The roles and unexplored potential of policy experimentation in climate adaptation governance: A systematic literature review

1 INTRODUCTION

Experimentation has gained momentum as a governance approach in adaptation (Bulkeley & Castán Broto, 2013; Bulkeley, 2021; Chan et al., 2015; Hoffmann, 2011; Huitema et al., 2018; Karvonen, 2018; Sengers et al., 2020). Adaptation governance refers to the governing activities of social, political, and administrative actors in the realm of climate change adaptation (Huitema et al., 2016). As per Driessen et al. (2012), governing activities can be pursued by different actors, in different institutional settings and with variations in terms of the substantive content that is being addressed. The governing of climate adaptation has been researched in the last years, but still continues to be fragmented and has failed to inform public policy (Huitema et al., 2016). Experimentation can provide new knowledge and induce broader change of technologies, policies and institutions (Kivimaa et al., 2017). Moreover, it encourages a wide range of public, private, and civil actors to develop and conduct experimental initiatives (Bulkeley, 2021; Bulkeley et al., 2014). Multiple experimental initiatives have emerged worldwide to tackle climate adaptation challenges (Bulkeley, 2021; Chan et al., 2015; Hoffmann, 2011). In the climate governance literature, experiments are conceptualized as initiatives to test technologies, policies and governance approaches on a small scale and under real world conditions (Laakso et al., 2017). They are often referred to as pilots, initiatives, test beds, demonstration projects, and urban laboratories (Connop et al., 2016; McFadgen & Huitema, 2018; Wamsler et al., 2016). For example, McFadgen and Huitema (2017b) studied an experiment, which tested a policy innovation in water management. The innovation was the coupling of multiple water issues into one solution. Huitema et al. (2018) argued that experiments should accomplish two conditions: test a theoretical hypothesis or assumption in the real world and include a novelty aspect. Along similar lines, Kivimaa et al. (2017) emphasized the novelty and the capacity of experiments in a limited time and spatial scales to disrupt the status quo of climate policies. Thus, change is enabled when deviating from ‘business as usual’ (Karvonen, 2018; Matschoss & Repo, 2018). However, “business as usual” in climate adaptation is resistant to change (Groen et al, 2023). Experiments are used because they are flexible and can adapt to climate change uncertainties. During an experiment, it can be decided who to involve, how to achieve this and how to deal with the challenges encountered (Kivimaa et al., 2017).

Besides understanding what a policy experiment and its aims are, it is also relevant to comprehend how they contribute to governance. Experiments can solve a problem, develop new practices or help us learn about how something “works in the real world” (Kivimaa et al., 2017). Experiments offer the opportunity for a learning environment as a result of new actors and networks who share knowledge, capacities and resources (Kivimaa et al., 2017; van Buuren et al, 2018). Knowledge and experiences acquired from experiments are valuable for reframing and adjusting the climate adaptation strategies and policies (Berkhout et al, 2009; Hinkel et al, 2010; van Popering-Verkerk & van Buuren, 2017; Driessen et al, 2018). Thus, learning is an essential outcome of experiments (Armitage et al., 2008; Bos et al., 2013; Farrelly & Brown, 2011; van der Heijden, 2014) but also a process. Learning outcomes can be divided into three types: (1) cognitive learning, which refer to new or improved knowledge; (2) relational learning is an example of changes in trust, the ability to cooperate and understanding other parties; and (3) normative learning involves acquiring a deeper understanding of an interactive policy process based on reflection and changes in perspectives, goals or priorities (McFadgen & Huitema, 2017b). Relational learning resonates with the social learning literature (Pahl-Wostl et al., 2008). Social learning is frequently defined as a convergent change in stakeholders' perspectives on a particular problem and its possible solutions. Such social learning goes beyond the individual, toward collectives and social networks (Gonzales-Iwanciw et al, 2020) whereas the cognitive and normative learning types resonate strongly with the policy learning literature (Haug et al., 2011). Policy learning is defined as the capacity to learn from experiences and modify actions based on the impact evaluation of previous actions (Gonzales-Iwanciw et al, 2020). Besides learning, policy experiments can also induce long-term normative changes. These changes start when a new idea is developed into a new policy framework (Schön & Rein, 1994) or image (Baumgartner & Jones 2002), alternative system configurations (Olsson et al., 2006), policy paths (van Buuren et al., 2018), new long-term visions and transition agendas (van der Brugge et al., 2005) or a new storyline (Hajer, 1997). The development of a new idea needs to be demonstrated within an experiment and be supported and accepted by actors to become a new solution (McFadgen, 2019). For achieving policy change, experimental outcomes need to be scaled up temporally and spatially (van Doren et al., 2018) through trialing, learning and rolling out (Evans et al., 2016). While there is existing literature to show how experiments emerge into social learning, less attention has been given to how experiments can lead to policy learning and how it translates into policy changes (McFadgen & Huitem, 2017a).

