2.1.6. Aggregation Rules

The fifth element is the control that participants in the action situation have, how power is played out, how participants can affect the outcome, and how they reach final decisions [47]. Aggregation rules determine, for instance, whether a decision by a participant is needed to proceed to action and the implementation phase. In other words, aggregation rules can be symmetric (e.g., unanimity or voting schemes) or nonsymmetric (e.g., a leader takes a decision on behalf of the others, such as a chairman or lead organization) [8].

Compared to the position rules, which explain how the positions and the approach are structured, this dimension, 'aggregation rules', describes how participants come to decisions in the action situation. The literature tends to focus on three distinct forms, unicentric (tight), multicentric (having more than one authority), and pluricentric (loose) [57–59]. The role of the coordinator in a unicentric form is not to gain or to monopolize power but to perform roles that 'normal' participants are unable to perform: foreseeing threats, disasters, and deadlock, as well as undertaking long-term planning (Mulgan, 1997, as cited in [60]).

#### 2.1.7. Payoff Rules

The cost–benefit analysis that the participant makes about the utility gained by either cooperation or defection is the sixth element of the IAD framework [8]. The 'payoff' can be either an extrinsic reward or sanction or an intrinsic valuation (e.g., joy, shame, or guilt [8]). In government organizations, the payoffs are highly institutionalized through accountability approaches. Accountability can influence policy coordination in different ways. Firstly, making organizations accountable places emphasis on goal formulation and achievement. Secondly, accountability systems can enable participants to comprehend the potential benefits of and pay attention to policy targets. Thirdly, accountability can aid in detecting and resolving overlaps and conflicts amongst policy objectives [61].

Fox [62] helped to operationalize this dimension by discussing two basic dimensions of accountability:


Types (2) and (3) in Table 1 are types of institutional answerability of the involved organizations to the coordinating actor: one type without an inspection, sanctions, or rewards, and the other type with these elements. The tightest form is what Fox [62] called the 'hard' accountability type: performance-based accountability to the coordinating actor with sanctions and the possibility to investigate actual institutional behavior.

#### 2.1.8. Scope Rules

Lastly, the scope describes the range of possible outcomes that could be affected, and it specifies the ultimate goal that must be achieved [8]. How the scope is seen affects what is needed to reach the outcomes. These are the envisioned outcomes—not on the level of the participants but on the level of 'the collective'. The scope is, in the majority of instances, explicated in general mission statements.

According to Alter and Hage [64], the scope is the extent to which participants frame and see the mission they want to achieve in the action situation in a comprehensive way. On the one hand, this can be done broadly and holistically by seeing the problem and solution in a multidimensional and multidisciplinary way. First of all, 'multidimensional' means that the solution is the extent to which the problem and solution are seen 'holistically', for instance, that an area is a social, natural, geographical, and a political place. Secondly, 'multidisciplinary' means that the expertise of different professions is needed to understand the problem and find a solution. A broad framing leads to a wide-ranging assessment about which services, programs, and objectives should be met by the participants to target a particular problem. On the other hand, the problem and the solution can be seen more narrowly. A narrowly defined scope leads to objectives being unambiguous and more straightforward [64], as well as monodimensional and monodisciplinary goals and outcomes. In such a situation, it is clear who is responsible, who should take action, and how the goal must be achieved. Hence, monodimensional monodisciplinary goals and outcomes often explicitly link lead organizations to a problem, in which the organizations have a (self-established) authority over the action situation, and subsequently, the relations in the action situation are more tightly structured. Table 1 presents the various rules, using the IAD framework of Ostrom [8], and the different gradations between loose and tight versions of these rules.

**Table 1.** Rules in a governance action situation (based on [13]).


#### **3. Results**

When we try to deduce the rules that structure the action situation sketched above, it becomes clear that the governance approach governing flood safety until 2008 was rather tight (see Table 2 second column). For a long time, the governance regime regarding flood safety can be characterized as a highly tight governance approach that consists of a small coalition of interconnected actors with clear jurisdictions, which uses a highly formalized and detailed set of rules in order to realize a clear objective. Thus, there is a robust system of legal norms for flood protection. This system is embodied in a quite independent network of autonomous organizations responsible for implementing these norms. Within this network, there is a strong sectoral focus upon one dominant idea, namely prevention by means of protective measures.

This approach resembles the idea that flood management is a complicated, but technical, issue that can be dealt with in a well-structured way (a problem that can be 'tamed' [21]). The National Ministry of Infrastructure and Environment is responsible for flood management. The National Agency of Public Works and the regional water boards are responsible for implementing the norms. They focus their activities on the dikes, and only in very exceptional cases, a more spatial perspective is used to think about flood management strategies. The Room for the River program (2003–2016) can be seen as a first attempt to apply a more integrated focus by including aspects of discharge capacity and water levels in the decision to replace or to enforce the dikes along the main rivers [65]. When it comes to implementation (e.g. dike-enforcement projects or maintenance activities), other actors are informed or consulted, but they are not in a position to alter or to veto what has to be done in the eyes of the water authorities, the scope they use, or the alternatives they select. The maintenance of the flood defense system is based upon periodical inspection reports (every six years). In Table 2 (second column) we summarize the various rules that characterize the traditional approach of flood management.

#### *3.1. The Delta Program and the (Envisioned) Concept of Multilayered Safety (MLS)*

In 2008, a second Delta Commission was appointed to prepare advice about how the Netherlands could deal with the consequences of climate change. The report of the Veerman Committee stimulated the Dutch government to rethink the current methods of flood management. In 2009 the National Water Plan put the quest for another flood-risk policy on the national policy agenda and introduced the idea of MLS. The concept of MLS is based upon the idea that flood management has to become more risk-based and thus must have an open eye regarding the following question: What if a flood actually does happen? The concept grants a more prominent role to measures mitigating the impact of a possible flood by emphasizing the importance of risk management and planning strategies. It distinguishes amongst three layers of safety. The first layer includes dikes and other means to prevent flooding (the traditional approach of flood protection). The second layer consists of spatial measures that can help to reduce the impact of a flooding event. The third layer has to do with all sorts of activities regarding crisis and disaster management. Figure 2 presents this idea of MLS.

**Figure 2.** Multilayered safety (MLS). Source: Foundation for Applied Water Research Meerlaagsveiligheid (2014) Waterrobuust bouwen in stedelijk gebied, June 2014.

In 2010, the Dutch Cabinet appointed the national Delta Commissioner, and he started in 2011 by preparing the national Delta Program. The aim of this Delta Program was to develop a strategic policy plan for the long term that has to safeguard the Netherlands from the consequences of climate change related to floods, extreme rainfall, and droughts. One of the main characteristics of the Delta Program was its openness for other than traditional actors to get involved in drafting long-term policy strategies regarding flood risk management. This openness came with a more spatial perspective on flood management because regional and local authorities added their (spatial) agenda to the discussion about flood risk norms and ways to achieve these norms. Opening up the policy arena thus meant a more inclusive problem definition on flood risk and management.

In late 2014, the main results of the Delta Program were laid down in the so-called Delta Decisions. As van Buuren et al. [1] stated, two of the elements of multilayered safety were included. One Delta decision was devoted to the issue of 'land-use adaptation' or water-robust planning (to come to more risk-neutral spatial development on a voluntary basis), and the Delta Decision on 'flood-risk safety' incorporated the possibility of 'smart combinations'. The latter is aimed to provide the opportunity for exceptional situations in which dike enforcement can be replaced by a combination of measures in the first, second, and third layer: partial dike enforcement, measures in the spatial domain (compartmenting, waterproof development), and evacuation or risk reduction [1].

This Delta Program also laid the foundation for a new set of flood-risk norms in which the element of risk was much more emphasized than in the former norms. This new approach also resulted in a search for alternative governance approaches at the local and the regional levels to give meaning to this new philosophy and to enable a search for solutions at all three layers. In the six regional sub-programs of the Delta Program, the possibilities for MLS were explored. More specifically, the potentials of the concept were explored in three regional pilot projects to see whether it was possible to replace dike enforcements with a 'smart combination' of measures in all three layers [30].

The new policy paradigm, as proposed in the Dutch Delta Program, thus implies a couple of significant adjustments of the 'old' paradigm. These adjustments are at least threefold. First of all, the norms for flood-risk safety were adjusted in order to make them really "risk-based". The norms

are now based upon two elements: the probability of a flood in relation to the expected impact of it in terms of economic losses and casualties. Secondly, by introducing the concept of MLS, the Delta Program ensures that flood-risk safety is no longer solely approached from the perspective of 'prevention'. It also draws attention to the so-called 'second layer', the spatial planning of an area, and the 'third layer', the possibilities for disaster and risk management. The Delta Program created the possibility to regionally explore the most suitable way of realizing the norm: an integrated package of measures of dike enforcement, spatial planning, and disaster management. Within the Delta Program, the importance of explorative searches at the local or regional level in order to find out which combination of measures fits best into the local context, the foreseen spatial developments, and the agendas of stakeholders was emphasized. This was framed as a joint responsibility of all regional actors.

The Delta Program also added the possibility of a 'smart combination'. Within a specific situation in which the norm necessitates dike enforcement, it became possible to exchange dike enforcement with a specific combination of measures in the first, second, and third layers. Such a combination should result in a lower norm for the dike and a formal commitment among the various responsible authorities to implement and maintain this combination of measures.

When we analyze the way in which the concept of MLS was used in the context of the Delta Program in the period between 2011 and 2014, we refer to the last column of Table 2, in which we sketch the various rules (in theory) the concept of MLS imply. In theory, the idea of MLS implies a fundamental shift toward a more loosely organized governance system. It implies that all actors that can possibly contribute to risk reduction at one of the three layers are welcome in a regional policy arena around a diked area (boundary rules: 4). It also presupposes that all these actors (e.g., the water boards, public safety regions, the municipalities, and other actors) internally develop a set of rules to set up a regional dialogue to discuss the way in which flood risk is dealt with (positions rules: 4). Regarding boundary rules, the idea of MLS can be seen as an open invitation to all actors within a certain region to bring in their ideas that could function as a building block at one of the three layers.

Developing a MLS strategy necessitates a regional dialogue in which all relevant stakeholders can contribute their ideas. It requires a very knowledge-intensive search in which an open and creative search can emerge toward innovative and integrative solutions. In various pilots, such a search was organized, and actors developed ideas about how to give meaning to this idea [30,66,67].

#### *3.2. The Concept in Practice*

However, in practice (the third column of Table 2), it proved very difficult to implement the concept of MLS as originally envisioned (the fourth column of Table 2). The concept was translated in such a way that it was more or less compatible with the current institutional practice. In the final Delta Decisions, the first layer was said to safeguard a 'basic level of safety'. The new norms for the dikes have to ensure that the mortality rate due to flooding is 1 in 100,000, thus 0.001%. A strong bias in favor of legal norms regarding the dikes dominated the ultimate Delta Decisions: The idea of flood-risk management was entirely translated in a system of new norms for the dikes and thus not for the second or third layer (which were deemed too difficult to control). The more conservative community of national policymakers, civic engineers, and legal experts played an important role in this translation process. In the implementation phase, the policy arena was much more closed compared to the policy formulation stage. Regional and local actors were not in the position to defend or enforce the original concept. As a result, the idea of MLS was severely downsized because the norms for flood-risk safety are entirely focused upon the first layer, and all measures in the second and third layers are framed as complementary but not substitutive for the first layer. They seemed as 'nice to have' but not necessary. After the Delta Decisions were made, the new norms were implemented and translated by the water authorities in a strategy for planning the necessary measures to meet the norms in time. In this implementation process, many elements of the original concept (aimed at including the possible consequences of a flood- which depend upon the spatial characteristics of an area - in the calculation

of the strength of the flood defense needed) were skipped in order to simplify the implementation challenge. At the moment, regional governance approaches around flood-risk management are less tight, but the focus remains on flood prevention by means of hard structures (the first layer).

Thus, although the new system enables the water authorities to opt for more tailor-made interventions to realize the norms, the actual resulting governance approach exhibits much more tight characteristics and thus a hybrid mix. Table 2 (column three) presents the various rules as they are currently in place regarding MLS. The current system of flood-risk management thus exhibits an interesting division between the strategic level (which is about programming measures) and the operational level (which is about implementing measures). At the strategic level, some regions invest in an intergovernmental dialogue between water authorities, provinces, and municipalities. The scope of such a dialogue differs in practice. In other regions, this dialogue is much less substantive and mainly used by the water authorities to present their intended measures regarding dike enforcement. Their plans are no longer only based on the technical specifications of the dike but also on the possible consequences of a dike breach. At the operational level, the water authorities organized a participation trajectory in order to fine-tune the way in which they implement dike enforcement. Within the domains of spatial planning and disaster management, other authorities (with different ambition levels) explore the possibilities for risk mitigation, but these trajectories are not entwined with the implementation efforts of the regional water boards. These attempts differ not only when it comes to how ambitious these are but also to what extent they are actually implemented. The latter mainly depends upon opportunities for coupling these types of measures to other spatial developments.




#### **4. Discussion**

Although the need for a more adaptive, integrated, and risk-oriented way of dealing with the issue of floods seems to be increasingly recognized, it remains very difficult to be freed from the classical idea of controlling floods by maintaining a system of dikes and dams only. The responsible actors in the Dutch flood-risk domain continue to tame the issue of flooding along three lines, which clearly reflect the taming strategies of Roberts [4].

Firstly, by relying on experts in one particular issue arena, the legal and regulative capacity remain in the hands of a few. The national government legally anchor new norms that are entirely translated into standards for dikes, and the very controversial idea of 'risk acceptance' is averted. As a result, all measures in the second and third layers are framed—by these actors who narrowly focus on flood safety as the one issue—as voluntary additions to the first layer, which is, in principle, sufficient to meet the norm. This is the second taming strategy of Roberts [4]. Thirdly, although a participative trajectory was organized, the original lead organizations decided on the strategy. By formally stating that a base level of safety is guaranteed by the first layer, the complexity of flood-risk management (in terms of scope, involved actors, available options, etc.) is reduced to the simplified question regarding the quality of flood defenses. This finding clearly supports the point of Daviter [3] that, via taming, problems are not dealt with in a holistic way but result in a solution that fits the functional specialization of a particular group of people. This taming strategy hinders reflexivity, as well as the inclusion of local experience and knowledge [2]. Certainly, the absence of a learning attitude can be explained by the rather strong path dependency that dominates this domain (cf. [1]) and the closed epistemic community that, for a long time, dominated the knowledge base of Dutch flood policies [68].

The introduction of the concept of MLS implies a search for more polycentric approaches in which the dominant position of the water authorities is reduced in favor of other actors who are responsible for planning and disaster management. Although the uncontrollable character of flood-risk safety is increasingly recognized, it is very difficult to acknowledge these unpredictable consequences in terms of how flood risk is dealt with. Given the risk-averse opinion of the public in the Netherlands, the strong normative consensus that the government is responsible for public safety, and the fact that people perceive themselves safe behind the dikes makes it nearly impossible to communicate that protection will never be perfect, that risks cannot be eliminated, and that it is important to think about more integrated strategies instead of focusing on the height and thickness of the dikes. Moreover, the consequences of a flood are, for most Dutch policymakers and citizens, too severe to accept. Therefore, it has proved to be highly controversial to untame the issue of flood-risk safety and to successfully implement MLS.

Two aspects make untaming very complicated. First of all, the current paradigm of flood-risk management by protection—essentially built on the simplifying idea that flood management is about protection—is seen as a very successful strategy that can be maintained even when the sea level rises by several meters [69]. With the help of tight rules, the choice menu for participants is limited, which leads to less uncertainty for those authorities responsible for maintaining flood-risk norms. Adopting MLS requires the courage to be open to a more ambiguous problem definition that necessitates mutual adjustment, exploratory strategies, and solutions that can only be realized with the help of all actors involved. Furthermore, the current governance approach is not only the simpler one; it is also a deeply embodied, path-dependent repertoire within the flood-risk domain [1]. There are strong self-reinforcing mechanisms in this domain. Not only high sunk costs (because of the existing infrastructure of flood defenses) but also learning and coordination effects make it difficult to leave the existing path [70]. The Dutch excel at flood protection, which causes a classical success trap [71].

#### **5. Conclusions**

In this article, we have developed a refined way to study governance approaches by distinguishing several dimensions, which together define and form tighter or looser approaches. The seven rules developed in this article, on the basis of the IAD framework [8], provide more insight into different elements of what can constitute a more polycentric and adaptive governance approach that can enhance resilient socio-ecological systems confronted with climate-related risks, and this approach proved to be a suitable analytical lens to study such issues. The seven rules give rich and in-depth insight into a complex governance arena. It has shown to be a holistic framework that can grasp the hybridity of governance arrangements and allows for comparison of approaches and over time.

By means of these rules, we have described the attempt to reinvent the governance of flood risk in the Netherlands. Following advice from the Dutch Delta Commission to reframe the issue, to re-configure the governance approach, and to have an open eye for new problem frameworks and solutions, an innovative policy concept was introduced: MLS. This concept, envisioned to enlarge the adaptivity of the Dutch flood-risk governance approach, was built on the idea that the core of the existing policy paradigm, focusing upon prevention, had to be exchanged for the idea of risk reduction, meaning that both the probability of a flood and its consequences are leverage points for measures. Policymakers saw MLS as a way to enlarge the scope of possibilities beyond the traditional focus on dike enforcement. Furthermore, it began a search for more polycentric and regional governance approaches, in which various actors can develop ideas or take measures that contribute to flood-risk reduction. In a couple of cases, this idea was piloted, and many interesting ideas were developed. However, the ultimate governance approach that resulted constitutes a much less adaptive and loose approach compared to the one envisioned.

