INTRODUCTION

One of the recurring challenges with social-ecological system (SES) analysis and governance is integrating the understanding of their multidimensional, complex nature into practical decision-making processes (Leenhardt et al. 2015). Studying SES requires a multi-disciplinary analytical framework, a scheme through which one gains a focus on the structure and the elements that underlie complex systems (Ostrom 2009). A variety of analytical frameworks has been developed to study SES, each one with specific benefits regarding the focus and the purpose of the study (Binder et al. 2013). Notably, it has become customary in the literature to represent SES dynamics through illustrations of causal relationships between system variables, as seen, for instance, in causal loop diagrams (CLDs) (cf. Forrester 1961, Inam et al. 2015, Halbe and Adamowski 2019). However, such approaches fail to acknowledge that interactions responsible for explaining emergent social-ecological phenomena cannot be fully predicted based solely on a deterministic approach. Because key dynamics in a system can have repercussions on the system as a whole, just as systemic social-ecological phenomena can also affect system variables (Schlüter et al. 2019), recognizing these phenomena as well as their non-deterministic nature is essential to understand the overall systemic state and to anticipate their potential effects. Emergent phenomena can be, for example, the collapsing state of a system or an ecosystem governance regime shift.

Emergent phenomena in SES are difficult to predict and even more difficult to govern. There is thus a need for multidisciplinary research that identifies and evaluates SES dynamics that lead to specific phenomena of interest. The recently developed social-ecological action situations (SE-AS) framework (Schlüter et al. 2019) focuses on the action situation (AS) as a unit of analysis, like its predecessors the institutional analysis and development (IAD) framework and the SES framework (Ostrom 2005, 2007, McGinnis and Ostrom 2014). An AS, as stated by Ostrom (2005:32), exists “Whenever two or more individuals are faced with a set of potential actions that jointly produce outcomes (...).” The main difference with the IAD and the SES frameworks is what produces outcomes in the AS of the SE-AS framework. In the SE-AS framework, the concept of an AS is broadened to treat social actors and ecological elements on the same level, and to place their interactions at the center of the analysis. The SE-AS framework thus explicitly considers interactions between social and ecological elements and enables the analysis of those interactions. The focus on interactions among and across social actors and ecological elements allows the study of dynamics within an SES responsible for generating emergent phenomena, which include and affect both humans and ecological entities.

Using such a framework for SES research requires gathering knowledge about complex systems by combining different types of data. Three databases exist in which scholars gather data and knowledge on variables of SES in a structured way (these are the IFRI and NIIS databases for the IAD framework, and the SES library and SESMAD for the SES framework; see Bernstein et al. 2019; https://seslibrary.asu.edu/). Despite the existence of these databases, systematically linking analytical frameworks to data collection practices is not yet customary in publications (Bernstein et al. 2019). Studies often do not address the nature of the underlying data and the actual data analysis process (Ratajczyk et al. 2016). When studies of SES do not make explicit how a framework is used to build a conceptual model, showing little to no coherence in how categories are used, measured, or derived from empirical data, they leave little support to guide scholars aiming to use or apply such a framework (Thiel et al. 2015). This strongly diminishes the potential reach of any given framework in SES research, especially for communication with decision makers or for consistency within the research field. We contribute to filling this gap in the literature by describing how categories of AS can be derived from empirical data, and we do this by using and comparing two analytical methods. We apply the recently developed SE-AS framework, thus contributing to the practicability of its use.

Our objectives are thus to (1) compare the suitability of an interview content analysis with that of a CLD analysis to identify action situations and their interactions within an SES, guided by the SE-AS framework, and to (2) evaluate the potential and limitations of the SE-AS framework for comprehending the dynamics of SES and the phenomena these create. The reason behind the proposed comparison is to evaluate whether the AS configurations based on the SE-AS framework can be successfully generated through different methods, or whether some methods might be more appropriate than others for such a purpose.

Our case studies focus on human-induced alterations within two north temperate shallow lake ecosystems in Germany and Canada: Lake Dümmer is a shallow lake with a surface area of 13.5 km² located in southeastern Lower Saxony, North-Western Germany; Lake St. Charles covers an approximate area of 3.6 km² and is located 20 km north of Quebec City, province of Quebec, Canada. Both lakes are used for recreation, offer a fishing site, are habitats for birds and other wildlife species, and are impacted by human use, i.e., road salts and sewage being the main source of pollution at Lake St. Charles and nutrients from agriculture causing eutrophication of Lake Dümmer. In our analyses, we focus on the lakes and their catchment areas, including the feedbacks between the ecological and the social systems, and we use data generated in the context of a participatory research project. The representations of SES and the studied emergent phenomenon, which in our cases is lake ecological integrity, defined as “unimpared structure, composition and function [of the ecosystem] and a capacity for self-renewal” (Schallenberg et al. 2018:367), thus derive from the social depictions of those same systems. These social depictions were themselves identified through collaborative efforts with relevant stakeholders. Furthermore, we develop more comprehensive representations of the studied SES than have SE-AS framework studies to date. Our analysis thus contributes to the discussion on how the SE-AS framework can help expose a more detailed understanding of SES and their emergent phenomena.

CONCEPTUAL BACKGROUND

Qualitative data analysis in social-ecological systems (SES) research

When analyzing qualitative data that inform SES representations, a coding protocol that keeps a record of the analytical criteria and steps guarantees consistency in the analysis process (Anderies et al. 2016). This is just one element in a set of methodological tools, also including selection criteria, codebook manual or intercoder reliability testing, that together ensure sound qualitative research and a transparent record of qualitative data analysis (Ratajczyk et al. 2016). A transparent record of the analytical process helps guide the analysis and interpretation of data in light of any given analytical scheme, such as the SE-AS framework. A coding and analysis protocol enables replicability and comparability across case studies and provides a common language for variables and terms.

The study by Leslie et al. (2015) provides an excellent example of how to analyze an SES with qualitative and quantitative data within an analytical framework while transparently reporting the analytical steps. The SES library (https://seslibrary.asu.edu/) at Arizona State University and the Social-Ecological Systems Meta-Analysis Database (SESMAD) at Dartmouth College (SESMAD 2021) represent endeavors enabling comparisons of SES case studies based on databases of the studied variables and underlying coding manuals. These databases provide the grounds for comparable research on SES dynamics by relying on studies with transparent data operationalization and analysis processes. We further emphasize and illustrate this need for transparent data analyses when using frameworks in SES research.

Causal loop diagrams (CLDs)

We focus on the CLD modeling technique to (1) collect data and (2) lay the foundation of interrelated system variables, upon which we identify AS in the St. Charles case. A CLD presents a schematization of a system under study to reveal solutions for a focal problem (Haraldsson 2004) by describing system variables and events and the connections among and between them. A CLD represents “a dynamic system’s causal structure: variables, causal links with a polarity and symbols that identify feedback loops” (Schaffernicht 2010:653). A CLD thus reveals a system’s relevant elements, their relationships, and their repercussions.

