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

#### **Ralf Nordbeck 1, Lukas Löschner 2,\*, Melani Pelaez Jara <sup>3</sup> and Michael Pregernig <sup>3</sup>**


Received: 28 June 2019; Accepted: 7 August 2019; Published: 13 August 2019

**Abstract:** This paper analyses the science–policy interactions in the field of flood risk governance against the background of climate change. By the example of three neighbouring Alpine regions (Switzerland, Southern Germany and Austria), the study strives to shed further light on how flood risk governance regimes embrace the possible impacts of climate change. It builds on the assumption that flood risk management, as a 'technical' policy field, is strongly influenced by scientific evidence and that differences in how countries incorporate climate change can be explained by the way science and policy are brought together in the respective national arenas. We structure the empirical analysis along three dimensions: (i) dynamics of knowledge creation; (ii) institutionalization of the science–policy interface; and (iii) pathways of influence of expertise on policy development. We find that there is a mixed, though increasing influence of climate change on flood risk governance in the three selected Alpine regions. Climate adaptation has become an important issue of flood policy in all three study areas, and this shift has been strongly supported by evidence-based arguments.

**Keywords:** science–policy interface; flood risk management; climate change; adaptation

#### **1. Introduction**

The influence of climate change on future flood risk in Europe is still highly debated. Analyses of past trends show no clear patterns of robust and ubiquitous climate-driven change in flood magnitude and frequency of high discharge in European rivers during the last decades. Rather in contrast, there are strong regional and sub-regional variations in the observed trends [1]. Similarly, the results of climate models are uncertain regarding the prediction of future changes in frequency and magnitude of floods, inter alia, because of uncertain information on phenomena such as an increased spatio-temporal variability of precipitation and changes in extreme rainfall events [1,2]. This provides extensive challenges for policy-makers in charge of flood risk management and climate change adaptation "because it is naive to expect availability of trustworthy quantitative projections of future flood hazard" any time soon [3] (p. 1). Dealing with these highly uncertain changes requires new approaches to policy-making that support the move from resistance-based strategies to more resilience-oriented approaches. While resistance-based approaches can provide effective protection and minimize costs up to certain design floods, they "do not cope well with uncertainty" [4] (p. 291). Resilience-based strategies, on the other hand, embrace adaptability and flexibility as a means to respond to the

climate-induced increase in flood discharge and thus better cope with the shocks and stresses of (extreme) flood events [5–7].

Climate adaptation policies emerge from the interplay of a varied set of actors, including policy-makers, scientists, engineers, and many other stakeholders, leading to numerous ambiguities, since successful policy-making is not about finding the objective truth in a scientific sense, but about finding a balance between different positions. While the interlinking between climate chance and flood risks has been widely studied in the field of engineering and the natural sciences, there are only a few studies on the influence of climate change on flood risk governance. In a European context, the project STAR-FLOOD [8] delved into this question. In a comparative study of six EU countries, the project found that climate change as a risk driver has only a fairly modest impact on flood risk governance overall [9]. For five of the six countries, the authors found that the discourse on climate change tends to stabilize existing governance arrangements with their emphasis on prevention and defense: In England, the climate adaptation discourse influenced the broader discussions on the financial sustainability and cost-efficiency of the current flood risk governance, yet the overarching risk approach did not fundamentally alter [10]. In Poland, the influence of climate change has been limited, both because climate change projections are considered uncertain for Poland and because of lacking political and societal support for policy changes [11]. A mentionable exception is Sweden where climate change has played a pivotal role in the observed shift of flood risk governance [12].

This paper strives to shed further light on the question of how flood risk governance regimes embrace the possible impacts of climate change. Thereby, we build on the assumption that flood risk governance, being quite a 'technical' policy field, is strongly influenced by scientific evidence and that differences in how countries incorporate climate change might be explained by the way science and policy are brought together in the respective national arenas. Against this background, we delve into the following analytical questions: What are the implications of scientific evidence on climate change for flood risk governance? How do policy-makers deal with those high degrees of uncertainty, both in terms of causal drivers and impacts? What is the role of science and expertise in this technical policy field? These are the questions we are going to explore in this paper. Empirically, we draw on a comparative case study of flood risk management in the context of climate change in three Alpine regions: Austria, Southern Germany, and Switzerland.

The remainder of this paper is structured as follows: Section 2 summarizes the state of science–policy research and presents the analytical dimensions along which we explore the science–policy interactions in the field of flood risk governance. Section 3 presents the empirical findings for the three study regions Austria, (Southern) Germany and Switzerland. Finally, in Section 4 we discuss our main findings based on the cross-case comparison.

#### **2. State of the Research and Own Conceptual Framework**

Flood risk management is a highly complex and technical policy domain. It entails understanding the underlying factors that drive flood risks, including the complex drivers related to land use changes, socioeconomic developments, and climate change [13]. Further, it requires modelling these dynamics to assess changes in flood risk [14], and devising instruments to mitigate those risks—from designing and constructing flood defenses and other technical infrastructure to developing and adjusting land use plans and building codes [15]. Consequently, flood risk governance—understood as the structures and processes in which public and private actions and resources are coordinated for solving collective problems [16]—is expected to be informed by an evidence-based approach. This expectation has been partly substantiated by research in the abovementioned STAR-FLOOD project. As part of this project, Wiering et al. [9] provided a typology of varieties of flood risk governance in Europe. For the six countries studied in detail, the authors found national approaches to flood risks and their governance to be marked by a dominance of public actors (especially in the Netherlands, Poland, France, and Belgium) as well as central authorities with strong legal powers (especially in Poland and France). Similarly, Liefferink et al. [17] reported on the strong and powerful role of expertise and expert bodies (especially in the Netherland and Poland), an observation confirmed by other authors, like Lange and Garrelts [18] for Germany, or Mostert et al. [19] and Bergsma [20] for the Netherlands.

The literature on flood risk governance is still quite fragmented [21,22], and there are only a few studies explicitly focusing on the question of what role science and expertise play therein. There is more research on science–policy interactions in the broader field of water governance; however, also these studies have a strong bias towards more instrumental, managerial perspectives. Many of those studies start from the observation—or rather assumption—that there is a "gap" between science, on the one hand, and management or policy, on the other hand. This is seen as a problem because, as Timmerman and Langaas argue, "[c]onsistent and relevant information on the status of water systems is indispensable for rational and cost-effective water management" [23] (p. 177). Similarly, Xu and Tung postulate that "[d]ecision-making in water management requires the delivery of accurate scientific information" [24] (p. 535). As these two examples indicate, studies often build on a rather simple, linear model of "information transfer" and a quite a-political, if not to say "naïve", understanding of the policy process (for a critique of the linear model in general, cf. Refs. [25,26]).

Recent studies have overcome some of this linearity in that they do not just call for the more effective transfer of information from science to management or policy, but rather first try to better understand the context in which management or policy decisions are taken. For example, Leskens et al. investigate how information from flood models "is exchanged among participants in flood disaster organizations and how this exchange affects the use of modelling information" [27] (p. 53). Similarly, Höllermann and Evers analyses "how practitioners perceive and handle uncertainties in their daily decision-making routines at the knowledge/decision interface and how they evaluate and integrate uncertainty information into their decision-making" [28] (p. 9).

All of the above-mentioned studies focus on the micro level of individual decision-makers, mapping for example the information that managers have available, describing the organizational contexts in which they operate, and assessing the decision support tools on which they can draw similarly [29–32]. In contrast to that, there are only few studies that look at the macro level and take a more policy-oriented perspective. In a European context, those policy-oriented studies have often been developed around the EU Floods Directive (2007/60/EC) and its implementation in Member States. Empirically, most of those studies took a rather narrow perspective by looking at the role of participation as one of the key governance principles enshrined in the Directive (e.g., Refs. [33,34] for Germany; Ref. [35] for the UK; Ref. [36] for Sweden).

A more comprehensive assessment of flood risk governance arrangements (FRGAs) was provided by the STAR-FLOOD project: The project was comparative in scope, analyzing and contrasting FRGAs in six European countries. It went beyond the EU Floods Directive and its national implementation by taking a more long-term perspective, analyzing FRGA dynamics from the 1970s up to the present, thus, trying to explain stability and change in the sector [17]. In addition to that, similar to our paper, STAR-FLOOD engaged in the question of whether and how climate change might have an impact on FRGAs. On that topic, Wiering et al. [9] reported that in five of the six countries studied, discussions on climate change showed a rather modest impact on national risk approaches; only in Sweden, climate change was seen to play a key role in observed shifts of risk governance.

Our paper deals with some of the questions also addressed in STAR-FLOOD, however, we strive to not only look at the macro level of national arenas but we also incorporate meso-level structures and dynamics. For that, we take an explicit, more fine-grained look at the role of science and expertise in flood risk governance, a perspective that is missing in all more policy-oriented studies in the sector (except for Ref. [20]). In our paper, we explore the question of science–policy interactions in the field of flood risk governance along three distinct analytical dimensions:

1. *Dynamics of knowledge creation*: The impact of climate change on flood risks became an important topic especially after the publication of the IPCC's Second Assessment Report [37]. Subsequently, in many European countries, one c seen extensive efforts of knowledge creation and the related consolidation of expertise on flood risks and related management options [3,38]. At the same

time, the state of knowledge is—as almost necessarily seen in any complex policy arena—still highly provisional, uncertain and incomplete [13,39,40]. Against this background, our case studies address the following questions: How did the state of knowledge in the field develop over time, and was there consensus on key insights or did experts disagree? What kind of policy recommendations, if any, did scholars provide?


We explore these research questions in a comparative case study for three neighboring countries/states in the wider Alpine region: Austria, Southern Germany (namely, the states/Länder of Baden-Württemberg and Bavaria) and Switzerland (see Figure 1). We selected these three cases for our comparison because they are all located within the wider Alpine region and thus share similar challenges with regard to flooding. For one, due to the mountainous topography, these regions are particularly exposed to a high frequency of heavy precipitation events, while the concentration of vulnerable land uses in valley basins results in high economic losses in the event of flooding [13,47]. Secondly, Austria, Germany and Switzerland are federal countries, where flood policies are marked by a complex distribution of responsibilities between the national, state and municipal levels, with the state level (i.e., Länder/cantons) assuming strong operational responsibilities in flood risk management and climate change adaptation. This also explains why we decided to analyze two Länder (Baden-Württemberg and Bayern) rather than the entire country for the case of Germany.

**Figure 1.** Overview of the case study regions Austria, Switzerland and Southern Germany (Bavaria and Baden-Württemberg).

Despite their parallels, the three regions, however, exhibit considerable differences in the way they deal with the possible climate-related influences on flooding. Methodologically, the paper is based on a qualitative analysis [48,49] of scientific and grey literature as well as relevant policies, strategies, and project documents. Complementary to the document analysis, we conducted a total of 24 semi-structured interviews with scientists, water policy experts, engineers, and policy-makers of the three regions (between June 2016 and June 2018) (for a list of interviews see Table A1 in the Appendix A). The collected data set has been anonymized, coded and interpreted using the qualitative data application MAXQDA 12.

#### **3. The Role of Climate Change in Flood Risk Management in Austria, Southern Germany and Switzerland**

This section explores science–policy interactions in the field of flood risk governance for the case study regions in Austria, Southern Germany and Switzerland. We structure the empirical analysis along the three dimensions outlined above, namely (i) dynamics of knowledge creation; (ii) institutionalization of the science–policy interface; and (iii) pathways of influence of expertise on policy development.

#### *3.1. Austria: Flood Risk Adaption without Climate Change Allowances*

Austria experienced a series of major flood events in 2002, 2005 and 2013. Those events caused flood risk management in Austria to shift to new priorities. A broad discussion about future flood risks and the impact of climate change thereon has emerged since 2002. Until today, Austria has not introduced specific adaptation instruments such as climate change allowances, but future uncertainties are incorporated into the planning practice by various alternative means.

#### 3.1.1. Dynamics of Knowledge Creation

The first studies on the possible effects of climate change for individual regions in Austria were published in the early 2000s [50–52]. Due to the different characteristics of the investigated catchment areas, but also because of the different methodological procedures and different input data, the results of these studies in individual catchment areas are not directly comparable. However, the main results of these studies were quite similar. Godina [52] showed—based on 120 annual series of observations—different trends for three rivers in Austria: rising trend on the Salzach, falling trend on the Enns and a slightly falling trend on the Inn. Nachtnebel and Fuchs [51] found no indication for a climate signal leading to higher flood hazards based on their scenario analyses. Blöschl and Merz [53] analyzed different flood events (period 1828–2005) on the Danube near Vienna. The overall time period showed an increase in small floods, but no trend in large floods and an increase in winter floods at the expense of summer floods. From all these studies, no climate-induced changes in the discharge peaks could be detected [54].

The extreme flood event of 2005 and two studies (commissioned by the World Wide Fund for Nature and the State of Upper Austria) by Refs. [54,55] renewed the discussion. In spite of being cautious in their conclusions, both studies argued that climate change would increase future flood risks: "In general, although it is not possible to quantify how ultimately the overall flood risk in Upper Austria will change as a result of climate change, changes must be expected in any case, in particular seasonal and regional shifts in risk" [56] (p. 21). This initiated a new wave of studies in the following years that took a closer look at the effects of climate change on flood hazards [57–60].

In the early 2010s, the Austrian Ministry of Agriculture, Forestry, Environment and Water Management (BMLFUW) and the Water Resources Divisions of the State Governments commissioned a comprehensive sectoral study on water management as part of the Austrian Climate Adaptation Strategy [61]. The aim of the study was to assess climate impacts on water resources based on the most recent data and methods, and to derive policy recommendations on climate change adaptation strategies for water resources management in Austria [62]. The study shows that during the period 1976–2007, increasing trends occurred in about 20 % of the catchments, particularly north of the main Alpine ridge. However, if one extends the analysis period to 1955–2007, the trends are much less clear because a number of major floods happened in the period 1955–1975. It also shows that winter floods have increased significantly more than summer floods. The authors conclude that the accumulation of flooding in recent decades lies within the natural variability of flood decades, but that an influence of climate change cannot be ruled out. Furthermore, they conclude that the introduction of a general climate allowance for design values is not required at the current state [61,62].

The findings of the so-called "Blöschl study" [61] settled the scientific discussion for the following years. The study was updated in 2018 [54]. The new findings confirmed many of the previous results, but showed a somewhat stronger influence of climate change. In the future (time horizon 2021–2050), regionally different changes in flood discharge are expected to be in the range of −4 to +10% (for one-year flood events). The updated study reiterates the conclusion that a climate change allowance is currently not required in Austria, but it points to a higher number of catchment areas with rising trends and highlights the increasing risk from pluvial floods.

#### 3.1.2. Institutionalization of the Science–Policy Interface

Scientific expertise is very important for decision-makers in Austrian flood policy. As two of our interview partners stated: "Science is certainly the basis for the flood risk management that we have developed in recent years" (Interview AT02), and there is "a great deal of openness towards scientific findings" (Interview AT03 (all direct quotes from interviews are translated from German)).

After the extreme flood event of 2002, a research program called FloodRisk was initiated. FloodRisk brought together scientists and public officials from a variety of fields such as meteorology, hydrology, geomorphology, natural hazards, damage assessment, law, spatial planning, and disaster protection. It followed a broad approach, which one of our interview partners describes as follows: "As a result of the 2002 events, actors suddenly realized that this has much, much wider dimensions and much of what went wrong in Austria, in my view, has to do with the fact that key players ( ... ) have not used science to reflect on how to improve flood risk management" (Interview AT04). Despite its broad approach, the FloodRisk program did not focus on climate change in particular. Shortly after the final

report of FloodRisk was published, in 2005, another extreme flood event hit Austria and a follow-up program (FloodRisk II) was established, this time including a sub-project on the impact of climate change on floods [57].

While FloodRisk stands for a disaster-driven type of science–policy interaction, there are also more structural, long-term venues in which decision-makers, professionals and scientists interact. An important network is the Austrian Water and Waste Management Association (ÖWAV) (Interviews AT02; AT05). The ÖWAV acts as an information and communication hub that provides early information on legal, technical and economic developments and supports the exchange of experience in the field of water management. It has established a working group on flood protection in particular to engage with municipalities and flood water associations. The aim of the working group is to provide a platform for decision-makers, municipalities, authorities, planners, universities and industry in flood risk management and to develop, coordinate and bundle activities for this target group. The ÖWAV has organized seminars in 2009, 2016 and 2018 to discuss the impact of climate change on water management in Austria [63]. In 2014, it published an expert paper to assess the impact of climate change on water management based on the ZAMG/TU Wien study and to discuss the need for further adaptation measures [64].