The Dutch case of flood safety governance serves as a poignant example of an attempt to introduce an alternative governance approach based upon the idea of Ostrom to promote polycentricity and adaptivity in governance approaches. However, as demonstrated, this attempt also presupposes that previously successful taming strategies and paths have to be abandoned once a new approach is decided upon. Our case illustrates that untaming is not only highly controversial (because of the current frame that the government has to safeguard the people against flooding) but also institutionally very difficult due to the high institutional density and opacity of the current flood-risk domain. Taming strategies are firmly anchored in existing institutional practices. Therefore, the lead actors—the small group of established actors—demarcated and reframed the problem again as a tame issue to ensure safety by means of protective measures.

Presenting only one single case study from a rather particular domain (flood management) in a country with a strong tradition in this respect (the Netherlands), we must be cautious about drawing strong conclusions regarding the possibilities for untaming complex issues. The question becomes how we can successfully 'untame' different issues in different contexts in such a way that people dare to adopt the new kind of governance approaches that allow for exploration and deliberation. The answer to this question will differ amongst policy domains. Future research can focus on the untaming issue in other domains, to see whether this approach is attainable in other ways, by other actors and other approaches. However, from our case, we can conclude that acknowledging the power of path-dependency mechanisms and thinking about a strategy to mitigate them are indispensable steps for successfully untaming issues that (apparently) were successfully tamed for a long time.

**Author Contributions:** Conceptualization, A.M.; methodology, A.M.; software, not applicable; validation, A.M. and A.v.B.; formal analysis, A.M. and A.v.B.; investigation, A.v.B.; resources, A.M. and A.v.B.; data curation, A.v.B.; writing—original draft preparation, A.M. and A.v.B.; writing—review and editing, A.M. and A.v.B.; visualization, A.M.; supervision, A.M. and A.v.B.; project administration, A.M. and A.v.B.; funding acquisition, not applicable.

#### **Funding:** This research received no external funding.

**Acknowledgments:** The theoretical work in this article is based on the PhD project of Astrid Molenveld, which was part of the Policy Research Centre on Governmental Organisation in Flanders (SBOV III – 2012–2016), funded by the Flemish government. The operationalization, based on the IAD framework, was developed in the manuscript and further elaborated on in this article.

**Conflicts of Interest:** The authors declare no conflicts of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Social Learning in Multilevel Flood Risk Governance: Lessons from the Dutch Room for the River Program**

**Jacomien den Boer 1, Carel Dieperink 1,\* and Farhad Mukhtarov 2,3**


Received: 2 September 2019; Accepted: 25 September 2019; Published: 28 September 2019

**Abstract:** Although social learning is a key element of multilevel flood risk governance, it is hardly studied. This paper addresses this knowledge gap. The paper aims to identify enabling conditions for social learning in multilevel flood risks governance arrangements.We first conceptualize social learning and draw up a conceptual framework consisting of enabling conditions for social learning, using the literature on adaptive co-management, sustainable land and water management, and integrated flood risk management. Next, we apply this framework to analyze social learning in the context of the Dutch Room for the River program. Our interview results reveal that social learning about integrated flood protection measures took place at multiple levels. We found that a strong personal commitment to learning and mutual interpersonal trust in working groups are key conditions for successful social learning. Based on our analysis, we conclude with some recommendations for enhancing social learning processes in future flood protection programs.

**Keywords:** social learning; integrated flood risk management; Room for the River program; multilevel governance

#### **1. Introduction**

Population growth and increased socioeconomic activity in floodplains, combined with more recent effects of climate change, are important factors in the increase of flood risks [1,2]. A gradual shift is visible in flood management across a number of countries: from flood defense based on 'fighting against water' to flood risk management and 'living with water' [3,4]. This shift in flood protection measures over the last decades is known as integrated flood risk management (IFRM) and is increasingly adopted as the basis for a number of flood protection programs, e.g. in the Netherlands, the United Kingdom, and other countries [5–7]. Successful coexistence of the natural phenomenon of flooding and the human population depends largely on the organization of the governmental system and the decision-making process among mandated stakeholders. For IFRM, multiple actors at multiple policy levels need to collaboratively design and implement new flood risk management policies [6,8]. An integrated approach to flood risk governance includes both engineering solutions and spatial planning projects that are combined and aligned, such as spatial development, housing, economic activity, as well as governance solutions such as aligning funding from different policy domains and collaboration between multiple stakeholders [9–11]. IFRM can therefore, be seen as a multilevel and multistakeholder governance practice [12–14].

Growing scholarly interest in collaborative modes of governance stresses the complexity and uncertainty inherent to such institutional arrangements [14–18]. Collaborative arrangements are network structures where participating organizations depend on each other to achieve tasks that reach beyond their individual capacities [19]. Such interdependence requires the capacity for social learning for success and endurance of collaborative and multilevel arrangements. This realization has led to blossoming literature on social learning for collaborative institutions [10,17,20–25]. This literature discusses social learning in a natural resources management context. However, studies on social learning in the context of multilevel flood risk governance are rare, and we have not yet been confronted with studies that specify the conditions that enable social learning in flood risk governance.

This article deals with this gap in the literature by presenting a conceptual framework to study social learning in the context of flood risk governance arrangements, which is then applied in a case study. The case study concerns social learning in the Room for the River program, a multilevel flood risk management program in the Netherlands. The case involves multiple and diverse governmental actors, working collectively on flood risk management in the Netherlands.

Sections 2 and 3 reflect on the literature and conceptualize social learning for multilevel flood risk governance. For the conceptualization of social learning, we draw from literature on learning and organizational theories, which are related to the natural resources management domain. Suitable literature was found using Google Scholar and Scopus, by searching for different combinations of keywords like 'social learning', 'multi-level water governance', 'adaptive co-management', 'integrated flood risk management', 'learning process', 'collaborative learning' and 'multiple-loop learning'. In Section 4, the methods applied in the empirical part of the paper are explained. In Section 5, we present the results of the case study, leading to a discussion and conclusion in Section 6.

#### **2. Conceptualizing Social Learning: Three Debates**

Social learning is a key concept for adaptive co-management, multilevel governance, and integrated flood risk management (see e.g., [10,13,15–18,20,21,26,27]). Both complementary and overlapping definitions for social learning are found in these strands of literature. Social learning appears to be a highly debated concept in literature. Three debates can be distinguished, which concern the relation between individual and social learning (e.g., [19,21,28], learning as an outcome or as a process [15,25,29,30], and learning-levels and learning loops [6,31,32]).

A first important debate in the literature is the distinction between social learning as different from individual learning [25,28,33]. Scholars have maintained a distinction between individual learning and social learning [14,19,21,34] in which the latter is based on deliberation and produces qualitatively different forms of decision-making as opposed to individual learning [35–37]. This line of research, for example, ventured to understand how individual learning may become social learning, as in the case of flood risk governance in the UK (e.g., [38]). As opposed to individual learning, the concept of social learning relates to notions of 'learning communities' and 'communities of practice' (e.g., [20]). These communities consist of people that share a common concern and pursue knowledge through regular interaction, based in practice [20]. It goes beyond individual learning (a football player improving his personal skills) or individuals learning together (football players who train their personal skills during team practice). Social learning here is understood to be a process of change at the group level [39] in which people learn from each other, which benefits the wider socioecological systems they operate in (e.g., river basins or community forests). For learning to be social, a change in understanding must occur at a larger scale, through social interaction [14,21]. Thomas and Allen [40] argue in a similar way for organizational learning, from which social learning is partially derived: 'while organizations learn through individual learning, organizational learning is not a cumulative result of individual learning. Rather, organizations learn when discoveries, evaluations, and insights by individuals are successfully embedded in the organization's mental models or cognitive systems and memories' (p. 125). To other scholars, however, collective learning may originate only when individual members of a collective learn and hence, is merely a sum of individual learning [41–43]. Heikkila and Gerlak [44] (p. 486), in turn, recognize the complexity of separating individual and collective learning in empirical settings and accept that learning may happen 'at multiple levels in collective settings,

from individuals, to groups, organizations, and networks, and these levels are linked by dynamic social processes of producing and sharing knowledge' [44]. According to these scholars, the focus must be placed on the processes and products of learning rather than the relationship between the individual and collective learning, a view which is supported by Suskevics et al. [25].

Another area where a consensus is yet to emerge is in the debate on whether learning should be seen as a process or as an outcome [21,25,29]. Learning outcomes, or products, 'can include new shared ideas, strategies, or actions among the parties involved in the learning process' [19] (p. 4). Social learning outcomes are numerous (e.g., [15,25,29,32]). Among others, increased understanding of key issues, trust-building, acquisition of factual knowledge, and social skills are the results of social learning processes. The learning process, subsequently, is 'the set of actions that allow new information or knowledge to be acquired, processed and shared, and transferred across individuals within a group' [19] (p. 3). Mostert et al. [15] (p. 1,2) perceive social learning as a process within a specific context, which includes both a natural context (ecology, geography) and a social context (economy, culture, governance system). A learning process emerges when stakeholders realize they are interdependent in pursuing a goal or policy, which means that individual action will yield less or no result, and collaboration is thus necessary. The learning process can be initialized by the stakeholders or by an external party. According to Mostert et al. [15], the learning process involves 'the development of trust, joint problem definition, joint fact-finding, the development and assessment of different alternatives, joint decision-making, and joint planning for implementation' (p. 2). Process outcomes are, e.g., better relations, increased trust, and empowerment of stakeholders. They emphasize the participatory character of the process. Similarly, Bos et al. [30] define social learning as 'a collective process enabling change in a situation, as opposed to social learning as an outcome or 'emergent property' of a process to change a situation' (p. 399). The notion of social learning as a process points to 'actors developing shared meanings, values, and understanding through interaction, which provides the basis for joint future action' ([29,45] are quoted from [31]).

The third debate in the literature concerns the relative importance of various contextual factors in influencing the process and outcomes of social learning, which is a growing body of empirical literature that discusses case studies in individual and comparative settings. Medema et al. [32] and Pahl-Wostl [31] argue that learning processes are largely influenced by the governance system in which they are embedded. The governance structure includes 'the pertinent legal and organizational framework as well as the cultural and socioeconomic environment' [31] (p. 4). Social learning can be described as a multiscale process, hence learning can occur at multiple levels of agent interaction [31,32]. The macro-level is the level of the governance structure at societal level. The meso-level is at the level of actor networks, or organized stakeholder groups. The micro-level is at the level of collaboration processes between stakeholder representatives of flood protection authorities (i.e., actual meetings). According to Pahl-Wostl et al. [31], the multiparty collaboration processes are the 'nuclei' of the learning process (p. 5). When transferring this multiscale learning process to a geographical level, Huitema et al. [16] advocate a bioregional approach (i.e., at watershed level) for water governance, reflecting the micro-level of regional multi-party collaboration processes. With regard to conditions that explain learning processes, they will also pertain to the micro-level.

In this paper, we aim to contribute to these three debates by analyzing an empirical case in order to reveal what factors influence social learning in the context of multilevel flood risk governance.

#### **3. A Framework of Factors that Influence Social Learning**

To study learning in the setting of multilevel flood risk governance, we adopt the following definition for social learning (based on e.g., [15,17,18,20,21,23,27,46]:

*Social learning is a process of change in understanding, that goes beyond the individual, at the micro-level of multi-party collaboration, aiming at collective action for integrated flood protection.*

By reviewing the abovementioned learning literature, we have found that most of the factors that will likely influence the learning process can be categorized into 4 categories—individual attributes, collaborative arena factors, organizational factors, and external context factors (see also [33]). Combined, these factors provide a framework that can be used to analyze learning processes (Figure 1).

**Figure 1.** A framework for analyzing social learning.

#### *3.1. Individual Attributes*

Individual attributes (see Table 1) are described as the 'the micro-level factors relating to the reactions to change efforts by stakeholders and individuals involved in the learning process' or 'the psychological predispositions of the individuals experiencing change' [32] (p. 26); [47] (p. 107). Individuals and their characteristics form the basis through which learning is manifested [48]. First of all, a commitment to ongoing learning is an important characteristic of learning [49,50]. Individuals with personality traits like tolerance of ambiguity and openness to new experiences are better able to learn from experiences and are less eager to have a confirmation bias, i.e., searching for information that confirms their beliefs. The extent to which a person is flexible and open-minded is an important attribute for learning: when the individual has little resistance to changing the way things are done, he or she is more eager to learn from this situation [51–53]. Related to that is the capability for self-reflection, which is crucial for multiloop learning [17,49,54–56] and the individual's perception and awareness of his own role and abilities to exercise control in certain situations, or locus of power and control [51,57]. Mostert et al. [15] stress that a legitimate process in which participants are willing to learn require high professional competence and motivation. Social skills, in the form of sticking to rules and principle of dialogue and interaction come into play when differing views and perceptions of parties may lead to conflict [15]. When participants address each other in a respectful manner, they are more likely to agree on core issues regarding the policy process, which improves learning. A factor that explains successful learning is previous experience with a multiparty approach [15]. A positive experience will make it more likely that participants are willing to learn from other participants.


**Table 1.** Individual attributes influencing the social learning process.

#### *3.2. Collaborative Arena Factors*

Collaborative arena factors (see Table 2) refer to the way the participatory process is organized and which actions and directions are taken during a policy planning and implementation phase. It is widely recognized that the structure of the institutional arrangement is an important explanatory factor in the occurrence or absence of social learning [19,34]. These factors determine if participants feel respected and heard [32]. Involving all relevant stakeholders in a participatory process for developing and implementing new policies is important [59]. With regard to collaboration, crucial factors in the participatory process are mutual trust, goodwill, and mutual understanding, as is acknowledged interdependency [58]. If lacking, learning is unlikely to occur as common understanding and goals will not be reached [60]. Openness and legitimacy of the participatory process through continuous feedback on the process (e.g., through distribution of meeting records, background documents, questionnaires) is important to create trust and goodwill [15,32]. The overall process architecture is an important condition in this respect. Frequent interaction among stakeholders is crucial for social learning [15]. Clear perspective and direction for feedback are crucial for multiloop learning, as the feedback navigates the process toward desired goals. It is also crucial that participants have clear expectations on the participatory process. A joint planning approach can facilitate this and also increase the transparency of the process, which in turn increases trust and goodwill [15,59]. Role and control of parties involved explaining what influence the role a party or representative has in the collaboration process and what the position of a party is relative to other parties. Related to that is having a neutral organizer or leader who facilitates and presides the participatory process so as to enable fair balancing of interests. This helps to build a democratic and enabling environment for learning processes, in which appropriate meeting formats and extended engagement (repeated and frequent interactions, field trips, bilateral contacts) facilitate trust, collective meaning, constructive conflict solving, that in turn help participants to create a sense of ownership and commitment toward the learning process [15,31,59]. Besides these rather intangible factors, available resources in the form of budget, time, and information are essential as well in order to facilitate a successful learning process. Feedback and reflection, provided there is a clear goal, create a loop in which participants can learn from their previous experiences, and consequently optimize the collaborative process. Framing and reframing of how a policy problem is perceived can open up individual perspectives, thereby creating win-win situations [61,62]. This results in improved mutual trust and better relations, factors previously marked as crucial for social learning. A final factor, communication, both internal (between members of the working groups) and external (to external parties or home organizations), is crucial for diffusion of information throughout the working group and to other organizations.


#### *3.3. Organizational Factors*

Organizational context factors (see Table 3) are preexisting forces in the governance system, which includes the internal context of the participants' home organization. Equality and balanced interests, or the lack thereof, as well as (a lack of) internal conflict on interests, are often mentioned as a prerequisite for social learning or frustrating learning [15], as equal treatment of and opportunities for participating and influencing the policy process will positively influence the learning process. Entering a long-term working relationship, thereby creating a horizontally and vertically integrated cooperation structure, brings the context of the participatory process and the home organization together [15,32]. It prevents contradictory decisions or recommendations in the internal context of an organization. Integration of knowledge and information sources through advanced information management is important in this respect. Bridging organizations, learning platforms facilitate institutional interplay, stimulate learning in the participatory process by providing connection between different scale and higher- and lower level organizations, thereby enabling the integration of knowledge [16,59,70]. According to Cash et al. [73], knowledge is stored and perceived differently at different levels, based on differing beliefs of what is credible and legitimate knowledge and the nature of the problem for which policy is developed and implemented. Cross-level linkages that allow access to validated information makes the information trustworthy and links the participants through the use of the information [74]. Organizational commitment to learning and overall involvement of parties in the process is an important condition for social learning. Continuity, facilitated by, e.g., transfer of information to other representatives of the same organization, fosters social learning [15]. This also creates a sense of belonging for the participants that feedback the process outcomes in their own organizations. Changing staff or participants during the participatory process hinders social learning, as knowledge and experience is lost. Consequently, being a reliable and consistent stakeholder in the policy formulation and implementation processes is important for a person's credibility and the extent to which other parties are willing to accept, e.g., new information [32].


**Table 3.** Organizational factors that influence the social learning process.

#### *3.4. External Factors*

External context factors (see Table 4) are preexisting conditions and forces, but ones that are difficult to control by a specific governance regime [32]. Abrupt periods of change or crisis events (e.g., flooding or elections) are important drivers for change and hence for learning [15]. Such crises may turn out to be windows of opportunity for bringing about political support and supportive regulatory changes, which in turn can alter lengthy administrative procedures that hinder social learning [32]. Vested interests may hinder social learning, since existing regulatory frameworks are supported, allowing little room for lessons in social learning processes [15]. Existing laws and regulation also affect learning processes due to inflexible working processes that leave little room for change.

**Table 4.** External context factors influencing the social learning process.


#### **4. Methods: Case Characteristics and Data Collection**

The framework presented in the previous chapter is used for studying social learning in the context of multilevel flood risk governance. An eye-catching example of multilevel flood risk governance is the Dutch Room for the River (RftR) program. Aim of the program is to increase the buffering capacity of the Rhine river branches (Waal, IJssel and lower Rhine) in case of high water levels. This is done by creating more room for rivers and their branches through making more space for water by, e.g., dike relocation and the lowering of floodplains (Figure 2). By doing this, the Netherlands can deal with higher water levels (16.000 m3/s instead of from 15.000 m3/s) that may enter the country at Lobith, where the river Rhine crosses the German-Dutch border [87].

