Abstract

Complex social-ecological interactions underpin many environmental problems. To help capture this complexity, we advance an interdisciplinary network modeling framework to identify important relationships between people and nature that can influence environmental conditions. Drawing on comprehensive social and ecological data from five coral reef fishing communities in Kenya; including interviews with 648 fishers, underwater visual census data of reef ecosystem condition, and time-series landings data; we show that positive ecological conditions are associated with ‘social-ecological network closure’ – i.e., fully linked and thus closed network structures between social actors and ecological resources. Our results suggest that when fishers facing common dilemmas form cooperative communication ties with direct resource competitors, they may achieve positive gains in reef fish biomass and functional richness. Our work provides key empirical insight to a growing body of research on social-ecological alignment, and helps to advance an integrative framework that can be applied empirically in different social-ecological contexts.

Introduction

Humans are a fundamental part of ecosystems and rely on them to support a wide array of their needs. The extent of environmental stressors connected to human activities thus makes understanding social–ecological linkages of central importance for the analysis of almost any action related to securing a sustainable future1. Recognizing this, research on the environment is increasingly focused on transcending traditional disciplinary boundaries and embracing an integrative, complex systems view to understand ecosystems from a perspective that incorporates theories and frameworks from both the natural and social sciences2,3. Even with this progress, studying complex systems involves inherent limitations, including a lack of common language and methods shared between the natural and social sciences4,5. Thus advancing tractable and informative frameworks and models that capture social–ecological linkages and can be applied empirically remains a defining challenge to address real-world sustainability issues.

A path forward that is gaining increasing attention in the literature is the development and application of social–ecological network approaches4,6,7,8,9. Network approaches offer a fruitful framework for theorizing and empirically investigating important social–ecological interactions and how they relate to sustainability outcomes for several reasons. First, social–ecological network approaches can capture important relationships both among and between social and ecological entities (Fig. 1), thus explicitly accounting for interdependencies (e.g., spillovers and feedbacks) that can have dramatic effects on social–ecological system behavior10. Second, social–ecological network approaches evoke language, methods, and models common to both the natural and social sciences11,12, thus providing one avenue to facilitate the cross-disciplinary engagement necessary for solving complex environmental problems. Yet, despite recent theoretical and conceptual developments of social–ecological network approaches13, empirical applications have struggled to move beyond individual case studies or explicitly link aspects of social–ecological structure to quantitative data on ecosystem conditions13,14,15. We advance this emerging research through a novel multi-case, comparative empirical assessment that demonstrates how certain social–ecological interdependencies relate to quantitative ecological conditions.

Fig. 1
Fig. 1

A coral reef fishery as a multilevel social–ecological network. An illustrative example of the integrative, social–ecological network modeling approach and key configuration of interest. The social network (A) captures key communication relationships between individual fishers. The ecological network (B) captures trophic interactions among target species. In reef fisheries, each fishing gear type catches a diverse and overlapping, but distinct assemblage of species in B. Individual fishers are thus linked to particular fish species (X; social–ecological ties) depending on the type of gear they use (depicted in the nodes in A). All nodes and links are representative of our empirical data. The multilevel structure (A, B, X) captures the dependencies that exist within the system, i.e., how features of social and ecological systems are interrelated both within and across levels. Full multilevel social–ecological networks can be disassembled into smaller building blocks, or key configurations (right), that form the foundation for the larger system structure4,68. Here a form of social–ecological alignment is emphasized, i.e., social–ecological network closure, which captures the tendency for actors tied to the same resource to form cooperative communication ties

Our research rests on the assumptions that (a) important aspects of social systems, ecological systems, and the interactions between them can be modeled and analyzed as nodes and links in a multilevel social–ecological network, and (b) social–ecological networks are themselves composed of precisely defined network configurations [i.e., building blocks or network motifs16] that reflect key relationships among social actors and ecological resources important for achieving particular outcomes (Fig. 1)4. Perhaps the most salient social–ecological network configuration highlighted to date17,18 is the closed, cross-level social–ecological triangle—where two actors connected to the same resource are also connected to each other (Fig. 1). This configuration captures a form of 'social–ecological network closure', i.e., fully linked and thus closed, network structures between social actors and ecological resources (which stand in contrast to open social–ecological network structures; e.g., where social actors are connected to common ecological resources but are not connected to each other). In social network science, network closure19 [often equated with bonding social capital20] emphasizes that tight coupling between actors facilitates trust, learning, and the establishment of common norms and sanctions while minimizing uncertainty21,22. Social–ecological network closure extends this coupling across the social–ecological divide, identifying specific forms of communication and cooperation that bind actors connected to the same (or interconnected23) resources, thereby better equipping them to learn from each other and agree on and address important environmental problems (Fig. 2).

Fig. 2
Fig. 2

Theoretical mechanisms linking social–ecological network closure and ecological conditions. A conceptual diagram illustrating key social processes theoretically supported by social–ecological network closure that may lead to improved ecological conditions in the commons. When direct resource competitors in settings characterized by strong and complex patterns of social–ecological interactions form cooperative communication ties, it lays the foundation for the emergence of trust, a shared vision, and sustained commitments27,29,30 regarding the management of shared resources. Two examples of such commitments include the development of conflict resolution mechanisms and agreement on rules. These social interactions and processes can ultimately lead to improved ecological conditions. It is important to note that this figure is only illustrative of key mechanisms linking social–ecological network closure to ecological conditions and does not include the full range of social–ecological interactions and feedbacks that can affect both ecological and social conditions in any given environmental system

The proposed utility of this type of social–ecological network closure is especially pronounced in the commons, where actors use shared resources for extractive purposes14. In this context, actors are faced with a ubiquitous social dilemma, i.e., the tragedy of the commons24, whereby each individual has an incentive to overharvest in order to maximize their own short-term gain due to the non-excludable and rivalrous nature of common resources. Privatization or third-party regulation and enforcement can help to solve this dilemma; however, these actions are not always feasible, preferable, or cost effective. In such cases, the ability of resource users to act collectively to devise and enforce commonly agreed upon norms and rules for sustainable resource use is critical25. Yet, how such cooperation emerges when faced with social dilemmas without oversight from a central authority has been of considerable interest among scholars for decades26. Though several explanations have been proposed and some have been supported through empirical research27, one of the most robust findings has been that communication is critical—when individuals engage in face-to-face communication, cooperation increases significantly28. Thus, if actors with a stake in the same resource have opportunities to communicate, there is strong theoretical evidence to support the notion that it can facilitate cooperation toward effectively managing shared resources, thereby leading to improved ecological conditions (see Fig. 2)4,27,29,30. This type of social–ecological network closure can also facilitate learning, which is critical for updating management strategies in the face of social and ecological change31. In common-pool resource settings, social–ecological closure is thus an important aspect of what is often referred to as social–ecological alignment (or social–ecological fit) where relationships between social actors are aligned with the characteristics of the underlying biophysical system14,32.