#### 3.1.3. Influence of Expertise on Policy Development

The close relationship between scientists and public officials, and the regular exchange among members of the flood community led to an evidence-based flood policy in Austria. The findings of the major studies on the impact of climate change are mentioned in every relevant policy document, often cited literally. The first document of the federal ministry that referred to climate change as a possible driver of flood hazard was published in 2006, and it was clearly influenced by the earlier studies of Refs. [49–51]. In this document, the federal ministry asserts that "whether the hydrologic extremes of 2002 and 2005 and the increasing frequency of local heavy precipitation in the last years are signs of a global climate change cannot yet be clearly determined. Developments are to be followed closely and—if necessary—need to be taken into consideration for flood protection" [65] (p. 7). After the "Blöschl study" was published in 2011, it became the new point of reference. The Austrian Climate Change Adaptation Strategy from 2012 states: "Due to the prognostic uncertainties about the actual effects of climate change on the different regions in Austria, a general increase in these design values by a factor X is currently discouraged by the scientific community" [66] (p. 172). The report of the Austrian Panel on Climate Change [67], the expert paper of the ÖWAV [64], and the National Flood Risk Management Plan [68] all refer to the Blöschl study and basically copy-and-paste the findings as well as the policy recommendations. The update of the Blöschl study in 2018 will probably become the new reference point. The latest publication by the federal ministry has already taken into account some of the findings: "The hydrological extremes of local intensive rainfall events currently still match the historical variation. However, it becomes ever more likely that they are a consequence of global climate change" [69] (p. 11).

Adaptation to likely future flood risks is a multi-causal process in Austria in which uncertain climatic factors are overshadowed by highly likely socioeconomic trends such as land-use change induced by population and economic growth [70,71]. For various reasons, Austrian flood policy makers are not willing to account for possible climate-induced changes in flooding, e.g., in form of a precautionary climate allowance to runoff calculations (as in parts of Germany, see Section 3.2). First, they point to the lack of scientific evidence suggesting that flooding will increase in Austria due to climate change (see Section 3.1.1). Second, since implementing a precautionary climate allowance would increase the costs of structural flood protection drastically (Interviews AT02; AT05; AT07), this would reduce the number of projects that can be implemented with the same budget (Interviews AT05; AT06). Third, some experts see such a precautionary orientation in structural flood defence to be in conflict with the federal prioritisation of non-structural measures, and they assume that a "levee effect" could increase the damage potential in protected areas (Interviews AT02; AT08).

Despite the fact that Austria has not added a climate change allowance to its design values, it would be short-sighted to conclude that until now Austria has not actively factored-in the effects of future climate change into its flood risk management strategies. Policy-makers and scientific experts stress that in many areas, potential climate change influences are already incorporated into planning practice, as extreme events lead to an adaptation of the design values for flood protection measures (Interviews AT02; AT07). The federal government also improved design standards such as the safety allowance (called "freeboard" in Austria), which is added to the calculated design of flood defence infrastructure as a "buffer" to accommodate for epistemic and modelling uncertainties. On the state level, water management authorities started to take the issue of local intensive rainfall more seriously. Facing an increase in heavy rainfall, as a likely result of climate change, for instance the Governments of Lower Austria and Styria have recently developed a state-wide hazard map for surface runoff, which could provide a comprehensive hazard information base for flood proofing measures in areas prone to pluvial floods.

#### *3.2. Southern Germany: Early Awareness and Institutionalization of Climate Change Impacts*

The Southern German states of Baden-Württemberg (BW) and Bavaria (BY) witnessed several major flood events with devastating losses: Rhine floods in 1993 and 1995; Alpine floods in 2000 and 2005; and Danube floods in 1999, 2002, and 2013. Scientists and professionals engaged in looking at possible connections between climate change and regional extreme flood events already early on, i.e., in the 1990s. Consequently, BW and BY became the first federal states in Germany to introduce a technical flood risk instrument explicitly focused on addressing the impact of climate change, namely the Climate Change Factor.

#### 3.2.1. Dynamics of Knowledge Creation

During the early 1990s, the hydrologists Bárdossy and Caspary published the first studies on the connections between climate change, European atmospheric circulation patterns, and extreme flood events for Southern Germany [72,73]. Their statistical analyses indicated for some catchments in BW a connection between increased temperatures and changes in the stationarity and frequency of flood events [73]. At first, those studies triggered negative reactions from policy-makers and some scientists of the Southern German water management community, who criticized the methodological approach and, consequently, cast doubt on the relevance of human-induced climate change for floods (Interviews DE01, DE02, DE05). Professionals argued that the sector always had to deal with climatic variability and natural climatic cycles and their related impacts on flood-proneness (Interview DE03). Additionally, flood risks were seen to be contingent on land use changes and other societal dynamics. Therefore, some experts argued that the sector was not dealing with a "fundamentally new situation" [74]. This early criticism and rebuff later turned into generally high interest, political and technical, to systematically examine the causality between climate change and flood risks, and to develop their own regional climate and hydrological models (Interviews DE01, DE02).

The ensuing studies produced during the late 1990s and early 2000s showed high degrees of variance and uncertainty with regards to a possible link between climate change and extreme flood events in Southern Germany. Study results varied among different catchments (especially depending on whether catchments are influenced mainly by precipitation or also by snowmelt), but they also varied along the technical parameters used in the different case studies, including the length of historical data series, or the chosen modelling methods and ensembles [75–77]. Long-term analyses of hydrological parameters showed for the period 1932–1998 no significant, area-wide trends of flood discharge increase. Only when reducing the period under consideration to the last 30 to 40 years of recorded data, most of the analyzed gauging stations showed increases, especially during the winter half year [74]. Regarding future scenarios, early KLIWA-studies presented estimates of an increase of mean annual floods and their frequency, again especially during the winter half year [74,78–80]. At the same time, studies on extreme and rare flood events under climate change influence showed a relatively low increase [79]. Overall, scientists largely agreed on their assessment that human-induced climate change would not have sweeping effects on flooding patterns in Southern Germany, however, it might lead to a locally differentiated increase in floods [77].

In the late 2000s and 2010s, KLIWA continued its monitoring program based on historical data while developing more complex climate and impact models and partly extending its research focus by paying larger attention to heavy precipitation [81–83]. Overall, the key messages resulting from those studies have not changed much, but one still sees some minor shifts: Studies of long term flood discharge behavior (1932–2015) now show positive trends for the majority of gauging stations as well as an increase in the significance of those trends, both for the whole year as well as for the winter half year [84–87]. Concerning future scenarios (2021–2050, 2071–2100), simulations of the mean monthly flood discharge for the winter half year indicate for BW an increase of flood discharge trends for the nival flow regime and a strong increase for the pluvial flow regime [86]. For BY, flood discharge projections show mixed results with strong spatial variations [82,86]. These findings are also supported by Hatterrmann et al. [88], who showed that a considerable increase in flood-related losses can be expected in Germany in a future, warmer climate.

Given the unclear trends and the high degrees of uncertainty inherent in early studies, it is interesting—if not to say surprising—to see that KLIWA scientists in 2004 explicitly suggested implementing a Climate Change Factor (CCF). They presented the CCF as a flexible non-regret adaptation instrument to be included in the planning of all new technical flood protection measures. As will be described in more detail in Section 3.2.3, the CCF was readily taken up by policy-makers both in BW and BY. In recent years, KLIWA has become more cautious with specific policy recommendations: Emphasizing the inherent uncertainties that come with future projections, scientists tend to suggest rather "soft" instruments such as guidelines for municipal management of heavy rainfall risk [83,89].

#### 3.2.2. Institutionalization of the Science–Policy Interface

In both fields of climate change policy and flood policy, South-German decision-makers have for a long time strongly relied on the technical expertise of their respective environmental agencies, that are the Baden-Württemberg State Institute for the Environment, Survey and Nature Conservation (LUBW) and the Bavarian State Office for the Environment (LfU) (Interview DE06). In our interviews, those agencies were described as the first-hand technical and scientific advisory bodies that connect, translate, and moderate between science and policy (Interviews DE01, DE04, DE07, DE09). Also, in institutional terms, both agencies occupy a hybrid space as they collaborate with scientists, private engineering firms, communities, and federal state policy-makers. They draw on external technical expertise in various ways, mostly under the formats of commissioned research and short-term project work, either through public tendering or by approaching research groups or scientists directly (Interview DE04). The water management sector is described as a closed community with strong ties between science, engineering and policy-making (Interview DE10).

A new space for science–policy interactions was introduced with the Cooperation Project Climate Change and Consequences for Water Management (KLIWA) in 1999. The formation of KLIWA was, among other reasons, a reaction to the scientific debate of the early 1990s publications mentioned above and the first and second IPCC reports (Interview DE01). KLIWA has been led by the respective environmental agencies of BW and BY in cooperation with the German Meteorological Service (DWD). With its explicit focus on climate change, KLIWA has been seen as a pioneer initiative in the water sector, not only for Southern Germany but also at the federal level (Interview DE01, DE02, DE03). Its purpose is to scientifically and technically examine the connection between climate change, extreme floods and other impacts on the water sector as well as to provide science-based recommendations for decision-makers (Interviews DE01, DE04, DE07) [90].

KLIWA has been organized along the areas of long-term observations of meteorological and hydrological data; modelling future developments; monitoring and assessing current impacts; adaptation strategies; and public outreach [90,91]. Central to the ongoing work of KLIWA has

been its communication to the expert community and decision-makers through different formats e.g., symposiums and experts' conferences, monitoring reports, project publications, and short informative posters.

#### 3.2.3. Influence of Expertise on Policy Development

The tight institutional coupling of science and policy-making, as most prominently enshrined in KLIWA, is also reflected on the policy level. In our interviews, both policy-makers and scientists approved and valued their close cooperation and they emphasized the benefits of evidence-based policy-making (Interviews DE03, DE08, DE10). The most direct impact of expertise on policy-making goes back to the early 2000s. In 2004/2005, both BW and BY introduced the Climate Change Factor (CCF) as a first concrete policy instrument to address possible future climate change risks. KLIWA played a crucial role in that: In KLIWA reports, the CCF was presented as a flexible, non-regret adaptation instrument to be used in design floods and the planning of new technical flood protection measures [74,92]. BW introduced the CCF based on rigorous scientific analyses of regional data by LUBW and the Karlsruhe Institute of Technology. LUBW, in cooperation with engineering partner companies, developed the "Guideline for Determining Design Floods for Technical Flood Protection". Those guidelines recommend the use of a set of regionalized factors (ranging from 0% to +75%) in design floods and corresponding technical flood protection measures. In BY, a CCF was introduced at the same time as in BW, even though regionalized results were not yet available for all BY catchments. In BY, decision-makers chose to adopt a general factor of +15% for all catchments. The political decision to adopt the CCF was announced at the KLIWA Symposium 2004 in Würzburg. With the adoption of the CCF, policy-makers claimed to follow the precautionary idea of taking action in spite of uncertainties, but allowing for future corrections of data and modelling ("no regret strategy") (Interviews DE01, DE04, DE07, DE05, DE02, DE03).

The CCF in BW and BY was an early instrument that was developed and adopted before Germany reformed its Water Act (2009), implemented the EU Flood Risks Directive (2007), and developed its Climate Change Adaptation Strategy Program (2008). The CCF was introduced as a non-binding recommendation to be adopted only under cost–benefit considerations in BW [93], and as an obligatory measure in BY, where exceptions can be considered in individual cases [94]. In practice, it is applied only on new linear technical flood protection measures of narrow scope, and not on larger technical measures such as dams, polders, and wide retention measures. The CCF can be seen as a rather stand-alone instrument, which is still technically implemented, but it has not been explicitly revised or updated, and its effectiveness, applied scope or impact have not been assessed (Interviews DE04, DE07).

Since the implementation of the CCF, the flood risk management sector in BW and BY has introduced several other measures, all of them driven by the precautionary principle. Yet, the sector is still oriented towards a more technical approach of flood protection. This can be seen by a recent initiative in BY: Based on studies of estimation, flood waves and in-depth impact analyses of the Bavarian Danube conducted by the Technical University of Munich [95,96], the State Government adopted the "Bavarian Polder Program". This program accounts for 12 potential sites for the construction of controlled polders along the Danube. The Polder Program has been presented as a means to increase the resilience of flood protection infrastructure by expanding retention volume and to reduce flood risk for downstream cities in the case of overload [96,97].

#### *3.3. Switzerland: Integrating Climate Change Adaptation into Flood Risk Management*

Following a "disaster gap" between the late 1880s and the late 1970s [98], Switzerland experienced an active period of flooding in the past three decades, including two particularly damaging events in 1987 and 2005 [99]. Swiss flood policies responded to the deficiencies of technical flood defenses by adopting a more integrated approach to flood management. Flood policy-makers at an early stage acknowledged the likely adverse consequences of climate change and integrated climate adaptation strategies into flood risk management.

#### 3.3.1. Dynamics of Knowledge Creation

In the late 1980s, a long-term study of flood discharge changes in Swiss rivers found no significant trends in the majority of catchments [100]. A sequence of damaging events in the 1990s (1992, 1993, 1994, 1999), however, prompted scientists to take a closer look at the observed changes in flood frequency. In one of the first encompassing studies, Frei et al. [101] explored the linkages between climate change and flooding for Switzerland. The study (i) suggests an interrelationship between changes in atmospheric circulation and the frequency of extreme events, and (ii) indicates that climate change will result in a "growing proportion of rainfall ( ... ) which will accelerate the runoff formation process" (p. 296). In a study on "Extreme Events and Climate Change" [102], scientists underline the difficulties to "identify or exclude a statistically valid trend in the frequency of rare extreme events". A follow-up study, however, confirms that, "as a result of the expected changes in the precipitation regime, more frequent and in part larger floods are expected (...), in particular in winter and the transitional seasons" [103] (p. 60).

In response to the 2005 flood events, Switzerland's Federal Office for the Environment (BAFU) commissioned two studies to explore the long-term changes in observed flood frequencies in Switzerland since 1850 and 1500, respectively [104,105]. The studies show that periods with frequent flooding historically alternated with quieter periods. While the analysis since 1850 indicates a large number of measuring stations with upward trends, from a long-term perspective (i.e., since 1500) it becomes evident that the recent flood-rich period (starting in the late 1970s) "is still in the range of formerly observed ones (...) and might continue for some decades under natural climate variation" [105] (pp. 1591–1592). However, the recent accumulation of large flood events in northern Switzerland "suggests that the (observed) changes in flood frequency in Switzerland are due to changes in atmospheric circulation" [104] (p. 6).

Climate projections are detecting increasingly clear signals that both seasonality and magnitude of floods in Switzerland will change in the future. With rising temperatures, more heavy rainfall events and a pronounced shift from solid (snow) to liquid (rain) precipitation is expected, which could increase the frequency of occurrence of extreme events, especially during the winter season [106–109]. These findings are supported by a more recent report by the Swiss Academy of Sciences, which points to the difficulties of projecting future change in extreme flood events but highlights that climate effects on the hydrological cycle lead to an increase in flood volumes and that an increase in flood risk is expected [110].

The latest Climate Change Scenarios for Switzerland (CH2018) substantiate the above projections with more robust climate evidence. They highlight that "heavy rainfall is projected to intensify in all seasons (...) for all event categories". Across Switzerland, flood events relevant for infrastructure planning (i.e., 10-year floods and 100-year floods) "may intensify strongly in all seasons". Compared to earlier reports "confidence in heavy rainfall intensification is now substantially higher", inter alia due to improvements in the resolution and accuracy of the new generation of climate models [111] (p. 10).

Regarding their advice for flood policy, climate scientists in Switzerland are rather cautious. For instance, Schmocker-Fackel and Naef [104] recommend that "future flood protection measures should (...) take into consideration the observed cyclical behavior of flood frequency" (p. 7), while Frei et al. [101] highlight the importance "to invest in reducing the negative impacts of extreme events, irrespective (and not knowing) whether these are of anthropogenic or natural origin" (p. 297).

#### 3.3.2. Institutionalization of the Science–Policy Interface

Switzerland's flood risk management builds on strong institutional ties with research institutions and scientists. This goes back to the origins of modern flood defense, when scientific experts were commissioned by governmental authorities to investigate the root-cause of the seminal flood events in 1868. Their expertise and recommendations provided the basis for the legislative organization of responsibilities in Switzerland's natural hazard management [112]. More than 100 years later, following the disastrous floods in 1987 and 2005, scientific expertise yet again provided impetus for

policy change. Together with policy officials and practitioners, scientists played a leading role in the post-event documentation and analysis [113], and thus, provided the necessary evidence base for the subsequent changes in flood policies (see Section 3.3.3).

Organizational integration of knowledge actors plays an important role in the institutionalization of science–policy interactions in Switzerland. This becomes evident by the example of BAFU, which is responsible for setting strategic priorities and for co-funding disaster risk reduction measures, and guides the federal governments' efforts to assess the impacts of climate change on natural hazards [114]. BAFU bundles expertise related to flood risk management in different organizational divisions and traditionally has strong personal and institutional ties to the leading research institutions in the field, which serve not only as partners in climate-related contract research but also as a recruitment pool for administrative officials (Interview CH02). Given its role as a pivotal administrative authority on the federal level, BAFU thus assumes an intermediary role between science and politics in flood risk management: whereas the latter provides the financial and legal frameworks for flood risk management, science delivers the knowledge base to further develop and implement flood risk management in practice (Interview CH01).