Bulkeley et al. (2014) not only emphasized the importance of scaling up but also the transformative power of climate adaptation experiments. Despite progress made globally, there are still implementation gaps remaining (Bulkeley, 2021; Chan et al., 2015; Hoffmann, 2011). Many initiatives prioritize immediate and short-term climate hazards, which reduces the opportunity for transformational adaptation (IPCC, 2022). However, there remains a lack of empirical evidence on how experiments can lead to transformative change in climate adaptation governance (Pahl-Wostl, 2009; Termeer et al, 2017). More specifically, it is not clear how policy experiments function as a strategy for change or how learning outcomes would lead to policy changes (McFadgen, 2019; van Popering-Verkerk & van Buuren, 2017; Wellstead et al., 2016). In response to the gap, this systematic literature review (SLR) was specifically conducted to explore the role of policy experiments and their potential to drive transformative change in climate adaptation governance, focusing on their empirical characteristics, purpose, and outcomes. This review aims to contribute to cumulative empirical knowledgebase and identify knowledge gaps for further research in this field.

This study is divided into the following sections. In Section 2, the analytical framework is presented, which was used to analyze the policy experiments. In Section 3, the methodological steps of the SLR are described. In Section 4, the results are explained and finally discussed in Section 5. Finally, the main conclusions and research agenda are presented in Section 6.

2 ANALYTICAL FRAMEWORK

Kivimaa et al. (2017) have carried out a systematic analysis of experiments focused mostly on climate mitigation and a few in climate adaptation. In order to do this, the authors used a generic framework that consisted of six categories with 25 sub-categories built upon the concepts of experimental governance, socio-technical transition (Bos et al., 2013; Bulkeley & Castán Broto, 2013; Bulkeley et al., 2014) and the policy evaluation literature (Vedung, 1997). Findings suggest that Kivimaa's framework helped to better understand the features and diversity of experiments in climate mitigation and adaptation, as well as guidance for future research.

The framework of Kivimaa et al. (2017) is quite general, including elements such as time and spatial scales that describe experiments, the involved actors, or outcomes. As it is so generic, we argue that it can be applied to analyze experiments in other fields, such as climate adaptation. Based on the available knowledge about climate adaptation experiments, we adapted a few of the elements to make the framework more tailor-made to the aim of this review. Some categories remain the same, while others were rephrased and new categories were added inductively in the course of the development of the methodological approach (Table 1). Table 1 presents the adapted analytical framework (left column), the original Kivimaa's framework (middle column) and the justification for changes made (right column). The original framework of Kivimaa et al. (2017) can be found in Figure A1, Appendix A. The adapted analytical framework includes 13 elements of experimental governance that describe the empirical characteristics of experiments such as the type of experiment, objective, climate change objective, sector, location, scale, duration, actors involved and experimental outcomes (Table 1).

3 METHODOLOGY

A systematic literature review (SLR) is a summary and assessment of the state of knowledge of a given topic area or research question. They are structured to rigorously summarize existing knowledge and evidence while identifying the gaps and new directions for future research and thus differ from generic literature reviews in three main ways: (1) they begin by defining a review strategy, (2) they explicitly identify inclusion and exclusion criteria, and (3) they aim to assess the largest amount of relevant and available literature possible. A SLR allows for a dynamic, flexible, and adaptable process to meet the scope and address research questions of a scientific study (Berrang-Ford et al., 2015). Since empirical studies on policy experiments in climate adaptation are fragmented in the literature (McFadgen, 2019; van Popering-Verkerk & van Buuren, 2017; Wellstead et al., 2016), this review provides cumulative knowledge and suggestions for future work on this topic. Thus, the methodology is tailored to the objective of this review and divided into four steps, from defining the search key to analyzing the empirical papers (Figure 2):

3.1 Step 1a: Development of the search key

The search key (Figure 1) was developed to fit the scope of the review using Scopus as the main search database. Inclusion and exclusion criteria were defined to cover the literature on policy experiments in climate adaptation.