Coral reef fisheries are an ideal common-pool resource system to investigate the potential utility of this form of social–ecological alignment. Reefs are one of the most productive and biologically diverse ecosystems on the planet33, providing critical services that support the livelihoods of millions of people34. Yet, reefs are rapidly degrading on a global scale33, in part due to unsustainable fishing35. All reef fisheries face (or have faced) the tragedy of the commons, and most are characterized by multiple species being targeted (or incidentally caught) by multiple gears (Fig. 1). This complexity in the resource base (network level B, Fig. 1) and associated harvesting strategies (network level X, Fig. 1) presents considerable challenges for sustainable management36. Most coral reefs are also located in regions that suffer from low institutional capacity for governance, high dependence on reef resources, and high rates of poverty37. Thus a better understanding of how social–ecological alignment relates to ecological conditions in coral reef fisheries could potentially have large implications for millions of people worldwide.

Here we test the hypothesis that social–ecological network closure is associated with positive ecological conditions in the face of the commons dilemma. We do so by examining whether cooperative communication relationships between fishers harvesting the same species (i.e., closed, cross-level social–ecological triangles, Fig. 1) mediate biomass and functional richness of fished resources across five coral reef fishing communities (sites) along the Kenyan coast (Methods). Our ecological indicators—reef fish biomass and functional richness—are strong predictors of reef ecosystem condition. Reef fish are key elements of reef ecosystems that drive processes linked to ecosystem condition and stability38. Fish biomass has been shown to be related to a wide range of information on reef fish functioning (e.g., herbivory, predation), trophic structure, life history composition, and benthic ecosystem state39,40. The magnitude of fishable biomass is highly sensitive to fishing and is commonly used to gauge the status of coral reefs globally41. Functional richness captures the roles species perform in an ecosystem by categorizing species based on a combination of key traits (e.g., diet, body size, and mobility), rather than taxonomy. As such, functional richness quantifies the number of unique trait combinations within a given sample and has been shown to predict ecological responses to disturbance, understand competitive interactions, and partly drive productivity42. Functional, as opposed to taxonomic, richness is fast becoming a much preferred measure of biodiversity in ecology as it captures more about the role of species in ecosystem functioning42,43.

To support our inquiry of the role of social–ecological network closure on ecological conditions in reef fishing communities, we accounted for biophysical, environmental, and human impact characteristics known to effect reef ecosystem conditions. We also evaluated other social and institutional conditions known to effect collective management of the commons to determine whether they provided alternative explanations for the relative ecological condition of some sites versus others (Methods). Finally, we conducted a preliminary assessment of indicators of the key social processes supported by social–ecological network closure (Fig. 2) across sites to explore whether they aligned with our theoretical expectations. Taken together, our results provide support to our hypothesis that social–ecological network closure is associated with positive ecological conditions. Specifically, our results indicate that when fishers facing commons dilemmas form cooperative communication ties with direct resource competitors, they may achieve positive gains in reef fish biomass and functional richness.

Results

Social–ecological ties

We constructed full, multilevel social–ecological networks akin to Fig. 1 for each reef fishing community, or site (Methods). Across sites, there were 71–232 fishers in each social network (Supplementary Table 1). On average, fishers had 1.52–3.49 contacts with whom they had formed cooperative communication ties specific to fishing and fishery management (i.e., social ties in A, Fig. 1). Social–ecological ties (X, Fig. 1) linked fishers to their respective target species via the primary fishing gear they used (Methods, Supplementary Methods, Supplementary Tables 24). We found at least three, but up to five different types of primary fishing gear in use, which included hook and line, gillnets, seine nets, spears, and traps (Supplementary Table 2). There was substantial—but not complete—overlap in target species across gear types, with the majority of catch from all gear types comprising a total of 36 species (Supplementary Table 3). Many individual fishers thus competed for the same resources, irrespective of their choice of fishing gear (Supplementary Table 4).

Social–ecological network closure

We tested whether and to what extent social-ecological network closure helped to explain the structure of our empirically observed social–ecological networks by leveraging advances in multilevel exponential random graph models44 (ERGMs; see Methods, Supplementary Methods). We found a significant positive effect of social–ecological network closure in three of our five sites: sites A–C, as indicated by the positive and significant parameter estimates for the closed, cross-level social–ecological triangle (Fig. 3). Thus, in sites A–C, fishers harvesting the same resources were significantly more likely to have formed cooperative communication ties, whereas in sites D and E, they were not. Aside from this effect, results from our ERGMs showed little to no difference across sites in endogenous and exogenous factors structuring the empirical social–ecological networks. In all sites, fishers had a similar baseline tendency to form social ties (social network density, Fig. 3). There was no consistent, significant effect of preferential attachment45 (centralization) in the social networks (Fig. 3). Fishers had a tendency to form ties with community leaders more so than others in all sites46, as indicated by the positive and significant parameter estimates for leader activity shown in Fig. 3. There was also a significant homophily effect47 on landing site in all of our study sites where more than one landing site was in regular use (Supplementary Methods), meaning that fishers tended to preferentially form ties with others from their community who visit the same location to land and sell their fish (Fig. 3). Lastly, we found a significant, positive effect of social network closure19 (i.e., closure in the social network A, Fig. 1), indicating that in all of our sites, there was a general tendency for fishers to form triadic social structures (i.e., a friend of my friend is also my friend; Fig. 3). Importantly, even when controlling for this general tendency for cooperative, triadic structures to emerge in the social network, fishers in only three of our five study sites (sites A–C) had specifically formed cooperative communication ties when they shared the same resource more so than expected by chance alone.