**Figure 2.** Measures applied to create more room for river in the Room for the River program. Adapted from [13].

The program consisted of 34 separate projects, with a total budget of €2.3 billion. It had a dual goal: both flood protection and spatial quality in the riverine area are increased, by means of, e.g., dike replacement and nature development. Hence, multiple disciplines and organizations are involved in designing and implementing the measures for making room for the river. It is exemplary for a multilevel governance program, as multiple actors, multiple disciplines, and multiple policy and decision-making levels are represented in the program both in the design and executive phase [13,88]. The program involved numerous formal and informal relationships among a group of organizations that agreed to achieve certain goals in a set period of time and a fixed budget. Nevertheless, competing demands and interests occurred.

RftR was initiated in 2000 and was finalized in mid-2017 [89]. From 2008/2009 until approximately 2012, the chosen projects were meticulously prepared by an extensive multilevel stakeholder collaboration process (including e.g., Rijkswaterstaat, provinces, water authorities, municipalities, citizens, societal organizations (e.g., nature conservation organizations, environmental groups)). Early involvement of politicians and non-governmental stakeholders was deemed crucial for establishing commitment and support [88]. The program had a long duration (2000–2017) which allowed for learning-by-doing [10], but it also requires an adaptive approach to deal with ongoing changes in both the internal and external context of the program, like changing stakeholder interests and configuration, socioeconomic developments (e.g., elections or economic crises), and new scientific findings [90]. Two government ministries (Ministry of Economic Affairs and the Ministry of Infrastructure and Environment, I&E), together with eight water authorities, seven provinces, and thirty municipalities, were responsible for the execution of the projects in the program [91]. Fifteen out of 34 projects are executed by Rijkswaterstaat, the remaining 19 are executed mainly by water authorities, and some by provinces and municipalities.

In order to find out what learning has taken place during the Room for the River process, two rounds of in-depth, semi-structured, face-to-face interviews (N = 16) were conducted with policy officers of governmental organizations involved in the RftR program (Rijkswaterstaat, Provinces, municipalities, water authorities). These governmental organizations were chosen since in the Netherlands they have the mandate and responsibility for flood risk management, as opposed to civil organizations or interest groups. The first round of interviews was instrumental in activating the respondents in recalling learning processes since they occurred over long periods of time. The second round of interviews (with the same interviewees) was based on a topic list that was adapted from the conceptual framework (Figure 1) including the possibility to add conditions that respondents felt were instrumental to social learning processes and were not mentioned in the initial framework. We assumed that respondents were able to clarify and distinguish the lessons learned and the different conditions that affected their learning processes. By asking follow-up questions during and after the interviews, we managed to get a better insight into the respondents' perceptions.

Within the program, respondents were selected based on their involvement in either an administrative guidance group (ABG) or a licensing authority working group (AWBG) or both. The first was mainly involved in designing projects, whereas the latter is responsible for the permit procedures for, e.g., soil removal. These working groups were chosen as the platform for selecting respondents because the multilevel governance arrangement is institutionalized in these structures, i.e., all layers of governmental parties in the Dutch water sector are represented in these working groups. Represented are water authorities, Provinces, municipalities, Rijkswaterstaat RftR program directorate (PDR), and Rijkswaterstaat regional offices.

Supplementary to interviews, mid-term and end-term reviews commissioned by the PDR were used to gain insight into the overall RftR program and its performance on, e.g., timing and budget [92–95].

The interview results were transcribed, arranged, and coded using an excel table, which corresponded with the interview topic list: a list of learning outcomes and a list of enabling factors and how they influence learning.

#### **5. Results**

#### *5.1. Learning Outcomes*

The learning outcomes we found can be clustered in collective products [19], development of relations, and new shared interests, ideas, and strategies.

Examples of *collective products* are new strategies for, e.g., permit procedures, the introduction of life cycle costing in the design phase of the projects, strict record-keeping of complex decision-making procedures, and the creation of a bridging organization that facilitates knowledge exchange between projects. Besides, the use of Integrative Project Management (IPM) was introduced during the RftR program and later used in other, non-flood related projects as well. These IPM teams perform five very specific roles with corresponding responsibilities related to major infrastructural projects. The *development of relations* shows in the frequent so-called 'milestone celebrations', frequent face-to-face meetings, both official and in the field and flexible attitudes in formal operations (e.g., the occurrence of shared debates). Shared meaning shows in the parties' recognition and acknowledgment of one another's interests. At some point, parties were able to bring their own interests to the table, without denying other parties theirs. *New strategies* show in the early involvement of a broad range of stakeholders in the projects to allow for a joint project, that is broadly supported. Showing the added value of the project and choosing the right party to lead the project in its different phases is a learning outcome that has been instrumental in ensuring the progress of the projects.

#### *5.2. Learning Process*

The main findings from the interviews suggest that social learning in the Room for the River project is highly influenced by individual attributes and collaborative arena factors, and less by external context and organizational factors, as shown in Table 5.

#### 5.2.1. Individual Attributes

Respondents confirmed that openness to new experiences, self-reflection, and flexibility add to a social learning process. Individuals that are open for new experiences are more likely to observe the lessons to be learned and are more willing to change. '*Learning is change*' (interviewee 1), '*Every now and then you should take a little distance from your own work, and reflect on what's going well, what you should do di*ff*erent. That is the most important thing. [..] Because if you're not open for learning and change, you won't change*' (1). Social skills or the ability to express yourself and formulate interests and arguments is paramount. This skill works in two ways: having the right social skills helps to formulate your own or organizations' point of view in the discussion, but also helps to recognize and formulate viewpoints from other parties or individuals in the collaboration process. It is seen by many as a precondition for collaboration and consequently for learning.



#### *Water* **2019** , *11*, 2032

#### 5.2.2. Collaborative Arena Factors

The role and control of different parties and reflection and feedback is seen by almost all respondents as a positive influence on the social learning process. Mutual trust and interdependency are also seen as preconditions for social learning, although some argue that trust is a precondition for collaboration but not necessarily for learning. Some even argue that a little distrust or suspicion has a positive influence on learning, *'because you learn most of someone who does things di*ff*erently than you do'* (interviewee 2). The opposite was also rhetorically addressed: '*Would you learn without trusting someone? I don't think so. So first of all you'll have to trust someone before you can learn anything. That is crucial. But trust is something you earn. And you earn trust by letting go of the details and checks. [* ... *] So giving trust in fact*' (interviewee 3). The importance of trust is explained by most respondents by linking it to a safe environment for learning, in which people trust each other and share their issues. A safe learning environment through mutual trust is seen as a precondition for social learning by many respondents. The score for both a negative and a positive influence on learning is explained by learning when collaboration is not successful. As a respondent explained: '*Collaboration does not always have to be successful in order to learn, you can also learn when collaboration isn't working out. And you might even learn more from that, but not as a group. You'll learn as an individual, but as a group you won't, because everyone has withdrawn in his own island again*' (interviewee 2). An additional condition that was mentioned in the interviews which influences the level of trust in a working group is stereotypical thinking. Due to parties having a certain image of each other, or ideas about each other's working processes, stereotypical thinking occurs, which negatively affects the collaboration process and consequently the level of trust in a working group. This is also related to previous experiences with other parties. When starting a new project with the same party, it frequently occurs that the participants start the project with a certain idea about the other parties, which frustrates an open start of the project and creates unrealistic or wrong expectations, which in turn affects the level of trust in a working group. Role and control of parties determine which position in the collaboration process you have, which also shapes the interdependency relation with the other parties. This relation determines what your attitude is toward other parties and delineates different tasks.

Reflection, feedback, and trust have a reciprocal relation, as trust is needed before individuals will give feedback and reflect on the collaborative process. Reflection is a joint activity in which the collaborative process so far or process outcome is reflected, whereas feedback is of a more personal nature, between individuals and their performance. Both yield the same results: explicit lessons to be learned to improve the collaborative process. For the other factors, results are mixed. The process architecture appears to be an important condition influencing social learning. The frequency of meetings, field trips, and extended engagement through bilateral contacts are helpful in formulating expectations and frameworks for the team to work with. It helps to delineate the process and creates the network in which the parties operate and the agreements on which the collaboration is based. Communication, both internal and external, is seen by many as part of the process architecture and a positive influence on social learning. Without communication in the working group, no information or feedback is fed back in the collaboration process, which would hamper social learning. Timing and what to communicate and what not were mentioned as well. Time was the only resource that was explicitly mentioned by respondents as a resource that influenced learning. Its effect is both positive and negative: getting time to do a task more than once to gain experience helps learning, but time pressure frustrates learning since there is only a little time for feedback and reflection.

#### 5.2.3. Organizational Context

The internal context of an organization enables learning to such an extent that the organizations that are involved in the working groups need to have a commitment to learning in order for working group members to learn and for social learning to occur. Organizational commitment to learning allows representatives to deviate from the business as usual, which is beneficial for a social learning process: *'Regarding commitment to learning, I think that has been one of the major strengths of the Room* *for the River program. [* ... *] On all levels, this commitment was created*' (interviewee 4). So, it is both a condition in the program and a result of the intensive collaboration processes throughout the program. Integration of knowledge and information sources is important in the context of sharing experiences, but less in sharing actual information and knowledge. The PDR would organize special training days for IPM-team members (e.g., technical manager, contract manager, location manager) to share their experiences and consequently learn from each other. The PDR also made sure that experts from the program office had cross-knowledge of all projects to prevent similar problems to occur and to transfer experiences and knowledge from one project to another. However, this points at sharing of experiences, and less to actual integration of information, which is therefore seen as less crucial. The negative score for this condition is explained by the fact that information and knowledge management was organized very well by the PDR, and there was a strong focus on facilitation of knowledge sharing to decentralized parties, so the necessity for parties to integrate their own information sources was small (see also [95] (p. 28)). Internal conflict on interests were deemed less important for social learning as they are mainly solved in the representatives' home organization, but aligning these different interests is a lesson in itself, although not a 'social lesson'. Involvement of stakeholders also contributes to learning in an ambiguous manner. Involvement of stakeholders was mainly seen as the continuity of staff and consistency in one's role and representation of the home organization. The influence of changing staff on social learning depends on the number of people that are replaced and their function; when too many people leave at once and from a specific team, the collective memory of the organization is affected, but a small change in staff every now and then results in new, fresh ideas that help to learn. It must be noted that this is also an indirect condition for social learning as this condition was researched for the internal context of an organization and not the multilevel governance structure.

#### 5.2.4. External Context

Overall, external context conditions seem less important conditions for social learning compared to individual attributes and process factors. Political support is seen as both a positive influence and a negative influence on social learning; on one side, political support improves the collaborative process and creates space for a project to get going, whereas the lack of political support will create tension in the collaboration structure. A respondent formulated the role of political support like this: *'The lack of political support creates an even more urgent need to have a successful learning process I think. You are regarded more critically, which means the story you bring about the project needs to be really good. [* ... *] when you just go with what the alderman says or wants, he'll like the ideas, and be more like 'we'll just do this', instead of 'why do we want this?', and 'is this really the best option?'* (interviewee 5). It shows the need to be more critical of your own work, which helps to learn. On the other hand, political support for the outcome of a design process will create commitment in the working group, which may positively influence learning.

Crisis events do not directly contribute to social learning, but project-hiccups tend to bring project teams closer together, which in turn may positively influence collaboration and social learning. Administrative procedures do not directly influence social learning, except when they slow down the process to such an extent that the collaboration process suffers from a loss of trust, which frustrates social learning. The same goes for existing laws and regulations.

#### **6. Concluding Remarks**

This paper aimed to identify enabling conditions for social learning in multilevel flood risk governance. We found that most of the factors identified in the conceptual framework are relevant in a direct or indirect way to the learning processes in the RftR program. So, learning processes in multilevel flood risk governance are influenced by individual attributes, collaborative arena factors, organizational factor, and external factors. However, our case study results indicate that a hierarchy of conditions can be identified.

Individual attributes and collaborative arena factors can be labeled as key, whereas organizational factors and external factors are deemed supportive. Individual attributes of participants in IFRM projects seem to play a major role in the social learning process, as a participants' individual attributes ultimately determine the commitment to learning.

The relative importance of individual attributes and collaborative arena factors in the social learning process is also reflected in recent work of Bentley Brymer et al. [95], who state that a change in understanding (see [21]) is foremost an individual step that takes place in multiple dimensions of their understanding. On collaborative arena or process factors, they state that factors that trigger social learning are mostly inclusiveness, extended engagement, opportunities for information exchange, and opportunities for dialogue and interaction, which correspond closely with the factors in our framework. Studies on the Quebec water governance system [96] and the Bangladesh Adaptive Delta Management [97] show findings that also point at some kind of hierarchy in the conditions for social learning. Our findings on the importance of external factors is to some extent contradicted by Johannessen et al. [98], who state that transformations in water governance are often triggered by crises. This can be explained by their focus on multiple-loop learning and the outcomes of social learning, as opposed to a focus on the process of social learning.

However, more comparative future research on conditions for social learning in different contexts should point out the validity of this hypothesis (e.g., [99]). It would, for instance, be interesting to find out whether studying social learning in the context of the German Room for the River project results in the same findings. Moreover, future studies should also pay attention to political indecisiveness and distrust and other factors hampering social learning. Such studies could result in a more validated theory on social learning in multilevel flood risk governance. A very relevant topic to do further research on is how the level of centralization of the government system, as for instance addressed by Bonasia and Lucatello [100], affects social learning processes. In the Netherlands, the governmental system is very centralized (see e.g., [101]), which leads to a clear distinction of tasks and responsibilities, whereas in, e.g., Mexico, the governmental system is much more decentralized, which affects decision making in disaster risk management [100]. Further application in different socioeconomic contexts would further help refine the framework and judge its usefulness in understanding and facilitating learning processes in flood risk governance.

The framework drawn up in this paper is based on literature on natural resources management, adaptive co-management, and social learning. As another avenue for future research, the framework may be enriched by adding additional insights or governance conditions based on stakeholder theory [102], communities of practice [103], and network governance [104].

Social learning appeared to be a process that lacks a starting point as well as a finish line. Learning itself may emerge from both success and failure. It is a constant process of balancing stakes, sharing information, and the creation of mutual trust and collective memory. Learning itself is an intangible process that may happen without the learner fully aware of it, and is mostly identifiable in retrospect. So multilevel flood risk governance asks for the setup of learning organization or platforms in which reflection and openness to change are core qualities. More specifically, we conclude that in order to enhance the changes on successful learning, such a learning organization can try to mitigate influencing conditions. Process factors and internal context can be organized as such that individuals committed to learning are attracted and are willing to participate in a collaborative process. A program directorate can trigger social learning processes by designing specific learning workshops, in which mutual trust and reflection are addressed. Besides, a learning organization can contribute to learning processes by actively taking up a bridging function in the sharing of knowledge and experiences throughout the program. Learning from past experiences is of great importance for the development of future flood risk governance programs.

**Author Contributions:** Conceptualization, J.d.B., C.D., F.M.; Methodology, J.d.B., C.D., F.M.; Validation, C.D., F.M.; Formal Analysis, J.d.B.; Investigation, J.d.B.; Writing-Original Draft Preparation, J.d.B., C.D. F.M.; Writing-Review & Editing, C.D, F.M.; Supervision, C.D.

**Funding:** This research received no external funding.

**Acknowledgments:** The authors would like to greatly acknowledge Cor Beekmans for his contribution in creating the opportunity to carry out this research project at the Rijkswaterstaat Room for the River Program and his valuable feedback and in-depth discussions on the research outcomes. We also would like to thank the anonymous reviewers for their valuable feedback on the article.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **City-To-City Learning for Urban Resilience: The Case of Water Squares in Rotterdam and Mexico City**

**Silvana Ilgen 1, Frans Sengers 1,2 and Arjan Wardekker 1,3,\***


Received: 11 April 2019; Accepted: 7 May 2019; Published: 10 May 2019

**Abstract:** Cities worldwide are building 'resilience' in the face of water-related challenges. International networks have emerged through which urban communities draw on each other's experiences and expertise in order to become resilient cities. Learning is a key principle in resilience-building, but thus far little empirical research is available on city-to-city learning and learning for urban resilience. This paper presents an analysis of how policy relevant knowledge on the notion of 'Water Squares' is exchanged between Rotterdam and Mexico City. We mobilize a framework composed of four distinct phases: exploration and marketing (phase 1), building pipelines (phase 2), translation and adoption (phase 3), and internalization and reflection (phase 4). Critical in first phase was introspective analysis of one's own systems, strengths and weaknesses, rather than an outward-looking search for knowledge or mentees. During the second phase, the cities reframed their own narratives to match those of their counterparts as a way to create a mutual understanding of each other's struggles and histories. This facilitated policy and knowledge exchange as equal partners on a basis of trust. In the third phase, strong local leaders were recruited into the process, which was key to anchor knowledge in the community and to reduce the risks of losing institutional memory in centralized, hierarchical institutions. For the fourth phase it should be stressed that by internalizing such lessons, cities might strengthen not only their own resilience, but also enhance future exchanges with other cities.

**Keywords:** city-to-city learning; policy transfer; resilient cities; water squares

#### **1. Introduction**

More than half of the world's population lives in cities and these numbers are still rising [1]. Cities are under intensive pressure to accommodate the needs of these rapidly growing populations, including providing adequate and safe housing, production activities and work, and resources such as energy, food, and water. The expansions of cities and the urban adaptations that were made have given rise to sprawling metropolises beset by a range of social and environmental problems [2]. Many current and future global challenges relate to the interaction of urbanization and its social and environmental impacts, and cities may therefore host both the problems but also the potential solutions to these challenges [3].

