Contributed Paper
Conservation Success as a Function of Good Alignment of Social and Ecological Structures and Processes ¨ ORJAN BODIN,∗ § BEATRICE CRONA,∗ ‡ MATILDA THYRESSON,∗ ANNA-LEA GOLZ,† ¨∗ AND MARIA TENGO ∗
Stockholm Resilience Centre, Stockholm University, 10691 Stockholm, Sweden †Department of Ecology, Environment and Plant Science, Stockholm University, 10691 Stockholm, Sweden ‡Global Economic Dynamics and the Biosphere, Royal Swedish Academy of Sciences, P.O. Box 50005, SE-104 05, Stockholm, Sweden
Abstract: How to create and adjust governing institutions so that they align (fit) with complex ecosystem processes and structures across scales is an issue of increasing concern in conservation. It is argued that lack of such social-ecological fit makes governance and conservation difficult, yet progress in explicitly defining and rigorously testing what constitutes a good fit has been limited. We used a novel modeling approach and data from case studies of fishery and forest conservation to empirically test presumed relationships between conservation outcomes and certain patterns of alignment of social-ecological interdependences. Our approach made it possible to analyze conservation outcome on a systems level while also providing information on how individual actors are positioned in the complex web of social-ecological interdependencies. We found that when actors who shared resources were also socially linked, conservation at the level of the whole socialecological system was positively affected. When the scales at which individual actors used resources and the scale at which ecological resources were interconnected to other ecological resources were aligned through tightened feedback loops, conservation outcome was better than when they were not aligned. The analysis of individual actors’ positions in the web of social-ecological interdependencies was helpful in understanding why a system has a certain level of social-ecological fit. Results of analysis of positions showed that different actors contributed in very different ways to achieve a certain fit and revealed some underlying difference between the actors, for example in terms of actors’ varying rights to access and use different ecological resources. Keywords: common pool resource management, environmental governance, scale mismatch, social-ecological fit, social-ecological systems El E´xito de la Conservaci´ on como Funci´ on de una Buena Alineaci´ on de Estructuras y Procesos Sociales y Ecol´ ogicos Bodin et al.
Resumen: C´omo crear y ajustar a las instituciones gobernantes para que puedan alinearse (adecuarse) con los procesos complejos y estructuras de los ecosistemas es una cuesti´ on de preocupaci´ on creciente en la conservaci´ on. Se discute que la falta de dicha adecuaci´ on socio-ecol´ ogica hace que la gobernanza y la conservaci´ on sean dif´ıciles, sin embargo el progreso en la definici´ on expl´ıcita y en probar rigurosamente qu´e constituye una buena adecuaci´ on ha sido limitado. Usamos un m´etodo de modelado novedoso y datos de casos de estudio de la conservaci´ on pesquera y de bosques para probar emp´ıricamente las supuestas relaciones entre los resultados de la conservaci´ on y ciertos patrones de alineaci´ on de interdependencias socio-ecol´ ogicas. Encontramos que cuando los participantes que compart´ıan recursos tambi´en estaban conectados socialmente, la conservaci´ on en el nivel de todo el sistema socio-ecol´ ogico era afectada positivamente. Cuando las escalas en las cuales los participantes individuales usaban recursos y la escala en la cual los recursos ecol´ ogicos estaban interconectados con otros recursos ecol´ ogicos se alineaban por medio de ciclos de retroalimentaci´ on
§ email [email protected] Paper submitted September 12, 2013; revised manuscript accepted January 20, 2014.
1 Conservation Biology, Volume 00, No. 0, 1–9 C 2014 Society for Conservation Biology DOI: 10.1111/cobi.12306
Social-Ecological Fit and Conservation
2
compactos, los resultados de la conservaci´ on eran mejores que cuando no se alineaban. El an´ alisis de las posiciones de los participantes individuales en la red de interdependencias socio-ecol´ ogicas fue u ´ til en el entendimiento de por qu´e un sistema tiene cierto nivel de adecuaci´ on socio-ecol´ ogica. Los resultados del an´ alisis de las posiciones mostraron que los diferentes participantes contribuyen en diferentes formas para obtener cierta adecuaci´ on y revelaron algunas diferencias subyacentes entre los participantes, por ejemplo en t´erminos de los derechos variantes de los participantes de acceder y usar diferentes recursos ecol´ ogicos.