Designing a CLD usually begins with the identification of a problem within a system, which gives the CLD its purpose (Richardson and Pugh 1981, Vennix 1996). The purpose can be to better understand the dynamics that surround the identified problem or to find potential solutions. Once the purpose is identified, the causes and consequences of the problem are identified as variables. Variables are linked through either a positively or negatively correlated relationship, signifying that a given first variable will affect a second variable in the same or opposite direction. Consequences of the problem are then linked back to causes either directly or through a series of variables. Loops in a CLD are formed when two (or more) variables are connected by links that form a circle comprising both variables (Fig. 1). Loops can be balancing (indicated with a “B”) or reinforcing (indicated with an “R”). A reinforcing loop means a variation in one variable leads to an intensification of the initial deviation of this variable throughout the loop; a balancing loop means that a variable within the loop dampens the intensification of a causal link, limiting an increase in the loop variables over time as a result of negative correlation between two variables (Vennix 1996, Haraldsson 2004, Inam et al. 2015). Causal loop diagrams are drawn on a researcher-specified temporal scale, and important delays are marked on the affected causal links (Haraldsson 2004).

The social-ecological action situations (SE-AS) framework

The SE-AS framework (Schlüter et al. 2019) expands on the concept of an action situation (AS) from the institutional analysis and development (IAD) framework (Ostrom 1990) and combines it with ideas from complexity science and resilience thinking (Folke et al. 2010). The concept of an action situation as a unit of analysis with which to study the behavior and interactions of social actors in the context of multilevel complex systems was elaborated within the IAD framework (Ostrom 2005). In the IAD framework, an AS consists of participants, their positions, potential outcomes, a function turning actions into outcomes, the control of this function, information and costs, and benefits related to these actions and outcomes (Ostrom 2005). The AS and the participants therein contained thus generate actions and outcomes that feedback into the system. The IAD suggests that so-called exogenous variables that are part of the system under study also influence AS and participants. Examples are biophysical conditions, attributes of community, and rules. The IAD thus strongly focuses on social dynamics and institutions to understand SES and how actors act and behave in those.

The SES framework, a predecessor of the SE-AS framework, was developed from the IAD as a response to natural scientists who found that the IAD was reductive by condensing all ecological processes into exogenous variables. The SES framework thus integrates the ecological part of an SES actively into the analysis: it centers the action situation, within which actors interact and create outcomes that shape the SES, in the middle of the resource system, resource unit, governance system, and actors (McGinnis and Ostrom 2014). In this framework, the exogenous variables of the IAD have switched to the center of attention and the ecological side, i.e., the biophysical conditions, figures more prominently in SES studies using the SES framework.

The SE-AS framework proposes recognizing the social and ecological spheres equally in an SES study, offering AS that consist of social actors and/or ecological entities, interacting with each other within a specific context. Action situations are thus defined by the very nature and quality of the social, ecological, or social-ecological interactions therein. Action situations themselves shape the system, influencing each other through outcomes and creating AS configurations that generate a system’s emergent phenomenon. Presented as a complementary approach to its predecessors, the SE-AS framework is thus able to capture dynamics, processes, and interactions within SES, on a variety of organizational levels (Schlüter et al. 2019). The framework has been created to analyze how emergent phenomena occur within SES.

The SE-AS framework maps configurations of AS, i.e., contextual interactions of system components that are hypothesized to underpin an emergent phenomenon of interest (Fig. 2). The SE-AS framework consists of (Schlüter et al. 2019):

• actors (A) and ecosystem elements (EE) that interact with each other at the micro level;
• purely social or ecological action situations (S-AS and E-AS, respectively) which, within the frame of a given analysis, define exclusively social or ecological interactions;
• social-ecological action situations (SE-AS) within which the human-ecosystem interactions of interest occur;
• outcomes, generated by AS, that are connected to and influence other action situations and that can also influence the very action situation that produced them;
• a configuration of action situations, situated at the meso level, resulting from the links created by different outcomes between different action situations. The configuration of AS represents an assumption about how interactions and outcomes have produced the emergent phenomenon of interest;
• the emergent social-ecological phenomenon, which can be observed at the macro level, and which feeds back to influence the AS configuration.

Schlüter et al. (2019) identified a non-exhaustive set of interactions that define the action situations and indicate their nature or characteristics. Consulting Pahl-Wostl et al.’s (2020) study of action situations in the context of water governance, we found that the list of the SE-AS framework’s social AS derived from Schlüter et al. (2019) could be further expanded by some of the categories used by Pahl-Wostl et al. (2020), namely, coordination, planning, application of measures, conflict resolution, and enforcement of rules. We thus added these categories to our list of AS (see Table 1; see also Appendix 1 for the coding protocol, which includes more details on AS categories). We also included in Table 1 the AS that we created based on the analysis of our case studies.

Based on this section’s concepts, we propose the following terminology:

• AS configurations: relate to the configurations of AS that are drawn for each case study, using the interview data and based on the SE-AS framework.
• Social representation: refers to a narrative of the SESs as perceived by the actors.
• SES representation: relates to the illustration of the SESs under study, presented through AS configurations and based on the actors’ social representations of the system.
• Framework: relates to the SE-AS framework, unless stated otherwise.

METHODS

To gather our data, we conducted interviews with relevant stakeholders in the two case study regions, Lake Dümmer in Northern Saxony, Germany, and Lake St. Charles in Quebec, Canada. The interview questions were structured along the logic of causal loop diagrams, asking about a central issue, its causes and consequences, and their repercussions to form questions (see Appendix 2, Table A2.1 for the interview questionnaire). We did so because the technique of revealing links between an issue, its causes, consequences, and potential feedbacks is easily explained, and thus more accessible to participants. Although we are aware that this approach differs from the one used by Schlüter et al. (2019), the participatory nature of our study called for such an approach. We considered that explaining the concept of AS and having participants understand and use it effectively within the timeframe imposed by the interview process did not seem realistic. That said, although the interview questions were based on CLDs structure, we used two distinct analytical methods to conceptualize AS configurations, using one analytical method for each SES, and focusing on lake ecological integrity as the emergent phenomenon (see Fig. 3 for a summary of our methodology).

The first analytical method is based on the content analysis of interviews. For this, we applied a qualitative coding scheme to the transcribed interviews with which we assessed social, ecological, and social-ecological AS within the SES that the participants had described (for the coding scheme, see Table 1, with more information on the categories in Appendix 1). For the other method, we constructed causal loop diagrams that were merged into a unique CLD as a preliminary analytical step (CLD analysis) before identifying AS. The rationale for using two different analytical methods for identifying AS and their configurations was to examine whether different approaches can lead to comparable representations when those latter are based on the SE-AS framework, or whether one approach may be superior to the other in certain aspects.

It should be noted that, compared to the AS configurations originally presented by Schlüter et al. (2019), we focused on more detailed representations of the studied systems, with many AS integrated into our configurations instead of only a few main ones. As Schlüter et al. (2019) wrote, “[t]he framework is flexible with respect to the level of abstraction and detail of an analysis (...).” We opted for a more detailed analysis for three reasons: (1) we were interested in exploring the SE-AS framework suitability for different adaptations, beyond Schlüter et al. (2019) original proposition, (2) the data collected through interviews permitted detailed representations, and (3) with stakeholders directly involved in both cases, it was relevant to communicate our results with more detailed SES representations instead of broad generalizations that might be of little practical use to them.

The two case studies

The watershed of Lake Dümmer covers an area of 346 km² (Hochschule Osnabrück 2013). Its waters are fed by the River Hunte, which runs from south to north through the lake. The western and southern shores, as well as the lowlands east and south of the lake, are nature reserves (Landkreis Diepholz 2020). The lake is administered by two districts in the federal state Lower Saxony, Germany, and around 8000 inhabitants live close to the lake (for the lake’s location within Lower Saxony, Germany, see Fig. 4).