Climate change is a key sustainability challenge to the future development of cities. Potential effects of climate change range from flooding to drought, impacts on water quality and availability, and impacts on critical infrastructures, health, food supply, tourism, biodiversity, housing, and communities [1]. These can result in short-term disruptions and disasters in cities, as well as have a broader and long-term influence on urban resource and population dynamics and environmental and

social equity. In cities worldwide, and in developing countries in particular, these impacts are expected to also exacerbate numerous existing problems [1,4,5]. The impacts and their interactions with other societal challenges are, however, complex and uncertain [6]. The need is widely recognized for societies at large, and cities in particular, to better promote and govern a transition towards sustainability [7–9]. This includes mitigation, reducing greenhouse gas emissions, but also making cities more resilient to climate change and other social, economic, and environmental challenges [8–10].

The notion of 'urban resilience' has become prevalent in urban thinking. The Habitat III New Urban Agenda [9], for instance, focuses considerable attention on the need to improve urban resilience, including to climate change. Resilience is one among a number of urban sustainability concepts, such as 'smart', 'low carbon', and 'eco cities' [11]. The term 'resilient city' refers an urban area that has the ability to withstand and cope with a wide range of shocks and stresses, including the ability to plan/prepare, absorb, recover, and adapt to such disturbances [12,13]. The literature on urban resilience encompasses a wide range of scientific fields, including urban ecology [14,15], urban hazards and disaster studies [16,17], urban and regional economics [18], governance and institutions [19], spatial planning [20–22], and climate change adaptation [12,23]. Resilience is particularly focused on living with change and coping with disturbances in complex adaptive systems, with multiple pressures, economic sectors, policy fields, and scientific disciplines involved, and high levels of uncertainty [22–27]. Urban water systems are a prime example of such a situation, combining highly interwoven social, ecological, and technical systems with changing and highly uncertain pressures and requirements due to climate change and a multitude of other urban trends [28,29]. In many places, water scarcity and uncertainty are already forcing a re-think about the way governments manage their water resource systems [19]. Urban water reforms and adaptations should result in resilient water systems that explicitly take into account complexity, uncertainty, and immediate and long-term change [30].

Many cities are now incorporating resilience thinking into urban planning at large, and specifically with regard to climate change adaptation. Different cities will (or may want to) opt for different strategies to improve resilience, and toolkits have been developed to design resilience strategies and policy options appropriate for a city (e.g., [31–35]), and to scrutinize choices and trade-offs made in that process (e.g., [36–39]). Several international organizations and networks have been established to stimulate resilience-building, as well as actively exchange policies and knowledge as a way to facilitate mutual learning. Examples include Local Governments for Sustainability (ICLEI), the Asian Cities Climate Change Resilience Network (ACCCRN), and 100 Resilient Cities. On the broader topic of sustainability, a wide range of such networks has emerged, often with a similar organization, large role for regionally-oriented cities, but irregular representation of developing regions [40]. The importance and potential of such networks for improving systems and building sustainability and resilience is widely recognized [40–44]. Policy exchange, knowledge transfer, and learning in these networks can help cities enhance policy innovation [44–46]. For urban policymakers, their peer network in other cities, and exchanges through conferences and city networks, are key sources of information and new ideas [47]. According to an UNESCO-IHE flyer [48], advantages of city-to-city learning include: accelerated transfer of knowledge and experience, joint knowledge creation, empowerment of local governments, reinforcing local networks, enhancing collective memory of participating cities, and increasing learning capacity. Learning is also described as a key characteristic of resilient socio-ecological systems, including cities [5,49]. However, individual cities learn in different ways [50], and full social learning can be hampered in situations of diverging perceptions, framing, values, interests, and levels of power [5]. Between cities, further difficulties may emerge due to differences in the culture and language, specifics of the local situation, institutions and regulations, and even the specific ways in which agents and persons involved in the knowledge and policy transfer act or handle the situation [40,46]. Consequently, while cities can learn much from each other, stimulated by the resources and reach of transnational city networks, such knowledge and policy transfer may not always run smoothly and requires constant dedicated work [51].

At the moment little is known about the mechanisms that underpin knowledge and policy transfer on urban resilience among cities, and about what barriers and stimuli that influence this process. The literature available on the topic is often prescriptive rather than descriptive [52]. Clearly, it is not simply a matter of cutting and pasting from one location to another. Cities are already actively trying to exchange ideas and best practices, and there is a need for more systematic knowledge on how this can be facilitated. This paper examines a case where such policy transfer is actively and intentionally pursued, using a preliminary analytical framework. Particular attention will be paid to the actors, processes, practices, and structures involved in the transfer. This study will focus on the transfer of resilience thinking and practices between Rotterdam and Mexico City, and will draw lessons from the observations made in this process. More specifically, we will investigate city-to-city learning around the notion of the Water Square—a best practice project already implemented Rotterdam and yet to be implemented in Mexico City.

#### **2. Materials and Methods**

To investigate city-to-city learning for the transfer of policy and adaptation measures to build urban resilience in the context of climate change, we conduct an empirical, in-situ analysis. Our approach can be described as an 'explanatory case study' because we do not only explore or describe in-situ learning but also develop theory to deepen our understanding of key processes or mechanisms and [53,54]. We examine the case of Mexico City and Rotterdam, two cities currently actively engaged in a process of policy and knowledge exchange, particularly on Rotterdam's concept of Water Squares (Section 2.1). Tangible best practices projects like water squares are of special interest here because they became focal points in collaboration efforts and a vessel that facilitated much wider transnational knowledge flows and broader lessons about urban resilience in general. We examine this process of city-to-city learning sparked by these water squares in action through semi-structured interviews, site visits, and participant observation (Section 2.2). This process is assessed using a preliminary analytical framework (Section 2.3).

#### *2.1. Case Study*

The resilience literature widely acknowledges learning as a key aspect in the resilience of complex adaptive systems, including cities, and crucial for policymaking under uncertainty and complexity [5,12,52,55–58]. In such cases knowledge is always incomplete and change and surprise are inevitable [48] Learning is related to capacities, such as innovativeness and resourcefulness, as well as to practices and relations, such as experimentation, knowledge sharing, and collaboration. Learning has been used loosely in the resilience literature, combining notions of individual, organizational, and social learning [57]. Bahadur and Thornton [52] observe that much of the resilience literature is prescriptive on learning—describing practices that cities employ to stimulate learning. They argue that a closer, more practical look at learning in empirical context is needed, exploring the situational aspects that determine whether learning occurs or faces barriers, in order to overcome the current idealized and prescriptive notions of knowledge and learning for urban resilience. In line with this reasoning, this paper will explore a case of learning for resilience from an empirical in-situ standpoint: as the process of learning unfolds.

To develop such an in-situ understanding, we investigate the case of policy and knowledge transfer between Rotterdam and Mexico City. This entails an effort to 'mobilize' knowledge from Rotterdam to enhance water resilience in Mexico City. Both cities have flooding-related vulnerabilities, and Rotterdam has been very active (and successful) in flood risk management. Both cities are still searching for options that will enhance their water resilience. Both have appointed 'Chief Resilience Officers' and recently published a 'Resilience Strategy' [59,60]. There are also notable differences, including geographic, cultural, political, and institutional, that could influence the successfulness and ease of policy and knowledge transfer. A coordinated effort has recently been started to exchange

experiences and knowledge, providing a valuable opportunity to study this process in action. We will briefly discuss the background of both cities and the water squares case.

Mexico City (Mexico) is a city of 8.9 million residents (city proper), with a subtropical highland climate. The city is located in the lower part of the Mexico Valley Basin around 2200 m above sea level, in a former lakebed surrounded by mountains. The lake system used to act as natural drainage for precipitation run-off, which was carried down by rivers and streams from the higher elevations that surround the basin. The hydraulic cycle of the lake system has been extremely and irreversibly transformed. Mexico City has made some achievements in water management, but continues to face challenges and vulnerabilities to flooding and water availability due to the structural features of the water system, changed hydrological cycle in the basin, and unequal access to supply and to drainage services [61]. Mexico City has been faced with wet years and floods alternating with episodes of drought. Average yearly precipitation in Mexico City is 846.1 mm. It is relatively concentrated in one half of the year. Key climate related challenges include flooding due to heavy precipitation and flash floods, heat waves (combined with a heavily urbanized area), and drought and water scarcity. Groundwater extraction from the aquifer below the city is also leading to soil subsidence.

Rotterdam (The Netherlands) is a city of 0.6 million residents (city proper), with a temperate oceanic climate. The city is located in Western Europe, at the Rhine–Meuse–Scheldt river delta and North Sea coast. The city is located on river banks, polders, and reclaimed land, and much of the city is below sea level (up to −6 m). Rotterdam hosts one of the largest harbors in the world, and its economy is heavily dependent on the careful control of water. The city has longstanding experience and expertise regarding water management, and is currently pairing urban renewal with water sensitive and resilient adaptation (e.g., [23,62]). Specific water challenges include precipitation runoff in the low-lying polders due to more intense rain showers, and riverine and sea-based flooding in unembanked areas and flood risks in the polders. Average yearly precipitation in Rotterdam is 855.6 mm, relatively spread out over the year. Key climate related challenges include flooding due to heavy precipitation, flood risks from sea and rivers, and heat waves (combined with a heavily urbanized area). Drought and soil subsidence are an issue, but less than in Mexico City, partly also because drinking water is extracted from the river rather than the aquifer. Drinking water scarcity is no issue.

The 'Water Squares' in Rotterdam were demonstration projects developed in the context of the Rotterdam Water City 2035 program, and the earlier Water Plan 2. The Benthemplein Water Square [63], completed in 2013, is the best-known, but several other squares have been developed. The squares core concept is that they are dry under regular conditions, and can be used as a public square, but can temporarily store rainwater during downpours. The squares are recessed (one or more aboveground basins) and connected to underground basins and sometimes groundwater infiltration devices. These provide buffer capacity, allowing rainwater from the square and surrounding buildings to be retained until the city sewage system can better handle the discharge. The water squares explicitly aim for multi-functional solutions, such as rainwater buffering, making water management more visible to the public (communication), and enhancing the quality of public space. Rotterdam and Mexico City have taken such squares as focal point in their collaboration. In Rotterdam, water squares have been developed to the north and to the west of the city center, and lessons learned from these early designs have been included in the Zoho Climate-Proof District (north of the center), which expands on the earlier designs with broader climate and social resilience efforts, such as explicit attention for community initiatives. All of these locations are highly urbanized areas, and slightly below sea level (0 to −1 m), protected by dikes. In Mexico City, proposals focused on the Iztapalapa district, a highly urbanized, low elevation area on the former lake bed of Lake Texcoco. For a detailed description of the design process in Mexico City, see [64,65]. See Figures 1 and 2 for locations of the water squares and climatic profiles of the Netherlands and Mexico. Supplementary Materials S3 contains further details on the climate in both cities.

**Figure 1.** Locations of the water squares in Rotterdam (**left**) and proposed locations in Mexico City (**right**) (map source: Google Maps).

**Figure 2.** Temperature and precipitation profiles for the Netherlands (**left**) and Mexico (**right**) (data source: World Bank Climate Change Knowledge Portal; data from: CRU, University of East Anglia).

#### *2.2. Data Collection and Analysis*

To explore the actors and issues relevant to the case study, two helicopter interviews were conducted with the architectural firm that designed the water squares and with an international governmental organization involved in city-to-city knowledge transfer.

A first set of ten in-depth semi-structured interviews was conducted with actors in Mexico City, Rotterdam, and international organizations involved explicitly in the transfer of knowledge and policies concerning the water squares. These were focused interviews, exploring the specifics of the collaborations. A second set of ten semi-structured, but more open-ended interviews was held with various actors involved in the broader collaboration and knowledge exchange on water and resilience between the Netherlands and Mexico, including other water resilience projects. These interviews provided information on the local, national, and international context of the case study and potential barriers and stimuli that might and did arise. Interviews were transcribed, coded in nVivo, and qualitatively analyzed. An anonymized list of the interviews and the coding scheme can be found in the Supplementary Materials.

Other data was collected using site visits and participant observation. The first author was embedded in the Dutch embassy in Mexico, which is highly relevant because for the case of the water square this organization fulfilled the role of an 'innovation intermediary' [66]. The embassy played an important role in coordinating and facilitating knowledge exchange between Mexico City and Rotterdam. Specific project sites were visited to gain a good understanding of the local situation. Furthermore, the first author participated in local project meetings and workshops on the water square project as well as five other collaborative water resilience projects and events. As participant observer (cf. [67,68]), the researcher had a moderate to active level of participation in the resilience projects and could observe the collaboration, decision-making, and social interactions in this network at close hand and in-situ. Observations were recorded in a logbook and qualitatively analyzed. They provided

valuable insight into the tacit aspects of city-to-city knowledge and policy exchange, such as social and cultural aspects, supplementing the more formal setting of individual interviews.

#### *2.3. Analytical Framework*

An empirical, in-situ analysis requires a process-based analytical framework, rather than a conceptual principle-based one. One challenge is that, as noted above, most literature on learning for resilience is prescriptive rather than empirical. Another challenge is that most sources focus on learning within communities and cities, not on learning between them. Orleans Reed et al. [5] do provide a process model for resilience learning in the context of the ACCCRN network. Steps include: gathering information and perspectives on hazards, socio-economic development and urban plans; vulnerability assessments; pilot projects and sector studies; developing a resilience strategy; city interventions; and learning from interventions, revisions, and further networking. However, this model also focuses on learning within cities ('shared learning' with local stakeholders), rather than knowledge and policy transfer between cities. Process frameworks like these are reminiscent of resilience analysis process methods (e.g., [23,69,70]), but with more focus on implementation design and testing, including pilots and strategy development. e.g., Wardekker et al. [70]: characterizing the system; characterizing the disturbances; inventorying options/plans; evaluating options/plans; follow-up analysis; and adjust plans/options and repeat if needed. Roughly, such frameworks contain an exploration/orientation phase on the context and issues, followed by inventory or design of pilots, options, plans or strategies (which may include implementing pilots or early measures), and an evaluation, testing, reflection, and refinement phase. While both in-city learning frameworks and general resilience assessment frameworks are relevant for city-to-city learning, they do not cover the full process of interaction that takes place between cities.

Contrasting this with the notions above, the framework would need to place more emphasis on the contact and interaction between cities (before detailed city-to-city subject-matter discussions and exchange take place) and the anchoring of the transferred knowledge (after the exchange of knowledge). Consequently, we have structured our analysis using a four-phased analytical model that describes different stages in the interaction and learning process between cities (see Figure 3 for an overview, elaboration per phase below).

**Figure 3.** A circular model of the process of city-to-city knowledge transfer. City A imparts its knowledge on City B.

We start with the idea that communication channels for knowledge exchange can be purposely created to foster resilient urban development. Before these channels between knowledge providers and users can be constructed, urban actors need to become aware of their current knowledge resource and assess that is value based on comparison to others [2,5,71]. This includes critical reflection on the knowledge available, knowledge culture, local setting (institutions, regulations, history, ecology, geography, and socio-political dynamics), and physical water infrastructure in their own city and in other cities [28]. In reality not all other places can be assessed in the same way to see if they are

viable candidates for knowledge exchange, and shortcuts are taken. Some actors are more adaptable than others in telling success stories and framing their city as a 'place of best practice' [51]. Notable examples include Copenhagen flaunting its expertise on cycling, Bogota peddling its knowledge on inclusive public transport across the globe and, most importantly for this paper, Rotterdam marketing its water know-how. Other cities are drawn to these cities and their best practice narrative in their exploration for policy relevant knowledge to reshape their urban environment in similar ways. We therefore refer to this first phase as the 'exploration and marketing phase', where knowledge providers in best-practice city A seek to market their knowledge and the knowledge users in city B seek to explore the wares that best-practice cities like A are selling.

Second, since the actors involved are anchored in different local settings, a concerted effort needs to be made by two cities to connect. The potential linkages dreamed up in phase 1, through the marketing and exploration of knowledge stocks, need to be translated into real connections and formalized collaboration agendas. This solidifies the transfer channels between city A and city B. To make this happen city A and city B will need to reflect on each other's situations and develop mutual expectations [72]. They also need to consider the transfer, translation, or transformation of policy models [46]. Similar to regular strategic program development [29], we expect that developing a shared vision and securing long-term commitment will also be important for knowledge and policy transfer programs. It is important to note that different kinds of policy-relevant knowledge travel through these transfer channels—not only codified knowledge (such as text in manuals and guide books about alleged best practices in city A) but also tacit knowledge (experimental and practical insights embodied in the actors of city A not ready made to be implemented in city B). Elements of both tacit- and codified knowledge can be exchanged locally and globally when the 'local buzz' of vibrant information exchange and interactive learning within local milieus of city A links up with receptive local milieus in city B. Knowledge is embodied in actors and actors engage in the building of 'pipelines' to communicate and harness extra-local knowledge flows [71,73]. Therefore, we refer to this as the 'build pipelines phase'.

Third, the policy-relevant knowledge from city A, which has travelled along with the relevant actors through the solidified channels, needs to be implemented in city B. This is not simply a matter of taking a policy from one location and implementing it in another. Rather, it involves adapting it to a specific local context, practice and policy community, with a focus on the political process [45,46,74]. Local adaptive capacity, in general and of local institutions, is a relevant factor in this stage (cf. [75]). In other words, a process of translation takes place before adoption. The eventual adoption of this knowledge results in concrete changes into the material and intuitional fabric of city B. We call this the 'translate and adopt' phase.