Fig. 3
Fig. 3

The importance of social–ecological network closure. Values shown are the coefficients (and SEs) of social–ecological network closure (shaded) and other key parameters from five multilevel exponential random graph models (ERGMs) fit to empirical social–ecological networks representing each of the five reef fishing communities studied (sites A–E). Shapes and colors in the conceptual graphical depictions follow Fig. 1. L indicates an actor in the social network who is also a leader, and the tie linking this leader to another social actor demonstrates the potential for leaders to have more ties on average than others; a indicates an actor in the social network who uses hypothetical landing site a, and the tie linking this actor to another whom also uses landing site a demonstrates the potential homophily effect on landing site. Note that the depictions for centralization and closure in the social network are only representative and do not explicitly capture the alternating nature of the specific parameters included in the model (termed ASA and ATA in MPNet; Methods, Supplementary Methods). Full models also included controls for activity in each landing site where a residual analysis suggested that fishers may be more active in forming and maintaining ties than would be expected by chance alone (Supplementary Table 5). Asterisk (*) indicates significance at P < 0.05

Ecological conditions

We found evidence that social–ecological network closure is indeed associated with positive ecological conditions (Fig. 4). Specifically, we found a significantly higher mean level of both reef fish biomass and functional richness in sites with a positive tendency toward social–ecological network closure (sites A–C) compared to those without [biomass: t(9.49) = 2.09, p = .03; functional richness: t(12.45) = 3.56, p < 0.01]. Effect size estimates suggest that these differences are meaningful (Cohen’s D, biomass = 0.89, 90% CI = 0.17, 1.71; Cohen’s D, functional richness = 1.55, 90% CI = 0.60, 2.50). Importantly, differences in ecological conditions across sites do not appear to be related to other biophysical, environmental, or human impact factors known to be important for driving reef ecosystem conditions (Table 1). Specifically, we found no significant difference between sites with and without social–ecological network closure in terms of sea surface temperature (SST), net primary productivity (NPP), coral cover, rugosity (a measure of structural complexity48), human gravity49 (a human impact measure that accounts for population size and reef accessibility50), or fishing pressure (Table 1). The potential differences in shared versus non-shared species comprising our biomass estimates also do not appear to explain these results; e.g., the majority of our biomass estimates are comprised of species that are caught by multiple competing fishers (Supplementary Methods, Supplementary Table 4). These results lend support to our hypothesis that social–ecological network closure can help to overcome commons dilemmas—indeed, where actors linked to the same resource had a significant tendency to form cooperative communication ties (i.e., sites A–C), we saw better ecological conditions.

Fig. 4
Fig. 4

Ecological conditions across study sites. Sites that have a significant, positive social–ecological network closure effect (sites A–C) are outlined in the gray box with the network icon. a Fish biomass observed in fished areas across each study site from underwater visual surveys compared to the expected level of pristine fish biomass (green line) for unfished reef ecosystems in Kenya, as reported by ref. 69. Black dots are individual data points; gray bars and text above bars report mean biomass observed; gray arrows denote closeness toward pristine biomass (1200 kg/ha); percentage difference between pristine and observed biomass is reported below the green line. b Functional richness of reef fish species (mean ± SE) in fished areas across each study site based on underwater visual surveys and a combination of abundances and trait values. Black dots are individual data points. There is a significantly higher mean level of both reef fish biomass and functional richness in sites with a positive tendency toward social–ecological network closure compared to those without [t(9.49) = 2.09, p = .03; t(12.45) = 3.56, p < 0.01; respectively]; and effect size estimates suggest that these differences are meaningful (Cohen’s D = 0.89, 90% CI = 0.17, 1.71; Cohen’s D = 1.55, 90% CI = 0.60, 2.50; respectively)

Table 1 Biophysical, environmental, and human impact characteristics

Key social processes

The results of our exploratory assessment of key social processes supported by social–ecological network closure (Fig. 2) partially correspond with our theoretical expectations. First, we found indicative evidence that sites D and E (which do not exhibit a predisposition for social–ecological network closure, Fig. 3) differed from other sites in regard to (1) trust, and (2) shared vision (i.e., resource users have a common understanding of how the system operates and how their actions affect it)51. Importantly, we did not find evidence that mean levels of trust differed between sites with and without social–ecological network closure. However, we did find that there was significantly more variation in trust in both sites D and E compared to other sites. This indicates that in sites D and E there is less agreement about whether others can be trusted, and the lack of social–ecological network closure in these sites suggests that there may be pockets of mistrust—or at least a lack of trust—between resource competitors who do not communicate28. On the flip side, there may also be pockets of trust. We also found that respondents in site D exhibited significantly more variation in their understanding of the state of coral reef fisheries resources (Table 2). Second, sites D and E also differed from other sites in terms of the commitments made regarding fishery management. For example, in terms of the rules in use, we found that all sites had instituted some form of access rights and designated an area that was closed for fishing. However, only sites A–C had also agreed on and successfully initiated gear restrictions, despite reports that internal conflict over gear use continued to be a problem in both sites D and E. Mechanisms to aid in conflict resolution had also not been designed and established in site E (Table 2).

Table 2 Key social processes

Social and institutional conditions

Success in managing the commons in the absence or failure of top–down governance is known to be associated with a set of social and institutional conditions25,51,52. Some of these conditions we argue here are directly supported by social–ecological network closure (e.g., trust, a shared vision; Fig. 2). Yet, others are not (e.g., dependence on common resources; organizational experience/leadership). Thus any variation in these conditions across sites may offer competing explanations for observed differences in ecological conditions. To account for these potentially confounding factors, we used data from our fisher surveys, interviewed community leaders, and drew on existing research53 (Methods). We found little to no differences across sites in these social and institutional conditions: all had high levels of dependence on fisheries resources, the rights to devise local institutions for management, and had prior organizational experience and local leadership (Table 3). All had developed rules adapted to the local condition, the ability to exclude outsiders, graduated sanctions, monitors that were locally accountable, and high levels of participation in decision-making (Table 3). Hence, none of these conditions could explain the observed differences in biomass and functional richness of fished resources.

Table 3 Social and institutional conditions

Discussion

Our quantitative and qualitative results provide evidence that closed social–ecological network structures among direct resource competitors may facilitate more effective cooperation that is associated with positive ecological conditions in coral reefs. In these multi-resource commons settings, the distinction between cooperation in a general sense and the more precise form of cooperation evaluated here that accounts for complex social–ecological interdependencies appears to be an important one. Indeed, results from our network models demonstrate that all study sites have a baseline propensity for cooperation among social actors (indicated by the significant, positive parameter estimates for social network closure, Fig. 3). This result supports recent research on the risk hypothesis20, which argues that social actors tend to form closed, triadic social network structures to manage high-risk cooperation problems due to their ability to help develop and sustain trust and exert social pressure to comply with rules. Yet, despite this baseline tendency for cooperation across all sites, our results demonstrate that only sites A–C have a propensity for cooperation that results in social–ecological alignment by directly binding those who are dependent on the same resources (social–ecological network closure, Fig. 3). Importantly, sites A–C also had higher levels of both biomass and functional richness of fished resources (Fig. 4), and these ecological conditions do not appear to be related to other network effects (Fig. 3); biophysical, environmental, or human impact characteristics (Table 1); or potentially confounding social and institutional factors (Table 3).