Currently, the lake stores water that is used to regulate the River Hunte levels, protecting downstream communities from floods. The lake further serves as a recreational area for many tourists throughout the year, while being home to a rich biodiversity, including a variety of iconic flora and fauna habit (FFH) species and migratory birds (Council of the EU 1992, Landkreis Diepholz 2018; FFH refers to the EU Council Directive on the conservation of natural habitats and of wild fauna and flora). A redevelopment project is underway for the lake to reduce nutrient loading resulting mainly from agriculture in the Hunte River Basin.

Lake St. Charles’ watershed covers an area of 165.8 km² (Légaré 1998; Fig. 5). St. Charles River flows from the lake into the St. Lawrence River, hence the lake also lies in that river’s watershed, which extends up to 550 km². The higher St. Charles River watershed, created by the water intake on the river downstream of the lake, covers around 348 km² over five municipalities: Stoneham-et-Tewkesbury, Lac-Beauport, Lac-Delage, Quebec City, and Saint-Gabriel-de-Valcartier, as well as over an urban territory of the Indigenous Huron-Wendat Nation, Wendake (APEL 2009, 2017). Although Stoneham occupies the greatest extent of land area in the watershed (41%), Quebec City owns most of the riparian area around the lake, with around 80% of the shoreline falling under its jurisdiction, while the remaining belongs to Stoneham (APEL 2009). The population in the watershed of the upper St. Charles River was around 19,000 people in 2016, mainly concentrated along the tributaries and around the lakes (APEL 2019).

The lake became a semi-natural reservoir, with a hydro-electric current capacity of 14.8 Mm³, after the construction of the Cyrille-Delage dam in 1934 and reconstruction in 1948, which raised the water level approximately 3.6 m above natural historical levels (APEL 2009, OBVC 2015). The northern part of the lake drains an important protected wetland known as the Northern Marshes, which together with the lake and the watershed are home to a rich bird diversity and provide a handful of ecosystem services, notably recreational opportunities. In addition, Lake St. Charles is the major water source (up to 60% during summer) for the drinking water filtration facility located 11 km downstream on the St. Charles River (APEL 2017). The distribution networks provide drinking water to a population of more than 300,000 residents living in the Metropolitan Community of Quebec.

Common basis: interviews

At Lake Dümmer, XX interviewed 18 key stakeholders in the spring and summer of 2019, while YY interviewed 22 key stakeholders at Lake St. Charles in the fall of 2019. Interviews were semi-structured, following a list of questions, but leaving the interviewer the freedom to ask additional questions (for the questionnaire, see Appendix 2, Table A2.1). Interviews lasted between 45 minutes and 3 hours and most of them were recorded for the Dümmer case, while all were recorded for the St. Charles case.

Using purposive sampling (cf. Palys 2008), interview participants were identified based on their organization’s importance for either the management, the use, the protection or the assessment of each lake, their natural resources, and their catchment area. In the St. Charles case, guidance from a local NGO was also provided to identify which actors were the most actively involved in the lake watershed management or use. The interviewees included: state authorities from the federal, regional, and municipal levels; conservation and other nature-oriented NGOs; collective actors from the tourism sector; private entrepreneurs; Chambers of Agriculture; consulting firms, and; water associations (for a list of the interviews of the Lake Dümmer case, see Appendix 3, Table A3.1; due to Quebec standards for maintaining participant confidentiality a similar list cannot be provided for the Lake St. Charles case).

Because CLDs are a convenient tool with which to represent a system’s variables and their connections, we decided to structure the interviews along the logic of CLDs to identify the stakeholders’ comprehension of the SES. We also explicitly asked stakeholders about the human-nature interactions they observed at the lakes. The purpose of the interviews was thus to construct each interviewee’s social representation of the main issue within the studied SES, causes and consequences associated with this issue, and the relations and feedbacks between identified variables.

Identifying action situations through text coding: case study 1, Lake Dümmer

The interview recordings of Lake Dümmer’s stakeholders were transcribed with the automatic transcription software AmberScript. To identify the interactions of actors and ecosystem components, three researchers coded the stakeholder interviews along the set of AS categories as defined by Schlüter et al. (2019) and Pahl-Wostl et al. (2020; see also Table 1; Appendix 1) and coded the outcomes these AS created.

The coding process was done via a qualitative content analysis using the qualitative text analysis software MaxQDA. The coders guaranteed intercoder reliability by reviewing each other’s coding and discussing subsequent differences. The addition of new types of AS to the list was permitted if needed. The identification of action situations and their outcomes was thus an iterative deductive and inductive process: researchers met regularly to discuss the coding process, the necessity of newly created AS categories, the interpretation of AS, and their connections via outcomes. The coding protocol supported the coding process by recording our observations of the process itself, the main issues of discussion, and the decisions made. The protocol also included a guideline of how to code and a glossary of the terms action situation, outcome, and phenomenon (for the protocol, see Appendix 1). The process helped identify 34 AS categories within the interviews.

To determine the relevant AS that explain the phenomenon of interest within the SES, we first identified a core set of AS: we started with the SE-AS that were more frequently mentioned by the interviewees, appearing to be the focus of the social representations. We then laid out the other coded AS that related to these initial ones and the connections among them, as described by participants, thereby creating a first version of AS configurations. In the second step, we checked whether the AS were constituent for the emergent phenomenon of interest. If that was the case, we analyzed their elements and their interactions to define their outcomes and subsequently related AS. We continued in a similar manner until we could not find any new necessary AS to explain the emergent phenomenon. For the Dümmer case study, 19 action situations were considered relevant to the analysis (for an overview of the action situations identified, see Table 1; Appendix 1).

Identifying action situations through CLDs: case study 2, Lake St. Charles

During the interview process, we offered to create a CLD together with the interview participants, using a large sheet of paper and sticky notes. When participants expressed a preference for an oral discussion instead, CLDs were created later on, with recordings of the interviews inserted in ATLAS.ti 8 for coding. We used deductive coding to identify the perceived causes and consequences associated with the main issue discussed (cf. Crabtree and Miller 1999), thus identifying CLD variables. Generated CLDs were then transferred into digital visualizations using the software Vensim (Ventana Systems Inc.). To obtain an overall impression of stakeholders’ social representations of the lake SES, we merged the individual CLDs, starting with the most comprehensive individual CLD and adding in elements from other CLDs until every relevant element had been incorporated (Appendix 4, Fig. A4.1; see Inam et al. 2015 for the six steps of how to merge CLDs, select relevant variables, and deal with controversies). Thus, a merged CLD does not embody one coherent, aggregated social representation of a given system, for there is no such thing as a single consistent representation of a system. A merged CLD is a bricolage of CLDs collected by a researcher and comprises the elements that constitute the most comprehensive representation of the system and its core challenges.

All variables were identified as either social, ecological, or social-ecological to prepare for the following analytical step, and all causal links between variables within the merged CLD were weighted based on the number of stakeholders who mentioned them. It should be noted that a merged CLD is not meant to simply aggregate all variables nor illustrate a consensual social representation of a system. It rather uses convergent perceptions between different social representations as a backbone to build upon, while divergent perceptions can be illustrated in several ways, such as through several conditional causal links.

Action situations and their outcomes were derived from the merged CLD, and a configuration of AS was hence constructed to explain the emergent phenomenon of interest, lake ecological integrity. When examined through the SE-AS framework lens, variables or groups of variables within a CLD can represent:

• the emergent SES phenomenon (macro level);
• configurations of several AS or some single AS (meso level);
• outcomes, interactions between and among actors and ecosystem elements, or single actors or ecosystem elements (micro level).