To facilitate climate-related research and to strengthen linkages between scientists and policy-makers, in 1988 the Swiss Academies of Sciences launched the ProClim initiative as part of the wider Platform Science and Policy. As an "interface for communication between science, public administration, politics, economy and the public", the platform plays a strong role in preparing the existing knowledge in climate research to support decision-making (Interview CH01). A more specific platform for the exchange of knowledge actors in natural hazard and flood risk management is the so-called FAN Association (Fachleute Naturgefahren Schweiz), which includes more than 460 members with different institutional and disciplinary backgrounds. The majority of members are private actors (44%), such as engineering consultants, but with a large share of actors from the administrative sector (21%) and academia (18%), the association assumes an important function to foster knowledge exchange between a range of actors in the field of gravitational hazards ([115]; Interview CH01).

Finally, the National Platform for Natural Hazards (PLANAT), established in 1997 by the Swiss Federal Council, provides an overarching framework for the interaction between science, public administrations and private actors. PLANAT has the mandate to set strategic priorities for an inter-sectoral, whole-of-society approach in risk management and to coordinate activities and foster knowledge exchange [114,116]. In that capacity, PLANAT also acts as an editor of guiding documents, including the national strategies for the integrated management of natural risk [117,118] but also for specialized policy reports e.g., concerning risk-based spatial planning [119].

#### 3.3.3. Influence of Expertise on Policy Development

Already at an early stage, Switzerland adopted a proactive stance regarding the possible climate-related effects on flooding. In the late 1990s, the founding document of PLANAT highlighted that "climate change and the tendency towards extreme weather events may further increase related risks" ([115]; translation from German). A few years later, a guiding policy document in hydraulic engineering suggested that "the risk of [flood] hazards could generally intensify in coming decades due to external influences (e.g., global climate change)" [120] (p. 7).

The above policy documents were issued before publication of the respective scientific studies (cf. Section 3.3.1), indicating that they were informed by the mounting awareness for climate change in the late 20th century, inter alia following the Second IPCC Assessment Report in 1996 [121]. In particular, following the succession of large flood events in the late 1990s and early 2000s, policy makers were eager to learn more about the climate-related effects on flooding (Interview CH03). In response to the 2005 floods, BAFU commissioned two scientific studies to build a stronger evidence base concerning the long-term changes in flooding and the possible climate influences [104,105]. Later policy documents explicitly refer to the scientific literature, although the statements concerning the climate effects on flooding are not necessarily more concrete [122–124].

Despite—or rather, given—the lack of hard evidence of climate change effects on flooding, scientific studies nevertheless supported the shift from hazard defense towards an integrated approach in flood risk management [125,126]. Following the 1987/2005 flood events, Switzerland's flood policies were oriented towards reducing flood discharge and increasing the "robustness" of technical defenses against flood overload [127,128]. The integrated approach in Swiss flood policy aims at planning for extreme events irrespective of the actual influence of climate on flood discharge. According to a leading policy-maker, the nascent paradigm may thus be described as "congruent" with Switzerland's precautionary stance in climate change adaptation (Interview CH01).

The synergies between this reorientation in flood policies and climate change adaptation are specifically reflected in the trans-sectoral pilot program "Adaptation to Climate Change", which was launched under the auspices of BAFU to support cantonal and municipal efforts in meeting climate adaptation challenges [129]. Within the framework of the program, a total of 31 projects were implemented between 2014 and 2016, thereof two directly related to the nascent flood policies. The pilot projects had an experimental format where scientific expertise played a lesser role. Although research institutions were eligible for funding, the selected projects mainly involved public authorities (cantons, municipalities) as well as professionals/practitioners (e.g., consultants, engineering bureaus) to support the practical implementation of the respective climate adaptation measures [129].

In sum, Switzerland strongly aligns climate change adaptation and flood risk management [122]. Flood policy's active stance in implementing Switzerland's Climate Change Adaptation Strategy as well as the involvement of administrative authorities in the assessment and monitoring of climate change demonstrates that policy actors in this field "consider adaptation increasingly on equal terms with other sectoral policy objectives" [130]. Climate science plays an important part in the reorientation of flood policy by providing the much-needed evidence to assess the climate effects on flooding. To support the implementation of future-oriented risk management strategies, flood policy, however, also relies on expertise from other disciplines (e.g., spatial planning), while non-scientific knowledge actors, in particular technical experts and municipalities, are important partners for implementing in practice the nascent paradigm of integrated flood risk management.

#### **4. Cross-Case Comparison and Discussion**

This paper analyzed the science–policy interactions in the field of flood risk governance against the background of climate change for the case studies of Austria, Southern Germany and Switzerland. We structured the empirical analysis along three dimensions: (i) dynamics of knowledge creation; (ii) institutionalization of the science–policy interface; and (iii) pathways of influence of expertise on policy development. Based on the current literature, we started with the assumption that climate change would not (yet) be widely and explicitly reflected in flood risk governance [9]. However, our case studies show that there is a mixed, though increasing impact of climate change on flood risk governance in the three selected Alpine regions. Climate adaptation has become an important issue of flood policy in all three case study regions and this shift has been strongly supported by evidence-based arguments (see Table 1 for a comparative overview of the case study findings).

Concerning our first analytical dimension, i.e., the *dynamics of knowledge creation*, our case studies indicate that there has not yet been scientific closure on the impact of climate change on flood hazards. In the 1990s and early 2000s, studies were often afflicted by methodological problems and data gaps. This created dissent in the Austrian and German scientific communities, when the findings of some studies were criticized. In the last 10 years, many of these problems were solved due to more comprehensive historical data and better climate and hydrological models. Nowadays, there is a broad consensus among scientists in all three regions about the potential impact of climate change. Nevertheless, epistemological and methodological uncertainties remain, in particular with regard to major flood events, which are difficult to predict in the Alpine region. This can, inter alia, be seen when comparing insights derived from historical trend analyses with those of modelling approaches: For Austria and Germany, those insights converge, however, content-wise they point to opposite

directions (with Austrian studies seeing few and German studies seeing numerous indications of climate change signals). In contrast to that, for Switzerland, historical trend analyses show a weak signal, whereas models indicate a strong(er) influence, particular concerning heavy rainfall events. This strong variation on the regional and sub-regional level in both the observed and expected changes in flooding is somewhat surprising, but can be explained by the topographic effects of the Alps and the countries' different susceptibility to cyclone tracks (cf. Ref. [100]).

The case studies further indicate that scientists adjusted their expertise to political expectations: When climate change emerged prominently on the political agenda, the production of application-oriented expertise followed suit. This can be interpreted as an effort of experts to ensure that they remain a valid source of knowledge for decision-makers. Strassheim and Kettunen [131] and Van Enst et al. [132] denote this phenomenon with the concept of "policy-based evidence"—as compared to the classical notion of "evidence-based policy" with a more linear understanding of science–policy interactions.

However, the case studies still show some variance in the degree to which experts lend themselves as advisors for policy: In Austria and Germany, experts provided quite explicit recommendations, with German scholars recommending the introduction of an explicit climate change instrument, i.e., the Climate Change Factor, and Austrian scholars explicitly recommending not to implement such an instrument. In Switzerland, scientists provided more cautious and partly ambiguous recommendations, thereby leaving the final decisions to policy-makers. These divergent patterns between countries are notable in various respects: On empirical grounds, it is interesting to see that even though the expert communities of the three regions regularly exchange views and experiences, we observe convergence neither on the level of recommended policy instruments nor on the level of advisory styles. With regards to the latter, the comparative science–policy literature would have made us expect that the three regions, because they share a similar neo-corporatist policy culture, would show similar patterns of interaction between science and policy-making [42,133–135]. Similarities can, for sure, be seen with regard to knowledge actors and organizational formats (see Table 1), however, not for the way scientists wriggle into politics.

With regard to the second analytical dimension, i.e., the *institutionalization of the science–policy interface*, we find that these are characterized in all three regions by a rather narrow set of actors consisting of scientists, public administrators, and political decision-makers. In the last decade, the scientific networks expanded and now include hydrology, meteorology, and spatial planning. Scientists play an important role in particular for risk assessments, while professionals and practitioners become more important when the discussion turns to risk management. In all three regions, we see a close cooperation between scientists and policy-makers, often including double roles and changes of roles. This is also reflected in hybrid platforms of knowledge exchange: In Austria and Switzerland, hybrid platforms mostly consist of scientists, administrative officials and political decision-makers, while in Germany those platforms are largely confined to scientists and administrative officials. The degree of institutionalization of those exchange platforms varies in the three regions, with the highest level of institutionalization in Switzerland, followed by Southern Germany, and the prevalence of more ad hoc, often event-driven, formats in Austria. Overall, there is a broad spectrum of science–policy interfaces: First, we have highly specialized organizations and research frameworks with a long continuity, such as ÖWAV in Austria, KLIWA in Germany, and PLANAT in Switzerland; second, we have specific research programs to generate focused research, such as FloodRisk in Austria and the pilot program on climate adaption in Switzerland, which are more event-driven and temporary; and finally, we have short-term contract research for ministries that deliver studies on the impact of climate change, such as the Blöschl study in Austria.


**Table 1.**Comparative overview of the case study findings.


**Table 1.** *Cont.* management

Concerning our third analytical dimension, i.e., the *influence of expertise on policy development*, our case studies show that experts had an important influence on flood policy-making in Austria, Southern Germany and Switzerland and that flood risk governance in these regions is to a great extent evidence-based. The close relationship between scientists and policy-makers described above creates policy frames that, by identifying the potential impact of climate change on flood hazards and specifying policy solutions for adaptation, constructed a coherent story around the "problem of future flood risks". Policy-makers in all three regions to a certain extent depend upon scientific expertise to act, which provides experts with an important source of influence in the policy-making process. At the same time, our case studies also reveal that policy-makers set limits on the influence of those experts: Scientists had to "tailor" their knowledge to the new political issue of climate adaptation to get their expertise recognized. This can most explicitly be seen using the example of the Climate Change Factor in Southern Germany, where a generally high political sensitivity for climate change questions led scientists to recommend the Factor in spite of a very uncertain scientific basis.

Against the above, we find that the influence of experts on flood risk governance can best be understood as being "socially embedded" [131] or "contextually embedded" [36]. The notion of "social embeddedness" [131] emphasizes that "expertise and evidence [need to be seen as] socially embedded in authority relations and cultural contexts" (p. 259). In the case of flood risk management, this embeddedness most prominently played out in the flood paradigms that the three regions embrace: Switzerland was an early mover towards integrated risk management ("risk-based spatial planning"), a shift primarily caused by the extreme flood event in 1987. Climate change became an issue in the early 1990s, but the explicit consideration of climate adaptation in the form of planning for flood overload was driven by the 2005 extreme event. Switzerland's risk-based approach to the management of natural hazard is also reflected in the recently updated PLANAT strategy [118]. Entitled "Management of Risks from Natural Hazards", the strategy not only differs semantically from the previous strategy "Protection against natural hazards" [136]. Against the likely climate-related increase in hazard potential and the accumulation in vulnerable assets, it importantly embraces the need to "adapt in a timely manner to changes in conditions" (p. 6), as a core principle of resilience-oriented flood risk management strategies [4]. In Germany, the flood community discovered climate change as a problem also in the early 1990s, which is embedded in a national debate in which climate change is of high importance. Based on the precautionary principle, Germany introduced an explicit instrument (CCF) for hazard mitigation already in 2004, yet the CCF was still embedded in the classical safety paradigm of technical flood protection. More integrated approaches were developed later on in Germany, and with that, the CCF took a back seat and became one among many other measures. In Austria, the new integrated paradigm of flood risk management was adopted somewhat later than in Germany and Switzerland, specifically after the extreme flood events in 2002 and 2005. The new flood paradigm has been consistently implemented since then through the avoidance of new risk and the reduction of existing risks. The discussion about the impact of climate change emerged also relatively late in 2006/2007 in Austria, and until today has yielded only a few specific climate change-related measures.

#### **5. Conclusions**

Our analysis of science–policy interactions in the flood risk management sectors of Austria, Southern Germany and Switzerland showed that there are clear imprints of climate change on the sectors' governance arrangements and dynamics, though we also observed marked differences between the three countries. Climate adaptation has become an important issue of flood policy in all three countries, and this policy shift has been strongly supported by evidence-based arguments. However, uncertainties, inter alia due to lack or incompleteness of trend data and/or discrepancies in flood hazard projections, still remain. This provides challenges for policy-making, both with regards to the question of how to technically deal with flood risk in the face of imminent but still unpredictable climate change impacts (an issue that our social-scientific paper cannot address) and with regards to the question of which role to assign to science and expertise and how to organize the science–policy interface

in an effective way. On the last-mentioned question, our comparative analysis of science–policy interactions in three Alpine regions provides insights that might be of interest for decision-makers in other countries as well.

Our reconstruction of the temporal *dynamics of knowledge creation* showed that, over the last three decades, one sees an increasing scientific agreement about the possible impacts of climate change on the frequency and magnitude of floods. However, as our case studies demonstrated, the progression of knowledge in the field has been far from linear; we rather observed some marked shifts in the construction and interpretation of scientific knowledge, partly due to methodological innovations or the availability of new data sets, partly due to changes in the political environment (such as increased public sensitivities for climate change). Against the background of those observed discontinuities in knowledge dynamics, it seems to be wise for policy-makers to opt for adaptation strategies that are flexible and robust enough to account for changes in flood hazards in the future. Ideally, decision-makers should identify and implement "no-regret" or "low-regret" adaptation measures that are effective under different climate change scenarios and that can be modified in the face of new scientific evidence. Incorporating the potential effects of climate change into flood design guidelines by an adjustment for uncertainties or a climate change allowance like in Austria and Germany is one example for that kind of adaption. Shifting flood policies from structural flood defense to a broader portfolio of vulnerability-oriented flood risk management measures like in Austria and Switzerland is another complementary option for adaptation.

Our comparative analysis of the *institutionalization of the science–policy interfaces* revealed a broad spectrum of approaches with some conspicuous differences between the three case study regions (e.g., with Switzerland showing the most formally structured science–policy landscape and with Austria hinging more towards informal, ad-hoc setups). What all three regions have in common—and what we see as prerequisite for effective knowledge translation—is that there are firmly established institutions that enable an exchange between scientists and policy-makers. Otherwise, there might be a danger that in the disaster-driven sector of flood risk management, scientific arguments only get heard in the direct aftermath of severe flood events and get sidelined as soon as the public attention has shifted to another exigent topic.

Finally, our study on the *influence of expertise on policy making* confirmed that scientists and other experts had quite a strong role to play in the formulation of flood risk policies and the implementation of flood risk management strategies and measures. In the three case studies, we could identify a large number of partly divergent factors that might help to explain the influence of expertise, however, there are also some overarching insights that might be singled out as success factors. The flood risk governance arenas in Austria, Southern Germany, and Switzerland are marked by various venues in which scientists and policy-makers can interact on a regular basis. The most long-lasting—and probably also the most influential—venues are the ones with diverse membership, that is, venues in which scientists, administrators, planners, and ideally politicians come together on equal terms. Given the variety of challenges that flood risk governance is facing, it seems to be desirable to have a spectrum of some more 'political' science–policy arenas and some more 'scientific' arenas. Eventually, what has proven helpful is the publication of policy-targeted scientific reports on the impact of climate change on floods, like the sectoral adaptation studies in Austria, KLIWA studies in Southern Germany, and the Climate Change Scenarios for Switzerland.

**Author Contributions:** This research article was developed in a collaborative effort based on the following author contributions: conceptualization, R.N., L.L., M.P.J. and M.P.; methodology, R.N., L.L., M.P.J. and M.P.; validation, R.N., L.L., M.P.J. and M.P.; formal analysis, R.N., L.L., M.P.J. and M.P.; investigation, R.N., L.L. and M.P.J.; writing—original draft preparation, R.N., L.L., M.P.J. and M.P.; writing—review and editing, R.N., L.L., M.P.J. and M.P.

**Funding:** This research was funded by the AUSTRIAN CLIMATE AND ENERGY FUND, grant number KR14AC7K11809.

**Acknowledgments:** The open access publishing was supported by BOKU Vienna Open Access Publishing Fund.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **Appendix A**


**Table A1.** List of expert interviews conducted in Austria, Germany and Switzerland.

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

#### **Md Ruknul Ferdous 1,2,\*, Anna Wesselink 1, Luigia Brandimarte 3, Kymo Slager 4, Margreet Zwarteveen 1,2 and Giuliano Di Baldassarre 1,5,6**


Received: 25 April 2019; Accepted: 10 June 2019; Published: 13 June 2019

**Abstract:** Bangladeshi people use multiple strategies to live with flooding events and associated riverbank erosion. They relocate, evacuate their homes temporarily, change cropping patterns, and supplement their income from migrating household members. In this way, they can reduce the negative impact of floods on their livelihoods. However, these societal responses also have negative outcomes, such as impoverishment. This research collects quantitative household data and analyzes changes of livelihood conditions over recent decades in a large floodplain area in north-west Bangladesh. It is found that while residents cope with flooding events, they do not achieve successful adaptation. With every flooding, people lose income and assets, which they can only partially recover. As such, they are getting poorer, and therefore less able to make structural adjustments that would allow adaptation in the longer term.