We proposed several theoretical mechanisms by which social–ecological network closure capturing cooperative communication among direct resource competitors might impact ecological conditions in this setting: i.e., the development trust, a shared vision, and the establishment of commitments among direct resource competitors toward sustainable resource management (Fig. 2). Our exploratory evaluation of these social processes was largely in line with our theoretical predictions. Specifically, we found that sites with a propensity for social–ecological network closure (sites A–C) demonstrated less variation in trust; a higher level of agreement on the state of reef resources; and a stronger commitment to sustainably managing reef resources, demonstrated by the establishment of a greater number of rules and avenues for conflict resolution (Table 2). This is important because reaching a consensus regarding what actions to take to manage common-pool resources such as reef fisheries and whether they will be effective is likely to be more difficult where there is less agreement about the state of the resource system and about whether people—especially direct resource competitors—can be trusted, e.g., to comply with devised rules51. Indeed, although our sites without a propensity for social–ecological network closure (sites D and E) had devised some rules at the time of data collection, previous research53 suggests that these rules were not easily established (e.g., McClanahan et al.53 found that they experienced substantial delays in designating areas closed for fishing after indicating initial interest compared to other sites). Moreover, sites without a propensity for social–ecological network closure had not agreed on and instituted gear restrictions, which play a key role in managing reef fisheries because they modify fishing behavior rather than trying to prevent it54. This distinction is important because many reefs are located in developing countries, where more stringent regulations can undermine livelihoods and be difficult to enforce54.

Practically, our results suggest that investments in building community capacity that specifically focus on establishing communication channels among direct resource competitors may improve reef ecosystem conditions. Yet given the competitive nature of many common-pool resource systems such as reef fisheries55, important questions remain regarding how these relationships can be built. Here key social–ecological interactions were defined as those that linked fishers to specific species based on their fishing gear (Fig. 1). Our results thus suggest that stimulating gear-based communication may indirectly lead to a greater propensity for social–ecological network closure since the same set of species tend to be targeted by the same gear (Supplementary Table 3, Supplementary Methods). These communication channels can be facilitated by creating communities of practice centered around gear and technology, which can act to stimulate learning, build trust, and enhance shared ecological understanding of factors important for resources to be sustained56. However, caution is warranted, as efforts to build such communities of practice could lead to the emergence of competing gear-based coalitions and a zero-sum game where the potential ecological benefits from restricting one gear are captured by users of another gear36. This is a genuine risk in multi-species, multi-gear reef fisheries and other similar common-pool resource systems, where gear competition is ubiquitous. Thus broader community-building strategies that seek to establish communication and trust across all direct resource competitors, including actors using different gear types but overlapping in target species, is critical for achieving long-term sustainability. Notably, this communication may not need to be maintained over the long term, as recent research suggests that communication can have a persistent effect on cooperation in social dilemmas even after it has been removed28. What is critical, however, is that communication occurs long enough to establish prosocial norms that can activate guilt if and when someone considers defecting28.

This study represents the first multi-site comparative analysis to examine how key aspects of social–ecological networks relate to quantitative ecosystem conditions. It therefore fills a critical gap in advancing integrative social–ecological network approaches for environmental problem-solving, which has been repeatedly advocated in recent years6,7,9. Applying this approach, we tested an important theoretical question regarding how social–ecological alignment relates to ecological conditions. Future research can extend this work to empirically test theory-driven hypotheses regarding other types of social–ecological interdependencies at various scales that may have important impacts on sustainability outcomes. For example, if coupled with dynamic or longitudinal data, this framework could be used to test explicit hypotheses about how changes in social structures drive the formation or dissolution of ecological links. The framework could also be used to explicitly capture social–ecological feedbacks, which have been difficult to study empirically.

Given the multitude and scale of anthropogenic drivers affecting the environment32 and the costs associated with cooperation55, understanding who should cooperate with whom in different contexts and to address different types of environmental problems is becoming increasingly important14. The benefit of the interdisciplinary social–ecological network approach described here is that it allows for a much more nuanced and precise understanding of the interdependencies between social and ecological components of ecosystems, allowing one to unpack the specific types of cooperative connections that facilitate or hinder effective action. Employing this approach, we provide evidence that social–ecological network closure—fully linked and thus closed, network structures between social actors and ecological resources—supports key social processes that may promote more effective collective management of shared resources, having demonstrable ecological impacts. Our results suggest that investments in building community capacity that focus on establishing communication, trust, and a shared understanding among direct resource competitors may improve ecological conditions in coral reef fisheries.

Methods

Summary of our empirical strategy

We studied five coral reef fishing communities along the Kenyan coast. To test our hypothesis, we used a combination of quantitative and qualitative interdisciplinary data collected via semi-structured fisher surveys, underwater visual census, observed fish landings, key informant and expert interviews, and published reports53. Specifically, we drew on information from our fisher surveys, observed fish landings data, published reports, and expert interviews to construct full social–ecological networks akin to Fig. 1 for each study site. We then tested whether and to what extent the closed, cross-level social–ecological triangle (i.e., social–ecological network closure, Fig. 1) helped to explain the empirically observed structural characteristics of these networks using multilevel ERGMs. Next, we tested for differences in ecological resource conditions within fished areas of sites with and without social–ecological network closure using underwater visual census data. We also tested for differences in key biophysical, environmental, and human impact characteristics known to affect reef ecosystem conditions. We then drew on information from our fisher surveys, conducted key informant interviews, and reviewed published reports to explore whether the key social processes we argue are supported by social–ecological network closure were present in each site (i.e., Fig. 2). We also used this information to assess whether other social and institutional conditions associated with effective management of the commons25,51 may have affected ecological resource conditions across sites.

Site selection

Sites were selected from a ~100 km stretch of the Kenyan coast (Supplementary Fig. 1) in collaboration with our partners at the Wildlife Conservation Society’s Coral Reef Conservation Program (TRM). We specifically chose sites (1) that were relatively close together to minimize differences in key biophysical and environmental conditions, (2) where fishing was the primary occupation of the majority of the population, (3) where our partners had been engaged in monitoring, and (4) where communities were considered to have achieved a range of success in managing reef fisheries resources collectively as a community in order to combat declining trends (Supplementary Methods). Each site selected was comprised of a social community of fishers and an associated fishing area adjacent to their community that they use and have rights to manage (see Supplementary Information for more details). All fishing areas sampled were shallow (<10 m depth), exposed to similar environmental conditions (Table 1), and have a similar disturbance history (e.g., coral bleaching).

Constructing the social–ecological networks

To capture cooperative communication relationships among fishers (i.e., the social network A, Fig. 1), we administered a semi-structured fisher survey from December 2015 to May 2016. A total of 711 fishers were originally surveyed, representing 75–84% of the total estimated population of fishers within each site (Supplementary Table 1). Eighty-one fishers were subsequently dropped owing to missing information (Supplementary Table 1). We used a name generator with qualifiers (Supplementary Methods), where fishers were specifically asked to nominate up to ten individuals with whom they exchanged information and advice with about fishing and fishery management (e.g., rules, gears, and fishing locations). Name qualifiers were checked daily with local guides while fieldwork was being conducted to ensure identification accuracy of all nominated individuals. Non-respondent network actors were dropped and ties were symmetrized and treated as binary. The corresponding social networks were thus undirected, with edges representing information and advice relationships between respondents Ai and Aj in each site (Supplementary Table 1, Supplementary Fig. 1). Fishers were also asked to report what type of fishing gear they used in addition to other sociodemographic characteristics that existing research suggests plays a role in structuring social interactions in fisheries, e.g., ethnicity, leadership, and landing site46 (Supplementary Table 2). Surveys were conducted via in-person interviews in Swahili.