The inclusion criteria were peer-reviewed articles written in English about individual policy experiments or pilots that focused on climate adaptation.

The exclusion criteria were theoretical and technical articles, conference papers, gray literature, book chapters, and empirical papers that did not study individual policy experiments.

Terms for three bodies of literature were used to develop a search key that would fit the objective of this review (Figure 1):

1. Climate change hazards such as flood, drought, salinization, erosion and heat (IPCC, 2023);
2. Experimental governance terms such as pilot, policy/governance experiment, learn, upscale and mainstream (Huitema et al., 2018; van Doren et al., 2018). The terms of the third body of literature were included to better understand the experiments in the field of climate adaptation governance.
3. Implementation and decision-making terms such as policy, strategy and adaptation (IPCC, 2023);

The search keywords were formulated into search key statements using the Boolean operators “AND,” “OR,” and PRE/0. We followed a thorough approach to scrutinize the search terms. Multiple combinations of two- and three-word combinations were developed to define the search key (Table B1 in the Appendix). The first trial helped narrow down the search words for the first body of literature (1.1–1.5, Table B1). In the second trial, keywords from the first and second body of literature were tested in Scopus (2.1–2.5, Table B1). In the third trial, search key terms of the three bodies of literature were used (3.1–3.15, Table B1). Different combinations of terms and Boolean operators helped the authors to develop the search key, step by step. During each step, the authors discussed the potential terms and combinations of the search key. For example, we learned that using the PRE/0 operator between terms reduced the number of hits, from 30.458 to 51 (2.3–2.4, Table B1). Likewise, we also decided to leave out terms such as ‘coast’ because it would reduce the number of hits too drastically (3.11, Table B1). In addition, the climate change hazards (flood, droughts, salinization, erosion and heat) included in the search key already referred to the main hazards expected in coastal areas. Thus, also using the term ‘coast’ could be a redundance in the search. In addition, scanning the empirical literature from the list of 177 hits, the term “experimental governance” was often mentioned by scholars. Therefore, an additional search was run by replacing the term “policy PRE/0 experiment” with “governance PRE/0 experiment” (3.13), resulting in 44 hits with overlapping papers. Ultimately, the two search terms were combined into one search key (3.16, Table B1) resulting in 216 well-known publications from the policy experimentation in climate adaptation literature (McFadgen, 2019; McFadgen & Huitema, 2017b, 2018). This was chosen as a definitive search key for this systematic empirical study. This process is further elaborated on in Appendix B and Table B1.

The ultimate choice for the in- and exclusion criteria and consequently, the search key, was the result of an iterative process in which the authors discussed potential search terms, assessed if these provided relevant results in Scopus, and subsequently revised them based on the outcomes as a result. In this way, we are confident that we did not miss any crucial studies.