**Keywords:** flooding; erosion; coping; adaptation; Jamuna River; Bangladesh

#### **1. Introduction**

The image of Bangladesh as a country that is adapting well stems from its long history of living with floods. Bangladesh is a riverine country and one of the most flood-prone countries in the world [1–3]. The country is in the largest delta of the world, which developed, and is continuously changing, through processes of sedimentation and erosion by the three mighty rivers Ganges, Brahmaputra, and Meghna. Bangladesh is also a very densely populated country with most people depending on agriculture and fisheries for their livelihoods [4]. Most of the population lives in floodplain areas, with varying degrees of exposure to riverine flooding and riverbank erosion [5]. In Bangladesh, floods can be both resources and hazards. In the monsoon season, 25–30% of the floodplain area is typically inundated by the so-called normal floods (*borsha*). This is considered beneficial as flooding increases the fertility of the land by depositing fresh silt on the soil and replenishing soil moisture without doing too much damage to property or disruption to traffic and commerce. However, an abnormal flood (*bonna*) is considered a hazard because of the damage it does to crops and properties, and the threat to human lives [6,7]. Riverbank erosion is another hazard in times of flood and occurs almost in every year, whether the flood is normal or abnormal. Abnormal floods and riverbank erosion not

only cause substantial damage to inhabitants' agricultural lands, crops, and properties, they also have wider socio-economic consequences, for example a decline in agricultural production [8] and a steady flow of migrants to urban areas [9–11]. Each year several thousands of people become temporarily or permanently homeless and/or landless and take refuge to nearby embankments or on neighbors' or relatives' land. The extremely poor people who live on the islands in the wide (up to 16 km) rivers (called chars) are most exposed to and affected by flood hazards and riverbank erosion. In the 1988 flood, more than 45 million people were displaced and over 3000 died [12,13]. Flood mortality rates have declined since 1988 because of better institutional support during floods [14]. Yet, the socio-economic disruption caused by flooding events is still considerable [3,15].

"Living in a low-lying and densely populated country on the front line of climate change, Bangladeshis are taking a lead in adapting to rising temperatures and campaigning to limit climate change. Bangladeshis will keep their heads above water, but at huge costs" [1]. This back cover summary of the recent book "*Bangladesh Confronts Climate Change: Keeping Our Heads Above Water*" [1] portrays a common, positive image of Bangladesh as being able to respond successfully, or adapt, to climate change (see also [12,16–19]). However, it also indicates that this response comes with huge costs. Floodplain inhabitants use a wide range of tried-and-tested strategies to cope with the conditions during and after the flooding. However, repeated exposure to flooding events often means impoverishment.

Only a few authors have paid sufficient attention to this effect, which can be seen as a sign of maladaptation [20]. Most research done on people's abilities to cope or adapt is based on qualitative statements from floodplain inhabitants, without looking at the cumulative effect of floods over a longer time period (i.e., across decades). As time series of quantitative data to substantiate these claims are not available from secondary sources, estimates of flood damage often rely on unverifiable proxies [21]. More specifically, various studies assert that people are getting poorer because of recurring floods [12,22,23], but most of these do not provide quantitative substantiation. One of the processes that may lead to impoverishment is diminished agricultural production in times of disastrous floods [8]. Rural people are most affected by these floods; they are also suffering from the persistent effects of labor market disruption and income deficiency [24]. Moreover, they must spend a large portion of their income on food and repairing or constructing new houses, as their old houses are often ruined by floods and riverbank erosion. Their savings are reduced to zero and the other necessary expenses to survive are only increasing their burden [25].

A few studies quantify the average household losses due to severe floods and the associated riverbank erosion. Chowdhury [26] presents a figure of 47 and 88 USD/year for the years 1986–1987, in two locations. Ten years later Thompson and Tod [27] estimate losses to homesteads to be 132 and 190 USD/year for the 1988 and 1991 floods, again in two locations. The most complete figures are given by Paul and Routray [3], who estimate income losses at 82 and 108 USD/year and asset losses 191 and 257 during the 2005 flood, again in two locations. Combined with direct loss of assets in floods, and loss of the key productive asset (land) to erosion, severe floods in the chars make vulnerable households and the community poorer [27]. Finally, they are falling into debt and impoverishment [12,22,23,28]. Further indebtedness leads to even more drastic regimes of "eating simpler food", increased malnutrition, increased levels of disease, increased numbers of wives left to feed children alone while their husbands are temporarily migrated for work, increased sales of remaining assets, increased rural landlessness, increased migration to the cities, and increased vulnerability to future floods.

Moreover, only a few authors investigate the distributional effects of flood impacts. Paul and Routray [3] show how the adoption of a particular set of adjustment strategies depends on people's socio-economic circumstances, such as education, income, and occupation. Islam et al. [29] look at adjustment strategies against flood and riverbank erosion of the char inhabitants of the Jamuna River. They also find that household's ability to cope with flood and river erosion depends on people's socio-economic and environmental conditions. Hutton and Haque [28] assert that impoverishment and marginalization in part reflect inequitable access to land and other resources. The likelihood of

impoverishment of the household is further increased not only by social and demographic factors (including gender, education, health and age), but also by underlying economic and social relationships which can increase human vulnerability to risk. Indra [30] investigates the forced migration due to riverbank erosion (and hence loss of land) in the Jamuna floodplain, and concludes that most people displaced by riverbank erosion are already poor and disempowered before being uprooted by the shifting channels of the Jamuna River.

This paper critically assesses the idea that Bangladeshis have a good adaptive capacity to changing environmental conditions. To this end, it examines the socio-economic effects of repeated floods and riverbank erosion on rural households, and assesses whether and how different households deal with this flooding. The aim is to uncover whether and how households succeed in adapting to flood conditions, as opposed to those who are merely coping. As defined by Berman et al. [31] coping refers to immediate responses to events, while adapting prepares households for expected future events. These definitions are extended as follows: coping includes long-term adjustments where the net result can be a decline in socio-economic conditions, while adapting means that people continue to live as well, or better than before, and are as well, or better than before, able to cope with future events. Both coping and adapting require the mobilization of a variety of technologies and social or institutional changes, which is defined as adjustments. The actions and strategies for adjustment may be the same, but have a different result: coping or adapting. In general, coping strategies are those that are possible within the current settings, and adaptation strategies are more likely to involve more fundamental changes in the type of livelihood activity or location [32].

#### **2. Case Study**

To understand how Bangladeshi cope with flooding, this study focuses on the floodplain area along a 30 km reach of the Jamuna River in the north of Bangladesh. The study area includes parts of Gaibandha district and parts of Jamalpur district (Figure 1). The total surface is about 500 km<sup>2</sup> and the total population is approximately 0.36 million people [4,33]. This is a unique case study since three adjacent areas are inhabited by people with similar socio-economic conditions, but different levels of flood protection and exposure to riverbank erosion (Figure 1).

**Figure 1.** Bangladesh map with study area [34].

To investigate the effect of floods and riverbank erosion on livelihood conditions, the area is divided into three socio-hydrological spaces [34]. The socio-economic situations are also classified based on income levels and assets (Section 2.2).

#### *2.1. Socio-Hydrological Spaces*

Figure 1 shows the braided riverbed, which includes many inhabited river islands (*chars*) that get flooded with varying frequency (some every year, some only with severe floods). The area to the west of the river is protected by a human-made embankment, while the area to the east of the river only has a natural levee deposited by the river. These different physical conditions create different flood protection levels, which in turn give rise to different socio-economic responses and conditions. To reflect this socio-spatial differentiation, the authors have elsewhere explained and motivated the classification of the study area into three socio-hydrological spaces, abbreviated to socio-hydrological spaces (SHS) [34] (Figure 1). These spaces are briefly described below.

SHS1 is an area protected by an artificial levee (the Brahmaputra Right Embankment, BRE) constructed in the 1960s to limit flooding and increase the agricultural production of that area. This area is protected from regular annual flooding, but the maintenance of the BRE in the study area has been sporadic, and breaches occur during abnormal floods, causing catastrophic floods and damages [35]. Also, two small rivers Ghahot and Alai inundate some parts of the area, and other parts are frequently inundated with excess rainwater, due to their low elevation and limited drainage capacity. The total area is about 74 km<sup>2</sup> with a population of approximately 111,000 [4].

SHS2 is the floodplain within the embankment on the west bank and the natural levee on the east bank. In SHS2, flooding occurs most frequently, essentially every year. This area includes extensive chars (river islands), where multiple channels crisscross within the outer boundary of the riverbed. These chars can shift in space due to continuous processes of deposition and erosion of river sediments. The stability of the chars depends on their age. Some older chars have higher elevations than the areas in SHS1 and remain dry during flood conditions. If a new char develops, homeless people analyze the stability of the new char and after 2–3 years start activities such as farming or living on the char. When the age of the char exceeds about 20 years it is called a stable char [36]. Nevertheless, all chars can shift during the flooding events. The total area is about 246 km<sup>2</sup> with a population of approximately 104,000 [4].

SHS3 is the area on the east bank which is without any man-made flood protection. Flooding occurs here more frequently than SHS1, around once in two years [34]. This area is sometimes flooded by adjacent small rivers, in this case the Old Brahmaputra and Jinjira, as well as by excess rainfall. Riverbank erosion is also prominent in this area. Inhabitants take the initiative to build small spurs and bank protection made from bamboo and wood, to try to stop erosion. However, while these encourage sedimentation at a local scale, they are not sufficient to stop large scale erosion. The total area is about 174 km<sup>2</sup> with a population of approximately 146,000 [4,33].

Decadal data from 1961 to 2011 show that the population density has increased from 300 persons/km2 to about 800 persons/km2 over the whole study area [4,33]. However, population density in the three spaces differs substantially. In SHS1 it has always been higher than in the SHS2 and SHS3 (Figure 2). The density in SHS1 has increased from 600 to 1500 person/km2, a rate of 18 persons/km2/year. In SHS2 population density has increased from 200 to 400 person/km2, a rate of 4 persons/km2/year, and in SHS3 has increased from 300 to 800 person/km2, a rate of 10 persons/km2/year.

**Figure 2.** Population density of the study area in the period 1961–2011.

#### *2.2. Socio-Economic Conditions*

Data on socio-economic activities are sparse. The only historic economic data (from 1971) pertain to the number of commercial establishments developed for non-farming activities. These data are aggregated by districts that are larger than the SHS considered here. An establishment is defined an enterprise or part of an enterprise that is situated in a single location and in which only a single (non-ancillary) productive activity is carried out or in which the principal productive activity accounts for most of the value added [37,38]. Table 1 shows that the socio-economic situation in Gaibandha district was a little better than in Jamalpur district in 1971, but they were almost the same in 2013. The growth in the number of establishments changed dramatically around the year 2000 (Table 1). To get information about the company size and its impact on the local economy, this research uses the number of employees as a proxy. This information is only available for 2003 and 2013 (Table 1).


**Table 1.** Non-farm economic activities: number of establishments and employees [37,38].

Table 1 seems to suggest that long-term economic trend in the area is upwards. Yet, agricultural farm sizes decreased in the study area since the 1960s (Figure 3). This is a potential sign of impoverishment. Figure 3 shows that the number of large farms decreased from 10% to only 1% during the period 1960s to 2016 and the number of landless households increased from 10% to about 37% during the same period.

**Figure 3.** Number of non-farming establishments and change in agricultural farms in the study area.

In the study area, more than 80% of households rely principally on farm incomes, so agricultural income and assets are key indicators for socio-economic status. The existing Bangladesh Government classification of farm sizes only accounts for land holding size [39]. However, households have other assets (homesteads, equipment, cash or jewelry) that help them survive in case of flooding or riverbank erosion. Moreover, differences in average income affect how households can cope with such events. Data of this research include current annual income, annual expenditure and total wealth (see Section 3 for details about data collection). Using these data, the surveyed households are divided into five socio-economic classes that combine current wealth and income (Table 2). To increase the statistical significance of comparisons between classes, the class boundaries are chosen in such a way that each class contains an almost equal number of households.



By comparing the current distribution of socio-economic classes in the three SHS, see Section 2.1 and Figure 1), one can see remarkable differences (Figure 4a). In SHS1, 10% of the households are poor and 35% are rich, while in SHS2 it is just opposite. SHS3 takes an intermediate position, with 15% poor households and 25% rich. Using the Government farm size classification for comparison (Figure 4b), it is observed that more than 35% of households are landless in the study area and only 1% of households are large (Figure 4b). Most of the landless farmers are found in the SHS2 as they are experiencing much more land erosion than SHS1 and SHS3.

**Figure 4.** Current socio-economic status of the households in the study area by socio-hydrological space. (**a**) Socio-economic status (author's classification); (**b**) Distribution of farm sizes (BBS classification).

#### *2.3. Adjustment Strategies in the Study Area*

Inhabitants of the study area employ a range of strategies to adjust with flooding. These can be categorized into those that allow survival in times of flood and those that allow long-term living with floods and riverbank erosion.

To survive during flooding, people eat fewer meals, borrow money or take a loan, or sell their labor cheaply in advance [25]. If necessary they also sell their land, livestock, housing materials and other personal belongings, including jewelry and household goods [27]. In char areas (SHS2, Figure 1) the inhabitants only leave their homes when their lives are at risk. When high floods erode their land, they dismantle their houses and transport them to another char which is less (or not) affected by flooding and erosion [40]. If necessary, people take shelter on the embankments, with their livestock, in the hope that they might return in the near future to the re-emerged land, where they have property rights [15]. In most cases, these hopes are not fulfilled because it may take decades for land to re-appear [41]. Others move to a nearby relative's or friend's house or migrate temporarily to other districts looking for temporary work [9,30,42–45]. It is noted that main roads and railways are built on embankments to raise them above high flood levels. Rural roads and paths between settlements generally follow the highest land available and are usually also built on embankments to raise them above normal flood levels. This somewhat limits disruptions during floods, and provide an emergency refuge.

To structurally improve their chances of survival during a flood, inhabitants adjust their homes. In the char areas (SHS2), houses are built in a way that they can be easily dismantled [30,42]. In SHS1 and SHS3, homesteads are generally built on natural elevations, or on artificial earthen mounds, the height of which is determined by local experience of previous high flood levels [17]. The plinth of houses is further raised by digging earth from local depressions. Especially on the chars (SHS2) people also build platforms inside their homes to take shelter using bamboo, straw, water hyacinth, and banana stalks during abnormal flooding years [12].

To improve livelihoods, rural inhabitants have developed farming practices that are adjusted to the height, duration, and timing of normal floods, i.e., that commence and recede in time and attain normal height [6,45]. Rasid and Mallik [46] observe that the most common adjustment of rice cropping to the uncertainties of the flood regime is evident from the practice of intercropping two rice varieties, *aus* and *aman*, together. This measure ensures that at least flood-tolerant *aman* would be secured during an abnormal flood regime, even if the flood-vulnerable *aus* is lost or damaged. During a normal flood regime both *aus* and *aman* would succeed, often resulting in a bumper crop. Farmers have made a careful selection of the best adjusted varieties of rice over the centuries, to enable them to face floods [16], and they select other crops to sow off-season to fit the land elevation [12]. Other adjustment strategies include early harvesting of *aus* in case of excess or early flooding, planting older and taller seedlings on lands liable to repeated flooding, re-transplanting salvaged seedlings, protection of rice plants from water hyacinth by floating bamboo fences, and the post-flood cultivation of lentils, pulses, mustard, as well as winter vegetables and wheat [46].

While temporary migration can be a survival strategy (see above), migration can also be a long-term adjustment to increase household incomes and livelihood security [45]. Very few people move permanently to towns and cities [10,28,41]; the majority of flood affected people try to stay near their homes because kinship ties mean that local inhabitants help each other in crisis situations [2]. In that case, migration often concerns one or more family members who migrate, usually men. This can be on a seasonal basis, to other rural areas as agricultural laborer, or to urban areas as unskilled laborer or rickshaw driver, or more permanently to work in factories. Permanent migration out of the area of origin is rare, and mostly related to the loss of land due to riverbank erosion, while loss of livestock and crop failure more likely leads to temporary migration [10,30]. However, migration often comes at a cost: working conditions tend to be challenging and dangerous. People who are forced to migrate permanently from rural to urban areas often end up in slums with difficult living conditions.