Palabras Clave: adecuaci´on socio-ecol´ogica, desigualdad de escala, gobernanza ambiental, manejo com´un de recursos acumulados, sistemas socio-ecol´ ogicos
Introduction Contemporary environmental problems abound on all spatial and institutional scales, are tightly intertwined with numerous anthropocentric processes, and therefore challenge the traditional separation of socioeconomic and ecological sciences (e.g., Rockstr¨ om et al. 2009; Stafford et al. 2009). Hence, research on conservation and sustainability is increasingly focused on transcending these boundaries and embracing a more integrative systems perspective that acknowledges the links between the socioeconomic and ecological domains (e.g., Berkes et al. 2003). A growing interest emerging through this endeavor is focused on the fit between ecosystem processes, ecosystem services, and institutions in place to govern them and the importance of misfit among these for explaining degradation of natural resources (e.g., Young 2002; Folke et al. 2007; Galaz et al. 2008). The problem of fit is multifaceted and has been characterized along several dimensions (e.g., functional, spatial, and institutional fit [e.g., Cash et al. 2006]), and several challenges in achieving fit have been identified (Galaz et al. 2008). Despite the increasing interest in the topic, limited work has been done to move from theoretical constructs of fit toward methodologies that can be used to analyze empirically both social and ecological data to assess social-ecological fit and its potential implications. In an attempt to address this gap, we devised a method for integrative social-ecological analysis of fit and used it to empirically study 2 broad and substantive concerns that are by nature interdisciplinary and that we argue can be conceptualized and analyzed from the perspective of social-ecological fit. These were stakeholder interaction and how it relates to successful management of common pool resources (CPR) and conservation in the face of ecological externalities and geographic displacement. The first issue relates to how well patterns of stakeholder interactions align with the ways in which stakeholders use different common ecological resources (Ostrom 1990) and is of particular importance in cases where formally sanctioned and enforced institutions are weak or lacking. The second issue concerns the level of fit between the scales at which key ecological processes occur and the scales of the patterns of resource management, use, and extraction. Misalignments of these processes can, for example, decouple costs and benefits of resource use, and Conservation Biology Volume 00, No. 0, 2014
thus reduce incentives for sustainable management (e.g. Prugh et al. 1999; Cumming et al. 2006). Although research addressing each of these 2 challenges abounds, little has been done to examine them from an explicit and integrated perspective of socialecological fit. The research approach we used builds on a recently proposed social-ecological modeling approach (Bodin & Teng¨ o 2012). The approach draws from the growing interdisciplinary field of network science and models and analyzes a coupled social-ecological system (SES) as a social-ecological network in which actors and ecological resources are conceptualized as nodes and interdependences among these entities are conceptualized as links. Using this approach, we disassembled 2 realworld SESs with contrasting conservation outcomes into separate sets of small and precisely defined SES building blocks. We then show how this deconstruction makes it possible to theoretically and empirically link patterns of social and ecological interdependencies within and between the social and ecological domains of an SES to the 2 broad and substantive concerns outlined above. The 2 study systems involve local-level and small-scale resource use, and in both cases there is in practice an absence of external third-party rules of enforcement. The first case is a coastal fisheries system in decline in Kenya (Crona & Bodin 2006) and the second is a locally based system for conservation of forest patches in an agricultural landscape in southern Madagascar which has been successful at maintaining critical ecological functions (Bodin & Teng¨ o 2012). By comparing the results from the disassembling processes of the 2 SESs, we examined how our approach can make it possible to test hypothesized relationships between governance and conservation outcomes and different patterns of social-ecological interdependencies. We also investigated how a novel positional perspective (i.e., focusing on what specific positions different actors occupy in the SES building blocks) can be used to explore the contribution of different actors to conservation outcomes.