3.2 Step 1b: Selection process of the empirical papers

The authors performed content analysis of the title, keywords, and abstracts of the 216 papers, applying the inclusion and exclusion criteria mentioned in step 1a. The publications were independently classified as “in scope,” “out of scope,” and “unclear” by the four authors to avoid bias (table available on request). The main author scanned and labeled all papers while four other authors each independently reviewed one-quarter of the papers. Afterwards, the main author combined all of the reviewed papers to make sure that everyone applied the inclusion/exclusion criteria consistently (step 1a). As a result, there were no fundamental differences among the authors and any discrepancies were remedied once everybody had internalized the inclusion and exclusion criteria in the decision-making, such as excluding papers on climate mitigation, technical experiments, or no individual studied experiments. Moreover, during the selection process, three of the five authors “in scope” papers were contacted to provide additional information on the individual experiments. As a result, the contacted authors provided more information about the 18 policy experiments investigated in the three respective empirical papers (McFadgen, 2019; McFadgen & Huitema, 2017b, 2018). The additional information provided was useful to consider these empirical papers in the analysis. The authors from Nair and Howlett (2018) suggested an additional paper (Nair, 2019) to complement information about the one experiment. However, authors from the paper by Vinke-de Kruijf et al. (2020) did not have additional information regarding the experiments mentioned in the study. Thus, this paper was discarded because it did not provide sufficient empirical details about the experiments mentioned. As a result of the steps mentioned above, 202 papers were considered “out scope” and 14 papers were selected as “in scope” by the authors of this review (Figure 2, Table 2a). Having only these 14 papers could be considered a limited number for a systematic literature review, therefore, an additional snowballing was applied to the reference of the selected papers.

TABLE 2a. Empirical papers as a result of the search key.

#	Authors	Title	No exp.
1	Butler et al. (2016)	Priming adaptation pathways through adaptive co-management: Design and evaluation for developing countries	1
2	Chu (2016)	The Governance of Climate Change Adaptation Through Urban Policy Experiments	6
3	Cloutier et al. (2015)	Planning adaptation based on local actors knowledge and participation in a climate governance experiment	1
4	Connop et al. (2016)	Renaturing cities using a regionally-focused biodiversity-led multifunctional benefits approach to urban green infrastructure	3
5	Doorn (2016)	Governance Experiments in Water Management: From Interests to Building Blocks	1
6	Huang et al. (2015)	Testing a participatory integrated assessment (PIA) approach to select climate change adaptation actions to enhance wetland sustainability: The case of Poyang Lake region in China	1
7	McFadgen (2019)	Connecting policy change, experimentation, and entrepreneurs: advancing conceptual and empirical insights	18a
8	McFadgen and Huitema (2017a)	Stimulating Learning through Policy Experimentation: A Multi-Case Analysis of How Design Influences Policy Learning Outcomes in Experiments for Climate Adaptation	18a (same as McFadgen, 2019)
9	McFadgen and Huitema (2018)	Experimentation at the interface of science and policy: a multi-case analysis of how policy experiments influence political decision-makers	18a (same as McFadgen, 2019)
10	Panditharatne (2016)	Institutional barriers in adapting to climate change: A case study in Sri Lanka	1
11	Rocle and Salles (2018)	“Pioneers but not guinea pigs”: experimenting with climate change adaptation in French coastal areas	1
12	Wamsler and Pauleit (2016)	Making headway in climate policy mainstreaming and ecosystem-based adaptation: two pioneering countries, different pathways, one goal	2
13	Wellstead et al. (2016)	Canada's Regional Adaptation Collaborative and Adaptation Platform: The importance of scaling up and scaling down climate change governance experiments	6
14	den Uyl and Munaretto (2020)	Experiment-based policy change over time: Learning from experiences in the Dutch fen landscape	7
		Total number of experiments in step 1b	48

• ^a Same 18 policy experiments mentioned in the 3 empirical papers, counted only once.

4 RESULTS

The findings are presented in three sub-sections: Subsection 4.1 presents the statistical analysis of the elements 1 to 9; Subsection 4.2 presents a synthesis table of the policy experiments' empirical characteristics. And Subsection 4.3 presents a qualitative analysis of the elements 10 to 12. When findings refer to specific papers, we use paper numbers from Tables 2a and 2b between brackets in the entire Results section.

4.1 Descriptive statistics

A total of 111 unique policy experiments were studied in the 27 empirical papers. Figures 3 and 4 show the empirical characteristics (elements 1 to 9 of the analytical framework).