To facilitate the conceptualization of the AS configuration from the merged CLD, we color-coded the merged CLD variables or groups of variables based on whether they represented social AS, ecological AS, social-ecological AS, the emergent phenomenon, external AS, or emergent outcomes. We used the preliminary classification of CLD variables to guide the identification of the different types of AS. Variables were considered to be part of an AS if we could identify actors taking part in the described variables, and emergent outcomes if the variables were rather an expression of an ongoing process without any identifiable actors or interactions relevant for the analysis (Appendix 4, Fig. A4.2). More detailed examples of this process are given in the discussion.

As with the Lake Dümmer case, we classified AS according to the categories from Schlüter et al. (2019) and Pahl-Wostl et al. (2020; Table 1; Appendix 1), while remaining open to new categories that might not have been identified in previous publications; thereby adopting a primarily deductive approach completed by an inductive one. The selection of relevant CLD variables for constructing AS was based on those most frequently mentioned during the interviews, as well as those with the greatest number of causal links.

To evaluate how the identified AS and outcomes interact, we considered only the strongest causal links between variables, as illustrated within the merged CLD. The strength of causal links needed to be evaluated comparatively to others for each variable, because not all variables shared the same number of mentions during the interview process. Thus, no causal loop strength “limit” was established for the entire merged CLD. Furthermore, in instances when two AS were linked without any identified emergent outcomes in between, causal links were considered as additional outcomes and described accordingly to better translate the merged CLD into a configuration of AS. Once all AS and outcomes were identified in the merged CLD, the AS configuration was redrawn as an SES representation based on the SE-AS framework.

ACTION SITUATION CONFIGURATIONS BASED ON THE SE-AS FRAMEWORK

The two case studies revealed different anthropogenic actions, social-ecological interactions, and ecological processes affecting lake ecological integrity. Furthermore, we identified distinctive feedbacks within the SES, in which actors were involved in regulating and managing emerging dynamics and phenomena. It should be noted that, contrary to the phenomena illustrated in Schlüter et al. (2019), we illustrate a phenomenon (lake ecological integrity) that is more representative of the overall system state compared to a more measurable phenomenon such as eutrophication. The reason for this choice is that we included both AS responsible for eutrophication or for perturbing natural ecosystems, as well as measures taken to counter such negative effects. Therefore, our emergent phenomenon can express either the maintenance or the deterioration of the lake’s ecological integrity.

Dynamics and feedbacks within the Lake Dümmer SES

The ecological integrity of Lake Dümmer is challenged by an oversaturation in phosphorus, which alters internal lake dynamics and affects the ecosystem services provided regarding recreation and repercussions on the system (Fig. 6; Table 2).

Overall, the AS configuration shows us that the eutrophication process of the lake, stemming mainly from agricultural activity nutrient input, comes with a range of negative consequences for the SES. Those are intensified by the artificial regulation of water levels. Both the source and the symptoms are counteracted by a range of measures. These measures are the result of a rule-making process and are continuously supported by ongoing coordination, planning, and deliberation of the actors involved in the lake’s management.

Dynamics and feedbacks within Lake St. Charles SES

The ecological integrity of Lake St. Charles has long been of concern for interested parties, but it only truly became an urgent matter in 2006 when significant cyanobacteria blooms began to appear in the lake (APEL 2009). However, because the governance system surrounding Lake St. Charles was in a transition phase at the time of the interviews, participants expressed social representations of both the past and the future envisioned state of the system (see Appendix 4, Fig. A4.2). For the sake of the present study, the AS configuration solely includes a recent past and present depiction of the system (Fig. 7; Table 3), thus excluding measures that were proposed by stakeholders as potential solutions yet to be implemented to solve issues related to the state of the lake, or so-called “vision measures” (see Appendix 4, Fig. A4.2).

To summarize, the cycle of inadequate governance, followed by social conflicts and even more residential development, which lead to increased pressure on the lake and its resources, appears to be central to the deterioration of the lake ecological integrity. However, as seen in Figure 7, governance could also be a catalyst for ensuring the maintenance of this ecological integrity, notably through more coordination among stakeholders, which would help in taking different measures to decrease the effects of human actions on the lake.

DISCUSSION

To discuss the practicability and relevance of the two analytical methods used to develop AS configurations based on the SE-AS framework, we start with an exploration of the comparability of the AS configurations identified for our two case studies. We then dive into the challenges of using qualitative data to construct such configurations and highlight the added value and limitations of the SE-AS framework for studying SES, as encountered in our two case studies.

Comparability of the two case studies

Developing AS configurations through the interview content analysis and the CLD analysis allowed us to identify key AS responsible for generating the emergent phenomenon under scrutiny, lake ecological integrity. Interviews were structured similarly for both case studies, and close collaboration between the teams helped to ensure that any differences in content between the two AS configurations successfully reflect real variation in the interview data rather than any differences in the methods used for developing the configurations. Such an observation is fundamental to highlight the practicability of the SE-AS framework because it underlines the fact that, despite being used in different contexts and with different analytical methods, it still allows for comparisons to be made between two different cases.

Although there are similarities between the two studied SES regarding ecosystem services (e.g., recreational activities), the handling of water levels, and the symptoms of water quality degradation (presence of cyanobacteria), the AS responsible for the emergent phenomena are different. The main activity responsible for water quality degradation in Lake Dümmer is agriculture, as observed in the AS Polluting (Fig. 6), reflecting the interaction between farmers and the land. In the Lake St. Charles watershed, residential development, including associated road networks and wastewater, is responsible for deteriorating the water quality, as seen in the AS Converting (Fig. 7). Measures taken to counter the decrease in water quality are also quite dissimilar, with a remediation plan, volunteer farmer actions, ecological manipulation, and hydrological alteration as the main solutions already in place for the Lake Dümmer case, while stakeholders from Lake St. Charles proposed an integrated governance system, better wastewater management, and ecosystem protection as measures that should be taken in the future.

The AS configurations also point out which of the three parts of the SES, i.e., social, ecological, or social-ecological, is most prominent in generating the emergent phenomenon. In the Dümmer case, ecological AS are limited to four while there are nine social-ecological AS and six social AS. In the St. Charles case, there are 3 ecological AS, 8 social-ecological AS, and 14 social AS. This reflects that, in terms of proportions, stakeholders from the Dümmer case focused more strongly on social-ecological AS, which they consider to be mostly responsible for influencing water quality, whereas in the St. Charles case, stakeholders attributed responsibility mostly to social AS, as a result of governance failure and of other potential social AS that were hampered by such failure.

Methodological challenges and comparison

The methods we used do have challenges.

The interview content analysis

When coding the AS from the interviews, several new questions arose. The first one was related to the level of detail to apply. For instance, while we initially decided to code implicit AS mentioned in the participant discourses, we eventually realized that they were not relevant for explaining the emergent phenomenon. In the course of the coding process, we also recognized the need for two new AS categories. The first was (Non) transporting, an ecological AS for which there was no equivalent in the list of AS extracted from the literature, to describe the movement of an ecosystem element transporting another element or substance within the system space. Secondly, Ecological manipulation was created to account for regular perturbations with potential effects for food web dynamics, e.g., by inhibiting pest species or supporting desired ones. This AS category was distinguished from the AS category Converting, which considers ecosystem change through technology or by restoring or converting use to protect ecosystems. Our novel category accounts for the more dynamic nature of ecological manipulation. Finally, we also realized that some interactions could be assigned to several AS categories because some of them represent broader categories. For example, interactions that describe an ecological manipulation or a conversion were often additionally coded as an application of measures. However, the former categories (either Ecological manipulation or Conversion) were favored in the AS configuration because they described those interactions more precisely. For instance, in Figure 6, Ecological manipulation 2 (mowing macrophytes) is, in fact, a measure applied by stakeholders. But stating that this is an application of measures would be less precise and less informative than stating that it is an ecological manipulation.