Method SESs as Sets of Simple Building Blocks At the core of our approach is the concept of minimal SES building blocks (i.e., small and precisely defined SES
Bodin et al.
configurations or motifs [Bodin & Teng¨ o 2012]). It is often difficult or impossible to, in a theoretically informed way, to infer direct relationships between patterns of interdependences and governance and conservation outcomes in a full-scale SES consisting of myriads of social and ecological components and interdependencies. A simple building block capturing some well-defined pattern of social and ecological interdependencies can, by applying appropriate social- and ecological theories, however, more easily be linked to specific governance challenges and associated conservation outcomes via hypothesized and theoretically informed causal relationships. We used this social-ecological network approach to formulate hypotheses related to the outlined governance challenges by constructing a set of precisely defined SES building blocks. CPR Management and Stakeholder Interaction CPR management is a central concept in studies of SES (Ostrom 2009). A CPR is, essentially, a resource that is extracted by several actors, and extraction by one actor means there is less of the resource available for others. Thus, in a social-ecological network context, 2 social actors sharing one ecological resource implies competition that can lead to overharvesting (Ostrom 1990). In the absence of clearly defined regulation that is officially sanctioned and enforced by formal authorities, these 2 actors would, in order to avoid overharvesting, in effect be reliant upon mutual agreement if they were to regulate resource extraction themselves. Such agreements would be easier to reach if the actors were communicating. In other words, we hypothesize that social connectivity is a prerequisite for the establishment of mutual agreement on resource regulations and that such mutual agreements are beneficial for conservation in the absence of other institutions regulating resource extraction. These assumptions can be captured by 2 different SES building blocks consisting of 2 social actors and one ecological resource as outlined in Figs. 1(a) and (b), and yield the following hypotheses: (1a) the SES building block in which there are 2 socially connected actors that share a resource (CPR with communication in Fig. 1a) is more prevalent than by chance in cases with sustainable conservation of ecological resources; (1b) the SES building block in which 2 socially unconnected actors share a resource (CPR without communication in Fig. 1b) is more prevalent than by chance in cases without sustainable conservation of ecological resources. Externalities and Displacement It is commonly argued that social and ecological processes need to be aligned in space and time; otherwise, such scale mismatches will make governance inherently difficult because management measures cannot, by def-
3
A. CPR w comm
B. CPR w/o comm
C. Scale fit
D. Scale misfit
Figure 1. Stylized configurations of minimal social-ecological system (building blocks) related to a set of governance challenges. Nodes (white circles) are actors and ecological resources (grey circles). In (a) the link between the 2 actors is realized (they are communicating [comm]); therefore, the potential for the actors to avoid overharvesting the common pool resource (CPR) is assumed to increase. In (b) the link between actors is not realized (they are not communicating), and CPR management is more difficult than in (a). In (c) the 2 links between the actor and the ecological resources are realized (both interlinked ecological resources are used by the actor); therefore, indirect ecological effect feeds back directly to the actor, creating a tight feedback loop (scale fit). In (d) the link between the actor and the resource on the right is not realized, and any indirect effect caused by using the ecological resource on the left, which affects the resource on the right, is not feedback directly to the actor (i.e., scale misfit). inition, embrace the full extent of the managed entity (i.e., the ecosystem). This in turn can lead to ineffective management (Cumming et al. 2006; Guerrero et al. 2013) and overall resource decline. Our concern is, therefore, to differentiate between situations where actors are linked (though use) to all or just to a subset of the constituent components of an ecosystem consisting of several interdependent ecological resources. As a social-ecological building block this would be expressed as follows. If 2 ecological resources are interdependent, and a specific actor is linked only to one of them, the other resource can, in the eyes of that actor, be perceived as external (Fig. 1d). Hence, the incentive to use the first resource in such a way as to minimize potential negative side effects on the second resource is reduced. In economics this would be referred to as externality (Prugh et al. 1999), and it creates a misalignment between how costs and benefits related to resource use are distributed and the extent of the underlying ecosystem that provides for the resources. The misalignment could have a negative effect on the actor using one of the resources, but this feedback could be long delayed (this is an example of a temporal misfit). However, if the actor in this example uses both resources simultaneously, the incentive to manage both resources as interdependent
Conservation Biology Volume 00, No. 0, 2014
Social-Ecological Fit and Conservation
4