#### **3. Data and Methods**

Primary data and secondary data were collected to explore how households in the study area adjust to regular flooding and riverbank erosion, as well as the long-term socio-economic effects of these adjustments.

Household surveys and focus group discussions (participants selected from the previously surveyed households) were performed during the dry seasons of 2015 and 2016. The principal set of primary data consists of 863 questionnaires dealing with several themes: general information (location of settlement and agricultural land, main occupation, age, income and expenditures, wealth and origin of the households), information on different flood experiences (depth of floods, frequency, duration, flood damages, effects on agricultural income and expenditures, other adjustment options such as migration) and experiences with river erosion (frequency, damages, migration, adjustment options etc.). For several of these aspects the respondents were asked to compile a historical record going back to 1960, which is why the households headed by older men or (occasionally) women were selected. This is a limitation of the study. Many old men were surveyed since they experienced many flooding events. Although other members of the family were typically present during the surveys, this choice may have introduced a selection bias. 12 focus group discussions were also set up in the study area to validate and contextualize survey data.

Since the respondents were asked to recollect their flood experiences going back to 1960, inaccuracies due to memory loss is one of the limitations of the household survey data. When asked to recollect major flood events and details of flood damages, respondents could easily remember events in the last two decades, but they were not so confident about the floods before 1980s. To handle this issue, the questionnaires were filled in in the presence of other family members, who helped the primary respondent to remember details about the past. The surveyors also used references to remarkable years, such as the year of independence of Bangladesh in 1971, or the construction period of embankments, to connect the respondents with years of major flood events.

To further check the reliability of the respondent's flood memory, the surveyed data are compared with both Government data and published journal articles. Based on flood duration, exposure, depth, and damage, the Flood Forecasting and Warning Centre (FFWC) of Bangladesh Water Development Board (BWDB) classifies flood events into three categories: normal, moderate, and severe. According to this classification, Bangladesh has experienced 9 severe floods since 1950s, namely in 1954, 1955, 1974, 1987, 1988, 1998, 2004, 2007 and 2017 [47]. However, according to household surveys a total of 33 major floods were experienced in the study area since the 1960s. A plausible reason for this difference is that the study area is very close to the Jamuna River, so households are facing huge damages almost every year. As such, almost all floods have severe consequences in the study area.

In the literature different lists of severe floods are presented. Hossain et al. [48] report major flood events in 1954, 1955, 1956, 1962, 1963, 1968, 1970, 1971, 1974, and 1984. Brammer [17] mentions that 16 major floods struck Bangladesh in the years between 1950 and 2000, with additional floods in 1977, 1980, 1987, 1988, 1998, and 2000. The respondents in the study could easily remember those flood events mentioned by Brammer [17] except the flood of 1977. According to Paul [16], Bangladesh has experienced 28 major floods in the past 42 years (1954–1996), of which 11 were classified as "devastating" and five as "most devastating". Summarizing all published flood information in the journal articles, 30 major floods were observed during 1962–2017. This flood count is very close to the numbers that the study collected from the household surveys. This suggest that the memory of the respondents is good enough to compile a record of historical flood events in the study area.

For the data on land loss due to riverbank erosion, all respondents were confident that they could remember the actual losses and the years they occurred. According to them, the loss of land is never forgotten. Respondents claimed to remember very well how much lands they had in 1960s, how much land had eroded since then, and in which year. They said that it is possible to recover from flooding, but that it is not possible to recover losses from riverbank erosion.

#### **4. Results**

Statistical analyses are performed for the whole study area, for each SHS (Figure 1), and for the different socio-economic classes (Table 2). This is done through statistical analysis of single variables and correlation between variables. ANOVA test (p < 0.05) and Chi-square tests are also performed to verify the significance of the outcomes. Statistical test summaries are given in the Electronic Supplementary Material (ESM).

#### *4.1. Adjustment Strategies*

Relocation means to settle permanently in another place with the whole household. Households in the study area do not normally relocate due to flooding events. In case of severe flooding, some people may temporally evacuate over a short distance, to return to their houses after the flood. When households were asked about permanent relocation, many of them have expressed that "it is very hard to live here but we are born here, our forefather lived here then why should we leave?" They relocate permanently only when they face riverbank erosion. About 49% of households relocated during the entire period 1962–2016, and more than 95% of households claimed that riverbank erosion is the main reason of relocation. As expected, the maximum number of relocations occurred in the char areas, i.e., SHS2 (Figure 5a). By considering the socio-economic status, the maximum number of relocations occurred in poor (76%) and moderately poor (70%) households (Figure 5b). More than 90% of respondents relocated within 5 km from their previous locations and about 40% poor and 30% moderately poor households are considering relocating again because of riverbank erosion. A statistical analysis was performed and it was found that there exists a significant difference in relocation by the respondents between SHS and socio-economic groups (with α = 0.05).

**Figure 5.** Households relocated due to riverbank erosion. (**a**) Information of origin of households. (**b**) Household relocated in lifetime by economic classes.

Only 5% of households mentioned that members had to change their occupation due to flood (Figure 6a). They were mainly from the poor and moderately poor classes (Figure 6b). They changed their occupation mainly from farmer to day laborer, rickshaw puller, and fishermen. The number of households reporting a change in occupation is low because alternative employment opportunities are limited. More than 90% of households mentioned that they have nothing to do during flood events and a scarcity of work arises during those periods. Normally, they look for day laborer work during flood events. A statistical analysis was performed and it showed that there is no significant difference in change in employment by the respondents between the SHS but there exists a significant difference in change in employment by socio-economic groups (with α = 0.05).

**Figure 6.** Change in occupation of households in the study area. (**a**) Change in occupation by SHS. (**b**) Change in occupation by economic class.

The results show that most households do not change their cropping pattern due to flooding. Using indigenous knowledge, they developed a unique technique to cultivate rice which they follow every year. However, in severe floods this technique does not work and they lose the whole harvest. About 20% of households cultivated fast growing crops after the severe flood in 1988, and about 15% of households did so after the flood events in 2007 and 2015. A few farmers mentioned that they keep their lands fallow during the flood season to avoid losses.

In the study area, inhabitants raise the plinth of their houses to prevent flood water to get into their home. Only 11% of household respondents have also raised their homestead platform. They all are from rich and moderately rich socio-economic classes.

Inhabitants of the study area are undertaking some other adjustment strategies such as storing food for flood event, constructing houses with movable materials, planting trees to avoid erosion. A few of them are also temporarily going to big cities to earn some money as day laborers to overcome the flood losses.

#### *4.2. Impacts of Flooding and Riverbank Erosion*

By analyzing the negative impacts of flooding and riverbank erosion, it was found that the average income is decreasing over time. About 98% of household experienced decrease in income in every year. Households claim to lose about 90% of their monthly income during severe flood events (Table 3). A statistical analysis was performed and it showed that there is no significant difference in income loss of the respondents between the SHS and by socio-economic groups (with α = 0.05).


**Table 3.** Decrease of income due to floods in the study area.

Moreover, 71% of households face an increase in monthly expenses in times of floods (Table 4). Such an increase is about 60% of the monthly income with little (but not significant, see ESM) differences between SHS and by socio-economic groups (Table 4). Decrease in income loss and increase in monthly expenses due to flood and riverbank erosion are generating the indebted situation for the households in the long run.


**Table 4.** Increase in expenses due to flood in the study area.

#### *4.3. Impoverishment*

Land loss caused by river erosion is the major loss for the inhabitants of the study area. Land loss data were used to better analyze whether impoverishment differs by SHS and socio-economic classes. On average, 0.9 ha of land per household was lost between 1962 and 2016. As expected, the inhabitants of SHS2 had the highest losses with 1.4 ha per household (Figure 7a). About 10% of households lost all their land since 1962 and 20% of households lost more than 80% of their land. Analyzed by socio-economic status, it is found that the poor households have lost the most, i.e., 1.5 ha per household (Figure 7b). Moreover, poor and moderately poor households lost the highest proportion of their land, many ending up with no land at all (Figure 7c,d). The households of SHS2 have lost about 80% of their lands during this period. A statistical analysis was performed and it showed that these differences in land loss between the SHS and socio-economic groups are significant (with α = 0.05).

**Figure 7.** Total land loss by SHS and economic classes in the study area from 1962 to 2016. (**a**) Average land loss per household (in ha) by SHS; (**b**) Average land loss per household (in ha) by socio-economic classes; (**c**) Land loss by socio-economic classes and SHS; (**d**) % of land loss by socio-economic classes and SHS.

The survey data also include time series on agricultural crop loss, homestead loss, and other asset losses caused by flooding and riverbank erosion. In terms of total asset losses, households in SHS3 have lost the most, around 2000 USD per year, while the households in SHS2 have lost the least, around 1250 USD per year. Households in SHS2 are better prepared, since they live in the char area, and have less to lose because they are, on average, poorer. By analyzing asset losses by socio-economic status, one can see that poor households have lost most assets (more than 2000 USD). Most poor people today claimed to be richer back in the 1960s, but that they became poorer because of repeated asset losses caused by flooding the riverine erosion. Yet, it should be mentioned that no significant correlations between flood severities and reported losses were found (details are given in ESM).

The respondents were asked about their recovery processes to explore how they recover from losses caused by consecutive events. "Flood and riverbank erosion have snatched everything from us, and we become destitute", is one of the most commonly heard expressions in the char area (SHS2). People in the areas also like to ease their situation by believing that "Almighty Allah (God) has sent this flood towards us and so we have to accept it. This is our fate". Most respondents informed us that it is almost impossible to recover from the riverbank erosion. About 53% of the respondents mentioned that they could not recover at all from the combined losses of flooding and riverbank erosion (Figure 8a). Only 5% of the households recovered a little but not enough to get back to their previous position. They recovered partially with hard working as agricultural laborer, other day laborer employment, and by selling properties or taking loans to do business, cattle farming, or send a household member to temporarily migrate to cities to earn some money. Also, more than 80% of the households in SHS2 could not recover at all, which is highest among the SHS. Looking at socio-economic classes, one can see that more than 70% of poor and moderately poor households could not recover at all (Figure 8b). A statistical analysis was performed and it showed that there is no significant difference in recovery from loss by the respondents between the SHS but there exists a significant difference in recovery of loss by socio-economic groups (with α = 0.05).

**Figure 8.** Recovery of loss from floods and riverbank erosion in the study area. (**a**) Recovery of loss from floods and river erosion; (**b**) Recovery of loss from floods and river erosion.

These statistics do not tell the whole story, since they present averages. During the field survey a few exceptions were found to the overall picture sketched above. While most of the people suffer, a few rich households benefited from flooding and riverbank erosion by giving loans with very high interest to the flood victims. Statistics do not show the richness of individual experiences either, where each household has its own story to tell such as the one presented in Box 1. The situation of Mr Abdul Baki Khan is typical for some char dwellers, but the number of times his family moved is also quite exceptional compared to other households.

**Box 1.** An example of a life course of impoverishment in the study area.

#### **Story of Mr Abdul Baki Khan, an example of living with floods in the study area**

Mr Khan, who is 63 years of age, lives in a char of the Jamuna River in Uria union of Fulchari upazila of Gaibandha district. He is a landless farmer whose household consists of seven members. They organize their livelihoods around the river. They grow maize, jute and rice on a field by the Jamuna River char. The occurrence of regular flooding events, associated with eroding land, forces them to relocate every two to three years. Over the period 1978–2018, the family moved 13 times. Because of growing population in the area, finding possible places to live and farm has become increasingly difficult. In the past, he used to have access to some 9 hectares of land (large famer according to the agricultural land size), but today he has become landless. Now he is working on the lands of others and his eldest son is also working with him. His income from agricultural land is often not enough for the family, so they limit their daily needs. From morning to afternoon, he and his son go to the field and work hard. Despite these efforts, their monthly income is only about 50 USD. Their house is made of locally available materials with earthen floor, wood, paddy straw with tin on the top of the house. Mr Khan is an example of many people living in the Jamuna floodplain and its char. Many people like him used to be richer earlier, but they have become poorer because of flooding and erosion.

#### **5. Discussion and Conclusions**

Primary and secondary data were analyzed to reveal how the three SHS and the socio-economic classes are affected by flooding and bank erosion, and their inhabitants adopt adjustment measures to face these hazards and reduce or mitigate the related risks. This study quantified the costs of living with floods and showed that this can lead to general impoverishment in the long term.

While the country (Bangladesh as a whole) has become more prosperous [14], the per capita gross domestic product (GDP) of Bangladesh is accelerating rapidly [49], and the percentage of the population living below the national poverty line is decreasing [50]. Yet, the prosperity of the study area is below the national level, partly because growing GDP in Bangladesh is mainly urban [39]. According to the National Accounts Statistics for the year 2016 [51], the per capita annual income of the study area is estimated at 1160 USD compared with the national average of 1544 USD. Based on the surveyed data, it is estimated that per capita annual income in SHS1 at USD 1200, in SHS2 at USD 700, and in SHS3 at USD 1000 (below BBS estimates). It is also estimated that the percentages of households below poverty line is about 18% in SHS1, 32% in SHS2 and 15% in SHS3. Given that the main source of income in this area is represented by farming [34], the occurrence of flooding in the area highly impacts income generation. In SHS2, inhabitants only cultivate one crop per year and experience flooding essentially every single year. In SHS1 and SHS3, flooding occurs roughly once every two years. In all three SHSs, over 90% of the respondents experienced over the years an average income loss of about 80 USD which they attribute to flooding, which corresponds to about 85% of their monthly income. In addition, riverbank erosion is responsible for loss of land and assets in the area. According to the surveys, households from all three SHS and all socio-economic classes have lost land over the past 50 years. SHS2 people have lost up to 80% of their land and many households in the poorest socio-economic class stated that they have lost all their land. Inhabitants that relocated several times became poorer more quickly than those who did not need to relocate as frequently. Bank erosion victims tend to relocate to nearby land, hoping that they can recuperate their land in the near future. There are no many alternatives to recover from income, land, and asset loss. They either change their occupation from farmer to day laborer, fisherman, or temporarily migrate to cities to earn. However, recovery rate is very low: income loss and asset loss recovery is marginal even for rich class people. Furthermore, due to the lack of employment alternatives in the area, many people need to take loans with high interest during the flood season. In an attempt to increase their loss recovery capacity, people try to save some money for the flood season; however, their savings evaporate rapidly during the flood season when monthly expenses can be higher than incomes. As a result, they move towards impoverishment over time. Household respondents are affected by flooding and riverbank erosion in terms of income losses (Table 3), increases in expenses (Table 4) as well as land losses (Figure 7). The household respondents have also mentioned that it is almost impossible to fully recover the loss (Figure 8). This study analysis showed that although flood mortality rates in Bangladesh have been significantly decreasing over time [52,53], the costs of adjusting to flooding and riverine erosion are very high and can negatively impact the livelihood of local people. While households mainly blame flooding and bank erosion, there can be other socio-economic factors, such as a lack of investments in alternative types of employment, which also can significantly contribute to the impoverishment of the area.

Figure 9 depicts the definitions of coping and adapting introduced in Section 1. This research shows that households in the study area are coping (but not adapting) with flooding and riverbank erosion, since the net result of their adjustments is a decline in socio-economic conditions (see orange trajectory in Figure 9).

**Figure 9.** Coping or adapting? Deviation from a hypothetical trajectory of socio-economic growth (black line) caused by flooding and erosion, in case of: (i) no adjustments (red line), (ii) coping (orange line), and (iii) adapting (green line).

Loss of homesteads forces people to move to new places without any option and puts them in desperate situations. Despite these extreme living conditions, the char inhabitants do not leave this area because of various obstacles, the lack of available land elsewhere being foremost. They try to stay as long as possible in their home during high floods and in the case of erosion, dismantle it and transport it to another char which is less or not affected by erosion at that moment. While the situation in the char area (SHS2) is extreme, in every location of the study area people are losing income and assets, since they can only partially recover from flood events. Overall, they are getting poorer and therefore less able to make the further adjustments for the next flooding event. Although there are most likely other factors at play, repeated flooding and associated bank erosion contribute to overall impoverishment (in relative terms compared with national figures of economic growth) of the floodplain people of Bangladesh, despite their use of multiple strategies to respond to flooding and riverbank erosion. Due to the scale of works, the flood control measures constructed by the Government are not likely to ever be sufficient to prevent flooding or riverbank erosion. This research posits that indigenous adjustments, such as the adoption of different types of crops to varied flood depths, should be reinforced to reduce the negative impacts of flooding on agriculture and therefore slow down the rate of impoverishment in the area.