The ecological network (B, Fig. 1) captures trophic interactions among target fish species comprising the majority of catch by all fishing gears employed in our five study sites (n = 36 species, Supplementary Methods; Supplementary Fig. 2). Target fish species for each gear type were identified using detailed landings data from 25 landing sites along the Kenyan coast collected continuously between 2010 and 2016 (Supplementary Table 3). Trophic interactions (i.e., predator–prey relationships) were estimated based on a combination of diet, relative body size, and habitat use18,57,58 (Supplementary Methods). The corresponding ecological network was thus undirected, with edges representing trophic interactions between fish species Bu and Bv. Social–ecological ties (X, Fig. 1) were identified by linking individual fish species to individual fishers via their primary fishing gear as identified in the fisher survey (Supplementary Table 4). In other words, if fisher Ai used gear type Gt as their primary gear, and gear type Gt targeted fish species Bu, a social–ecological link would exist between fisher Ai and fish species Bu.

Multilevel network models

We used multilevel exponential graph models (ERGMs) (Supplementary Methods) to test the prevalence of the closed, cross-level social–ecological triangle configuration representing cooperative communication among direct resource competitors within each site. ERGMs are statistical models of networks based on explicit hypotheses about network dependence59. ERGMs model network ties explicitly by treating each tie as a random variable and specifying the probability of observing the network (Y) with n nodes as a function of various local network processes. These network processes are expressed as micro-level network configurations (e.g., edges, stars, and triangles) where all ties are assumed conditionally dependent. The dependence assumption is key because it captures the idea that, rather than forming at random, empirical network ties self-organize into various patterns arising from underlying social processes60, e.g., preferential attachment45 and transitivity19. The observed network structure is thus seen as one possible outcome of these stochastic network processes. Multilevel ERGMs can be seen as an extension of ERGMs that account for networks linked across multiple levels44. Here, network ties are considered interdependent not only within levels but also across levels, enabling the interpretation of cross-level interactions and configurations (e.g., Fig. 1). In this study, we employed an extended version of multilevel ERGMs that builds on social selection models61 to incorporate nodal attributes as exogenous covariates in order to account for their ability to effect network structures (Supplementary Methods).

We tested for social–ecological network closure—i.e., the closed, cross-level social–ecological triangle depicted in Fig. 1—while controlling for nodal attributes known to shape social interactions among fishers and other well-known mechanisms involved in shaping social networks59. Nodal attributes included were (1) leader activity (the propensity for leaders to be active/have more ties in the network) and (2) landing site homophily (homophily among fishers using the same landing site), as these have been shown to affect social tie formation in small-scale fisheries46. Full models also included controls for activity in each landing site where a residual analysis62 suggested that fishers associated with that landing site were more active in forming and maintaining ties than would be expected by chance alone (Supplementary Table 5). To control for endogenous mechanisms in the social network, we included (1) the edge parameter to capture density, which corresponds to the baseline propensity to establish ties; (2) centralization parameters (the alternating star and a two-star parameter where appropriate; Supplementary Methods) to capture preferential attachment; and (3) the alternating triangle parameter to capture transitive closure.2

Because the focus here was on social processes, and particularly the propensity for fishers to form ties with direct resource competitors, the X and B level networks (Fig. 1) were fixed and treated as exogenous; in other words, their structure was treated as given and therefore ties within these levels were not explicitly modeled. Goodness-of-fit tests and residual analyses demonstrated that nearly all graph characteristics were well accounted for by our final models (Supplementary Methods, Supplementary Table 6). Note that, in multi-level ERGMs, the parameter estimates for cross-level effects (e.g., social–ecological network closure) cannot be directly compared to the parameter estimates for within-level effects (e.g., social network density). Mahalanobis distances for each model indicated a better model fit with the inclusion of the cross-level social–ecological triangle (Supplementary Methods). All models were run in MPNet63, which implements a Markov Chain Monte Carlo procedure to estimate model parameters using maximum likelihood estimation64.

Assessment of ecological conditions

We used detailed underwater visual census data collected between 2010 and 2015 that surveyed fish in replicate 500 m2 transects at each site (Supplementary Methods, Supplementary Table 7) to generate our estimates of biomass and functional richness of fished resources. Using this data, we tested for mean differences in reef fish biomass and functional richness between sites with and without social–ecological network closure using a one-sided, two-sample t test and effect size estimates (Cohen’s D). We conducted identical tests on all available data (2010–2015) and on data from 2014 only (which most closely matches when our social data was collected) and found no difference in our results (Supplementary Table 8). Satterthwaite’s formula65 was used to approximate the degrees of freedom for all tests where the data was found to have unequal variance across groups.

Identifying key social processes

To explore the presence of, and variation in key social processes theorized to be supported by social–ecological network closure (Fig. 2, Table 2), we drew on our fisher survey, community leader interviews, and existing research53. Specifically, we examined trust using a five point Likert-scale variable in our fisher survey, where fishers were asked to report how much they trusted other fishers. To assess whether fishers had a common understanding or shared image, we asked how they perceived the state of the resource system in our fisher survey (i.e., was there more, the same, or less fish on the reef than 5 years ago?). We compared the variation in trust and fisher’s perceptions of the state of the resource system across sites using Levene’s test for the equality of variance. We compared mean levels of trust using Mann–Whitney U non-parametric test. The results of this test were insensitive to whether one applied this procedure; a two-sided t test accounting for unequal variance using Satterthwaite’s approximation for degrees of freedom; or a linear mixed model with a fixed effect for group (i.e., sites with and without significant social–ecological network closure effects) and a random effect for individual, which accounts for the non-independent nature of observations from the 45 fishers (out of 648) who identified themselves as part of two of our study communities. To assess the level of commitments made within each site regarding the management of fishery resources, we interviewed community leaders to examine the rules in use and whether conflict resolution mechanisms had been established. Reports of within-community conflict were described in the existing research53.