increases, and the overall level of fit would be higher (Fig. 1c). This situation yields the following hypotheses: (2a) the SES building block in which an actor and 2 interdependent ecological resources are fully connected (scale fit, Fig. 1c), is more prevalent than by chance in cases with sustainable conservation of ecological resources; (2b) the SES building block in which an actor is only linked to one of 2 interdependent ecological resources (scale misfit, Fig. 1d), is more prevalent than by chance in cases without sustainable conservation of ecological resources. Deconstructing the Full Scale SES A real SES is naturally much more complex than any of the defined building blocks. To be able to infer relationships between conservation outcomes and the socialecological fit of a given system, we therefore needed to proceed in 2 steps. First, we disassembled the fullscale social-ecological network, encompassing all actors, ecological resources, and interdependencies of the studied SES into sets of building blocks. In other words, the full scale SES was deconstructed into sets of irreducible building blocks. Second, we zoomed in on the patterns of occurrences of each of the types of building blocks. In that way, we had a precise measure of how common the occurrence of each specific building block was in the SES. This measure should however be related to a baseline measure (i.e., the measure needs to be grounded). To that end, we compared the empirical measure with the distribution of measures acquired from a large number of randomly generated social-ecological networks. These random networks each had the same number of social and ecological nodes as the empirical network. The amount of social-to-social, social-toecological, and ecological-to-ecological link were also the same. The randomly generated social-ecological network thus provided a baseline estimate (i.e., the null model), and the degree to which the distributions of building blocks deviated from the baseline informed us as to whether these different building blocks were over- or underrepresented in the full scale social-ecological network. Conceptually, this could be seen as a sort of spectral analysis of the real social-ecological network (cf. Fourier transform). Our approach makes it possible, in general and beyond this study, to empirically test theoretically informed relationships between patterns of social-ecological interdependences and specific governance and conservation outcomes of an SES. Hence, this approach is not about searching for any type of pattern of social-ecological interdependences; it is about searching for specific patterns that could be linked, theoretically, to different governance and conservation outcomes. Also, the socialecological networks could be interpreted more explicitly from a perspective of multilevel network interdepen-
Conservation Biology Volume 00, No. 0, 2014
dency. The social and ecological networks, respectively, would from this perspective be seen as directly dependent on each other. For example, the structure of the social network could be seen to be the result of various social processes constrained both by the structure of the underlying ecological network itself and by the structure in which actors are linked to the ecological network. In other words, these structures would be seen as given, and it would only be the links between the actors in the social network that could vary. In contrast to a multilevel network interdependency perspective, we based our hypotheses on the assumption that actors can choose which ecological resources and which other actors to connect to. That does not mean that creating a link is effortless and will never involve negotiation or engaging in conflict with other actors or that there will not be other kinds of costs. Rather, it means there are no imposed constrains that would fundamentally prevent actors from succeeding in creating certain links (or to abandon certain links). Given the nature of the studied cases, we argue this assumption essentially holds true. In other cases, the actors’ degrees of freedom might be more limited, and in such cases the issue of multilevel network interdependency becomes more pronounced. Our disassembly approach is conceptually similar to recent work on multilevel exponential random graph models (MLERGM) used for analyzing social networks (Wang et al. 2013). Such models are fairly complex, and further description is beyond the scope of this work. However, MLERGM could provide, among other things, a more statistically robust way to separate different building blocks that to some extent overlap. It also explicates the perspective of network interdependencies. MLERGM currently supports quite a few different building blocks, and could be extended to support even more. Finally, our approach has some similarities to recent methods applied in studying motifs in food webs (e.g., Stouffer et al. 2012). Thus, there is high potential for methodological synergies across different research field. Positional Analysis The disassembling approach is not limited to investigating characteristics at the level of the entire socialecological network; it also allows for investigations of the extent to which individual actors and resources are engaged in certain building blocks, as well as what specific positions they tend to occupy (cf. Gould & Fernandez 1989; Stouffer et al. 2012). This provides a means for linking the patterns of interdependencies of the larger SES with the characteristics of individual actors and ecological resources. This can provide information on whether certain actors more commonly occupy positions in the SES building blocks characterized by good or bad fit, thereby suggesting that certain actors might have a
Bodin et al.