The CLD analysis

We found that producing a configuration of AS based on a CLD was a relatively fast process. We also found that, once the merged CLD based on interview data was completed, it could be reused relatively efficiently to develop AS configurations based on different emergent phenomena under scrutiny. Still, because we had to construct CLDs before identifying the AS configuration, we were faced with all known limitations associated with creating and merging CLDs. However, these limitations have been extensively covered in the literature (e.g., Haraldsson 2000, Schaffernicht 2010, Inam et al. 2015, Bureš 2017), and we could rely on this accumulated knowledge to ensure proper use and merging of CLDs.

Apart from elaborating CLDs, the main challenge to developing an AS configuration based on CLDs was the dissimilarity in structure. Converting a series of diverse variables and causal links into interactions between actors or elements and the emergent outcomes was not a straightforward process and was somewhat subjective. However, using a primarily deductive approach helped reduce subjectivity because AS categories described in the literature could be found relatively easily in the merged CLD by isolating relevant variables or grouping them with their causal links when necessary. Only when important variables failed to fall into previously described AS categories did we use an inductive approach to describe a new type of AS.

Another challenge came from translating the CLD variables and links, which were mostly devoid of actors, into AS involving mainly actors and their interactions. This required an additional level of interpretation in which we had to bring appropriate actors back into the relevant AS by referring to the interview data when necessary. For instance, the CLD variables “residential development” and “road network” were combined to express the AS “Converting” in which developers were brought back as actors who interact with the land, the ecosystem element. One of the outcomes of this AS, represented by the CLD variables “outmigration from the city center” and “attractiveness of the living environment” were combined as a single emergent outcome from AS “Converting.” That is because, even though migration does imply some actors, no relevant interactions existed within these latter variables that would have helped analyze the lake ecological integrity. Rather, the other AS that influenced outmigration, which are social conflicts and recreation, were more relevant for understanding the dynamics affecting the phenomenon of interest. They were thus described as AS, whereas migration remained an outcome, described as “citizens leaving urban center.”

Comparing the two analytical methods

Although no analytical method was preferred over the other because both methods were used for analyzing a given case study, comparisons can be drawn from our study. The most obvious advantage of the interview content analysis over the CLD analysis is that it allows for translation of the interview data directly into AS configurations. It is thus less subject to information loss that may happen when transferring raw information (such as interview data) into a model (such as the SE-AS framework). In the interview content analysis, this process of translation is done only once, whereas it needs to be done twice for the CLD analysis, i.e., from the interview data to the merged CLD, and from the merged CLD to the AS configuration. This second analytical method might thus be more prone to information loss during the analysis.

The CLD analysis also has benefits compared to the interview content analysis. First, we found the CLD concept to be easier to explain to participants in an interview context than the SE-AS framework. Thus, we could structure the interview a priori in a way to make the interview data easy to integrate into CLDs. Second, it was easier to define links between AS when those AS were based on CLD variables because the latter were already structured in causal relationships. This is especially true when compared to the interview content analysis for which establishing links between AS, solely based on the interviewees’ responses, added another level of complexity during the analysis.

Despite these comparative strengths and weaknesses in our two analytical methods, what our two-pronged methodology demonstrates is that the SE-AS framework can be used successfully with different analytical approaches to link empirical data with the stylized models proposed by the SE-AS framework. According to our experience, we found no real “superior” analytical method to identify AS and create AS configurations. The need and familiarity of researchers with specific methods might thus be the most important element to consider for identifying AS. Indeed, despite the high level of complexity revealed by our case study analyses, they were made comparable through the use of AS configurations, as described in the SE-AS framework.

Methodological limitations

A first limitation comes from the fact that variables mentioned as important during the interviews were likely more representative of the better-known variables rather than those with greatest influence on the systems. Although these variables could be the same, we cannot assume this. To work around this, we engaged with diverse involved stakeholders to ensure that different types of knowledge were included in our system representations, and we also remained transparent that our AS configurations reflect the stakeholders’ social representations of the SES and are thus not meant to be comprehensive nor consensual. Other limitations relate to data gathering methods and the nature of AS.

Data gathering

The fact that interviews were structured around the generation of a CLD could have limited, to some extent, the data obtained. Participants were asked about human-freshwater interactions they observed, but they were mainly asked to think in causal links rather than in terms of interactions between elements and actors of the SES. Hence, subtleties of any given interactions might have been lost. However, having stakeholders think through causal links helped identify a diverse array of system variables, producing a complete picture of their social representation of the system’s main issue. Also, based on our experience, we estimated that CLDs would be more accessible to participants, being easily explained and generated within the time allowed by the interview process. In comparison, it would have been difficult to structure interviews around the concept of AS and their outcomes, while not knowing beforehand what phenomenon we were looking for and which interactions were responsible for its manifestation.

The nature of action situations

We mentioned at the beginning of the discussion how social and social-ecological AS were perceived as more important by interviewees for generating the emergent phenomenon in our case studies. Still, it should be highlighted that regime shifts in our case studies are produced by ecological AS and their outcomes, hence by ecological non-linearities, which can trigger a change in the ecological integrity of the lakes under study. The SE-AS framework is an attempt to bring the social-ecological into focus because these interdependencies are of great importance (see Lade et al. 2013, 2015, Martin et al. 2020). However, what is considered most relevant in an AS configuration of an SES depends on both the purpose of the analysis and, in our case, on the point of view of interviewees. If the purpose is to find ways to manage or govern an environmental problem, interactions that can be manipulated or influenced will be of more importance than those that are out of human’s control (e.g., the phosphorous in the sediments). Similarly, because most of the stakeholders involved in our study are interested in the management of the lake and are less knowledgeable of its ecological aspects, it probably influenced their stronger focus on social elements of the systems.

General strengths and limitations regarding the use of the SE-AS framework for studying SES

As stated by Schlüter et al. (2019), AS configurations of an emergent phenomenon open space for discussing the mechanisms underlying the phenomenon, challenging common sense knowledge about alleged causes, generating insights about formerly unknown factors influencing the SES, or juxtaposing different AS conceptualizations of a phenomenon. The framework advises to “(...) clarify different understandings as a basis for further exploration and learning.”

Strengths

First, in our cases, the SE-AS framework was useful for illustrating the main dynamics underlying an emergent phenomenon in a simple yet comprehensive manner. As mentioned, although we used different analytical methods to develop AS configurations from our two case studies, we concluded that the two AS configurations permitted comparisons of the cases. The SE-AS framework also helped in pointing out which of the three AS archetypes, i.e., social, ecological, or social-ecological, were perceived as the most important in each case to generating the emergent phenomenon, lake ecological integrity.