Accounting for potentially confounding factors

We assessed differences in key biophysical, environmental, and human impact characteristics known to effect reef ecosystem condition between sites with and without social–ecological network closure using a two-sided, two-sample t test and effect size estimates (Cohen’s D; Table 1). Satterthwaite’s formula65 was used to approximate the degrees of freedom for all tests where the data was found to have unequal variance across groups. Biophysical variables were hard coral cover66 and rugosity, a measure of the structural complexity of the habitat48. Environmental variables were SST and NPP. Human impact measures were fishing pressure and human gravity49, a metric that accounts for human population and reef accessibility (including travel time50) that aims to capture both market and subsistence pressures on reefs. Data sources are further detailed in Supplementary Table 7. To assess relevant social and institutional conditions within each site (Table 3), we examined the prevalence of, and variation in Ostrom’s25 institutional design principles shown to support robust management of the commons52. Specifically, we interviewed community leaders to determine whether each site had the ability to exclude outsiders, if rules were adapted to local conditions, whether graduated sanctions were in place, and if conflict resolution mechanisms existed. We drew on existing research41 to determine whether monitors were locally accountable and whether communities had rights to devise their own institutions without being challenged by external governing authorities. We used our fisher survey to assess mean levels of participation in decision making about resource management issues using two approaches: a two-sided t test for equality of means, and a linear mixed model with a fixed effect for group (i.e., sites with and without significant social–ecological network closure effects) and a random effect for individual, which accounts for the non-independent nature of observations from the 45 fishers (out of 648) who identified themselves as part of two of our study communities. The results were insensitive to the approach used. Using information from our fisher survey and published reports41, we also examined two attributes known to be positively related to collective action in the commons: (1) salience, i.e., the majority of resource users are dependent on the resource system to support their livelihoods, and (2) prior organizational experience and local leadership51.

Limitations

Common to empirical inquiries attempting to uncover network effects67, our comparative analysis is not without limitations. First, owing to the high data demands of our approach and the intensive nature of collecting detailed and complete, empirical social networks, we were only able to study five communities. Despite this, the results of our multilevel ERGMs and ecological conditions provide support for our hypothesis, and we were able to further support our inferences by incorporating a range of additional data characterizing key social processes; biophysical, environmental, and human impact characteristics; as well as the social and institutional conditions in each community. Second, because we collected detailed social network data in addition to data on fishing behaviors and other social factors, the amount of time spent on each topic in our interviews had to be carefully considered in order to avoid respondent fatigue. Thus we were only able to gain preliminary empirical insights into the mechanisms by which social–ecological network closure can affect ecological conditions (Fig. 2). Mechanisms—particularly those that involve human behavior—are difficult to isolate and study empirically in field settings. As an example, we assessed variation in perceptions over the state of reef resources to gauge whether fishers had a common understanding or shared image of the resource system and how it operates (Table 2); yet it is possible that variation in fisher’s perceptions of the state of the resource could potentially be due to more complex or less obvious resource dynamics. Still, the mechanisms proposed here have strong theoretical support27,28,30. Our empirical assessment of these social processes should thus be seen as exploratory in nature, and only one part of a triangulation effort to more thoroughly test our claims linking social–ecological network closure to ecological conditions. Third, our approach relied on cross-sectional network and socioeconomic data, preventing us from establishing clear temporal trends and causality between social–ecological network closure and ecological conditions. This is a common limitation in empirical social–ecological research due to high data demands and is particularly pronounced with empirical network research. However, our inquiry was grounded in well-established theories of communication and cooperation, giving us a high level of confidence that our results point to social–ecological network closure as a predecessor to improved ecological conditions, rather than the reverse. More firmly establishing casual links would require integrative, interdisciplinary social and ecological data collected at multiple points in time—a task likely to require a career of work but could be more efficiently facilitated by long-term collaborative endeavors.

Ethics statement

Research protocols were approved by the Institutional Review Board of the Office of Research Compliance Human Studies Program at the University of Hawaii at Manoa and the Human Ethics Research Committee at James Cook University. Informed consent was obtained from all respondents.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability

Summary ecological and social data that support the findings of this study are available within the paper and its Supplementary Information files. Raw ecological network data have been deposited in the Tropical Data Hub and can be accessed at https://doi.org/10.25903/5c89d99f5d654. Raw social, social network, and social–ecological network data are available upon request from the corresponding author M.L.B. with reasonable restrictions, as these data contain information that could compromise research participant privacy and consent.

Additional information

Journal peer review information: Nature Communications thanks Jeffrey Johnson, Albert Norström, and other anonymous reviewer(s) for their contribution to the peer review of this work. Peer reviewer reports are available.

Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

References

  1. 1.

    Folke, C., Pritchard, L., Berkes, F., Colding, J. & Svedin, U. The problem of fit between ecosystems and institutions: ten years later. Ecol. Soc. 12, 30 (2007).

  2. 2.

    Ostrom, E. A general framework for analyzing sustainability of social-ecological systems. Science 325, 419–422 (2009).

  3. 3.

    Levin, S. et al. Social-ecological systems as complex adaptive systems: modeling and policy implications. Environ. Dev. Econ. 18, 111–132 (2013).

  4. 4.

    Bodin, Ö. & Tengö, M. Disentangling intangible social–ecological systems. Glob. Environ. Change 22, 430–439 (2012).

  5. 5.

    Blythe, J., Nash, K., Yates, J. & Cumming, G. Feedbacks as a bridging concept for advancing transdisciplinary sustainability research. Curr. Opin. Environ. Sustain. 26–27, 114–119 (2017).

  6. 6.

    Dee, L. E. et al. Operationalizing network theory for ecosystem service assessments. Trends Ecol. Evol. 32, 118–130 (2017).

  7. 7.

    Cumming, G. S., Bodin, Ö., Ernstson, H. & Elmqvist, T. Network analysis in conservation biogeography: challenges and opportunities. Divers. Distrib. 16, 414–425 (2010).

  8. 8.

    Sayles, J. S. & Baggio, J. A. Social–ecological network analysis of scale mismatches in estuary watershed restoration. Proc. Natl Acad. Sci. 114, E1776-E1785 (2017).

  9. 9.

    Baggio, J. A. et al. Multiplex social ecological network analysis reveals how social changes affect community robustness more than resource depletion. Proc. Natl Acad. Sci. 113, 13708–13713 (2016).

  10. 10.

    Liu, J. et al. Systems integration for global sustainability. Science 347, 1258832 (2015).

  11. 11.

    Janssen, M. A. et al. Toward a network perspective of the study of resilience in social-ecological systems. Ecol. Soc. 11, 15 (2006).

  12. 12.

    Turnbull, L. et al. Connectivity and complex systems: learning from a multi-disciplinary perspective. Appl. Netw. Sci. 3, 11 (2018).

  13. 13.

    Bodin, Ö., Barnes, M. L., McAllister, R. R., Rocha, J. C. & Guerrero, A. M. Social–ecological network approaches in interdisciplinary research: a response to Bohan et al. and Dee et al. Trends in Ecology &. Trends Ecol. Evol. 32, 547–549 (2017).