propensity to engage (or not) with others in certain ways that can disproportionally affect the overall level of fit. This propensity can be caused by individual characteristics, social norms, or other social structures. Comparison of 2 Contrasting Study Systems To test the applicability of our proposed approach, we analyzed 2 contrasting real world SESs. The study systems differed in several respects which are further described below, but most importantly they differed with regards to how well the actors in the 2 systems had managed to conserve some ecological resources on which they depended for their livelihoods. As such, the 2 cases allow for a comparative approach when evaluating the results from the social-ecological network deconstruction. Based on the theoretically informed link between each building block and conservation outcomes, we hypothesized that the study system where resources are more conservatively used would exhibit a higher frequency of building blocks hypothetically associated with a good fit (hypotheses 1a and 2a). In the case with less successful conservation building blocks associated with poor fit should be overrepresented (hypotheses 1b and 2b). The number of study systems is far too low to test the hypotheses in a statistically rigorous way; however, the comparative approach at least allowed us to make an early assessment of the potential of the approach. The first study system was a rural, coastal fishing village, approximately 50 km south of Mombasa, Kenya (Crona & Bodin 2006). The village had approximately 200 households and an estimated 1000 inhabitants. Approximately 44% of the households were directly involved in a small-scale multispecies fishery, and overharvesting and declining fish stocks characterized the fishery in the region. Despite a broad recognition of the decline, the actors involved had initiated no substantial efforts to remedy this apparently unfavorable situation. Describing this study system as a social-ecological network is a nontrivial task, and any meaningful interpretation of the revealed patterns of interdependencies is inevitably linked to how this has been done. Bodin and Teng¨ o (2012) suggest a sequential heuristic approach generally applicable to basically any SES consisting of multiple actors and resources. The approach aims to guide the conceptualization of the SES as a social-ecological network in such a way that the nodes and links are defined in a transparent way, that nodes and links are comparable in size and presumed effect, and that no possible interrelationships are defined a priori as impossible. Also, defining SES boundaries should be given careful consideration. The approach consists of a series of steps that should be reiterated with the aim to construct cross-case comparable social-ecological networks. How comparable any 2 social-ecological networks are also depends on how
5
generically the hypotheses related to the relationship between structure and function have been formulated. We formulated our hypotheses broadly; therefore, they should be broadly applicable to many different conceptualizations of SESs as social-ecological networks. The small-scale fishery study system was modeled as a social-ecological network with actors (social nodes) defined as groups of fishers using the same type of fishing gear and therefore targeting the same fish species. The way we defined the study system as a social-ecological network is described in detail in Supporting Information. Links between the fisher groups represented exchange of ecological information, and the ecological resources (ecological nodes) were the most important targeted fish species. Social-ecological links connected fisher groups to their targeted fish species, and the ecological links represented trophic interactions in the fish community. The data on trophic interactions in the fish community were based on theoretical estimates and not explicit observations, and for all fisher groups, except the ring net fishers, secondary data were used to assess the targeted fish species (Supporting Information). Because some of the data for this study system were based on assumptions, we performed a sensitivity analysis in which the number of trophic links varied and the number of targeted fish species included in the network varied systematically (Supporting Information). The second study system was a rural agricultural landscape in Androy, southern Madagascar (Bodin & Teng¨ o 2012). Forest patches, ranging in size from 3 to >90 ha, are scattered across the landscape of small fields and pastures. The forest patches are protected by taboos restricting access and use, they generate essential ecosystem services such as microclimate regulation and crop pollination, they are culturally important as ancestral burial grounds, and they serve as sites for ceremonies and as symbols of the link between people and the land (Teng¨ o & von Heland 2011). The landscape has been remarkably well preserved over a substantial amount of time despite strong pressures on land and forest resources (Teng¨ o& von Heland 2011; Bodin & Teng¨ o 2012). The social-ecological network in our study system was defined by Bodin and Teng¨ o (2012), and we used it without modification. The actors were defined as clans residing in different parts of the landscape who used or benefited from the ecosystem services (including cultural services) stemming from one or several of the forest patches (the resources). The social links were based on interclan relationships such as joint agreement on forest use and management and on kinship and intermarriages that had implications for access to forest services. The ecological links between the forest patches were based on species dispersal, which is essential for the ability of the forest patches to exist over time (Bodin & Teng¨ o 2012).