Scholars used multiple terms to refer to policy experiments in climate adaptation (Figure 3a). The terms “pilot projects” and “policy experiments” were used as interchangeable terms. Pilot projects were also defined as a form of policy experimentation to introduce, test, or evaluate new practices, concepts or technologies before implementation (23, 24, Table 2b). Compared to policy experiments, governance experiments allow the involvement of diverse actors to improve or make structural and procedural changes over time (1–3, 5, and 16, Table 2b, b). Comparing the experimental objectives (Figure 3b), almost half of the experiments (49 out of 111) aimed to implement governance or policy strategies while a third (38 out of 111) tested technical innovation. A new governance strategy would be to involve and empower local communities in urban water management (Chu, 2016). Likewise, a policy strategy could be the development of alternative flood risk management strategies (e.g., multi-layer safety) (van Popering-Verkerk & van Buuren, 2017). Moreover, an example of technical innovation could be experiments testing nature based solutions for urban water management (Frantzeskaki, 2019; McFadgen & Huitema, 2017a, 2017b, 2018).

Policy experiments in climate adaptation involve multi-hazards (3c), multi-sectors (3d) and multi-actors (4e). Likewise, they have a short duration (up to 4 years, Figure 4d) and they address local issues (Figure 4b) in urban areas (4c) of European countries (Figure 4a). The most-studied climate change hazards are flooding, followed by heat, drought, sea level rise and water scarcity (Figure 3c). It is notable that many of the experiments focused on flooding (17 out of the 38) and sea level rise (13 out of 20) are located in urban areas of The Netherlands, which is a low lying country vulnerable to future climate change. The multi-sectoral approach (Figure 3d) includes water, land use planning, recreation and community development dealing with issues related to health, poverty and the empowerment of local communities in developing countries. The multi-actors approach (Figure 4e) includes a variety of local, regional, and national public actors (e.g., governmental authorities), private actors (e.g., consultancies) and civil actors (e.g., NGOs, citizens). It is clear that civil actors are the larger group of actors involved in the policy experiments focused on climate adaptation (Figure 4e).

5 DISCUSSION

In this review, empirical studies were examined to fully grasp the potential of policy experimentation to advance climate adaptation. In order to reach this objective, we analyzed the empirical characteristics, role, and outcomes of policy experiments. These findings reveal that, although policy experimentation has gained relevance in the climate adaptation literature, it is still fragmented in practice. In this section, we will first highlight the key findings, second we bring forward the limitations of the study and third, we reflect on the usefulness of the framework used.

The empirical findings show that policy experiments are characterized as being multi-hazard, multi-sector, and multi-actor (Figures 3c,d and 4e) but limited spatially and geographically. They are predominant at the local level in urban areas of European countries (Figure 4a–c). Because of this, the involvement of citizens in the experimentation process was high (Figure 4e). One unexpected observation was that one third of the urban policy experiments were located in The Netherlands (den Uyl & Munaretto, 2020; Frantzeskaki, 2019; McFadgen & Huitema, 2017a, 2017b, 2018; McFadgen, 2019). This might be explained by the fact that The Netherlands is a low-lying country, which makes cities vulnerable to future climate change impacts such as flooding and sea-level rise (Oppenheimer et al., 2019). Other studies are also consistent with the finding that policy experimentation in urban areas is a well-researched topic in the climate governance literature, more for climate mitigation than adaptation (Bulkeley, 2021; Kivimaa et al., 2017; van Doren et al., 2018). Another observation is that more than half of the experiments were developed in Europe, while fewer were from other continents (Figure 4a). The reason might be that the European Commission made funding available to research and develop knowledge development on climate adaptation topics through the collaboration of state members. Moreover, the EU funded projects could also explain that a considerable number of experiments had a duration of three to 4 years (Figure 4d). Within 3–4 years, these pilots or experiments are able to test a solution or strategy but not to enable formal policy change. The short duration might also be influenced by the election period of decision makers, which takes place every 4 years. However, this contradicts the findings of Kivimaa et al. (2017), who stated that even though experiments are limited by time and spatial scales, they are still able to disrupt the status quo of climate policies.