  14. 14.

    Bodin, Ö. Collaborative environmental governance: achieving collective action in social-ecological systems. Science 357, eaan1114 (2017).

  15. 15.

    Barnes, M. L. et al. The social structural foundations of adaptation and transformation in social-ecological systems. Ecol. Soc. 22, 4 (2017).

  16. 16.

    Milo, R. et al. Network motifs: simple building blocks of complex networks. Science 298, 824–827 (2002).

  17. 17.

    Guerrero, A. M., Mcallister, R. R. & Wilson, K. A. Achieving cross‐scale collaboration for large scale conservation initiatives. Conserv. Lett. 8, 107–117 (2015).

  18. 18.

    Bodin, Ö., Crona, B., Thyresson, M., Golz, A.-L. & Tengo, M. Conservation success as a function of good alignment of social and ecological structures and processes. Conserv. Biol. 28, 1371–1379 (2014).

  19. 19.

    Cartwright, D. & Harary, F. Structural balance: a generalization of Heider’s theory. Psychol. Rev. 63, 277 (1956).

  20. 20.

    Berardo, R. The evolution of self-organizing communication networks in high-risk social-ecological systems. Int. J. Commons 8, 236–258 (2014).

  21. 21.

    Robins, G., Bates, L. & Pattison, P. Network governance and environmental management: conflict and cooperation. Public Adm. 89, 1293–1313 (2011).

  22. 22.

    Granovetter, M. Problems of explanation in economic sociology. Netw. Organ. Struct. Form. Action 25, 56 (1992).

  23. 23.

    Guerrero, A., Bodin, Ö., McAllister, R. & Wilson, K. Achieving social-ecological fit through bottom-up collaborative governance: an empirical investigation. Ecol. Soc. 20, 41 (2015).

  24. 24.

    Hardin, G. The tragedy of the commons. Science 162, 1243–1248 (1968).

  25. 25.

    Ostrom, E. Governing the Commons: The Evolution of Institutions for Collective Action (Cambridge University Press, Cambridge, 1990).

  26. 26.

    Axelrod, R. The Evolution of Cooperation (Basic Books, New York, 1984).

  27. 27.

    Kollock, P. Social dilemmas: the anatomy of cooperation. Annu. Rev. Sociol. 24, 183–214 (1998).

  28. 28.

    Balliet, D. Communication and cooperation in social dilemmas: a meta-analytic review. J. Confl. Resolut. 54, 39–57 (2010).

  29. 29.

    Dawes, R. M. Social dilemmas. Annu. Rev. Psychol. 31, 169–193 (1980).

  30. 30.

    Sally, D. Conversation and cooperation in social dilemmas:a meta-analysis of experiments from 1958 to 1992. Ration. Soc. 7, 58–92 (1995).

  31. 31.

    Olsson, P., Folke, C. & Berkes, F. Adaptive comanagement for building resilience in social–ecological systems. Environ. Manag. 34, 75–90 (2004).

  32. 32.

    Cumming, G. S., Cumming, D. H. M. & Redman, C. L. Scale mismatches in social-ecological systems: causes, consequences, and solutions. Ecol. Soc. 11, 14 (2006).

  33. 33.

    Hughes, T. P. et al. Coral reefs in the Anthropocene. Nature 546, 82–90 (2017).

  34. 34.

    Teh, L. S., Teh, L. C. & Sumaila, U. R. A global estimate of the number of coral reef fishers. PLoS ONE 8, e65397 (2013).

  35. 35.

    MacNeil, M. A. et al. Recovery potential of the world’s coral reef fishes. Nature 520, 341–344 (2015).

  36. 36.

    McClanahan, T. & Kosgei, J. Redistribution of benefits but not defection in a fisheries bycatch‐reduction management initiative. Conserv. Biol. 32, 159–170 (2017).

  37. 37.

    Cinner, J. E. Social-ecological traps in reef fisheries. Glob. Environ. Change 21, 835–839 (2011).

  38. 38.

    Graham, N. A., Jennings, S., MacNeil, M. A., Mouillot, D. & Wilson, S. K. Predicting climate-driven regime shifts versus rebound potential in coral reefs. Nature 518, 94 (2015).

  39. 39.

    McClanahan, T. R. et al. Critical thresholds and tangible targets for ecosystem-based management of coral reef fisheries. Proc. Natl Acad. Sci. 108, 17230–17233 (2011).

  40. 40.

    McClanahan, T., Graham, N., MacNeil, M. & Cinner, J. Biomass‐based targets and the management of multispecies coral reef fisheries. Conserv. Biol. 29, 409–417 (2015).

  41. 41.

    McClanahan, T. R., Maina, J. M., Graham, N. A. J. & Jones, K. R. Modeling reef fish biomass, recovery potential, and management priorities in the Western Indian Ocean. PLOS ONE 11, e0154585 (2016).

  42. 42.

    Mouillot, D., Graham, N. A., Villéger, S., Mason, N. W. & Bellwood, D. R. A functional approach reveals community responses to disturbances. Trends Ecol. Evol. 28, 167–177 (2013).

  43. 43.

    McGill, B. J., Enquist, B. J., Weiher, E. & Westoby, M. Rebuilding community ecology from functional traits. Trends Ecol. Evol. 21, 178–185 (2006).

  44. 44.

    Wang, P., Robins, G., Pattison, P. & Lazega, E. Exponential random graph models for multilevel networks. Soc. Netw. 35, 96–115 (2013).

  45. 45.

    Barabási, A.-L. & Albert, R. Emergence of scaling in random networks. Science 286, 509–512 (1999).

  46. 46.

    Alexander, S. M., Barnes, M. L. & Bodin, Ö. Untangling the drivers of community cohesion in small-scale fisheries. Int. J. Commons 12, 519–547 (2018).

  47. 47.

    McPherson, M., Smith-Lovin, L. & Cook, J. M. Birds of a feather: Homophily in social networks. Annu. Rev. Sociol. 27, 415–444 (2001).

  48. 48.

    Luckhurst, B. & Luckhurst, K. Analysis of the influence of substrate variables on coral reef fish communities. Mar. Biol. 49, 317–323 (1978).

  49. 49.

    Cinner, J. E. et al. Gravity of human impacts mediates coral reef conservation gains. Proc. Natl Acad. Sci. 115, E6116–E6125 (2018).

  50. 50.

    Maire, E. et al. How accessible are coral reefs to people? A global assessment based on travel time. Ecol. Lett. 19, 351–360 (2016).

  51. 51.

    Ostrom, E., Burger, J., Field, C. B., Norgaard, R. B. & Policansky, D. Revisiting the commons: local lessons, global challenges. Science 284, 278–282 (1999).

  52. 52.