Fourth, looking back on this process over the previous phases, the hopeful expectation is that actors in city B—but also city A—have increased their knowledge pool and deepened their understanding of infrastructures and processes that foster urban resilience and that both are now more motivated and better equipped to transfer their knowledge to even more cities for the next round in the cycle. But equally important, the hope is also that this newfound knowledge is firmly anchored in material infrastructures and local practices of city B. Too often, unfortunately, experimental projects with innovative green infrastructure turn out to be isolated events that fade into oblivion without any effect on incumbent infrastructures and practices and the seemingly solid network connections that facilitated these projects might turn out to less robust when ties are not continually maintained or when a key individual is no longer part of a relevant organization [76,77]. To make this knowledge become robust, it is important that both cities 'reflect and internalize' the experiences in previous stages and to anchor these it in local practices and into vibrant and well-maintained knowledge transfer pipelines. We call this the 'reflect and internalize' phase.

#### **3. Results**

In this section we will apply the four-phased analytical framework laid out above. For each of these four phases, we will interpret the events that took place, specifically examining water system variables such as the context, system, actors, processes, and structures [29]. This analysis sometimes also highlights the cognitive and institutional proximity or distance between the cities involved [78,79].

#### *3.1. Phase 1: Explore and Market*

The Dutch Embassy was setting up a project called Holland Branding to increase business opportunities for the Netherlands in Mexico. This long-term Holland Branding strategy is a tool for the embassy and Dutch business and knowledge institutions to materialize the opportunities into concrete projects. It started with brainstorm sessions within the Embassy and interviews with multiple companies and organizations located in Mexico City to understand the (current) reputation of The Netherlands. The end result was a plan for a shared vision on Holland Branding that was discussed with all the employees involved in the focus sectors of the embassy. Several observations could be made during this process.

First, according to the actors involved, it is important to understand the strengths and weaknesses of The Netherlands before it is possible to search for market opportunities in Mexico. A strength of The Netherlands is its water systems. The sinking of coastal lands has continually increased the Netherlands' vulnerability to floods and made it more difficult to discharge drainage water. However, the Dutch have been able to cope with these problems through the centuries by increasing their control over water flows. This they have achieved through collective action, often involving very many people [80].

Second, research was needed on the Mexican market. Attention was given to drivers for Mexico to set up new projects in the focus sectors. Regarding the water sector, these are mostly preventing floods, droughts and shortages of water. Besides the drivers for Mexico, it was necessary to set up the identity of the Netherlands, in other words, the Dutch approach. A SWOT analysis of the Netherlands in Mexico was made based on the information of the interviews to match the identity with the drivers of Mexico. The information was formulated in general messages and themes were set up, in which the focus sectors were combined for a cross-sectoral approach. In each of these themes urban resilience plays a role.

The cross-sectoral approach was also a focus point in the concept of the water square in Rotterdam. The engineers developed it to benefit public space and water management. The idea was to bring the water back into the city and keep the water local instead of drainage. This one local system created international interest. An organization in Mexico City noticed an article published on the water squares, while they were searching for a project that combined public space with sustainable water management.

*'Given the vulnerability in terms of water in Mexico City, I thought that it would be interesting to integrate water but in a more responsible manner. So that is when I started with a research and that is why I was asking my colleague about a kind of project which mixed water management or responsible use of water on spaces* ... *'*

This quote showed that the organization knew what it needed to enhance its current systems. On the opposite side, Rotterdam was showing the strengths of their concept and uniqueness. The Dutch Embassy was setting up a more holistic approach to connect concepts like the water square with the market of Mexico. The focus on strength and weaknesses and connecting them is also highlighted by observations made by participants. They indicated that it is key to understand your own systems first.

*'One of the key things is that each organization recognizes their strengths and their weaknesses. And when you understand which are your weaknesses then you understand that you need help on those weaknesses.'*

In short, before cities can search for complementary knowledge or market their knowledge they need to focus on their own systems. It is a time-consuming process to understand the strengths and weaknesses of a city's systems before marketing or exploring knowledge, namely research on the city's systems is necessary to form a framework of strengths and weaknesses. This is a step that leads to the possibility of building channels of communication to selected providers outside the local milieu to enhance knowledge.

#### *3.2. Phase 2: Build Pipelines*

Following the initial search and connection between the cities involved, broader contacts will need to be established between relevant actors in those cities. This contact will also need to be solidified into joint commitment towards the collaboration.

In case of the water squares, the process started when the Spatial Planning Office in Mexico City contacted the Dutch Embassy, late 2014, to explore interests and potential collaboration. A first meeting was held in February 2015 to discuss the knowledge that Mexico City was looking for in more detail. Following this meeting, the Embassy established contact between the Spatial Planning Office and the architectural firm that designed the water squares in Rotterdam. A series of phone and Skype calls followed, and a first face to face meeting was held in December 2015 in Mexico City with the Spatial Planning Office, Embassy, and architectural firm. This meeting helped to build trust, in the sense that the organizations established openness towards each other's views and willingness to cooperate. The actors shared ideas and established an agreement towards further collaboration (resp.9).

In February 2016 a workshop was organized to connect a wider base of key actors from both cities. The workshop was in the form of a round table discussion, with a more targeted focus on the potential of creating a multi-purpose water square project in Mexico City. This four-day "Water Workshop" gathered seven core actors, four Mexican and three Dutch (see Figure 4). During the first day, all actors gave presentations on their background, organization, and activities. Many participants left directly after the presentations, for unclear reasons. Only most Dutch participants, one participant of the Mexico City Resilience Office and two of the spatial planning agencies remained.

**Figure 4.** Impression of the water workshop in Mexico City (photo by: Víctor Manuel Rico Espínola) [64,65].

In the following two days, the Dutch participants attempted to gain further understanding of Mexico City by means of a more in-depth analysis of the city. The city was divided into five areas, each with its own environmental and social needs, and potential water-related solutions were explored using the 'CAPP Climate Adaptation App'. The information was then spatially structured and the potential and usefulness of water squares was explored per area. During the final day, this analysis was presented to the Mexican organizations, leading to further discussion. The effort and ideas presented

were very well received and seemed to lead to an enhanced will to cooperate. Agreements were made on further steps, and the Dutch partners produced a report of the shared vision in collaboration with stakeholders.

Observations during the workshop and follow-up interviews suggested several factors that led to this success. First, the approach of exploring community needs per area in the city showed the Mexican participants that the Dutch were making a conscious effort at trying to understand their situation and specific needs, enhancing trust. This was important to overcome the cognitive distance between the participants due to the large geographical distance and differences, e.g., (resp.7):

*"Knowledge from rich countries to poor countries does not translate easily. Because we have a very di*ff*erent political and cultural situation; we are in di*ff*erent stages. [* ... *] For example The Netherlands is very small and we have a lot of mountains. And then you get a lot of 'buts'. And it is not that it can't be done. It is just that it is so far away."*

Second, the analysis was framed from the perspective of similarities between Rotterdam andMexico City, revealing shared struggles and histories, rather than highlighting Rotterdam's accomplishments. Interviewees noted that this enhanced mutual understanding, as it highlighted mutual struggles with water, rather than implicitly suggesting superiority. As one interviewee described an experience with another country (resp. 7):

*"So when I spoke, I said: We are very corrupt, just like you. And they were very happy to hear that, because that is an understanding. [* ... *] And I know that due to the Second World War, countries were very poor. However, when they come, they don't talk about how they were poor. They talk about how they are now the rich countries of the world."*

In sum, for the second phase it was important to create a shared vision by involving a relatively broad set of key actors, more than just the linking-pin organizations. Some actors were less committed to the process initially. However, these differences were bridged by a strong effort by the visiting party and hosting linking-pin organizations to jointly explore the system and relevant context, i.e., the situation on the ground, the requirements for resilience and adaptation in different neighborhoods, and potential options. In addition, the geographic proximity was observed to be fairly low between the cities; one city a huge landlocked metropolis in a developing country and the other a medium-sized coastal city in a developed country. It was particularly valuable to highlight shared struggles between both cities to build trust and mutual understanding. This helped bridge the geographic distance and the willingness of all actors to collaborate, which enabled the actors to engage in building of 'pipelines' to communicate and harness extra local knowledge flows. As a result, the water squares became a potential lighthouse project.

#### *3.3. Phase 3: Translate and Adopt*

The agreements that were made in the first meetings, were written in a shared report. The intention was that this would lead to a joint project on water.

The differences in the organizations involved in the process provided some challenges. Particularly, the differences between the Dutch and Mexican government systems caused a lack of organizational proximity along with a cognitive distance between the actors. Much of the technical knowledge within the Dutch actors resided in privatized companies or (public or semi-public) knowledge institutions (Respondent 9). Dutch urban water management systems involve a number of different organizations that need to collaborate due to the spread of responsibilities and expertise, i.e., the government system is more decentralized. The Dutch government system is also known for the strong collaboration between different public and private actors (Observation 4). Mexican government organizations often involve a combination of technical and policy departments. For example, SACMEX is the water organization in Mexico City that contains governmental and technical knowledge. SACMEX, therefore, is almost independent and as a result more powerful and does not share its knowledge and expertise easily (Respondent 9). As a result, the organization was under less pressure to contribute to a collaborative project such as the water squares, and was difficult to reach by other participants during the process of designing them. Unlike the Dutch system, the Mexican government system is highly centralized.

A clear example of the challenges of the lack of organizational proximity presented itself when there was an unexpected change of director in the Mexican organization. Many new leaders are reluctant to continue with projects that started under the previous leadership due to the importance of personal image and reputation in Mexico. The important decision makers within the organization mostly change with a change of a governmental leader. Consequently, parties who are working together with the government need to set up new contacts all over again within the governmental organization. The combination of a centralized governance and changeable leadership makes it hard to set up projects that involve a longer time frame (Respondent 10).

However, in case of the water square, a local leader was directly involved in the organization supporting this project. The person actively promoted the project to the new director, and setup a meeting between the director and the Dutch Embassy to further emphasize the importance of the project and the Mexican–Dutch collaboration (Respondent 9). The report showcasing the water squares as a lighthouse project was completed and could help convince the new director. This showed that local leaders are necessary, both external to promote interaction with the communities and internal (within actors) to push the process within their organization.

The interviewees also elaborated on this need for local leaders for a successful implementation of an innovation. Local pressure is often needed to overcome obstacles, i.e., much practical progress comes from bottom-up initiatives. Local actors and communities are necessary to carry through on long-term, multi-purpose projects (Respondent 9, 10; Observation 4). As interviewees mentioned:

*'So first we need to inform all these government agencies but at the same time we need to approach communities and we need to make sure to incorporate them into the design of the space. So in the end we learned that we design a process more than just a public space project or an architectural project.' (Respondent 6)*

*'And people, they create pressure. Awareness of the problem is also very important. If that is not there then it is very di*ffi*cult to do something.' (Respondent 1)*

*'We have a strategy for that. What we are doing is, we are trying to establish a committee that is depended on the project. So we call people from the academy, and comparable important leaders and personalities and we are going to invite them to take part in this committee to keep this project alive for years to come.' (Respondent 5)*

One of the challenges within the case was identified as the lack of organizational proximity that could negatively influence the policy learning and implementation of the water innovations. A lack of organizational proximity means that organizational flexibility will be required to tailor and adopt innovations [78]. Knowledge was also highly centralized, i.e., key experts from other organizations were not always available or left the office due to a change in leadership. This created a cognitive distance between the Mexican and Dutch actors that made it difficult for them to collaborate. These obstacles within the process show that city-to-city learning is not simply a matter of taking a policy from one location and implementing it in another. Rather, it involves adapting it to a specific local context, practice and policy community, with a focus on the political process [45,46,74].

A secondary challenge, potentially reinforcing the former, was that few people involved were able to connect different disciplines. In Mexico City, there were many people who were highly educated and experienced in their specific discipline, but few people with an interdisciplinary focus (Respondent 1, 2). Therefore, it proved difficult to translate the knowledge of individual disciplines towards interdisciplinary, multi-purpose projects, i.e., combining sector specific knowledge towards themes for projects (Observation 1).

In sum, during the translation and adoption phase, we observed strong gaps in the organizational and cognitive proximity between the actors. Partly, this related to the different actors involved on both sides: primarily public for Mexico City; mixed public-private for Rotterdam. More important, however were differences related to the structures of the governance system, such as the institutional environment (e.g., modes of governance [81], styles of leadership, perspectives on fair governance) and the knowledge system (e.g., disciplinary-specialist versus interdisciplinary). Process-related differences played a role as well, such as the tendency of key decision-makers and experts to leave and organizational priorities and support for projects to shift strongly following a shift in leadership. The shared vision developed in phase two helped mediate these lacks in institutional and cognitive proximity. Key to overcoming the differences, however, was the presence and mobilization of local leaders, in the form of community leaders, broad societal committees that took ownership of the project, and leaders within key organizations that helped the project transition changes in leadership and kept parties on both sides activated.

#### *3.4. Phase 4: Internalize and Reflect*

The case of the water square is still in development and currently in phase three. In addition, the 100 Resilient Cities network reached its goal of a hundred cities in 2016 and the network is still developing (Respondent 3). So, unfortunately, there is not a lot of information on this phase yet. Nonetheless, it can be said that the intensity of the interaction between the cities decreases in this phase. Actors in Rotterdam will reflect on the whole process and incorporate the newly gained knowledge and experiences for the next round of knowledge transfer with another city. In its turn, actors in Mexico City need time to internalize the policy relevant knowledge that was transferred so that it can be incorporated in other systems of the city or be used in a next round of knowledge transfer with another city. In other words, both cities have increased their knowledge pool and deepened their understanding of infrastructures and processes that foster urban resilience. And both are now more motivated and better equipped to transfer their knowledge to even more cities for the next round in the cycle. Additionally, the increasing number of knowledge transfer rounds on urban resilience helps with defining the framework around the concept more precisely. Cities gain knowledge by experiencing more knowledge transfers on urban resilience. Both cities need to market the project, individually and together, to create new projects—new knowledge transfers—on the knowledge they gained. However, it can be expected that the highly centralized Mexican government system will influence the internalization process negatively due to organizational inflexibility. One interviewee already mentioned an aspect for progress that can be reached by reflecting on cases and implementing the new knowledge:

*'It is totally important to exchange information and knowledge although it is also critical taking into account the particularities of each city. Even though the resilience initiative is of course a very remarkable international initiative, some people still think that it is still lacking precisely that adaptation layer so that the strategies make sense in each city. So it is a real challenge for the resilience strategy to come up with a toolkit that is really adaptable to the realities of each place.'*

The quote shows that conceptual development is still highly needed in the field of resilience, in which networks can allow a faster learning experience due to knowledge transfers and combining the existing and new knowledge. The knowledge transfer model is a tool for supporting this process.

To summarize, several cultural, economic and geographical factors shaped the various phases of the city-to-city learning process between Rotterdam and Mexico City. To provide an overview of the dynamics discussed in this section, Figure 5 below present our empirical findings in terms of the four-phased analytical framework introduced in Section 2.

**Figure 5.** Representation of the process of city-to-city knowledge transfer between Rotterdam and Mexico City related to water squares.

#### **4. Discussion**

#### *4.1. City-To-City Learning for Urban Resilience*

The existing literature on learning for resilience is rather prescriptive and primarily focused on in-city learning. In contrast, our analysis is empirically descriptive and focused on city-to-city learning. This connotes a particular conceptualization of learning related to notions of organizational learning [82] and social learning [83]. Organizational learning can start in anticipation of ('front loop' learning) or reaction to ('back loop') disturbances. It can range from practical to more fundamental: from changing skills, routines and practices ('single loop' learning) to questioning and adjusting assumptions and policies ('double loop') to fundamental debate about values and norms ('triple loop') (e.g., [5,57,82]). Judging by the case study, city-to-city learning might help cities move from single loop learning, focused on fixing the errors in routines based on status quo thinking, to double loop learning. This holds particularly true in international urban resilience networks, and could be stimulated by actively exploring and comparing good (and less successful) practices in cities worldwide. Different cities will have a different take on how to adapt their water systems, providing novel ideas to other cities. In our case study, the idea of storing water in public space and using excess water and urban adaptation options to actively increase public spatial quality was a novel idea that provided an impetus for Mexico City to change their current approach. It may relate to triple loop learning as well, since it provides a different take on the value of public space. However, no explicit debate on underlying values was observed. In general, there may be a risk that policies and practices are taken up by a city based on novelty value, rather than an actual reflection on and matching with public values and norms.

More practically, the literature on learning for resilience discusses tools such as monitoring and experimentation, and the importance of broad participation of stakeholders in the learning process, the need for long-term collaborative learning that is able to withstand short-term politics, and the need for inclusion of multiple types of knowledge (e.g., in addition to formal scientific knowledge, also practical, tacit and experiential knowledge) and co-production of knowledge (e.g., [5,57,84–87]). The influence of power, politics, and institutional dynamics is also mentioned as a problematic aspect in social learning. In the Mexico City case, we could observe that many of the factors mentioned above were problematic. The stakeholders and even individuals involved in the learning process were a select group of mostly representatives of authorities and scientists, with only one Dutch private partner involved. There was no broader participation of citizens, NGOs and companies. The group involved was highly sensitive to the local institutional, political and power dynamics: the learning parties were those that were formally responsible, and the process was complicated by a shift in formal leadership in Mexico City, which risked removing key people—and with them the institutional memory—from office. Nonetheless,

efforts were observed to anchor the knowledge through local leaders that were less vulnerable to such shifts in power and could help bridge the knowledge gained from the formal parties to other local actors. We would suggest that collaborative co-production, with attention for local power, political, and institutional dynamics, is a particularly sensitive and critical point in city-to-city learning. Formal city-to-city learning events such as local workshops and delegation visits easily result in a potentially vulnerable and technocratic form of learning rather than broader and well-anchored and internalized social learning. This calls for explicit attention in designing the city-to-city learning process.