Second, the SE-AS framework provided an essential tool to make sense of the information provided by Lake St. Charles’ merged CLD. Indeed, the SE-AS framework allows disentangling variables for which causal relationships are undefined or non-linear, by treating and interpreting such situations as single ASs. The framework is thus especially useful to help deal with highly complex and dense CLDs, which can result from merging many different CLDs. An especially eloquent example of the SE-AS framework added value for analysis and communication can be observed by comparing the AS configuration for the St. Charles case (Fig. 7) to the equivalent system CLD (Appendix 4, Fig. A4.1). Although the former can be understood fairly rapidly, including necessary measures to counter the deterioration of the lake ecological integrity and involved actors, the latter is much harder to grasp. Similar analytical interpretations about the system dynamics might be derived from both CLD and SE-AS system representations but it is much harder to derive insights from CLDs when they are overly complex.

Third, one of the main arguments we see for using the SE-AS framework is its potential for interdisciplinary and cross-sector communication. It provides a tool allowing the combination of ecological and social elements into a coherent whole while focusing on the interactions between those elements instead of treating them as isolated entities. Its basic unit of analysis, the action situation, helps communicate the role of stakeholders and ecosystem elements in generating the phenomenon under scrutiny. The framework thus highlights areas in the SES in which future actions need to be taken, thereby informing decision-making processes. The added value for policy communication is that using the SE-AS framework enables rapid understanding of an SES state and dynamics and the location of the collective actors in the system responsible for them. Action situation configurations may facilitate sustainable resource governance by indicating whose actions need to be altered and where in the system measures should be implemented. The SE-AS framework is thus also superior to CLDs in this regard because it helps position actors and highlight their roles in the management of SESs, hence it communicates a clear message for decision makers.

Fourth, the SE-AS may help to explain governance failure within a system. Because the AS configurations of our case studies are the result of analyzing SESs’ social representations of interview participants, these configurations reflect a subjective understanding of the system rather than an objective one. Thus, the SE-AS framework is a suitable tool to show how social representations of the lake and scientifically described social-ecological processes converge or differ. These differences may reveal to resource managers and resource management scholars how their view of common SES compares to different perspectives with differing emphases. Such divergence may lead to suboptimal decisions and may eventually explain pitfalls of lake governance and decision making in the face of social-ecological issues.

Limitations

Despite the strengths of the SE-AS framework, we did encounter limitations that should be acknowledged in future research. First, the SE-AS framework cannot, in itself, elucidate which AS are the most crucial for generating any given phenomenon, nor was it created for such a purpose. This must be well understood by researchers interested in using it. In our case, identifying relevant AS was accomplished as a result of data analysis, as well as through researchers’ knowledge and understanding of the focal systems. However, our analytical methods presented their own challenges. For instance, in the Dümmer case, identifying AS based on the interviews led to a large number of observed AS, illustrating a diverse array of interactions between different actors and ecosystem elements within the SES. Defining the emergent phenomenon at stake and the AS relevant to explain it was not an intuitive process and required a great understanding of insights provided by stakeholders. A coding protocol helped to guide the analysis and avoid differing analyses of the interview data.

Second, the distinction between AS and outcome is often unclear. The definition of outcome by Schlüter et al. (2019) remains rather vague: “Outcomes of one AS may affect rules, structures, or participants of another AS.” In some of the outcomes we identified, their elements intermingled with elements of other action situations or could be interpreted as interactions themselves. To overcome this, we discussed the specific cases among researchers to decide whether we would interpret the observed action (or interaction) as an outcome of an AS or as a new interaction. One example is the AS Transporting in which river and lake water interact. We discussed whether the “outwash of nutrients”, which is the outcome of AS Polluting, would be sufficient to describe the process of nutrients entering the lake water. It became problematic when we identified the AS Converting (re-naturalizing the tributary) which benefits the river water quality. In this case, we found that transporting of river water into lake water needed to be identified as an AS itself. The AS Converting affects the AS Transporting through the outcome “reduces nutrients”. This aspect may reflect that in an SES, the elements and their relations to each other are in a continuous state of change, implying that the interactions and their interrelations through outcomes are in an evolving process. From such a perspective, what one defines as an AS, or process of an interaction, and what one defines as an outcome then depends on the researcher’s perspective and might be relative to time, other AS, or even, as in the example below, a question of the resolution the researcher chooses to capture the AS and their interrelations.

Third, regarding the assignment of the AS categories to the identified interactions, we observed that some interactions could be interpreted through several AS categories. The Dümmer council, for instance, represents a deliberating as well as a networking interaction of actors. The researcher needs to decide which of the interpretations better express the AS contributing to the emergent phenomenon, or whether both should be included as two separate yet interlinked AS. The same accounted for the social AS categories Information sharing and Deliberating, which were often coded together, and the ecological AS Facilitation and Species-habitat interaction. A thorough discussion of the specific interactions in question is needed to assign them to their AS category.

Fourth, interactions in an SES may happen at different organizational levels (local, regional, national). The framework, however, does not easily allow to distinguish the levels at which AS occur in its SES representation. Schlüter et al. (2019) suggested how to reflect on the local and the global level in their representation of the potential triggers of a global outbreak of H5N1 avian influenza. However, that representation is less detailed than what we present. Other AS configurations may comprise more levels and more AS. Researchers focusing specifically on different levels may want to consider this point and find a workaround for using the framework.

Fifth, we found that, as is the case with CLDs, the time component was hard to express within the SE-AS framework, even more so than in CLDs. The framework simply does not cover time spans. If one studies an SES in which AS occurred at different points in time but are still affecting each other, it is difficult to illustrate this time difference in the representation. For instance, in the St. Charles case, interview participants had described both past and future variables in their representation of the system. Hence, different time periods were expressed in the merged CLD, and we found no convincing way to translate this idea into the AS configuration. Therefore, we chose not to include so-called “vision measures”: variables representing potential measures proposed by participants as ways to improve the overall lake SES. Components that happened in the past in the Dümmer case study were indicated as external factors (see Fig. 6). One possible workaround for such a limit could be to have different AS configurations for different time spans.

Finally, the SE-AS does not propose a set of prescribed AS meant to be comprehensive in explaining emergent phenomena. It is, on the contrary, open to new AS categories that may be relevant for any given research needs. This might be seen as a weakness because it reduces the analytical capabilities of the framework, but it gives it more flexibility, which can also be seen as a strength. Either way, it needs to be understood by researchers, so they can take full advantage of the SE-AS framework and use it accordingly. For instance, in our case, we observed the lack of a social AS that explicitly described the establishment of trust, social coherence, and mutual understanding; aspects important in resource management, which do not necessarily manifest in an AS. However, they are relevant to understanding social dynamics. In the St. Charles case, the trust variable, as well as lack of trust, were considered to be part of the AS Conflict resolution and Conflicts, respectively; social coherence was included in the AS Networking. For future research, we encourage SES scholars to explore possibilities for including AS that describe the state of mind of actors within interactions, regardless of whether those be conflictive or consensual.

CONCLUSION

We compared two analytical methods to use interview data and develop AS configurations based on the SE-AS framework developed by Schlüter et al. (2019), examining two case studies centered on lake SES in Germany and Canada. We illustrated how different SES dynamics could be observed within simple representations of the case studies’ SES. Using the SE-AS framework to do so revealed several benefits. The framework helps disentangle variables for which causal relationships are ill-defined or non-linear, thereby creating clarity in the analysis of complex systems. It serves interdisciplinary research of the social-ecologically intertwined nature of SES through the possibility of combining ecological and social knowledge and variables. Moreover, it could potentially provide a communication tool outside academia that illustrates the role of stakeholders and ecosystem elements in an SES, while stimulating discussions on the underlying mechanisms of SES.