We found that policy experiments primarily led to social learning outcomes rather than policy learning. Social learning resulted into changes in understanding and the formation of new networks. The changes are assumed to be incremental in climate adaptation (Karvonen, 2018; Kivimaa et al., 2017). Thus, we assume that the role of experiments has shifted from research to governing strategies (Bulkeley, 2021; Nair & Howlett, 2015; Sanderson, 2009). The process of experimentation has become a new governing approach in climate adaptation, which is observed by other scholars as well (Bulkeley, 2021; Bulkeley & Castán Broto, 2013; Chan et al., 2015; Hoffmann, 2011; Huitema et al., 2018; Karvonen, 2018; Sengers et al., 2020). The experimentation process and its outcomes are influenced by various enabling factors and barriers, as identified in the literature. Factors such as champions, networks, trust-building, and policy entrepreneurs facilitate success, while barriers include limited time for conflict resolution, rigid governance and lack of political commitment. Van Buuren et al. (2018) concluded that the same factors that enable a successful experimentation might also hinder the upscaling of its outcomes. In this study, the authors made a distinction between factors that influence the experimentation process and the upscaling of the experimental outcomes process. For example, factors such as time scale, planning or monitoring refer to the experimentation process while stakeholders involvement, conflict resolution and crossing [administrative, sectorial] boundaries refer to the upscaling process. This differentiation might be relevant for developing learning approaches to overcome barriers for policy change in climate adaptation.

Besides the key findings, this review found a few additional limitations in the research field. First, there is no agreed-upon terminology in referring to policy experiments. Scholars interchangeably use multiple terms such as “pilot project,” “policy experiment,” “experiment,” and/or “governance experiment” (Figure 3a). Likewise, scholars use a variety of terms to refer to experimental outcomes like social or policy learning, radical change vs normative change or scaling up vs mainstreaming, among others. Scholars conceptualize policy experiments and their outcomes in different ways than practitioners, governmental authorities, businesses, or citizens. Van der Heijden (2015) concluded as well that there is a mismatch between academics and other actors regarding the understanding of policy experimentation. We can assume that there is lack a common language in literature. Second, policy experiments are limited geographically and spatially (Figure 4a,c). More than half of them were developed in Europe and only a few in Asia, Australia, and Africa (Figure 4a). Likewise, most policy experiments documented in the empirical literature were located in urban areas (half of the policy experiments) and little attention was given to other spatial areas such as coastal, rural or nature conservation areas (Figure 4c). The IPCC report (2021) states that human and natural ecosystems require urgent actions to adapt to climate change globally. This limitation raises questions such as: why are policy experiments not developed in coastal, rural or nature conservation areas? Is policy experimentation limited to developed countries? Is this linked to specific funding programs only available for Northern European countries? Third, it is still ambiguous how policy learning happens in practice and how this is translated into broader policy change. For example, Wellstead et al. (2016) found that there is an absence of political and policy research on the mainstreaming of climate governance experiments. Furthermore, Butler et al. (2016) concluded that there is limited evidence of institutional change due to the absence of policy windows. Previous studies (Patterson et al., 2017; van Bommel et al., 2016; Wolfram et al., 2019) also agreed that co-learning for systemic governance transformation remains poorly understood. For example, van Bommel et al. (2016) concluded that transformative change in climate adaptation relies on the co-production of knowledge across the complex boundaries of the social, institutional, and cultural systems. Another limitation is the lack of empirical evidence regarding the success factors and/or barriers that influence the implementation of the outcomes of the policy experiments into climate adaptation. A possible explanation for this might be that enabling factors or barriers are dependent on the context of the experiment (Matschoss & Repo, 2018; Vreugdenhil, 2010). Another explanation could be that there is a lack of a systematic framework on how to overcome them (van Doren et al., 2018). For example, van Doren et al. (2018) developed a framework to systematically understand the drivers to scale-up experiments in climate mitigation, but it has not been verified in the context of climate adaptation.

Regarding the limitations in the research field, there are a few additional methodological limitations. The findings of this review are restricted to one search key, one research engine (Scopus) and the inclusion/exclusion criteria (Methodology). To cover a larger number of papers within Scopus, we applied the snowballing method to the selected empirical papers. In relation to the inclusion/exclusion criteria, gray literature or book chapters written in languages other than English were excluded because we aimed for peer-reviewed articles to ensure the scientific validity and significance of the empirical data studied. Although English is one of the most accepted language for international scientific publications, it might also be considered as a limitation because we might have missed relevant papers from the Global South. For example, experiments located in South America (Spanish language) or South Africa (local languages) were not found or they were limited in our findings (Figure 4a). The Global South might provide valuable insights such as building community resilience and how they can enable transformative change in climate adaptation (Burch et al., 2017).