In exploring the dynamics of city-to-learning we identified four phases. The first phase was described as the exploration and marketing of knowledge. It was initially expected that the main focus would be on the Rotterdam marketing existing knowledge of resilient systems by touting its pioneering water square and of Mexico City exploring by looking for best practice examples of innovative resilient systems around the world before stumbling on Rotterdam's example. In this conceptualization both transferring parties would be looking outward ('know the other'). However, this phase was much more introspective and to a large extent the actors tried to understand the strengths and weaknesses of their own city's systems during outreach ('know thyself'). This is in line with some of the work on Sustainability Transitions, which investigates how actors can break the lock-in of incumbent systems of water provision [88]. In particular with the approach of Transition Management, which stresses that a small group of agents who want to change the status quo needs to start with the introspective and mental work of developing shared conceptualizations and visions before linking up and build coalitions with other actors [89]. For instance, Ferguson et al. [29] use the Transition Management approach to develop a diagnostic framework to facilitate transformative change in urban water systems, which stresses the need for these kinds of introspective efforts by the actors in the early stages. For the second and third phase, we found that the differences in geographical and institutional environment in Mexico City turned a seemingly clear-cut one-directional transfer much more challenging endeavor. Successful transfer, then, is a two-way street not only in the in the sense that the transferring party also takes relevant knowledge back, but also in the sense that it requires alignment of the knowledge base between the two parties. This took place in the second phase, where we observed a conscious framing and reframing of the history, socio-political-cultural contexts, and challenges in each city. Participants attempted to rephrase their own positions in terms of that of the other, to match each other's situation in many aspects. This serves to better align the two cities. It should provide sufficient 'sameness' (cf. [90]) to establish mutual understanding and build trust—to argue that both cities' experiences and goals are sufficiently aligned to be a meaningful partner. Yet, there needs to be sufficient 'difference' as well [90], in order to highlight specific aspects in which the cities can learn something from the other. Ideally, this process also takes into account different framing of what urban resilience means, because different cities and different actors within a city can have very different approaches to resilience, leading to very different implementations into policies and practical interventions [91]. In the third phase, such further translation into practical policy projects (e.g., new policy designs or white papers, or pilot projects and interventions) is made. During this practical experimentation and implementation, learning can take multiple forms: technical, formal (hard data, e.g., documents, projects, official authorities) and informal (soft data, e.g., implicit/tacit knowledge, professional and social networks) (cf. [44,50]). In the context of a single tangible project, much broader policy-relevant knowledge can be said to 'hitch a ride'. For instance, Dutch ideas about interdisciplinary approaches and a less rigid and hierarchical institutional organization of the water sector also reached Mexican ears. The Dutch in turn might have appreciated the technical expertise available in the Mexican government organizations (which is quite often outsourced in Dutch governments). In another way broader policy-relevant can be said to be encapsulated within the tangible artefact of a single project. For instance, a water square doubles as a buffer in a large technical system and as a way to create vibrant public space, which makes the actors involved rethink the single-purpose and efficiency logic of urban infrastructure. Similarly, the local leaders recruited in the process serve both as a source of redundancy or omnivory for the knowledge learned, and as a source of polycentricity that facilitates alignment of the values and goals

for resilience-building between authorities and communities and citizens. This can in turn be used to stimulate a broader social learning process (cf. [92]). As for the fourth phase, at the time of our analysis the process has not yet progressed to this stage. As we showed, actors can reflect on the process and internalize part of the lessons during earlier phases. It would be valuable to conduct an ex-post analysis once there is 'closure' around the project and its associated activities. Particularly, it would be interesting to analyze if the water square project has engendered changes broader institutions and organizational routines. It would also be interesting to see whether the actors involved can be enrolled as advocates for the further spread of water squares or if they have to some extent become disillusioned with this idea. Similarly, many other aspects might have strengthened, including the ability to work across disciplinary boundaries, the strength of knowledge networks, and the ability to better align and anchor resilience-building with the interests of local communities.

#### *4.2. Avenues for Further Research*

Our analysis city-to-city learning zoomed in on the transfer dynamics in case of one tangible project. The pioneering water square in Rotterdam, which was to be transplanted to Mexico City, is best conceived of as a single 'socio-technical experiment'—defined as a "practice-based and challenge-led initiative, which is designed to promote system innovation through social learning" [93]. The literature on socio-technical experiments suggests that individual projects in and of themselves are not enough to change the underlying principles of the broader water system, but that they are a key unit of analysis and that their wider learning and demonstration effects do provide a crucial starting point for such a transition [93]. Furthermore, for both the actors involved in the transfer and for us as academics, in conceptualizing this process the advantage of providing center stage to individual experiments such as water squares is that it provides a very tangible precursor and a of clear sense of direction. Another way to get to the essence of city-to-city learning would be to look at larger collections of projects and at broader processes of translation of policy-relevant knowledge. Whilst we believe that our approach of describing one concrete project as a vessel for policy translation as a way to study has certain important merits, we certainty recognize that this would be a worthy agenda for future research.

One consideration for expanding the research on city-to-city learning for urban resilience is the potential impact of situational factors and the progression of projects over time. Policy transfer risks resulting in policy failure, if the transfer does not take into account these situational differences (e.g., [94]). Geographic factors can influence the success or shape that projects take in a city. Rotterdam and Mexico City experience similar amounts of yearly precipitation and both developed water squares in low elevation locations that function as a 'bathtub'. However, Rotterdam's precipitation is relatively spread out over the years compared to Mexico City, and Mexico City has hills that drain into the valley as well. Drought and water scarcity are also much more an issue in Mexico City. Consequently, the shape that water squares take will likely differ. Rotterdam's are designed to drain on the main water system within days, while Mexico's squares might focus on retaining water for a longer period. Cultural factors can also play an important role. As discussed above, Mexico City has a culture of strong leadership, disciplinary expertise and concentration of expertise into a single management organization, whereas Rotterdam is more focused on collaboration, interdisciplinarity, and an ecosystem of multiple organizations with different expertise. Cultural factors are important in learning processes and transfer of policy proposals, but often difficult to pinpoint within and between cases. As the case showed, however, successful learning can take place when such factors are taken into account. Political factors are also highly important. Research in the Dutch context showed that the political interest of local aldermen and councilors is of key importance for the success of local adaptation [95], and the Mexican case in this paper showed similar processes, where leadership changes could strongly impact specific projects when the previous leader strongly identified with a project. Socio-economic factors are also important. Actors might push for very different locations and implementations of water squares, to strongly diverging effects. One risk of any pilot project, for example, might be that it becomes a prestige project that might be implemented in high visibility locations, such as tourist centers,

business centers or richer neighborhoods, whereas they might be most needed in high density poorer neighborhoods. Similarly, governments might opt for high-tech implementations that involve the city's strategic business partners, potentially neglecting higher maintenance costs or the wishes of the neighborhoods in which projects are implemented. Rotterdam's first attempt at constructing a water square failed due to similar reasons: connections with that neighborhood were poor and the square was not accepted by the population. In next iterations, co-design with residents was a key point, leading to successful projects. In the case of Mexico City, the efforts to establish water squares has however specifically and intentionally focused on a poorer district, in the realization that such squares can also improve the spatial and social quality of the district [64,65]. In Rotterdam, the squares are also located in poorer areas. Such situational factors could impact the successful implementation of water squares, but could also determine their long term survival. For example, high-tech options might involve pumps and filtration systems that could be vulnerable to failures, neglect of maintenance (e.g., during periods of austerity), or debris and waste in runoff that enters the square. More low-tech and community-involved projects on the other hand could run the risk of deterioration when attention and community interest fade. Little is currently known of such long term survival factors for water squares and similar urban pilots and experiments, and this should be further investigated.

Another interesting point for future research would be to see if the four phases described here can also be identified in other cases studies of city-to-city learning or urban policy translation and whether describing these findings accordingly leads to productive insights. Attempts to do this could focus on other cases studies in water resilient cities or other related efforts. For instance, it would be interesting to compare actor coalitions and strategies of city-to-city learning in urban water projects with a 'resilient city' focus with those that are framed as contributing achieving 'smart city'. It would also be interesting to compare policy translation learning in the context of resilient water systems with resilient waste management or transportation systems. At least conceptually, there is nothing inherent in our four-phase framework that would limit the application to a focus on resilience or to the water domain.

#### **5. Conclusions**

We presented a policy practice analysis of the city-to-city exchange of policy relevant knowledge for building urban water resilience between two cities that are currently highly active on this topic: Mexico City and Rotterdam. Mexico City faces numerous water-related challenges and expects increasing pressure due to climate change [96]. Rotterdam is a city with well-known for its water expertise and experience. It is an often-used example of Dutch delta planning approaches, taken as 'good practice' in water management and climate adaptation worldwide [90,97]. Learning is seen as a key principle in resilience-building and cities worldwide are actively pursuing knowledge exchange through experimentation and collaboration with consultants, architects, and academics in bilateral or transnational networks. Many cities face urgent and worsening challenges today while public budgets are under pressure as they navigate through these slow and costly learning processes. Moreover, strong differences exist in the institutional, geographic, and cultural context of cities, which might limit the possibilities for knowledge and policy transfer. The analysis in this paper provided several lessons on barriers and stimuli for successful city-to-city learning in the context of water systems and urban resilience. To present these findings we mobilized a framework composed of four distinct phases.

During the first phase of the city-to-city knowledge transfer (exploration and marketing), a key success factor was that both cities did not start with an external search for sources or potential recipients of knowledge. Rather, they started with introspection. It was important to have clear picture of one's own systemic strengths and weaknesses, for both cities involved. This introspection helped the search for complementary partner cities, and it may also provide sufficient self-awareness and material for the reframing needed in phase 2.

In the second phase (building pipelines), it was important to create a mutual understanding of struggles and histories, to overcome geographical and cultural differences. Creating a shared vision helped reduce misunderstandings in the process. It appeared particularly critical to heavily negotiate and reiteratively reframe one's own and each other's reality. The perceived situation of both cities needed to be both similar (to build trust and develop a shared vision) and different (to have something to offer to the other; to provide something that could be learned). This process also served to establish the cities as equal partners, rather than a mentor–mentee relationship which can be a detrimental framing for leaning [98]. While negotiation and reframing are normal in multi-actor settings, the tension between matching and differing in framing and situation is very high in city-to-city learning and warrants further research.

For the third phase (translation and adoption) we observed that differences in the institutional environment (e.g., modes of governance, level of interdisciplinarity) could hamper the implementation of 'imported' ideas. These barriers were reduced by recruiting strong local leaders, external to the official organizations, into the process. This required organizational flexibility of the official organizations. However, they pursued this course because it would also reduce the risk of losing institutional memory, which was high during leadership changes in the centralized, hierarchical organizations involved. Anchoring through local leaders provided some redundancy or omnivory in the knowledge system. These issues relate to resilience principles not necessarily associated with learning: in this (and the next) phase, the new knowledge and the local knowledge system themselves are made resilient.

In the fourth phase, which follows the actual exchange, cities will need to internalize the lessons and resilience thinking in general (internalization and reflection) so that this can be applied in other projects. This applies to lessons regarding resilience-building, but also to the learning process itself. Particularly pronounced lessons for city-to-city learning seem to be that the starting phases and their introspection and reframing processes are critical, and that in the later phases it is important to make the knowledge itself resilient by building internal and external networks. By internalizing such lessons, cities might strengthen not only their own resilience, but also enhance future exchanges with other cities.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2073-4441/11/5/983/s1, Document S1: List of interviews, Document S2: Coding scheme, Document S3: Temperature and precipitation profiles for Rotterdam and Mexico City.

**Author Contributions:** Conceptualization, S.I., F.S. and A.W.; Data curation, S.I.; Formal analysis, S.I., F.S., and A.W.; Investigation, S.I.; Methodology, S.I., F.S. and A.W.; Supervision, F.S. and A.W.; Validation, S.I.; Writing—original draft, S.I., F.S. and A.W.; Writing—review and editing, F.S. and A.W.

**Funding:** This research received no external funding.

**Acknowledgments:** We thank the Netherlands Embassy to Mexico for hosting one of the researchers for the duration of the study and we thank various practitioners throughout Mexico and the Netherlands for their time and effort to share their insights on the dynamics of knowledge exchange on urban resilience and water squares.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Pluvial Flooding in Utrecht: On Its Way to a Flood-Proof City**

#### **Romy C. Brockho**ff **1, Steven H. A. Koop 1,2,\* and Karin A. W. Snel <sup>1</sup>**


Received: 24 May 2019; Accepted: 17 July 2019; Published: 19 July 2019

**Abstract:** Downpours are increasing in frequency and severity due to climate change. Cities are particularly susceptible to flooding from downpours because of their large share of impervious surfaces. Minimising pluvial flood risk requires all involved stakeholders to collaborate and overcome various barriers. Although an increase in citizen engagement in climate adaptation is generally preferred, experiences with inclusive decision-making are often limited. The aim of this paper is to obtain a deeper understanding of how the capacity to govern pluvial flood risk can be developed through citizen engagement. We scrutinised the capacity of local actors to govern pluvial flood risk in the city of Utrecht, the Netherlands. For the analysis of Utrecht's problem-solving capacity, the Governance Capacity Framework provided a consistent assessment of the key governance components. The results indicate that Utrecht's capacity to govern pluvial flooding is relatively well-developed. Collaboration between public authorities is advanced, sufficient financial resources are available, and smart monitoring enables high levels of evaluation and learning. However, citizen awareness and engagement in policy making is rather low. Accordingly, citizens' willingness to pay for flood adaptation is limited. Stimulating flood risk awareness by combining financial incentives with more advanced arrangements for active citizen engagement is key for Utrecht and other cities.

**Keywords:** citizen engagement; flood risk governance; governance capacity; climate adaptation

#### **1. Introduction**

Extreme weather events, such as heavy rainfall, are likely to increase in frequency and intensity as a consequence of climate change [1]. In the past few decades, physical, societal, and economic damages of natural disasters have increased considerably [2]. In particular, floods are expected to substantially threaten the quality of urban life in the near future [3,4], demanding sound flood risk management. Urban areas are particularly vulnerable to downpours due to their impermeable surfaces—such as roads, parking lots, and roof tops—that prevent rainwater from infiltrating and, as a consequence, generate increased surface-runoff and thus increase the pluvial flood risk of urban areas [5]. Pluvial urban flooding may lead to large-scale economic damage and traffic congestion. It may also induce irregularities in the provision of electricity [6–8]. In 2011, for instance, the city of Copenhagen (Denmark) was hit by a downpour of 150 millimetres in less than three hours. The concomitant damage was estimated at around one billion US dollars [9]. Therefore, making cities more flood-resilient is an urgent challenge for sustainable urban living.

Urban expansion and insufficient water storage capacity regularly lead to rainfall runoff peaks that exceed the water system's drainage capacity, resulting in pluvial flooding [5,10,11]. This is a pressing issue in many Western-European cities because the water infrastructure in these places is becoming increasingly obsolete and requires costly refurbishments [12,13]. These drainage systems are generally not designed for the climate-change-induced increase in frequency and intensity of storm events. Moreover, these systems are typically a combined drainage of storm water and sewerage (i.e. Combined Sewer System; CSS). This type of drainage system is more vulnerable to surface water flooding [4,14]. Thus, growing precipitation extremes together with a large percentage of impermeable urban surfaces and an increasingly obsolete drainage system, call for more advanced urban flood adaptation.

In many countries in Europe, governmental institutions have been solely responsible for flood risk management [15–17]. Their main objective is to ensure that floods do not affect economic growth, national security, or welfare standards [18]. However, the intensity and frequency of storm events is changing and affecting all types of land use. Accordingly, the division of responsibilities related to flood risk management is changing. A decentralising trend in flood risk management has been recognised [19], which results in a greater role for non-governmental actors [20]. These transformations are related to a more general trend, namely the shift from government to governance. This implies a relocation of power and authority both among governmental organisations, such as delegating certain tasks from the national government to local authorities, as well as from governmental organisations to private actors [21]. This trend towards governance is widely adopted in, for instance, the European Union's (EU) Flood Directive, the EU Water Framework Directive, and the Aarhus Convention [22]. These policies mandate the engagement of non-governmental actors in flood risk management [23]. The involvement of non-governmental stakeholders, such as citizens, project developers, housing corporations, and businesses, in local flood risk management is crucial in fostering climate adaptation in cities [21,24]. Citizen engagement is increasingly important for adapting to climate-related risks, including pluvial flooding [25–27]. However, the specific responsibilities borne by public and private actors in climate adaptation and flood risk management are often unclear [8].

Even though citizen engagement in flood risk management is encouraged and acknowledged by global organisations (e.g., Intergovernmental Panel on Climate Chance [1] and Organisation for Economic Co-operation and Development (OECD) [22]), it remains a challenge to effectively engage citizens in climate adaptation projects and decision-making of local governments [25]. To start, municipalities appear to have limited experience with citizen engagement in climate adaptation [28]. Wamsler [29] analysed city-citizen collaboration for climate change adaptation in eight German municipalities and concluded that this cooperation is 'practically non-existent' as individuals are insufficiently aided by city authorities and urban policy does not support collaboration. Accordingly, Brink and Wamsler [30] observed that Swedish municipalities rarely involve citizens in local flood or climate change adaptation. Moreover, a cross-country comparison between the United Kingdom, Italy, and the Netherlands shows that overall citizen engagement is limited when examining the respective types of interactions between citizens and authorities and the impact of citizen engagement on decision-making [23]. In the Netherlands, citizens are held responsible by the law for managing rainwater on their own property. In practice however, it has been found that Dutch residents often rely on local governments [31,32]. The downside of this national commitment to flooding is that citizens' initiatives in the implementation phase are considered as a 'backup strategy' in addition to collective flood risk measures [17]. Another consequence is that citizens lack awareness of their responsibility regarding rainwater on their own property [6]. The lack of clarity in duties, good examples, and experiences with this more inclusive form of decision-making and implementation may explain the slow progress in citizen engagement in climate adaptation that has been observed [8,33]. For example, citizens' motivation to participate does not only depend on their risk perception but also on their sense of self-efficacy and the financial means made available to genuinely let citizens influence the end-result of decision-making processes [34]. Thus, active citizen engagement in urban flood adaptation seems to be challenging in practice, although it is often claimed to be essential for implementing climate adaptation measures.