This research work does not only contribute to advance the knowledge about socio-hydrological dynamics in Bangladesh, but also provides more general insights for flood risk management in low-income regions of the world. There is an ongoing discussion about the need for a shift from hard (fighting floods) to soft (living with floods) approaches [54]. It has been argued that low-income countries such as Bangladesh should not implement hard engineering work to increase levels of structural flood protection, as it has been done in most Western countries, but stick to their traditional softer approach of living with floods [55]. Indeed, some of the polders that were constructed in 1970s had negative impacts on ecosystems and livelihoods and are currently being revised to re-establish a workable sediment and water balance [56]. However, the Jamuna case study demonstrates that living with floods has enormous costs and it prevents socio-economic growth in the areas. As such, there are no clear-cut answers to the question of how low-income countries should deal with flooding. This research argues that universal recipes do not exist. Instead, there is a need to find trade-offs between hard and soft options depending on the values given by local communities, experts, practitioners, and governments to environmental, social, and economic benefits and costs of alternative strategies. A better understanding of socio-hydrological dynamics, such as the one provided with the Jamuna case study, can help identify these trade-offs by shading light on both the positive and negative effects of living with floods.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2073-4441/11/6/1238/s1, Document S1: ESM Living with flooding in the Jamuna floodplain in Bangladesh: limits to adaptation, Table S1: EMS Literature review: Living with flooding in the Jamuna floodplain in Bangladesh: limits to adaptation.

**Author Contributions:** Conceptualization: M.R.F., A.W., L.B., K.S., M.Z. and G.D.B.; Data curation: M.R.F.; Formal analysis: M.R.F., A.W., L.B. and K.S.; Investigation: M.R.F., A.W., L.B. and K.S.; Methodology: M.R.F., A.W., L.B. and K.S.; Supervision: A.W., L.B., M.Z. and G.D.B.; Writing—original draft preparation: M.R.F., A.W. and L.B.; Writing—review and editing: M.R.F., A.W., L.B., M.Z. and G.D.B.

**Funding:** The research is funded by NWO-WOTRO grant W 07.69.110 "Hydro-Social Deltas: Understanding flows of water and people to improve policies and strategies for disaster risk reduction and sustainable development of delta areas in the Netherlands and Bangladesh". Giuliano Di Baldassarre is also supported by the European Research Council (ERC) within the project "HydroSocialExtremes: Uncovering the Mutual Shaping of Hydrological Extremes and Society", ERC Consolidator Grant No. 771678. Luigia Brandimarte is supported by the Swedish Strategic research programme StandUP for Energy.

**Acknowledgments:** Special thanks to Md Mahabubar Rahman, Md Enamul Haque, and Md Shahrier Islam for collecting household data from the study area. Special thanks to Abdul Baki Khan for allowing us to use his photo and life story in this paper. Special thanks to Engr. Md Waji Ullah, Executive Director, CEGIS for providing satellite images of Jamuna River for the study.

**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* **Adaptive Capacities for Diversified Flood Risk Management Strategies: Learning from Pilot Projects**

#### **Flavia Simona Cosoveanu \*, Jean-Marie Buijs \*, Marloes Bakker and Teun Terpstra**

Department of Technology, Water & Environment; HZ University of Applied Sciences, Het Groene Woud 1, 4331 NB Middelburg, The Netherlands; m.h.n.bakker@hz.nl (M.B.); t.terpstra@hz.nl (T.T.)

**\*** Correspondence: coso0002@hz.nl (F.S.C.); jm.buijs@hz.nl (J.-M.B.)

Received: 17 October 2019; Accepted: 10 December 2019; Published: 14 December 2019

**Abstract:** Diversification of flood risk management strategies (FRMS) in response to climate change relies on the adaptive capacities of institutions. Although adaptive capacities enable flexibility and adjustment, more empirical research is needed to better grasp the role of adaptive capacities to accommodate expected climate change effects. This paper presents an analytical framework based on the Adaptive Capacity Wheel (ACW) and Triple-loop Learning. The framework is applied to evaluate the adaptive capacities that were missing, employed, and developed throughout the 'Alblasserwaard-Vijfheerenlanden' (The Netherlands) and the 'Wesermarsch' (Germany) pilot projects. Evaluations were performed using questionnaires, interviews, and focus groups. From the 22 capacities of ACW, three capacities were identified important for diversifying the current FRMS; the capacity to develop a greater variety of solutions, continuous access to information about diversified FRMS, and collaborative leadership. Hardly any capacities related to 'learning' and 'governance' were mentioned by the stakeholders. From a further reflection on the data, we inferred that the pilot projects performed single-loop learning (incremental learning: 'are we doing what we do right?'), rather than double-loop learning (reframing: 'are we doing the right things?'). As the development of the framework is part of ongoing research, some directions for improvement are highlighted.

**Keywords:** adaptive capacities; diversified flood risk management strategies; pilot project; governance networks; learning

#### **1. Introduction**

Climate change will result in increased exposure of low-lying coastal areas to risks associated with sea level rise. Human and ecological systems will be faced with increased saltwater intrusion, flooding, and damage to infrastructure [1]. In response to the increasing risks, EU countries are making efforts to diversify their flood risk management strategies (FRMS) by combining flood risk measures spanning the whole disaster risk management cycle (pro-action, protection, mitigation, preparation, and recovery). This includes dyke reinforcements, compartments, flood proof houses, retention areas, and crisis management [2–5]. The diversification of FRMS has enabled more options for flexibility and adaptability of flood risk management [3]. The potential for diversification of strategies (e.g., emergency plans, zoning of flood prone areas, etc.) in response to climate change depends on the adaptation space and capacity of institutions [6]. Adaptive capacities enable a flexible response, learning, and adjustment by governance networks [7–10]. However, there is a gap of theoretical formulations that connects adaptive capacity and adaptation outcomes [11], such as the role of adaptive capacities to accommodate expected climate change effects [12].

In addition, more empirical research is needed to learn from ongoing attempts to diversify FRMS and investigate how governance challenges are addressed [4]. In particular, a better understanding of the required governance arrangements and how these are formed is lacking [13]. Transforming existing or forming new arrangements depends upon a larger variety of skills and capabilities of governance networks [14]. Herein, governance networks are defined as 'the set of conscious steering attempts or strategies of actors within governance networks aimed at influencing interaction processes and/or the characteristics of these networks'. When applying these strategies, actors produce outcomes such as changes or new solutions, policies, or services [15] (p. 11). Working towards these changes/transformations in FRM is typically done in pilot projects which provide an opportunity for experimenting and learning [16,17]. Knowledge and experiences acquired from pilot projects are often valuable lessons for upscaling of pilot project results [16,18]. Moreover, by learning from new insights and experiences, actors foster their capacities in governance networks to cope with uncertainty and change [14].

This research work attempts to fill these knowledge gaps by analysing the role of adaptive capacities for the development and implementation of diversified FRMS. The study was conducted by analysing the adaptive capacities of governance networks in two pilot projects in The Netherlands and Germany to better understand the required adaptive capacities of governance networks for implementing more diversified FRMS. The two pilot projects in this study are the 'Alblasserwaard-Vijfheerenlanden' located in the downstream area of the river Rhine in The Netherlands, and the 'Wesermarsch' located at the German coast. Both pilot projects are part of the EU Interreg project Flood Resilient Areas by Multi-layered Safety (FRAMES, [19]). Traditionally, both Dutch and German FRM is primarily based on flood protection through dykes and barriers. In addition, both countries are working on further diversification of FRMS by investing in preparedness. This includes the development of evacuation strategies, raising risk awareness, and stimulating preparedness among citizens.

The following sections present the theoretical framework (Section 2), research questions (Section 3), and the methodology for analyzing adaptive capacities (Section 4). Sections 5–7 present the results, the discussion of the outcomes, and the conclusion of this research, respectively.

#### **2. Theoretical Framework**

The theoretical framework aims to understand and learn from pilot projects regarding capacity development in the transition towards more diversified FRMS. Pilot projects are spaces where learning processes occur [20,21], and new insights arise related to transition dynamics [20]. The theoretical framework is grounded in the Adaptive Capacity Wheel (ACW) [22] and the Triple-loop Learning approach [23].

#### *2.1. The Adaptive Capacity Wheel*

The ACW is an analytical framework that enables the evaluation of institutions' ability to foster the adaptive capacity of governance networks over time and across scales. It defines six main dimensions divided into 22 adaptive capacities (Table 1). The ACW has been applied to a myriad of water related topics, including climate change [24], droughts and floods [25], sustainability of water governance systems [26,27], and policy content analysis [28] but has not yet been applied to the diversification of FRM.

Previous studies, dealing with adaptive capacities, encountered various challenges rooted in understanding the capacities' definitions [25,28], and weighting can lead to misinterpretations [27]. Because of the downsides encountered by scholars, for this study, the ACW is applied only as a qualitative method, not as a quantitative or measurable approach.


**Table 1.** Dimensions and adaptive capacities definitions [22].

#### *2.2. Learning Loops*

Pahl-Wostl [23] argues that social learning is essential for developing and sustaining the capacity of different authorities, experts, interest groups, and the public to manage water resources effectively and translate goals into actions. Social learning occurs through interactions between actors within social networks [29]. Learning can be seen as a feedback process consisting of three iterative learning loops. Single-loop learning refers to improving the performance of established routines, i.e., are we doing what we do right? [23], for example, should the dyke height be increased by 10 or 20 cm? [30]. Double-loop learning reframes the goal, the problem and assumptions (e.g., about cause–effect relationships) within a value-normative framework, i.e., are we doing the right things? For example, what strategies might facilitate more effective future transboundary flood management? Or how should vulnerability to other climate change impacts be included in FRM? [30]. Learning outcomes include, for example, changes in the organization's knowledge base, new objectives, or new policy frames [31]. Triple-loop learning refers to a transformation of the structural context and factors that determine the frame of reference and reconsiders underlying values, beliefs, and world views [23], i.e., how do we decide what is right, for example, should resources be allocated toward protecting the existing built environment, or should these assets be relocated or abandoned once certain risk thresholds are crossed? [30]. Learning outcomes

include changes to defining principles—for example, underlying governance protocols and structures or new learning strategies [32].

The learning process can also be linked to pilot projects and their outcomes in relation to diversification of FRMS. Change in water governance is conceptualized as a stepwise approach from single to double and to triple-loop learning. It can be explained as a feedback loop between the expected outcomes within a specific governance context and considering the structural changes, which result in an iterative cycle (Figure 1). In this cycle, triple-loop learning is directly related to transformations, while single- and double-loop learning play an indirect role in these processes. Although social learning is key to learn from and support changes/transformations in water governance, empirical evidence is lacking [33]. More empirical research is needed that addresses the underlying triggers of double- and triple-loop learning processes and links them to broader governance mechanisms and structures [34].

**Figure 1.** The concept of triple-loop learning applied to governance regimes [23].

#### *2.3. Evaluating Adaptive Capacities and Learning Processes in Pilot Projects*

In order to assess adaptive capacities and learning, the concepts of the ACW and the three loops of learning are combined in one analytical framework (Figure 2). Herewith, the framework aims to identify and analyze the adaptive capacities of governance networks and their importance for developing more diversified FRMS. Based on the Dynamic Adaptive Policy Pathways approach [35,36], six relevant steps for pilot projects were identified as part of adaptive planning processes. These steps are further divided into the situation before, during, and after the pilot projects. New diversified strategies, which include new actors, institutions, policies, and other aspects, require new knowledge, types of collaboration, and other capacities. Therefore, the goal is to understand how the concept of adaptive capacities supports the identification of requirements for diversified FRMS.

**Figure 2.** Framework for the analysis of adaptive capacities in pilot projects.

'Before' refers to the current FRMS as a result of the historical co-evolution of FRMS at the national and/or regional level within a specific governance setting [36]. It is important to have insight into this situation to understand the pilot outcomes within a specific governance context. Adaptive capacities that result in improvements of actions in the current strategy are considered as single-loop learning and have no direct effect on diversification of the current FRMS. 'During' refers to the implementation of actions during the pilot project and learnings from this implementation process. Employed adaptive capacities in this process can result both in single-loop learning and double-loop learning. The key question is whether developed adaptive capacities result in improvement and/or reframing of the FRMS. 'After' refers to outcomes of the pilot projects and upscaling of outcomes in order to change or transform the FRMS. Developed adaptive capacities lead to new frames (double-loop) or lock-in of the current strategy, with or without improvements of current actions (single-loop).

#### **3. Research Questions**

The main research question of this paper is: Which adaptive capacities support the learning process in pilot projects to achieve diversification of FRMS? The following three sub-questions are addressed:


Missing capacities refer to capacities that are not present, or present but not applied in the governance network of the pilot project. Employed capacities refer to the capacities present and applied throughout the pilot project. Developed capacities refer to the capacities improved or emerged in the governance network as a result of the pilot project.

#### **4. Materials and Methods**

Data were collected as part of the Interreg FRAMES project [19] comprising 16 FRM pilot projects in five countries: Belgium, Denmark, England, Germany, and The Netherlands. Pilot projects varied substantially in terms of their content (e.g., risk analyses of critical infrastructure, implementation of nature-based solutions, increasing community resilience, see Table A1 in the Appendix A). For the purpose of comparison, two pilot projects with a similar focus were selected. In order to gain a broad understanding of the context and focus, both pilots were visited and explained by the involved stakeholders (Figure 1). During the pilot project, empirical data were collected through questionnaires, focus groups, and semi-structured interviews.

The questionnaires provided the initial and the expected state of flood resilience before and after the pilot project, respectively. Interviews were conducted with the pilot managers to get a more in-depth understanding of the specific lacking, employed, and developed capacities in the pilot projects. The transnational focus groups (TFG) were organized to gain insight into the most needed adaptive capacities for flood mitigation and flood preparedness actions in the pilot projects. Additionally, this data was complemented additional information from FRAMES project meetings and documents to make the finding more robust.

#### *4.1. Pilot Projects*

Two pilot projects were selected as case studies: the 'Alblasserwaard-Vijfheerenlanden' in The Netherlands and the 'Wesermarsch' in Germany. These two cases were selected from a total of 16 pilot projects in the FRAMES project (see Appendix A for an overview of pilots and selection). These pilot projects were selected because flood risk governance is comparable in both countries (Table 2). In both countries, defence/protection strategies are dominant but are also looking into more integrated strategies: Multi-layered safety [37] in the Netherlands and the LAWA approach in Germany [38]. This is part of a paradigm shift from a safety to risk-based approach [39–42]. Traditionally, responsibilities

in Dutch flood control are divided between the centralised Rijkswaterstaat and decentralised water boards [43]. Local actors are involved only when traditional defence approaches are not feasible anymore. In Germany, the federal states (Länder) have the main responsibility for all water issues and civil protection [44]. Actors from spatial planning play an important role in zoning plans or mitigation and flood risk management strategies in Germany [40]. However, the use of spatial planning instruments has increased in The Netherlands, as well as within flood risk management [43]. As part of developing the risk-based approach, both pilot projects aim to enhance the integration of mitigation and preparedness measures, including the development of evacuation strategies, raising risk awareness, and stimulating preparedness among citizens.



#### *4.2. Questionnaire*

For each pilot project, a questionnaire (Appendix B) was completed by the pilot managers together with local relevant stakeholders. The pilot managers selected key stakeholders of their pilot project. In Alblasserwaard, the questionnaire was filled by four relevant stakeholders and in Wesermarsch, it was filled by seven local stakeholders (see Table A2, Appendix B). The actors filled out the questionnaire by organization prior to the start of the pilot project (October 2017) to identify the current (i.e., before the pilot) and the expected future (i.e., after the pilot project) level of diversification in the FRMS, and the perceived current and future level of flood resilience among authorities and communities. Closed questions were presented on rating scales 1 (not at all) to 10 (to a great extent). Open questions were asked to explain the indicated ratings. Finally, the answers to the open questions were aggregated, and scores to the closed questions were averaged.

Diversification. Using the disaster management cycle—pro-action (1), protection (2), mitigation (3), preparation (4), recovery (5)—stakeholders were asked 'To what extent is ( ... ) a strong characteristic of the pilot area?' on a 1–10 rating scale. This question was asked for the current situation (i.e., before the pilot project) with respect to the main pillars of FRMS in The Netherlands and Germany; i.e., for pro-action (1) and protection (2). For the diversifying elements mitigation (3), preparation (4) and recovery (5), stakeholders answered this question both, for the current and the expected situation, after the pilot project. In total, stakeholders responded to eight items. This information was used to qualitatively estimate the ambitions of authorities in diversifying FRMS. Open questions were used to further tap into these ambitions for elements 3–5 by asking 'What will be done in the pilot with regard to ( ... ) that improves the (a) 'physical resilience in the pilot area?', (b) 'capacities of local organisations/institutions in the pilot area?', and (c) 'capacities of local communities (citizens, businesses) in the pilot area?'

Resilience of authorities and communities. Stakeholders were asked to name the institutions and citizen groups that would be involved in and/or informed about the pilot project. Subsequently, stakeholders were asked to respond to the items: 'In general, to what extent is ( ... ) embedded in policy and practice of these organisations, in your opinion?' and 'In general, to what extent is ( ... ) embedded in the behaviour of these communities, in your opinion?' Both items were presented on a 1–10 rating scale. These items were presented separately for mitigation (3), preparation (4), and recovery (5), and in

the situation before the pilot project and the expected situation after the pilot project. Thus, overall stakeholders responded to twelve items. For each item, a written clarification was requested.