Finally, a reflection on the analytical framework applied is necessary. This review proves that an adapted version of Kivimaa's framework (2017) is suitable for analyzing climate adaptation experiments. The iterative process of coding and analysis helped to adapt or add new elements to the analytical framework. For example, “Spatial location” (element 7) was added as a new element to the framework, which helped to confirm the preliminary hypothesis that policy experiments are limited in coastal areas (Garschagen et al., 2024). Likewise, the “Role of policy experiments” (element 10) was added inductively in the coding process of elements “2.2 Objectives of the experiment” and “3.4 Outputs/outcomes” of Kivimaa's framework. This new element provided valuable information about the function or role of policy experiments in climate adaptation governance and thus helped to answer the main research question of this review. We argue these two elements could be added to Kivimaa's framework to investigate other type of experiments, such as mitigation. Thus, the general character of Kivimaa's framework was practical to adapt its applicability. It could be applied as a checklist in the analysis of experiments. Additionally, the “Objective of the experiment” (element 2) was useful to summarize the empirical characteristics of policy experiments in climate adaptation (Table 3) and gain a deeper understanding about their potential for transformative change. We found out that most policy experiments were employed as governing approaches to test and implement policy goals and strategies in climate adaptation. These findings could guide future research and generate cumulative empirical findings in the context of climate adaptation. Moreover, both reviews (Kivimaa et al. 2017 and this review) show that more academic attention is needed for [policy] experiments in the climate adaptation context. All in all, the application of this framework and its findings complement similar systematic reviews on experimental governance (Kivimaa et al., 2017; Laakso et al., 2017; van Doren et al., 2018).

6 CONCLUSION AND RESEARCH AGENDA

This paper addresses the existing insights in peer-reviewed literature regarding the roles and unexplored potential of policy experiments in climate adaptation. Through a systematic literature review, this paper examined empirical insights on the potential of policy experiments in the transformation of climate adaptation governance. It contributes to literature with cumulative empirical knowledge and provides suggestions for further research on policy experiments for climate change adaptation.

First, the findings contribute to a better understanding of the empirical characteristics of policy experiments in climate adaptation literature. This review analyzes the empirical characteristics of experiments dealing with climate adaptation issues. Although most experiments involved multiple climate hazards, sectors and actors, they are also limited in space (local, urban areas), time (short duration) and geographically (primarily in Europe). An additional finding is that policy experiments focused on climate adaptation required the involvement of civil actors among governmental actors.

Second, it provides insight into the role and outcomes of policy experiments as a means to induce change in climate adaptation governance. Regarding the role of policy experiments, most of them are used as governing approaches to test the technical or governing innovations and to implement climate adaptation strategies or policies. They are less important in developing new policies and evaluating policy actions/programs. Based on the policy experiments' outcomes, they are able to result in social learning rather than policy learning. To produce broader changes in policies and governance, the knowledge that emerges needs to be upscaled. Specifically, the tailor-made strategies could help in upscaling such knowledge (McFadgen, 2019; Wamsler & Pauleit, 2016). Nevertheless, it is still ambiguous how policy or social learning in practice result in policy change and what the most important enabling and hampering factors that influence this process are.

Third and finally, a few suggestions are offered for a future research agenda. It is necessary to gain more clarity on a common language that is used to define policy experiments and their influence at the policy level. This is valuable to better comprehend and compare insights on the matter studied by multiple scholars. Moreover, it is also relevant to systematically research how policy learning takes place and how to translate this learning into policy changes. It might help to develop policy learning approaches with practical examples from areas vulnerable to climate change, such as coastal or rural areas. A comparative analysis of multiple policy experiments could be a good way to draw sound empirical knowledge from practice. For example, such analysis could be conducted as a post-evaluation of the policy experiments a few years after they have ended to determine if and how upscaling happened in practice. Likewise, there is a need to further research how collaborative networks could facilitate transformative change using policy experiments as a governing approach in climate adaptation. Finally, this research could be complemented with a clear overview of context factors or barriers that either facilitate or disrupt the policy experimentation, its outcomes and the upscaling process.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.