The overall capacity of stakeholders to collaborate and address water-related challenges together, such as pluvial flooding, may be much more decisive than the capacity of individual organisations and stakeholders [35–37]. From this more holistic perspective, it becomes essential to scrutinise how citizens can contribute in formulating and implementing policies and objectives related to pluvial flooding.

In our study we apply the governance capacity analysis. This methodology is based on Likert scale descriptions of indicators that together are argued to form the capacity to govern water challenges. Although this method is well-embedded in existing literature on adaptive management, co-management, and water governance [32], it is important to note there is a plethora of frameworks developed to assess the key conditions that together constitute governance capacity (e.g., [22,38,39]). The Governance Capacity Framework (GCF) is selected because it is one of the most standardised approaches in terms of definitions, operationalisation, research approach, and geographical scope, which enables high levels of scientific reproducibility and falsifiability of the empirical results. An overview of the operational indicators and key references to literature can be found here [40]. A second reason for selecting the GCF relates to its (graphical) design, which aims to be intelligible for a variety of non-experts such as policymakers, operators, and citizens. Thirdly, although other governance assessment frameworks have been developed for institutions at various scales such as social-ecological systems, bioregions, countries, river basins, or organisations (e.g., [22,39,41,42]), there are not that many frameworks focussing on the urban context, particularly with respect to governance capacity. However, cities are important and well-established institutional entities where integration of water with different sectors, objectives, and interests is perhaps most prevalent and concrete. In fact, the direct interaction between citizens, governments, and smaller and larger private stakeholders may be most widespread in cities [43,44].

The GCF consists of nine key conditions for good governance such as awareness, useful knowledge, continuous learning, stakeholder engagement, and implementing capacity. This paper specifically addresses how citizen engagement can effectively contribute to each condition, and thereby improve the overall capacity to govern pluvial flood risk. In this way, both the engagement of citizens in decision-making processes, as well as the implementation of (individual) adaptation measures, are scrutinised in the case study of Utrecht. Accordingly, the aim of this paper is to obtain a deeper understanding of how the capacity to govern pluvial flood risk can be developed and further improved through citizen engagement. We first analyse Utrecht's capacity to govern pluvial floods and second, we scrutinise the role of citizen engagement in strengthening the governance capacity. In this paper, we use citizen engagement as a conceptual umbrella that captures both the participation of citizens within the local decision-making process and an active involvement in the implementation phase by taking climate adaptive measures. Additionally, 'Utrecht' will be used to refer to the local network of stakeholders (including local authorities and citizens), i.e., 'governance structure', within the administrative municipal area of Utrecht, the Netherlands.

Section 2 provides the conceptual framework, research methodology, and case study description. Section 3 presents the results of the governance capacity assessment of Utrecht and specifically addresses the role of citizen engagement. Finally, Sections 4 and 5 cover the discussion and conclusion, respectively.

#### **2. Conceptual Framework**

#### *2.1. Governance Capacity Framework*

To assess the capacity of Utrecht to govern pluvial flood risk, we apply the Governance Capacity Framework (GCF), developed by Koop, et al. [45]. The framework consists of three dimensions and nine conditions and is supported by 27 indicators (Table 1). The dimension *knowing* refers to the need to be aware (e.g., [46]), understand (e.g., [47]), and learn (e.g., [48]) about the risks and impacts of environmental challenges and policy. *Wanting* alludes to the willingness and motivation of various actors to cooperate (e.g., [34]), act upon ambitions (e.g., [49]), and devote oneself to finding solutions (e.g., [50]). *Enabling* refers to the network's ability to collaborate (e.g., [51]), coordinate, and implement action plans through various policy instruments and available resources (e.g., [38,52]). The selected nine conditions are based on a literature review in the field of water governance, environmental governance, and adaptive management [45], and are well aligned with the much-accredited principles for water governance proposed by the Organisation for Economic Co-operation and Development [22]. For detailed references to key literature for each of the framework's 27 indicators, we refer to [40]. The GCF provides a diagnosis of urban water challenges. These challenges generally require different organisations to collaborate and align their activities. The framework's indicators are consistently scored according to an indicator-specific Likert scale ranging from very limiting (− −) to very encouraging (+ +) to the governance capacity. The GCF has been applied to assess 41 water-related challenges in 15 cities across the globe [32,45,53–58].


**Table 1.** Overview of the Governance Capacity Framework (GCF) adopted from Koop et al. [45].

#### *2.2. Method*

The 27 indicators are scored according to three consecutive steps:


accomplishment of that activity [59]. The importance/influence matrix consists of four classes: (1) *Subjects* (high importance, low influence), (2) *Key players* (high importance, high influence), (3) *Crowd* (low importance, low influence) and (4) *Context setters* (low importance, high influence). For an in-depth understanding of the local urban context, this study focussed on key players and subjects for the interview selection. The nine interviews lasted approximately one hour and were recorded after permission was given. This ensured accuracy and enabled easy comparisons of specific indicators.

3. *Score determination:* Finally, the preliminary score of the policy review and the results of interviews were compared and led to a final score per indicator.

A coding system is applied in this paper to refer to interviewees while guaranteeing their anonymity. Accordingly, codes such as [SR01], [SR02], [SR03] refer to the individual interviews respectively. The interviewees included stakeholders that participate in collaborative regional networks and can be classified in the groups 'key players' and 'subjects'. As *key players*, we selected two policy advisors on urban water and public green spaces (Municipality of Utrecht), a spatial adaptation expert (Province of Utrecht) and representatives of the regional water authority (HDSR; in Dutch: Hoogheemraadschap De Stichtse Rijnlanden). For flood risk management in the city of Utrecht, the regional partnerships Winnet (in Dutch: Water Innovatie Netwerk), Coalition Spatial Adaptation (CSA; in Dutch: Coalitie Ruimtelijke Adaptatie), and Nature and Environment Federation Utrecht (NEFU; in Dutch: Natuur en Milieu Federatie Utrecht) form the *subjects.* Winnet is a regional cooperation in Utrecht, consisting of 14 municipalities and the regional water authority HDSR, and aims at a sustainable and efficient waste water cycle. Similarly, CSA is a regional platform facilitated by the engineering consultancy Sweco that addresses drought, heat stress, and flooding by joining forces with the Province of Utrecht, six municipalities, HDSR, and Safety Region Utrecht (in Dutch: Veiligheidsregio Utrecht). Finally, NEFU unites various stakeholders (e.g., citizens, local authorities, businesses, housing corporations) to achieve a sustainable province and to tackle climate adaptation, including pluvial flooding.

#### *2.3. Case Study: Utrecht (the Netherlands)*

In July 2014, the city of Utrecht was hit by the heaviest rainfall ever recorded, with measurements ranging from 75 to over 100 millimetres in 24 hours [60]. Utrecht has limited capacity to store such downpours as only 21.8% of the city centre is green (vegetation) or blue (water) [4]. Besides, the city is characterised by an ageing sewer system, and has only 384 kilometres of stormwater sewers and 630 kilometres of combined sewers (both rainwater and sanitary water) [61]. The combined sewer system is common in many Dutch cities and as risks of pluvial flooding increase [6], the exposure to Combined Sewer Overflows (CSOs) increases as well. This may result in urban surface water pollution that may negatively affect both the environment and human health [14,57].

The municipality of Utrecht has approximately 352,941 inhabitants (1 January 2019) and prognoses are that it will reach over 400,000 citizens by the year 2025 [62]. When comparing the four largest Dutch municipalities, Utrecht grew most rapidly from 2010–2018 (13.16%) and it is expected to continue growing at this rate [63]. The involvement of citizens in addressing pluvial flooding is important, as they own, together with businesses, approximately 60% of Utrecht's land surface [64]. Citizens can thus make a large contribution, for instance by installing green roofs on their property or removing pavements in their gardens, to respectively store and infiltrate rainwater. This clearly shows the potential of citizens' actions to help address downpours. In 2018, 3716 square metres of green roofs were installed in Utrecht, financially stimulated through a municipal subsidy programme [65]. Although this coverage has more than tripled compared to 2017, the installation of green roofs in Utrecht by citizens is still in its infancy.

Urbanisation, in combination with extreme rainfall and the aforementioned limitations regarding sewerage and water storage capacities, calls for more understanding of how to adequately govern these challenges. Knowledge will help local policymakers and other stakeholders to implement climate adaptive policies. As many other Western-European cities face the challenge of pluvial flooding and share the same characteristics as Utrecht (e.g., ageing water infrastructures, urbanisation, and sealed urban surfaces), our lessons may also benefit other cities.

#### **3. Results**

Figure 1 shows the capacity profile that indicates how well stakeholders work together to govern pluvial flood risk in Utrecht. Overall, the governance capacity is well-developed (Figure 1). However, note that all neutral (0) or encouraging (+) scores can still improve substantially. Section 3.1 provides the key results of the governance capacity analyses, which is structured according to the framework's three dimensions knowing, wanting, and enabling, and ends with a synthesis of these results. Section 3.2 focusses on the role of citizen engagement, which turned out to be a priority for future efforts to mitigate pluvial flood risk in Utrecht.

**Figure 1.** Results of the Governance Capacity of Utrecht to address pluvial flood risk. The indicators are arranged clockwise from very limiting (− −) to very encouraging (++). Key message is "the bluer, the better".

#### *3.1. Utrecht's Capacity to Govern Pluvial Flood Risk*

#### 3.1.1. Dimension 1: Knowing

The city of Utrecht performed a mandatory 'climatic stress test' in 2018. This test [66] contributed to identifying locations that are vulnerable to floods, heat stress, and water scarcity issues [SH02, SH09]. Moreover, sewer systems are adequately monitored, and precipitation prediction models are upgraded by a collaboration of the municipality of Utrecht, cooperation Winnet, and the regional water authority (indicator 3.1 [SH02, SH03, SH07]). Utrecht's current strategy is, however, not aimed at sewer pipe dimensioning to store excess water in case of a heavy rain event. Sewer pipes will only be enlarged when standard precipitation norms are exceeded [SH05]. This emphasises the need for alternative solutions. In addition, cross-stakeholder learning (indicator 3.3) is well-embedded in Utrecht, for instance in the form of knowledge sharing between many networks and cooperations [SH06]. Knowledge sharing with a broader audience than specialist networks is somewhat limited, specifically the citizens of Utrecht are largely overlooked.

Despite awareness campaigns such as 'Waterproof030 and 'Water-friendly Garden', a widespread sense of urgency about pluvial flood risk (indicator 1.2) has not been established yet. However, a sense of urgency does exist in flood-prone neighbourhoods: Lombok and Zeeheldenbuurt [SH01]. It seems that a more profound sense of urgency requires a downpour, as SH07 describes: "What we actually need, is another cloudburst as a kind of wake-up call to raise the urgency of the water issue." Citizens of Utrecht seem to be informed about the impacts and probabilities of pluvial floods (indicator 1.1). In addition, some communities are starting to engage in flood alleviation initiatives such as placing rain barrels in the street (indicator 1.3 [67]). However, in general, people do not feel an urgency to change their behaviour by taking precautionary measures (indicator 1.3). In fact, most people do not act because they perceive such adaptation measures as a primary responsibility of local authorities (i.e. the regional water authority and municipality) [SH01, SH02, SH06]. These results are in line with the OECD study [31] that observes a water awareness gap amongst Dutch citizens who take water services for granted. Contrary to this awareness gap, the availability of transparent and intelligible information about pluvial flood risk is well-organised through various channels such as websites, newspapers, television, or policy documents (indicator 2.1 and 2.2 [SH01, SH03, SH05, SH07]). For example, the municipality published an online manual for citizens on how to make dwellings and gardens waterproof [68]. In short, citizens in Utrecht know about the risk of pluvial flooding, yet do not consider this issue as a priority and do not seek for information until they experience 'wet feet' themselves.

#### 3.1.2. Dimension 2: Wanting

Stakeholder engagement (condition 4) is important for joint problem framing, gaining access to resources, and creating support for successful implementation of measures and policies. Although stakeholder engagement is an integrated part of governing pluvial flood risk related issues in Utrecht, its current application is rather limited. In fact, for pluvial flooding specifically, stakeholder engagement is hardly considered [SH07]. More generally, stakeholder engagement in Utrecht consists merely of consultation sessions where people can ask for amendments to proposed policy plans. In a number of cases, these consultations occur at the end-stage of the decision-making process (indicator 4.1-[SH07]), resulting in a low influence of stakeholders on the end-result and arguably low stakeholder engagement in the implementation phase [34]. In addition, only public parties and one consultancy company are represented in the main regional partnerships CSA and Winnet, whereas citizens and housing corporations have yet to be included.

Utrecht's sustainability ambition (condition 5) is found to be well-embedded and goals for water policy and green policy on the municipal level are more or less aligned and are thus enhancing cohesion (indicator 5.3 [SH07, SH08]). Besides, Utrecht has adopted the seven ambitions of the national Delta Programme [69], which aim at making the Netherlands water-resilient and climate-proof. However, the pathways to reach this goal are yet to be formulated by local authorities [SH09]. The role of local citizens who promote initiatives, bring actors together, and mobilise required local resources, can be improved (condition 6). In Utrecht, such agents of change are rather limited to small-scale neighbourhood initiatives as individual initiatives to install rain barrels [67]. Though limited in scale, these types of initiatives may spur neighbours to do the same [SH05, SH07]. As SH05 argues: "It is crucial to have examples in practice. If your neighbours take measures, this may encourage other residents to take action as well."

At the municipal level, the city mayor for instance, can be considered a visionary agent of change regarding sustainability initiatives, but he does not (yet) perceive pluvial flooding as a priority. By contrast, municipal representatives of the neighbouring smaller city of Houten are more engaged with pluvial flood risk adaptation [SH07]. Utrecht cannot fully rely on local agents of change but could facilitate more initiatives when the municipality recognises pluvial flooding as a priority.

#### 3.1.3. Dimension 3: Enabling

The results show that stakeholders who participate in collaborative regional networks (e.g., CSA and Winnet) have sufficient room to manoeuvre and find solutions to pluvial flood risks (indicator 7.1). However, these cooperations and local authorities are not the only stakeholders who bear responsibility,

as multiple interviewees acknowledge that citizens have to make an effort as well [SH06, SH07, SH09]. To enable actors to implement their ambitions and ideas concerning flood resilience, sufficient financial resources are crucial. For citizens in Utrecht, taking climate adaptation measures to cope with pluvial flooding is financially supported by the regional water authority and the municipality through multiple subsidy schemes [SH02, SH04, SH06, SH08]. This financial support enhances the affordability of various adaptation measures (indicator 8.1), such as the replacement of pavements by greenery in private gardens. According to SH04, there is, in general, a willingness to pay among citizens for taxes levied by the regional water authority. However, the willingness to invest in pluvial flooding solutions is found to be moderate among citizens in the flood-prone neighbourhood Lombok (indicator 8.2). The municipality realised a separate drainage of rainwater in this low-lying part of Utrecht and connected 68 semi-based dwellings to this system [SH07]. As these houses are private entities, homeowners bear responsibility as well. However, not every household was willing to invest, as SH07 explains: "About half of the 68 homeowners in Lombok signed an agreement with the municipality to contribute in implementing pluvial flooding measures on their property."

In fact, these limitations in the willingness to pay are a recurring pattern for Dutch municipalities. For example, a survey conducted by the Dutch Broadcast Foundation among 1700 Dutch citizens that experienced serious pluvial flooding issues, showed that the community would like to see the municipality invest more in the sewer system, while only 25% of them are willing to pay more municipal sewer taxes [70]. A study on Dutch water governance recommends strengthening the finance system by implementing polluter-pays principles, such as abstraction charges [31]. Following this report, a special commission appointed by all Dutch water authorities investigated optimising the regional water authority's tax system [SH04]. At present, rainwater accounts for approximately a third of the water treatment costs [71]. To minimise this share, the commission suggests increasing incentives to decouple rainwater pipes from the drain to relieve the sewer system and reduce treatment costs [71]. The commission's proposal is hitherto not implemented in Utrecht or elsewhere in the Netherlands.

Nonetheless, monetary aid or financial incentives are no guarantee for successful adaptation by citizens. For instance, the municipal subsidy for green roofs has had minimal effect because many people do not yet fully grasp the added value of having a vegetated roof [SH07, SH08]. To date, stimulating rather than implementing sustainable behaviour through binding guidelines has been preferred by local authorities [SH07]. Citizens are financially supported through various subsidy schemes to take climate adaptive measures (e.g., removing pavements, installing green roofs, or building climate-proof playgrounds), yet do not take advantage of this. This may be explained by the low sense of urgency and limited awareness that has been observed.

#### 3.1.4. Synthesis

Overall, Utrecht can considerably improve its capacity to govern pluvial flood risk. In particular, the following indicators and conditions showed the most room for improvement and therefore should form the core focus for future action. First, there is a relatively low willingness to pay (indicator 8.2) for climate adaptive solutions such as infrastructure augmentations (i.e. separate rainfall runoff from the sewer systems). Accordingly, local communities and the private sector show limited efforts to understand, react, and anticipate risks of pluvial flooding such as applying green roofs (indicator 1.3). Limitations in awareness among citizens and private stakeholders (condition 1) and a suboptimal use of policy instruments (indicator 9.1) both require additional effort to better address the increasing downpours that Utrecht is projected to have. Governmental bodies, such as the municipality and the regional water authority, are aware and are actively initiating action through multi-level collaborative networks (condition 7). However, with respect to private actors and citizen engagement, considerable progress is required to effectively address pluvial flood risk (condition 4). To achieve this, Utrecht may need to formulate an action plan in close collaboration with its citizens and local enterprises (indicator 9.3). In this way, stakeholder engagement (condition 4) can be improved to better serve both the policy development and implementation phase.