#### *4.3. Interviews*

Based on the theoretical framework, a comprehensive interview guideline (Appendix C) was developed and validated by the four involved knowledge institutes. The definitions of the adaptive capacities (Table 1) were integrated into an interview guideline through open questions to gain detailed insight into the opinions and arguments, but also to avoid social desirability bias. Interviewees were asked to reflect on past and current FRMS in the pilot project, any struggles encountered, the main accomplishments, the role of actors involved, and how these factors could contribute to mainstream the pilot project outcomes into the governance regime. All pilot managers were interviewed, and two interviews were used for the selected pilot projects (i.e., Alblasserwaard-Vijfheerenlanden, 28 February 2019; and Wesermarsch, 23 January 2019). The interviews were transcribed and checked by the pilot managers. Data analysis was performed using systematic coloured coding [45] to determine the adaptive capacities that were lacking, employed, and developed before, during, and after the pilot project. In order to facilitate data analysis, colours were assigned to each dimension of the ACW, and the criterions or adaptive capacities were numbered (see Table 1) to differentiate these in the interviews' transcripts and in the Results section. The results are presented in narratives to provide detailed storylines of the case studies [46]. The narratives of both pilot projects were reviewed by the pilot managers.

#### *4.4. Transnational Focus Groups*

Three transnational focus groups (TFG) were organized to gain insight into adaptive capacities before, during, and after the pilot projects and their relation to FRMS. This was done in parallel with the interviews of the pilot managers. Each TFG focused on a different FRM action (Table A3, Appendix D), that was typical for mitigation (via spatial planning), preparedness (integrating emergency response in FRM), and community resilience to address the actions by inhabitants in relation to mitigation, preparedness, and recovery. Each group was requested to select and discuss the five most relevant adaptive capacities (Table 1) needed for a specific action in an FRMS.

The three TFG were organized on 27 March 2019 in Oldenburg (Germany) and included 32 participants from five countries, representing (mainly regional) authorities with responsibilities in water management, spatial planning, crisis management, and community resilience. In order to facilitate transnational learning, each TFG consisted of participants from the 16 pilot projects and countries (Appendix D). Each group was moderated by an author of this paper.

#### *4.5. Additional Information*

In addition to the interviews held with pilot managers, presentations by other stakeholders of the pilot projects were attended during visits to the pilot projects (Alblasserwaard-Vijfheerenlanden, 22 February 2019; Wesermarsch, 28–29 March 2019). The presentations and discussions provided more background and insights into the role that these actors played in the pilot projects and FRMS.

#### **5. Results**

#### *5.1. Diversification of FRMS in Two Pilot Projects*

Both pilot projects are located in low lying areas that are exposed to flood risks. Results from both questionnaires (Table 3) confirm that current FRMS lean on flood protection by dykes. Protection levels are regarded as 'high', but further enforcements are required to meet the legal flood protection standards. Flood preparedness can be regarded as the second pillar in the current situation as it has been embedded already in current policy and practice of crisis management authorities. However, both pilot projects aimed to make small improvements in flood preparedness. Mitigation is less developed, and both pilot projects aimed for improvement. Recovery is least developed and is not particularly focused on in the pilot projects.

To achieve the improvements in mitigation and preparedness among authorities, the Dutch pilot project expected that stronger involvement of local authorities would increase awareness and lead to improvements in spatial planning (mitigation) and better evacuation plans (preparedness). Currently, stakeholders rely strongly on protection measures, increasing their knowledge and awareness will take time. Likewise, embedding flood risk mitigation and preparedness in current policies and practices also requires long term planning. Similarly, the German pilot project focused on increased involvement of actors and cooperation aiming to improve the incorporation of flood risk in regional planning (mitigation) and to improve contingency planning by reviewing plans collectively (preparedness).

In order to improve mitigation, especially preparedness among communities, both pilot projects focused on providing better information about the warning systems and increasing their knowledge and awareness.

**Table 3.** Perceived diversification of risk management strategies (FRMS) before and expected after the pilot projects. Scores on scale 1–10 (\* left number refers to before and right number after the pilot project).


*5.2. Lacking, Employed, and Developed Adaptive Capacities in Both Pilot Projects*

Table 4 summarizes the main characteristics of both pilot projects. Neither of the pilot projects had upscaling as its objective, but through interviews, several capacities needed for upscaling were identified that were not present.




**Table 4.** *Cont.*

#### 5.2.1. Case Study 1: Alblasserwaard-Vijfheerenlanden Pilot Project

The goal of this pilot project was to achieve more diversified FRMS by linking spatial planning and emergency management. The main characteristics of the pilot project, the pilot goal, the leading organization, the actors involved, and the duration of the project can be found in Table 4. The specific actions of this pilot project were to ensure that dyke roads are suitable for emergency vehicles, to build social capacity for evacuation involving local groups, and to improve evacuation management plans in case of flooding.

The pilot project faced a number of difficulties during set-up, that were identified as 'lacking capacities before the pilot project' (Table 5, indicated with - in column 'before'). The pilot manager determined that there was a lack of variety in FRM solutions (1.3) in the Alblasserwaard area because the FRMS rely mainly on protection. The second difficulty encountered was that local mayors had a lack of information about the diversification of FRMS. They thought that FRM beyond hard infrastructure was not possible or limited (e.g., existing evacuation plans are not clear). Moreover, there was low social capital (3.3) because there were not enough volunteers to improve preparedness and heavy reliance on governmental flood protection. The third issue was a lack of collaboration (4.3) between crisis and water management authorities. For example, prior to this pilot project, the safety region, water board, and local municipalities did not collaborate on integrated flood risk management projects (combining evacuation and spatial development). The fourth difficulty was limited human resources. Local municipalities have limited staff (5.2), and sometimes, one person is responsible not only for FRM but many other subjects.

In order to define the goal of the pilot project and implement the planned activities, multiple capacities were employed during the pilot project (Table 5, indicated with + in column 'during'). The implementation of diversified FRMS required the involvement of multiple actors (1.2) from a variety of levels and sectors (the safety region, the province, the water board, and municipalities) at an early stage of the pilot project. They met regularly to discuss problems and needs (1.1) related to the current FRMS and together came up with a diversity of solutions (1.3). For example, one of the problems was that emergency vehicles might not be able to use the dyke roads during evacuation because the water board needs it for equipment or because of the risk dyke failure due to instability under extreme conditions. Based on the discussions among the actors involved, the agreement was to align their needs in case of future dyke reinforcements integrating protection, infrastructure planning, and emergency response. The actors involved gathered in regular project meetings to define the (long term) goals (4.1) and specific objectives of the pilot project. The pilot manager highlighted that collaboration (4.3) was positive during the pilot project, and early collaboration started with existing initiatives in the area. Problems, needs, solutions, and next steps were discussed (2.4) with the actors. As a result, learning from the past and the current actions in FRMS (2.2) during the pilot project is an essential capacity. Moreover, pilot projects offer a higher availability of resources when human capacity (5.2) and finances (5.3) are combined to experiment with innovative solutions in FRM. For example, the safety region had the human capacity for modelling the evacuation routes while the province had the financial resources for it.

**Table 5.** Adaptive capacities before, during, and after the pilot projects and their perceived relevance for FRMS.


6.4 Accountability

Albl. = Alblasserwaard pilot project, Weser. = Wesermarschpilot project. - refers to lacking capacities; + refers to employed capacities; # refers to developed capacities; • refers to 5 most important adaptive capacities selected by the transnational focus groups (TFG). The grey colour reflects the lacking capacities before but developed as a result of the pilot projects, which were also emphasized as important capacities by the TFG.

Based on the pilot project outcomes, several capacities were developed (Table 5, indicated with **#** in the column 'after'). The first and the most important developed capacity was that more innovative solutions were applied in the pilot project to diversify the FRMS (1.3) by combining evacuation with spatial development. The second capacity that was developed as a result of the pilot project is the collaboration between actors (4.3) with respect to crisis management improved. For example, actors (from the Safety Region, the water board, and municipalities) who did not collaborate on this subject prior to this pilot project, currently they understand the significance of linking evacuation and spatial planning in relation to FRM. A relevant outcome of this pilot project is providing more detailed information (3.1) about the applicability of diversified FRMS in the area and making it accessible for everyone. For example, when this information was provided to the municipalities, they expressed more interest to participate in the project. Thus, sharing information resulted in capacity building (3.3) and awareness of the role of citizens and local authorities in evacuation management. Another developed capacity is the learning from pilot projects when actors share knowledge and learn from each other. For instance, during a meeting with local entrepreneurs, it became clear that several local entrepreneurs had already taken actions to protect themselves against floods in the current FRMS (2.2). Moreover, at the end of the pilot project, the lead authority provided a final report that included pilot project results and policy recommendations. This was an evidence-based document (2.5) that can be used by other actors to replicate the pilot project.

Finally, two capacities were identified as 'lacking capacities for upscaling of the pilot project' (Table 5, indicated with - in column 'after'). On the one hand, the pilot manager pinpointed that visionary leadership (4.1) is required in the process of diversification of FRMS. However, this can be hampered, for example, by a change of government representatives every four years. Likewise, it is not clear who takes the lead (4.2) to adapt or change current policies based on the pilot project outcomes. On the other hand, human resources (5.2) are lacking because generally, governments do not have enough human capacity to cooperate in projects related to the diversification of FRMS.

#### 5.2.2. Case Study 2: Wesermarsch Pilot Project

The goal of the Wesermarsch pilot project was to develop a more balanced FRMS, taking actions in spatial planning (mitigation) and emergency management (preparedness). The main characteristics of the pilot project, the goal of the pilot project, the leading authority, the actors involved, and the duration of the project are described in Table 4. The actions of the pilot project are to develop informative products (brochure for both individual farmers preparedness), to improve the current database with flood risk maps, to develop an app to organize volunteers during a disaster, and to organize a 'flood risk awareness day' and a 'flood partnership event'.

'Lacking capacities before the pilot project' were identified by using data from a previous EU project [47], and the outcomes of interviews conducted by Jade University (Table 5, indicated with - in column 'before'). Traditionally, the main focus in the Wesermarsch is on flood protection through hard infrastructure and far less on flood mitigation through spatial planning, and preparation and recovery from a flood event. Therefore, it was determined that there were insufficient varieties of solutions (1.3) in FRMS. Moreover, when the EU previous project ended [47], the communication and further actions between the actors involved (water management actors) also stopped. However, generally in Germany, there is a shortage of collaboration (4.3) among authorities responsible for diverse aspects of the current FRMS. Likewise, there are no local groups who take the lead (4.2) to prepare volunteers in case of an evacuation. Furthermore, it was also identified that citizens had limited access to information (3.1) about flood risk preparedness and thus, they had a low social capital (3.3) in order to prepare themselves in case of flooding. An additionally lacking capacity was an unsuitable procedure (5.2) for the use of spatial databases (e.g., elevation maps, evacuation routes, etc.) to improve decision making for spatial planning at a local level.

During the pilot project, different capacities were employed (Table 5, indicated with + in the column 'after'). First, the pilot manager involved multiples actors (1.2) in the pilot process (local and regional actors from disaster, water, flood risk management, and aid organizations, see Table 4). The actors met and discussed the diversification of FRM problems (1.1) to provide diverse solutions (1.3) for these problems. For example, flood risk maps used for evacuation exercises did not consider the topography behind the dyke, and no local group existed to organize volunteers in case of a disaster intervention. Secondly, leadership was promoted through the implementation process of the pilot project. Regional fora were organized (4.3) where key actors discussed problems and solutions. Afterwards, a priority list of activities with a long term vision (4.1) was made. Six activities were selected, and voluntarily actors took the lead (4.2) to implement them. Thirdly, the previous project [47] in the area was utilized as a learning example (2.2) to find out what did or did not work in the past and why. Likewise, the actors' discussions about undefined challenges (2.4) also contributed to the learning process.

As a result of the pilot project, multiple capacities were developed (Table 5, indicated with # in the column 'after'). The pilot outcomes resulted in a diversity of solutions (1.3) for FRM in the Wesermarsch area. For example, increased self-preparedness of local citizens and an improved spatial database for spatial emergency management planning. The development of informative products, regional fora to exchange information, and experiences resulted in diversified FRMS. Therefore, making information more accessible (3.1) for citizens leads to improved flood risk awareness and social capital (3.3). On top of that, the collaboration (4.3) between local and regional actors from disaster and water management authorities was enhanced. As a result of the collaborative capacity, the pilot manager observed an increase of trust (2.1) and improved communication about diverse FRMS. In addition, the pilot results showed improvements in preparedness (2.2) due to the actions taken. For example, farmers and other inhabitants are better informed about FRM and evacuation planning; flood risk maps at the local level are improved, and these will be used by the German state agency to support emergency planning decisions. The pilot project actions and outcomes are publicly accessible on paper and online (database, reports, flyers). This evidence base (2.5) can be used to learn from this pilot project and replicate it.

The interview with the pilot project manager revealed two capacities that were identified as 'lacking capacities for upscaling of the pilot project' (Table 5, indicated with - in column 'before'). The first one is the lack of entrepreneurial leadership because the leading authority of this pilot project has no formal responsibilities in FRM. Moreover, none of the other organizations involved have shown the initiative to take on responsibilities (4.2) regarding upscaling. The second lacking capacity is insufficient human (5.2) and financial (5.3) resources to update the developed pilot project outcomes when needed by the responsible authorities.

#### *5.3. Perceived Importance of Adaptive Capacities in Diversifying FRMS*

Three TFG focused on a different type of strategy to diversify the current FRMS, and each TFG selected five adaptive capacities that were evaluated as having the highest relevance for implementing these strategies. TFG1 focused on mitigation, TFG2 on preparedness, and TFG3 on emergency response. Table 5 presents an overview of the capacities that were linked to specific strategies.

The TFG1 (mitigation) focused on spatial zoning in flood prone areas to mitigate flood risk. Visionary long-term planning (4.1) was highlighted as having a high relevance. All pilot projects in FRAMES face long term challenges considering flood risk and other climate related issues. The TFG stressed that sharing information (3.1) about future projects between spatial planners and water managers is needed. Stakeholders perceived that the attention for long term goals is hampered by short term gain, and often, economic benefits on the short term prevail. Authority by law via procedures (5.1) and financial resources (5.3) were seen as important resources to arrange mitigation measures. In addition, they stressed the need for a variety of stakeholders (1.2) who provide a diversity of solutions (1.3).

TFG2 (preparedness) looked into flood preparedness by integrating emergency response planning in flood risk management. The group indicated that available human resources (5.2) are key adaptive capacities. There is a need for specialized staff that has the expertise to perform an impact assessment of flooding to aid preparedness and emergency response. Moreover, equity (6.2) is considered important in flood preparedness. Different societal groups have different needs in case of emergency, for example, less self-reliant (elderly, disabled) people depend on the assistance of authorities. This focus group also highlighted that emergency response could be supported by spatial planning measures, for instance, by building public shelters and safe evacuation roads. A diversity of actors (1.2) responsible for spatial planning and crisis management should collaborate (4.3) and share information (3.1) about their current policies and projects.

TFG3 (community resilience) focused in more detail on empowering communities to take action for local flood mitigation and response measures. The TFG emphasized it is important to encourage collaboration (4.3) and build trust (2.1) between responsible authorities and communities. Stronger collaboration capacity contributes to higher availability volunteers (5.2) in emergency preparedness. The capacity to provide relevant stakeholders with access to information (3.1) about mitigation and preparedness measures for communities appears to be an important issue. Raising awareness about the measures that communities can take by themselves fosters social capital (3.3).

#### **6. Discussion**

This paper aimed to identify the adaptive capacities that support the learning process in pilot projects to achieve a diversification of FRMS. Through questionnaires, focus groups, and interviews, we focused on the adaptive capacities that were lacking, employed and developed before, during, and after two pilot projects in the Netherlands and Germany. The results showed that in both pilot projects, the current FRMS leans traditionally on flood protection, with flood preparedness as a secondary pillar. Flood mitigation, but especially recovery strategies, were hardly present. The learning process in both pilot projects focused on strengthening flood preparedness and mitigation, for instance, by involving new stakeholders, sharing knowledge, reviewing contingency plans, and by providing information to citizens to increase knowledge and awareness.