In terms of analytical limitations, we found that different organizational levels of AS (local, regional, national) are difficult to represent and that the time factor cannot be illustrated easily in a single diagram. Further aspects to consider when using the SE-AS framework are that the definition of “outcome” remains vague and that identifying outcomes properly in the SES can be challenging. Also, several AS categories may describe a single interaction in the studied SES. It is up to the researcher to decide which category is more significant to include in the AS configuration. Finally, future studies may need to include social AS that reflect actors’ trust, social coherence, and mutual understanding to better illustrate social dynamics, all of which, it should be noted, are aspects included in Ostrom’s questions and variables used throughout her studies. Adding some of these variables to the SE-AS methodology would enrich the analytical capturing of interpersonal relations.

The need for research that goes beyond coupling social and ecological data, by putting social-ecological interactions at the center of data gathering and data analysis, is increasing rapidly as climate change and its resulting uncertainties and imbalances manifest in a plethora of SES around the globe. Future research could study how AS configurations can inform computational models to design simulations of the systems under study, as suggested by Schlüter et al. (2019). Translating the identified interactions and their respective outcomes into quantifiable measurements would be indispensable as a next research step. The methodology introduced here can help comprehend complex systems and organize knowledge of diverse stakeholders of the system in an elegant manner, simple enough for decision makers to grasp their essential parts, yet comprehensive enough to provide a fundamental understanding of dynamics behind any given phenomenon of interest. The SE-AS framework can help inform decision makers on the governance structure in place, which can become key to developing greater resilience in any SES system, such as freshwater bodies, facing increasing climatic disturbances and uncertainties.

LITERATURE CITED

Anderies, J. M., M. A. Janssen, and E. Schlager. 2016. Institutions and the performance of coupled infrastructure systems. International Journal of the Commons 10(2):495-516. https://doi.org/10.18352/ijc.651

Bernstein, M., M. del M. Mancha-Cisneros, M. Tyson, U. Brady, C. Rubiños, H. C. Shin, S. Vallury, S. Smith-Heisters, and E. Ratajczyk. 2019. Mapping Ostrom’s common-pool resource systems coding handbook to the coupled infrastructure systems framework to enable comparative research. International Journal of the Commons 13(1):528–552. http://doi.org/10.18352/ijc.904

Binder, C. R., J. Hinkel, P. W. G. Bots, and C. Pahl-Wostl. 2013. Comparison of frameworks for analyzing social-ecological systems. Ecology and Society 18(4):26. https://doi.org/10.5751/ES-05551-180426

Bureš, V. 2017. A method for simplification of complex group causal loop diagrams based on endogenisation, encapsulation and order-oriented reduction. Systems 5(3):46. https://doi.org/10.3390/systems5030046

Council of the European Union. 1992. Richtlinie 92/43/EWG Des Rates Vom 21. Mai 1992 Zur Erhaltung Der Natürlichen Lebens-räume Sowie Der Wildlebenden Tiere Und Pflanzen. Council of the EU, Brussels, Belgium. https://eur-lex.europa.eu/LexUriServ/LexUriServ.do?uri=CONSLEG:1992L0043:20070101:DE:PDF

Crabtree, B. F., and W. M. Miller, editors. 1999. Doing qualitative research. Second edition. Sage, Thousand Oaks, California, USA.

Diercke Weltatlas. 2021. Diercke Weltatlas - Kartenansicht - Physische Übersicht. Westerman Group, Brunswick, Germany. https://diercke.westermann.de/content/physische-%C3%BCbersicht-978-3-14-100754-1-3-1-0

Folke, C., S. R. Carpenter, B. Walker, M. Scheffer, T. Chapin, and J. Rockström. 2010. Resilience thinking: integrating resilience, adaptability and transformability. Ecology and Society 15(4):20. https://doi.org/10.5751/ES-03610-150420

Forrester, J. W. 1961. Industrial dynamics. MIT Press, Cambridge, Massachusetts, USA.

Halbe, J., and J. Adamowski. 2019. Modeling sustainability visions: a case study of multi-scale food systems in Southwestern Ontario. Journal of Environmental Management 231:1028-1047. https://doi.org/10.1016/j.jenvman.2018.09.099

Haraldsson, H. V. 2000. Introduction to systems and causal loop diagrams. LUMES, Lund University, Lund, Sweden.

Haraldsson, H. V. 2004. Introduction to system thinking and causal loop diagrams. LUMES, Lund University, Lund, Sweden. http://cmap.crs.org.pl:4444/rid=1244140954250_1167059429_1461/Introduction%20to%20Systems%20and%20Causal%20Loop%20Diagrams.pdf

Hochschule Osnabrück. 2013. Der Dümmer und sein Einzugsgebiet in Nordwestdeutschland - eine ökologische Perspektive. Hochschule Osnabrück, Fakultät Agrarwissenschaften & Landschaftsarchitektur, Osnabrück, Germany. https://docplayer.org/75236855-Der-duemmer-und-sein-einzugsgebiet-in-nordwestdeutschland-eine-oekologische-problemregion.html

Inam, A., J. Adamowski, J. Halbe, and S. Prasher. 2015. Using causal loop diagrams for the initialization of stakeholder engagement in soil salinity management in agricultural watersheds in developing countries: a case study in the Rechna Doab watershed, Pakistan. Journal of Environmental Management 152:251-267. https://doi.org/10.1016/j.jenvman.2015.01.052

Jeppesen, E., M. Meerhoff, T. A. Davidson, D. Trolle, M. Søndergaard, T. L. Lauridsen, M. Beklioglu, S. Brucet, P. Volta, I. González-Bergonzoni, and A. Nielsen. 2014. Climate change impacts on lakes: an integrated ecological perspective based on a multi-faceted approach, with special focus on shallow lakes. Journal of Limnology 73(S1). https://doi.org/10.4081/jlimnol.2014.844

Lade, S. J., S. Niiranen, J. Hentati-Sundberg, T. Blenckner, W. J. Boonstra, K. Orach, M. F. Quaas, H. Österblom, and M. Schlüter. 2015. An empirical model of the Baltic Sea reveals the importance of social dynamics for ecological regime shifts. Proceedings of the National Academy of Sciences 112(35):11120-11125. https://doi.org/10.1073/pnas.1504954112

Lade, S. J., A. Tavoni, S. A. Levin, and M. Schlüter. 2013. Regime shifts in a social-ecological system. Theoretical Ecology 6(3):359-372. https://doi.org/10.1007/s12080-013-0187-3

Landesamt für Geoinformation und Landesvermessung Niedersachsen (LGLN). 2019. Dümmerforum 2019 II Auf Dem Schäferhof. LGLN, Hanover, Germany. https://www.arl-lw.niedersachsen.de/duemmerforum/dummerforum-2019-ii-auf-dem-schaferhof-203569.html

Landkreis Diepholz. 2018. Verordnung Über Das Landschaftsschutzgebiet “Dümmer” in Der Samtgemeinde “Altes Amt Lemförde.” Landkreis Diepholz, Germany. https://www.diepholz.de/downloads/datei/OTAwMDA3ODYyOy07L3d3dy92aHRkb2NzL2Ntcy9sa2RoL21lZGllbi9kb2t1bWVudGUvbHNnX2RoXzE0X2R1ZW1tZXJfbmV1LnBkZg%3D%3D

Landkreis Diepholz. 2020. FFH-Gebiet Nr. 065 Dümmer: Kurzfassung Des Meldebogens Des FFH-Gebietes Sowie Verfah-rensstände. Landkreis Diepholz, Germany. https://natura2000.diepholz.de/schutzgebiete/steckbriefe-aller-ffh-gebiete/duemmer/