#### *3.2. Citizen Engagement*

Despite serious efforts made by the municipality and regional water authority (e.g., through campaigns and provision of information and advice), the level of awareness among citizens on pluvial flooding in Utrecht is limited. In general, they lack a sense of urgency to act as they hold local authorities responsible for taking climate adaptation measures to alleviate the risk of urban floods. If they do feel accountable, citizens show reactive behaviour (i.e. taking measures after pluvial floods occurred) rather than proactive. This reactive behaviour is mainly visible among citizens who are exposed to the negative effects of extreme rainfall in their garden or inside their dwelling, as SH05 explains: "A sense of urgency among citizens does not occur until they are confronted with pluvial floods themselves. They purely react upon pluvial flooding issues."

To change this reactive behaviour into (pro)active behaviour regarding pluvial adaptation, both the municipality and regional water authority in Utrecht make an effort to support its inhabitants by providing various grant schemes. Despite this, citizens' willingness to pay still appears to be low. Taken together, the combination of information provisions (e.g., through policy documents, campaigns, manuals, guest lectures at schools) and financial aid (e.g., grant schemes) provided by local authorities does not yield the desired result, namely, citizens taking climate adaptive measures to minimise the adverse effects of pluvial flooding.

What is largely missing is an active involvement of citizens in (municipal) decision-making. Citizens are expected to be actively engaged in addressing pluvial flooding, yet they have little influence on municipal flood-related policies. At present, the municipality is only obliged to ask for consultation from the regional water authority and province [SH07].

To stimulate citizens to adapt to pluvial flooding, an important incentive is to actively engage them in the development and implementation of flood adaptation policy plans. To do so, their level of influence should go beyond being informed or consulted. The opportunity to be actively engaged and coproduce policy plans may be essential in motivating citizens to take part. Active engagement usually takes much more time than more unilateral decision-making. However, many authors argue that this is generally more than offset by time gains in the implementation phase, not the least because citizens become more aware of the relevance and their role in flood mitigation [34,72,73]. Our results indicate that the stakeholder engagement process (condition 4; Table 1) of Utrecht can be improved for flood decision-making. Stakeholders should be given the opportunity to be actively engaged so that the municipality can structurally stimulate their active engagement. More precisely, additional effort may be required to engage all relevant stakeholders in an early stage of policy coproduction processes. In these processes, it is crucial that stakeholders (e.g., citizens and local experts) develop a range of different alternatives and, when all alternatives are considered, commit themselves to a final decision. In addition, clear and realistic procedures with clear exit moments may ensure sufficient progress for stakeholders to continue their initial engagement and ensure that they feel confident that their core values are not being harmed (i.e. creating trust).

On another note, the policy instruments which are currently applied in Utrecht have a suboptimal effect. The municipal subsidy which is supposed to stimulate citizens to implement green roofs, for instance, has been adopted by citizens on a rather limited scale [SH08]. In addition, the municipal sewer levies, which are mandatory for all citizens, are currently not related to the discharge quantity of wastewater into the sewer system. This indicates that the polluter-pays principle is not implemented, and therefore, producing less wastewater is not rewarded by tax reductions. This demonstrates that Utrecht is implementing soft policies (e.g., providing information and subsidies) over hard policies (e.g., binding rules or punishment, such as charging citizens if over 70% of their garden consists of impermeable pavements). Although the latter strategy requires considerable paperwork (and thus resources), it is likely to have a substantially better result than the current package of non-binding soft policies. These stricter baseline instruments are an important contribution to spur active citizen engagement and may simultaneously contribute to improved water quality and drought alleviation.

#### **4. Discussion**

The adverse effects of extreme rainfall on urban areas demand for adequate water governance to prevent pluvial flooding. We used the GCF [45] to assess the water governance capacity of all water-related stakeholders within the city of Utrecht to govern (the effects of) pluvial flooding. Our results demonstrate that the overall capacity of Utrecht to govern pluvial flooding is relatively well-developed. However, there is also room for improvement. In particular, the engagement and behavioural change of citizens seems to be essential for addressing pluvial flood risk. This result seems to be well-aligned with other studies of Utrecht and the main municipalities in the Netherlands (e.g., [6,31,74]).

#### *4.1. Method Validity and Limitations*

The GCF method integrates a wide range of governance gaps to assess a city's capacity to adequately manage water challenges [45]. This plethora of divergent aspects of water governance offers the opportunity to identify barriers and enablers, and thus reveals a city's current position on governing a specific water challenge. The applied methodology is comprehensive, enables reproducible results and includes both a policy review of local authorities and organisations as well as in-depth interviews with various local stakeholders. The results provide relevant insights for city planners and policymakers at the local level and can thus help the urban network in place to implement sound strategies and policies to alleviate the risk of pluvial flooding.

However, this study has also revealed limitations. The outcomes of the governance capacity analyses emphasised the role of citizen engagement in addressing pluvial flood risk. Since this study is based on a literature review and expert interviews, an assessment of how citizens consider their role in addressing flood risk is not accounted for. As such, a suggestion for future research is an in-depth study that explicitly includes citizens, for example through surveys. This will be relevant to further substantiate our findings related to citizen engagement in urban flood risk management.

There are also other frameworks that assess the governance capacity of environmental challenges, most notably, the adaptive capacity wheel [39] and the framework proposed by Mees and Driessen [75] to assess the governance capacity for green urban areas. Like the GCF, these frameworks rely on similar literature related to adaptive capacity, environmental governance, and water governance. Hence, elements such as social learning, division of responsibilities, resources, leadership, or accounting for uncertainty, are included. Moreover, the starting point of these frameworks is the comprehension of different values, viewpoints and alternatives in problem framing, decision-making, and implementation. Gupta, et al. [39] identify six dimensions: Variety, learning capacity, room for autonomous change, leadership, availability of resources, and fair governance. These dimensions and their 22 criteria assess the adaptive capacity of institutions (including their ideological values and norms), predominantly at the national level. This approach differs from the GCF's focus on how well different stakeholders collaborate and jointly address urban environmental challenges. In this respect, Mees and Driessen [75] have a similar approach as the GCF but focus specifically on green urban areas. They identify five key capacities: Learning, legal, managerial, political, and resources. Their framework also includes legal capacity and stakeholder engagement. Likewise, the GCF specifically includes indicators such as policy ambition, institutional compliance, smart monitoring, implementing capacity, and stakeholder engagement. The adaptive capacity wheel is, however, less focussed on the stakeholders and legal or policy aspects. When considering the scoring system, Gupta, et al. [39] apply a generic Likert scoring, which implies that there is more room for a researcher's interpretation when compared to the GCF's indicator specific Likert descriptions and standardised data gathering procedure. Moreover, Mees and Driessen [75] apply a more descriptive analysis of each component of their framework. Since this paper aimed to provide insights that may apply beyond the case study itself, a more standardised method has been selected that enables a better reproducibility and comparability of different cases. Given the similarities between the scope of this paper and that of Mees and Driessen [75], the frameworks can be considered complementary and may be combined in future research about the capacity of cities to

govern pluvial flood risk. In particular, the indicators of Mees and Driessen that specifically address urban green areas may strengthen the more generic scope of the GCF and thus enable a more accurate analysis of pluvial flood risk.

#### *4.2. Promising Multi-Sectorial Linking Opportunities*

This study revealed barriers (e.g., limited citizen engagement) and opportunities (e.g., many local partnerships working on the issue of heavy rainfall) that require action by the entire network of stakeholders in Utrecht. Although the city is generally adopting sustainable pathways, it hitherto insufficiently recognises the multifaceted benefits of implementing integrated climate adaptation plans. For example, improving soil permeability, adding green spaces, adapting underground water infrastructures, installing green roofs, and even relocating buildings may reduce pluvial flooding and urban heat island effects. Such measures have additional benefits, such as better air quality, urban surface water quality, biodiversity, human health, and the overall attractiveness of the city [4,13,22]. For instance, green roofs offer multiple environmental benefits, such as efficient temperature control of buildings (using less energy), retaining rainwater (reducing pluvial flood risk), restoring biodiversity, and enhancing air and stormwater runoff quality [76]. The benefits of these 'linking opportunities' [70] may outweigh their costs and may ultimately improve the attractiveness and liveability of the city of Utrecht.

#### *4.3. The Role of Citizen Engagement in Municipal Water Management and Climate Adaptation*

The importance of the involvement of both public and private actors in climate adaptation and flood risk management has been frequently stressed (e.g., [1,21,22,24,26]). For example, leading concepts such as integrated water resources management and adaptive management include principles of engagement of a variety of stakeholders. Influential organisations include International Water Management Institute, the Food and Agriculture Organization, the World Bank, and various regional authorities. In addition, the United Nations Sustainable Development Goals and the European River Basin Approach also emphasise the authentic participation of a variety of public and private stakeholders [77–79]. However, how to engage stakeholders and citizens is a key challenge in many contexts. Through our case study of Utrecht, we found that citizens are hardly involved in the local decision-making process on pluvial flooding. Similarly, recent studies on the engagement of local stakeholders (e.g., citizens and/or other private actors) in climate adaptation and flood risk management show that involvement of local (private) stakeholders tends to be limited (e.g., [25,27,80]). Moreover, city-citizen collaborations on climate adaptation are scarce [29,30]. We found that Utrecht's public actors' (i.e. municipality and regional water authority) current strategy is primarily focussed on supplying information about climate adaptation to spur civic action. Through an extensive study of 402 urban areas, Klein, et al. [27] found similar results as they argue that local authorities steer citizens solely through information provision. These residents are, in turn, expected to use this information to implement adaptation measures [27]. In addition to solely providing information in a one-way direction (i.e. from government to citizens), local authorities may consider citizens' capability to collect data or information themselves. With respect to this, the concept of 'citizen science' is repeatedly referred to. Citizen science is defined as a practice in which individuals voluntarily participate in data collection or observations for scientific purposes and viewed as a form of collaborative research [81,82]. Sy et al. [83] emphasise that citizens play a crucial role in flood hazard assessment through various techniques, such as monitoring rainfall or analysing messages on rainfall on social media. Moreover, citizen science contributes to an increased understanding of the investigated subject by all involved actors [82]. In turn, a better understanding results in a higher level of awareness [84]. Five recent citizen science projects in the Dutch surface and drinking water sector show promising results on the effects of citizen participation; 70% of these projects' participants indicated that their level of awareness regarding water had increased, and even 87% of them described participation in the project as a 'learning experience' [84]. Thus, citizen science can be seen as a valuable form of citizen engagement

(condition 4) through which awareness, knowledge, and transparency (conditions 1 and 2) on pluvial flooding can be enhanced.

In addition to citizen science, citizens may be further engaged in addressing pluvial flooding by taking measures on their private properties. These initiatives include, for example, the decoupling of rainwater from the sewer system, placing rain barrels, removing pavements, or installing a green roof (e.g., [25,27,85]). As our study shows, the implementation of these property-level protection measures (i.e. 'coproduction') often takes place after the occurrence of pluvial floods or solely in flood-prone areas. To address pluvial flooding more effectively, it is crucial to shift towards proactive actions, instead of reactive behaviour.

It has been suggested that involving citizens in decision-making processes is time-consuming and involves higher costs for the government [30,86]. However, the costs do not outweigh the positive effects of citizen participation, such as gaining legitimacy of decisions, trust-building, and learning from citizens [86]. Moreover, Mees, et al. [17] argue that 'coproduction' (i.e. interaction between citizen and public authorities during decision-making processes and in practice) can be seen as a way to reduce additional governmental investment in flood risk management. If citizen engagement becomes business as usual in governing pluvial flood risk, this may have a positive impact on the financial viability (condition 8) of dealing with the specific risk.

Furthermore, we found that citizens' willingness to pay (indicator 8.2) for flood protection measures in Utrecht is limited. This may be related to the observed limited risk perception (sense of urgency; indicator 1.2). In addition to this, Owusu, et al. [87] conclude that the scale of flood events and their impacts also relate to the extent to which people are open to adaptation measures. In other words, a large-scale flood event results in more citizens who might consider implementing adaptation measures on their property. Furthermore, Torgersen and Navrud [88] stress that citizens in high-risk flood areas have a greater willingness to pay for adaptation measures. Besides, Henstra, et al. [89] found that willingness to pay for property-level flood protection measures has a positive relationship with age, housing type, and level of education. However, the present study shows that living in a flood-prone neighbourhood does not automatically lead to investments (i.e. adaptation measures) to reduce pluvial flood risk. This might relate to the perception citizens of Utrecht have regarding the flood risk they face. This is in line with Bubeck, et al. [33], who argue that the supposed positive relation between flood risk perceptions and taking private adaptation measures is found to be limited in current empirical studies.

The available financial aid (e.g., subsidies) provided by Utrecht is currently suboptimal (indicator 9.1). The results of this study indicate that the dissemination of information in combination with financial incentives (i.e. 'soft policies') does not yield the desired effect of taking adaptive action. With respect to this, Dai, et al. [6] suggest that more binding rules instead of soft policies may be a valuable contribution. These regulations may contribute to the engagement of citizens in the implementation of climate adaptation measures. For example, if local authorities decide to levy taxes on heavily paved gardens (for example when >70% of a private garden is paved), citizens have a stronger incentive to take action. Likewise, Mees, et al. [90] conducted a comparative study on the installation of green roofs and concluded that hierarchical arrangements (steering through regulations) are the most effective. However, local governments should play a facilitating role in supporting citizens [85]. Hence, a well-balanced use of both soft and hard policy instruments seems key. For instance, Kamperman and Biesbroek [91] advocate for a combination of 'hard' and 'soft' modes, because the existing Dutch regional water authorities' strategy of soft policies seems to be insufficient to spur climate change adaptation.

This research may support this finding. One way to achieve an improved governance capacity includes regulations such as levies or taxes on heavily paved gardens or large wastewater discharges (according to the polluter-pays principle). Another way to enhance the overall governing capacity to address pluvial flooding is through an increased engagement of citizens in local decision-making processes. Providing sufficient examples of good local practices (i.e. adaptation measures of fellow citizens) may contribute to getting citizen engagement off the ground in practice. Further research may be required through various neighbourhood experiments where different combinations of 'soft' and 'hard' measures are applied. In such an experimental setting, behaviour influencing tactics, other than just information transfer, should be tested to develop an optimal package that triggers citizen engagement in climate adaptation measures both in home and within their city. In such an experimental approach, different citizen groups could be identified based on their socio-economic characteristics, behaviours, and attitudes with respect to climate adaptation. Engagement of each of these groups could be effectively triggered by a different approach. Such experimental knowledge seems promising for obtaining valuable knowledge to effectively engage citizens and trigger climate adaptive behaviour.

Raising awareness is often perceived as crucial to realise more citizen engagement. However, a more nuanced balance between effective policy instruments, stakeholder engagement processes, and the development of local private initiatives is needed to effectively engage citizens to adapt to urban flood risk. To obtain a better insight into how to engage different citizen groups, further empirical research is needed to examine citizen engagement in urban flood risk management in practice.

#### **5. Conclusions**

The aim of this study in Utrecht (the Netherlands) is to obtain a deeper understanding of how the capacity to govern pluvial flood risk can be developed through citizen engagement. We applied citizen engagement as an umbrella term for the participation of citizens in the local decision-making process and for active involvement in the implementation phase by taking climate adaptive measures. The results of this study indicate that solely providing information and subsidies, i.e., 'soft policy instruments', does not yield the desired effect of citizens taking climate adaptive measures to protect themselves against pluvial flooding. Residents in Utrecht are currently insufficiently engaged in the local decision-making processes, which may explain the limited flood risk awareness among citizens. Their limited awareness, in combination with a low willingness to pay, may explain why they rarely take climate adaptation measures to alleviate the risk of pluvial flooding. The city of Utrecht might consider: (1) Including citizens more explicitly in the decision-making process regarding (pluvial) flood risk management; and (2) broadening the scope of its policy instruments by implementing more binding rules, such as taxes on heavily paved gardens. In doing so, residents are expected to become more aware of and more engaged with pluvial flooding. Improved citizen engagement can also be enhanced through citizen science projects. By creating such initiatives to establish more meaningful citizen engagement, Utrecht's capacity to govern pluvial flood risk can be strengthened substantially. Because other cities in the Netherlands and Europe face similar challenges of increasing downpours, aging infrastructure, and inexperience with citizen engagement, these lessons may be of value for them as well.

**Author Contributions:** Conceptualization, R.C.B., S.H.A.K., and K.A.W.S.; methodology, R.C.B. and S.H.A.K.; validation, R.C.B., S.H.A.K., and K.A.W.S.; formal analysis, R.C.B.; investigation, R.C.B.; resources, R.C.B., S.H.A.K. and K.A.W.S.; data curation, R.C.B.; writing—original draft, R.C.B.; writing—review and editing, R.C.B., S.H.A.K., and K.A.W.S.; visualization, R.C.B.; supervision, S.H.A.K. and K.A.W.S.; funding acquisition, S.H.A.K.

**Funding:** This research was funded by the European Commission, in the POWER H2020 project under grant number 687809.

**Acknowledgments:** We would like to express our gratitude to all interviewees who have participated for providing valuable information for this research. Moreover, we would like to thank C.J. van Leeuwen for his suggestions during the writing process of this paper. The Governance Capacity Framework is part of the City Blueprint Approach, developed at KWR Research Institute in the context of Watershare (http://www.watershare.eu). The City Blueprint Action Group is part of the European Innovation Partnership on Water of the European Commission (http://www.eip-water.eu/City\_Blueprints). The European Commission is acknowledged for funding POWER in H2020-Water under grand agreement no. 687809.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