We found three adaptive capacities that were stressed as important for developing more diversified FRMS and that were also lacking before but had developed as a result of the pilot projects. First, a greater 'diversity of solutions' was regarded as important, especially for developing flood mitigation strategies but not for flood preparedness and community resilience. The reason for this is that flood mitigation is currently underdeveloped and requires a balanced mix of cost-effective spatial planning actions. Finding cost-effective spatial planning measures is difficult since flood defences act as a 'front door' which make any investments in the area behind this front door redundant [48,49]. In the current frame of FRM, clear added benefits first need to be identified to gain political support for these investments, which seems difficult. Increased opportunities for integrating spatial planning in FRMS, therefore, requires reframing of current FRM policy and practice. In the Netherlands, such reframing has partly taken place with the adoption of the multi-layered safety concept (protection, spatial planning, crisis management) in FRM policy [50]. However, because the basic question 'are we doing what we do right?' (single-loop learning) has not changed to 'are we doing the right things?' (double-loop learning). Pilot projects have not succeeded in putting more diversified FRMS into practice [16,48]. For instance, there is currently little urgency to consider the meaning of a wider set of challenges originating from long term processes such as soil subsidence and sea level rise [51]. Since such challenges are not yet fully incorporated into the current FRMS, also at the level of pilot projects they are hardly considered. Reframing the problems and goals of FRM, therefore, requires the inclusion of a 'variety of perspectives over problems/needs' beyond FRM. The processes that steer this type of fundamental reframing require learning and governance capacities, which were not prioritized as important capacities in the pilot projects due to the incremental improvements (single-loop learning) that were aimed for.

Second, to create room for autonomous change, authorities and communities require greater access to information. Although this may sound obvious, the challenge is in making the right information accessible for the different stakeholder groups. Information preferences may differ substantially between actors in terms of information type, detail, and ways of receiving information (e.g., channel, format), for instance, a step-by-step checklist for farmers to prepare themselves and their livestock (Wesermarsch) and how entrepreneurs can protect their businesses (Alblasserwaard-Vijfheerenlanden) in case of flooding. Making the information on emergency planning available for the actors resulted in enhanced mutual understanding of interests, actions, and information needs.

Third, type of leadership was regarded as an important antecedent of diversifying FRMS. Actors agreed that collaborative leadership, encouraging the collaboration among actors, is currently needed to further develop preparedness and community resilience. The literature supports that collaborative networks are essential for performing adaptive management [52,53]. Alignment across sectoral boundaries is key in governance arrangements for adapting to climate change [54], which is also observed in both cases. Boundary spanning interactions, including cherishing boundaries for clear allocation of responsibilities [55], is required for collective action in diversifying FRMS. Since mitigation strategies are underdeveloped and complex, visionary leadership seemed more important for developing cost-effective spatial planning strategies. These strategies were employed during the pilot projects; however, the capacity to develop 'long term goals and strategies' did not result from the pilot projects. This aligns with their focus on incremental improvements.

Furthermore, the adaptive capacity dimension 'resources' received some importance ratings for diversifying FRMS. Law, procedures, and policy development, as well as financial resources, were regarded as important for developing mitigation strategies. Human resources, such as knowledge, expertise, and availability of volunteers, were regarded as important for developing preparedness and community resilience. However, during the pilot projects, none of these capacities were developed, which can be seen as a risk for further uptake of the outcomes of the pilot projects [16].

Additionally, two capacity dimensions hardly received importance ratings. First, within the 'learning capacity' dimension only, trust was regarded of some importance for building community resilience. This is remarkable because the interviews with pilot managers did show that learning capacities were employed and improved as a result of the pilot projects. The reason that learning was not identified as an important capacity may be explained because, as stated previously, the pilot projects rather focused on single-loop learning (incremental improvements of established routines) instead of double-loop learning (reframing of the FRMS) or triple-loop learning (transformation of FRMS). This is also supported by the pilot managers ambitions of diversifying the FRMS through the pilot projects (i.e., small incremental improvements were expected in mitigation and preparedness). This aligns with planning literature, which emphasizes that planning practices are more adaptive (adjust to changing circumstances) and incremental (gradual changes) than often assumed by scholars proposing 'new' planning approaches [56,57]. The interviews in our study showed that stakeholders learning capacities improved as a result of the pilot projects. Second, none of the capacities related to the dimension 'governance' emerged from the interviews, and governance was hardly regarded as an important antecedent for mitigation, preparedness, and community resilience. Governance may have gained little attention because most of the governance dimensions are already well institutionalized in the current arrangements of FRMS and, therefore, little action is needed to improve governance capacities in the current FRMS. Again, because current FRMS are well developed and institutionalized, improving weak links in the current governance regime is challenging. For instance, in a review of Dutch water governance, the OECD has pointed to a lack of awareness and preparedness among citizens and the large distance between water institutions and the general public [58]. Since society has a high level of trust in FRM, there is little urgency to bridge this gap, neither by the institutions nor by members of the general public. As a result, governance capacities become a passive part of FRM and fall short in gaining public support, responding to (implicit) information needs in society and taking responsibility for providing information about preparedness and response strategies. The lack of importance ratings for governance shows that there was little awareness for this underlying mechanism, likely because the pilot projects did not fully enter the process of double or triple-loop learning. In addition to more urgency [51], more research is also needed about the role of pilot projects in transitions processes. The pilot projects studied in this paper appeared to be examples of incremental change in the diversification of FRM. Considering wicked problems like sea level rise, these pilot projects can be considered as small wins [59]. Taking the contextual dynamics of experiments into account, the studied pilot project matched best with a seedbed lens [60]. The protective environment of the Interreg project FRAMES creates an environment to develop new FRM actions and learn from these. The propelling mechanism framework by Termeer and Dewulf [59] is relevant for future research to evaluate the transformation potential of various small wins. Recent expectations about sea level rise [51] can result in a change of the contextual dynamics of flood resilience pilot projects, in which battleground experiments [60] could become more relevant.

The Governance Capacity Framework (GCF) developed by Koop et al. [61] and applied by Brockhoff et al. [62] has many similarities with the ACW framework applied in this paper (see [62] for a comparison of both frameworks). The main difference between the GCF paper [62] and this paper is in the application aim. We have applied the ACW to assess the capacity development of practitioners in pilot projects, while the GCF aims to assess the governance capacity of society to solve specific challenges [61,62]. This results in differences in the applied methodology. In this paper, we have combined the ACW with Triple-loop Learning and applied this as a qualitative approach without scoring the adaptive capacities. With case narratives and focus groups, we have aimed to gain insight into the development of adaptive capacities by pilot projects over time and identification of key capacities for diversified FRM. In the GCF approach, Brockhoff et al. [62] scored the current governance capacity of cities and prescribed what steps involved practitioners need to take. The combination of both methods can be complementary in future research by combining scores to assess the current status and development of governance capacity. The indicator scoring of capacities is valuable for comparing scores of multiple cases. The qualitative approach, as applied in this paper, provides a more in-depth insight into the development of adaptive capacities in the context of specific actions for flood resilience.

#### **7. Conclusions**

In this study, an analytical framework was proposed combining the ACW and Triple-loop Learning to assess capacity development in pilot projects. The combination of these two approaches is a unique outcome of this paper. It acknowledges the development of adaptive capacities as a result of pilot projects and enables to link this with three types of learning. The findings contribute to theories about niche–regime interactions [20] and policy transfer via pilot projects [63]. The ACW within the framework was used as a qualitative approach without scoring the adaptive capacities [22]. The narratives allowed to pinpoint the development and interdependencies between adaptive capacities over time [22] in the phases (before, during, and after) of the pilot projects. Therefore, this analytical framework is practical to assess the development of capacities of stakeholders in pilot projects that aim to diversify FRMS. Likewise, it also identifies lacking capacities that are needed to ensure successful pilots and uptake in policy.

Since the proposed framework is the product of ongoing research, much room for improvement exists. Here, we mention a few avenues needing improvement. First, the framework misses clear guidance to evaluate the success of pilots and upscaling of pilots in the policy regime. By assessing pilot goals and outcomes more explicitly, the evaluation process can be improved. In particular, we regard the 'pilot paradox' [16] as a valuable approach because it defines the conditions underlying this process. Interestingly, the pilot paradox argues that the same conditions that make pilots successful often hamper their uptake in policy. Propelling mechanisms can help to assess the transformation potential of pilot projects as small wins in the domain of climate change [59]. Second, the methods of identifying adaptive capacities can be improved by incorporating multi-item rating scales to increase the reliability of measurements. This is a common approach in questionnaire research and provides a strong asset for further validation of the framework in future studies. Lastly, in order to get a better understanding of the capacity development in the pilot projects, it is necessary to consider different actors' perspectives. Therefore, it is recommended to conduct interviews with a variety of stakeholders of the pilot projects. This can strengthen the framework and application in future studies.

To conclude, this study has shown that the analytical framework is valuable for assessing pilot projects and learning about capacity development in their transition towards diversified FRMS. This methodology is a unique outcome of the FRAMES project, and its applicability to this study contributes to the existing literature about diversification of FRMS.

**Author Contributions:** Conceptualization: F.S.C., J.-M.B., M.B. and T.T.; data curation: F.S.C., J.-M.B., M.B. and T.T.; formal analysis: F.S.C., J.-M.B., M.B. and T.T.; investigation F.S.C., J.-M.B., M.B. and T.T.; methodology: F.S.C., J.-M.B., M.B. and T.T.; supervision: F.S.C., J.-M.B., M.B. and T.T.; writing—original draft preparation: F.S.C., J.-M.B., M.B. and T.T.; writing—review and editing: F.S.C., J.-M.B., M.B. and T.T.

**Funding:** This research was funded as part of FRAMES [19], an Interreg project supported by the North Sea Programme of the European Regional Development Fund of the European Union.

**Acknowledgments:** We want to give thanks to the knowledge institutes (Ghent University, University of Oldenburg, Jade University of Applied Sciences) for their contribution in developing the interview guideline. Moreover, special thanks to the pilot managers to provide valuable input and reviews from their pilots. Additionally, we want to thank all FRAMES project partners for their participation and input through the questionnaires and focus groups discussions.

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

#### **Appendix A Selection of Cases, Based on Intended FRM Actions in Pilot Projects**

Table A1 below provides an overview of the pilot projects in FRAMES and their intended actions considering diversification of flood risk management strategies (FRMS). All pilot projects, actions, and implementation processes are more detailed described on www.frameswiki.eu.

The selection of the Alblasserwaard and the Wesermarsch pilots was based, firstly, on that the idea that traditional flood management relies mainly on flood defence with hard infrastructure. Selection of the two cases for this paper was based on the intended integration of Mitigation (via spatial planning) and Preparedness (awareness raising and evacuation planning) (see Table A1). In addition to the two selected cases, the Sloegebied pilot project also intended to focus on the same aspects, but this pilot project just started when we started with the analysis for this paper.

**Table A1.** Flood risk management actions implemented in FRAMES all pilot projects.

Light grey colour represents the diversified FRM actions implemented in each pilot project while dark grey colour shows the FRM actions' resemblance between the selected pilot projects for this study.

#### **Appendix B Monitoring Survey (FRAMES)**

Baseline measurement Pilot: ... ... ... ... ... ... ... . Important:

We kindly ask the pilot manager to complete this questionnaire in consultation with relevant experts/stakeholders in the pilot/region. The pilot manager can send this questionnaire and ask these experts/stakeholders to complete (certain) questions, or ask them to review answers. Per pilot we would like to have 1 questionnaire returned.

Who contributed to completing this questionnaire? Table A2 provides the organisations and the function of the stakeholders who filled the questionnaire for both pilots, Alblasserwaard and Wesermarsch.


**Table A2.** Stakeholders who filled the questionnaires for Alblasserwaard (left) and Wesermarsch (right) pilot projects.

#### Short Explanation

FRAMES [19] is about improving flood resilience by taking different types of actions. A common and well-known typology in flood risk management is the disaster management cycle, as shown in the diagram below. This survey first focuses on the flood resilience of areas by considering the five elements of the Flood Risk Management cycle (Figure A1): pro-action (1), protection (2), mitigation (3), preparation (4) and recovery (5). Hereafter, a few questions will be asked with regard to the flood resilience of communities and authorities.

**Figure A1.** Flood Risk Management cycle.

*Appendix B.1 Diversification of FRMS:*

Appendix B.1.1 Pro-Action/Prevention

Negative consequences of flooding can be avoided by pro-active spatial planning or land use policies ('keeping people away from water'), aimed at building only outside areas that are prone to flooding.

To what extent is pro-action/prevention currently a strong characteristic of the pilot area?


Please shortly explain your answer:

Appendix B.1.2 Flood Protection/Defence

Keeping water away from people by (combinations of) hard infrastructural works (dykes, dams, etc.) or softer (nature based) solutions (dunes, retention in nature areas, etc.).

To what extent is flood protection/defence currently a strong characteristic of the pilot area?


Please shortly explain your answer:

Appendix B.1.3 Flood Risk Mitigation

Consequences of floods can be mitigated by a smart design of the flood-prone area including spatial orders, constructing flood compartments, or (regulations for) flood-proof building. To what extent is flood risk mitigation a strong characteristic of the pilot area?


What will be done in FRAMES with regard to flood risk mitigation, that improves the


#### Appendix B.1.4 Flood Preparation

Consequences of floods can be mitigated by preparing for adequate response to a flood event. Measures include flood warning systems, disaster management and evacuation/rescue plans, and managing a flood when it occurs by taking last call emergency measures (e.g., sand bags).

To what extent is flood preparation a strong characteristic of the pilot area?


What will be done in FRAMES with regard to flood preparation, that improves the

1. physical resilience in the pilot area?


#### Appendix B.1.5 Flood Recovery

Facilitates a good and fast recovery after a flood event. Includes plans for draining/pumping away flood water and restoring safety and security, plans for reconstruction or rebuilding critical infrastructure, damage compensation/insurance systems, return of evacuated communities and social-psychological support.

To what extent is flood recovery a strong characteristic of the pilot area?


What will be done in FRAMES with regard to flood recovery, that improves the

1. physical resilience in the pilot area?


#### *Appendix B.2 Resilience of Authorities and Communities*

Appendix B.2.1 Flood Resilience of Authorities

Please name the organisations/stakeholders that will be involved in your pilot:


In general, to what extent is flood risk mitigation embedded in policy and practice of these organisations, in your opinion?


Please shortly explain your answer:

In general, to what extent is flood preparation embedded in policy and practice of these organisations, in your opinion?


Please shortly explain your answer:

In general, to what extent is flood recovery embedded in policy and practice of these organisations, in your opinion?


Please shortly explain your answer:

Appendix B.2.2 Flood Resilience of Local Communities

Please name the communities (e.g., neighbourhoods, municipalities) that will be involved in/informed about your pilot(s), and how many citizens they consist of:


In general, to what extent is flood risk mitigation embedded in the behaviour of these communities, in your opinion?


Please shortly explain your answer:

In general, to what extent is flood preparation embedded in the behaviour of these communities, in your opinion?


Please shortly explain your answer:

In general, to what extent is flood recovery embedded in the behaviour of these communities, in your opinion?


Please shortly explain your answer:

#### **Appendix C Interview Guideline**

The interview guideline was developed between December 2018 and January 2019 by 7 project partners from the knowledge institutes: 4 from HZ Delta Academy, 1 from Oldenburg University and 2 from Ghent University.

General information Pilot area: Name of the pilot manager: Organization: Interviewers: Date: Objective:

This interview consists of 14 semi-structured questions to facilitate/guide an open discussion with the pilot coordinators/managers about the implementation process of the MLS approach in the pilot area and the impact/influence of the expected outcomes in the pilot region. This input will be used for the development of the FRAMES Decision Support System and resilience toolkit. The bullet lists relevant aspects to follow and bring into the discussion if they are mentioned by the pilot manager.

Introductory questions:


#### Key questions:

	- -Specify if the goal(s) is/are for short/medium/long term (<5/5–25/>25 years)
	- -Reason of the pilot, was it completely new compared to the region's FRM approach?
	- -Follow up steps towards new/improved strategies in FRM
	- -Tools/methods used to perform and monitor these steps/decisions
	- -Current and missing actors (level, sector)
	- -Tools/methods to involved and keep actors engaged (special attention of local communities)
	- - Drivers/barriers of change: stakeholders, time, resources, uncertainties of climate change, power

Ending questions:


Thanks so much for your time and contribution in this interview

#### **Appendix D Transnational Focus Groups (TFG)**

Table A3 lists the 12 FRM actions derived from the 15 pilot projects of FRAMES.


#### **Table A3.** List of FRM actions, derived from pilot activities.

\* Actions used during the TFGs discussions.

The total number of 31 project partners participated during of the 3 TFGs held on 27th March 2019 in Oldenburg: 12 from The Netherlands (3/HZ Delta Academy, 4/Province of Zeeland, 3/Province of South Holland, 1/Rijkswaterstaat and 1/Reeleaf); 8 from Germany (2/Jade UAS, 3/Oldenburg University, 1/Küste & Raum Consultancy, 1/Oldenburg-East-Frisian Water Board and 1/Jade HS-citizens science-); 1 from Denmark (1/Danish Coastal Authority); 3 from Belgium (1/Ghent University, 2/Provincie Oost-Vlaanderen); and 7 from UK (1/Kent County Council, 2/National Flood Forum, 1/Trent River Trust, 2/South East River Trust and 1/Tees Rivers Trust).

#### **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/).