L’Association pour la protection de l’environnement (APEL). 2009. Étude limnologique du haut-bassin de la rivière Saint-Charles, rapport final. Association pour la protection de l’environnement du lac Saint-Charles et des Marais du Nord, Quebec City, Canada. https://agiro.org/wp-content/uploads/etude_limnologique.pdf

L’Association pour la protection de l’environnement (APEL). 2017. Faire autrement pour une protection des milieux naturels et des ressources en eau: recommandations règlementaires et non règlementaires visant à améliorer la gestion des milieux naturels et des ressources en eau à l’intérieur du bassin versant de la prise d’eau de la rivière Saint-Charles. Association pour la protection de l’environnement du lac Saint-Charles et des Marais du Nord, Quebec City, Canada. https://agiro.org/wp-content/uploads/Developper_protection_milieux_naturels_eau_APEL-201702.pdf

L’Association pour la protection de l’environnement (APEL). 2019. Diagnose du lac Saint-Charles 2016. Association pour la protection de l’environnement du lac Saint-Charles et des Marais du Nord, Quebec City, Canada. https://agiro.org/wp-content/uploads/Diagnose-du-lac-Saint-Charles-2016.pdf

Leenhardt, P., L. Teneva, S. Kininmonth, E. Darling, S. Cooley, and J. Claudet. 2015. Challenges, insights and perspectives associated with using social-ecological science for marine conservation. Ocean and Coastal Management 115:49-60. https://doi.org/10.1016/j.ocecoaman.2015.04.018

Légaré, S. 1998. Étude limnologique du lac Saint-Charles 1996-1997. Université Laval, Quebec City, Quebec, Canada. https://agiro.org/publications/etude-limnologique-du-lac-saint-charles-1996-1997-stephane-legare-1998/

Leslie, H. M., X. Basurto, M. Nenadovic, L. Sievanen, K. C. Cavanaugh, J. J. Cota-Nieto, B. E. Erisman, E. Finkbeiner, G. Hinojosa-Arango, M. Moreno-Báez, S. Nagavarapu, S. M. W. Reddy, A. Sánchez-Rodríguez, K. Siegel, J. J. Ulibarria-Valenzuela, A. H. Weaver, and O. Aburto-Oropeza. 2015. Operationalizing the social-ecological systems framework to assess sustainability. Proceedings of the National Academy of Sciences 112(19):5979-5984. https://doi.org/10.1073/pnas.1414640112

Martin, R., M. Schlüter, and T. Blenckner. 2020. The importance of transient social dynamics for restoring ecosystems beyond ecological tipping points. Proceedings of the National Academy of Sciences 117(5):2717-2722. https://doi.org/10.1073/pnas.1817154117

McGinnis, M. D., and E. Ostrom. 2014. Social-ecological system framework: initial changes and continuing challenges. Ecology and Society 19(2):30. https://doi.org/10.5751/ES-06387-190230

Moss, B. 2012. Cogs in the endless machine: lakes, climate change and nutrient cycles: a review. Science of the Total Environment 434:130-142. https://doi.org/10.1016/j.scitotenv.2011.07.069

Niedersächsisches Ministerium für Ernährung, Landwirtschaft, Verbraucherschutz und Landesentwicklung (ML) and Niedersächsisches Ministerium für Umwelt, Energie und Klimaschutz (MU). 2012. Dümmersanierung: Rückblick und Ausblick. Niedersächsisches Ministerium für Ernährung, Landwirtschaft, Verbraucherschutz und Landesentwicklung and Niedersächsisches Ministerium für Umwelt, Energie und Klimaschutz, Hannover, Germany.

Organisme des bassins versants de la Capitale (OBVC). 2015. Bassin de la rivière Saint-Charles/Lacs - Organisme des bassins versants de la Capitale. OBVC, Quebec City, Quebec, Canada. http://www.obvcapitale.org/bassin-de-la-riviere-saint-charles-lacs#cplacstcharles

Ostrom, E. 1990. Governing the commons: the evolution of institutions for collective action. Cambridge University Press, Cambridge, UK.

Ostrom, E. 2005. Understanding institutional diversity. Princeton University Press, Princeton, New Jersey, USA.

Ostrom, E. 2007. A diagnostic approach for going beyond panaceas. Proceedings of the National Academy of Sciences 104(39):15181-15187. https://doi.org/10.1073/pnas.0702288104

Ostrom, E. 2009. A general framework for analyzing sustainability of social-ecological systems. Science 325(5939):419-422. https://doi.org/10.1126/science.1172133

Pahl-Wostl, C., C. Knieper, E. Lukat, F. Meergans, M. Schoderer, N. Schütze, D. Schweigatz, I. Dombrowsky, A. Lenschow, U. Stein, A. Thiel, J. Tröltzsch, and R. Vidaurre. 2020. Enhancing the capacity of water governance to deal with complex management challenges: a framework of analysis. Environmental Science and Policy 107:23-35. https://doi.org/10.1016/j.envsci.2020.02.011

Palys, T. 2008. Purposive sampling. Pages 697-698 in L. M. Given, editor. The Sage encyclopedia of qualitative research methods. SAGE, Thousand Oaks, California, USA. https://doi.org/10.4135/9781412963909.n349

Ratajczyk, E., U. Brady, J. A. Baggio, A. J. Barnett, I. Perez-Ibarra, N. Rollins, C. Rubiños, H. C. Shin, D. J. Yu, R. Aggarwal, J. M. Anderies, and M. A. Janssen. 2016. Challenges and opportunities in coding the commons: problems, procedures, and potential solutions in large-N comparative case studies. International Journal of the Commons 10(2):440. https://doi.org/10.18352/ijc.652

Richardson, G. P., and A. L. Pugh. 1981. Introduction to system dynamics modeling with dynamo. MIT Press, Cambridge, Massachusetts, USA.

Schaffernicht, M. 2010. Causal loop diagrams between structure and behaviour: a critical analysis of the relationship between polarity, behaviour and events. Systems Research and Behavioral Science 27(6):653-666. https://doi.org/10.1002/sres.1018

Schallenberg, M., M. D. de Winton, and D. J. Kelly. 2018. Indicators of ecological integrity. Pages 367-393 in D. P. Hamilton, K. J. Collier, J. M. Quinn, and C. Howard-Williams, editors. Lake restoration handbook: a New Zealand perspective. Springer International, Cham, Switzerland. https://doi.org/10.1007/978-3-319-93043-5_11

Schlüter, M., L. Haider, S. Lade, E. Lindkvist, R. Martin, K. Orach, N. Wijermans, and C. Folke. 2019. Capturing emergent phenomena in social-ecological systems: an analytical framework. Ecology and Society 24(3):11. https://doi.org/10.5751/ES-11012-240311

SESMAD. 2021. Social-ecological systems meta-analysis database. Dartmouth College, Hanover, New Hampshire, USA. https://sesmad.dartmouth.edu/

Thiel, A., M. E. Adamseged, and C. Baake. 2015. Evaluating an instrument for institutional crafting: how Ostrom’s social-ecological systems framework is applied. Environmental Science and Policy Part B(53):152-164. https://doi.org/10.1016/j.envsci.2015.04.020

Vennix, J. A. M. 1996. Group model building: facilitating team learning using system dynamics. John Wiley and Sons Ltd., West Sussex, UK.