    Cinner, J. E. et al. Comanagement of coral reef social-ecological systems. Proc. Natl Acad. Sci. 109, 5219–5222 (2012).

  53. 53.

    McClanahan, T., Muthiga, N. A. & Abunge, C. A. Establishment of community managed fisheries’ closures in Kenya: early evolution of the tengefu movement. Coast. Manag. 44, 1–20 (2016).

  54. 54.

    Cinner, J. E. et al. Gear‐based fisheries management as a potential adaptive response to climate change and coral mortality. J. Appl. Ecol. 46, 724–732 (2009).

  55. 55.

    Barnes, M. L., Arita, S., Kalberg, K. & Leung, P. When does it pay to cooperate? Strategic information exchange in the harvest of common-pool fishery resources. Ecol. Econ. 131, 1–11 (2017).

  56. 56.

    Wenger, E. Communities of Practice: Learning, Meaning, and Identity (Cambridge University Press, Cambridge, 1998).

  57. 57.

    Hobson, E. S. Feeding relationships of teleostean fishes on coral reefs in Kona, Hawaii. Fish. Bull. 72, 915–1031 (1974).

  58. 58.

    Froese, R. & Pauly, D. FishBase, www.fishbase.org (2019).

  59. 59.

    Lusher, D., Koskinen, J. & Robins, G. Exponential Random Graph Models for Social Networks: Theory, Methods, and Applications (Cambridge University Press, Cambridge, 2012).

  60. 60.

    Wang, P., Robins, G. & Matous, P. in Multilevel Network Analysis for the Social Sciences Methodos Series, Vol. 12 (eds Lazega, E. & Snijders, T. A. B.) Ch. 6 (Springer International Publishing, Oxford, 2016).

  61. 61.

    Robins, G., Elliott, P. & Pattison, P. Network models for social selection processes. Soc. Netw. 23, 1–30 (2001).

  62. 62.

    Lubell, M., Robins, G. & Wang, P. Network structure and institutional complexity in an ecology of water management games. Ecol. Soc. 19, 23 (2014).

  63. 63.

    Wang, P., Robins, G. & Pattison, P. PNet: Program for the Simulation and Estimation of Exponential Random Graph Models (Melbourne School of Psychological Sciences, The University of Melbourne, 2009).

  64. 64.

    Snijders, T. A. Markov chain Monte Carlo estimation of exponential random graph models. J. Soc. Struct. 3, 1–40 (2002).

  65. 65.

    Satterthwaite, F. E. An approximate distribution of estimates of variance components. Biometrics 2, 110–114 (1946).

  66. 66.

    McClanahan, T. R. Kenyan coral reef lagoon fish: effects of fishing, substrate complexity, and sea urchins. Coral Reefs 13, 231–241 (1994).

  67. 67.

    Barnes, M., Lynham, J., Kalberg, K. & Leung, P. S. Social networks and environmental outcomes. Proc. Natl Acad. Sci. 113, 6466–6471 (2016).

  68. 68.

    Moreno, J. L. & Jennings, H. H. Statistics of social configurations. Sociometry 1, 342–374 (1938).

  69. 69.

    McClanahan, T. R., Graham, N. A., Calnan, J. M. & MacNeil, M. A. Toward pristine biomass: reef fish recovery in coral reef marine protected areas in Kenya. Ecol. Appl. 17, 1055–1067 (2007).

Download references

Acknowledgements

This project was supported by an NSF Social, Behavioral, and Economic Sciences Postdoctoral Research Fellowship Grant to M.L.B. (#1513354), the ARC Centre of Excellence for Coral Reef Studies, the Darwin Initiative (#20–017), and the Marine Science for Management Contract No. MASMA/OP/2014/04. Ö.B. was supported by the Swedish Foundation for Strategic Environmental Research (Mistra) through a core grant to the Stockholm Resilience Centre at Stockholm University. We thank G. Robins, G. Cumming, J. Cinner, and our reviewers for helpful comments on earlier drafts, the Wildlife Conservation Society’s Coral Reef Conservation Program for data access and logistical support, Stephen Wanyonyi and Innocent Muly for their assistance in the field, Emmanuel Mbaru for help with data processing and graphics, and all of the fishers who participated in this project.

Author information

Affiliations

  1. ARC Centre of Excellence for Coral Reef Studies, James Cook University, Townsville, QLD, 4811, Australia

    • Michele L. Barnes
    • , Andrew S. Hoey
    •  & Nicholas A. J. Graham
  2. Department of Botany, University of Hawaii at Manoa, 3190 Maile Way, Honolulu, HI, 96822, USA

    • Michele L. Barnes
    •  & Orou G. Gaoue
  3. Stockholm Resilience Centre, Stockholm University, 106 91, Stockholm, Sweden

    • Örjan Bodin
  4. Wildlife Conservation Society, Marine Programs, Bronx, NY, 10460, USA

    • Tim R. McClanahan
  5. Center for Oceans, Conservation International, Honolulu, 96821, HI, USA

    • John N. Kittinger
  6. Center for Biodiversity Outcomes, Julie Ann Wrigley Global Institute of Sustainability, Arizona State University, Tempe, 85281, AZ, USA

    • John N. Kittinger
  7. Department of Ecology and Evolutionary Biology, University of Tennessee, Knoxville, TN, 37996, USA

    • Orou G. Gaoue
  8. Faculty of Agronomy, University of Parakou, 01 BP 123, Parakou, Benin

    • Orou G. Gaoue
  9. Department of Geography, Environmental Management and Energy Studies, University of Johannesburg, APK Campus, Johannesburg, 2092, South Africa

    • Orou G. Gaoue
  10. Lancaster Environment Centre, Lancaster University, Lancaster, LA1 4YQ, UK

    • Nicholas A. J. Graham

Authors

  1. Search for Michele L. Barnes in:

  2. Search for Örjan Bodin in:

  3. Search for Tim R. McClanahan in:

  4. Search for John N. Kittinger in:

  5. Search for Andrew S. Hoey in:

  6. Search for Orou G. Gaoue in:

  7. Search for Nicholas A. J. Graham in:

Contributions

M.L.B. designed the integrated social–ecological research. T.R.M. designed the research on ecological conditions and biophysical characteristics. M.L.B., T.R.M., A.S.H., and N.A.J.G. performed the research. M.L.B., Ö.B., and T.R.M. analyzed data. M.L.B., Ö.B., T.R.M., J.N.K., A.S.H., O.G.G., and N.A.J.G. wrote the paper.

Competing interests

The authors declare no competing interests.

Corresponding author

Correspondence to Michele L. Barnes.

Supplementary information

About this article

Publication history

Received

Accepted

Published

DOI

https://doi.org/10.1038/s41467-019-09994-1

Comments

By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.