Conservation Biology Volume 00, No. 0, 2014
Social-Ecological Fit and Conservation
6
Spear gun
Ring net
Handline
Gill net Seine net
Figure 2. Estimated social-ecological network of the African small-scale fishery. Circular nodes at the top are gear-defined groups of fishers (Crona & Bodin 2006), dark grey diamonds in the middle are targeted species of predator fish, and light grey diamonds at the bottom are targeted species of prey fish. Dark grey lines with double-headed arrows are links that represent an exchange of ecological knowledge and information (i.e., communication) among groups of fishers, light grey arrows are extraction of particular fish species by certain groups of fishers, and black arrows are trophic interactions.
Results
Figure 3. Estimated social-ecological network of the Madagascar agricultural landscape (adapted from Bodin and Teng¨ o [2012]). Grey circles are nodes representing different clans residing in the landscape, polygons are nodes representing forest patches, black lines show which clans are benefiting from ecosystem services stemming from which forest patches, light grey lines show patch interconnectivity through seed dispersal, and dark grey lines are social ties between clans. The clans with a white cross reside outside the area shown.
The social-ecological networks representing the 2 study systems both exhibited nontrivial patterns of socialecological interdependencies (Figs. 2 and 3). To provide a baseline measure for the frequency analysis of the studied building blocks, we created 10,000 random social-ecological networks for each study system. The frequency analysis showed that for the CPR governance challenge, the building block corresponding to resource competition without actor communication was significantly underrepresented in the agricultural system (Fig. 4b). Resource competition paired with actor communication was, however, significantly overrepresented in the agricultural system (Fig. 4a) and was significantly underrepresented in the small-scale fishery system (Fig. 4e). For the governance challenge of scale mismatch a similar pattern was found. The building block corresponding to scale-match was significantly overrepresented in the agricultural system (Fig. 4c) and marginally significantly underrepresented in the small-scale fishery system (Fig. 4g, P = 0.053), whereas the frequency of the building block corresponding to scale mismatch was significantly underrepresented in the agricultural system (Fig. 4d).
The sensitivity analysis (Supporting Information) indicated that results for the fishery system were fairly robust in regards to trophic link uncertainties (Supporting Information). The significant patterns of underrepresentation of the building blocks associated with good fit (Fig. 4e & g) were, however, less pronounced when only the 70% most targeted fish species were accounted for. When a larger pool of species were accounted for (90%), the underrepresentation of the building blocks associated with good fit instead became more pronounced, and the building block corresponding to resource competition without actor communication became significantly overrepresented. Hence, the overall social-ecological fit decreased when more targeted species were included in the analysis. How often the different actors in each case occupied all the structurally unique positions in the studied building blocks varied greatly among actors (Table 1). The variability could, for example, range from 0% to 55% (agricultural system, scale fit, Table 1).
Conservation Biology Volume 00, No. 0, 2014
Bodin et al.
7
CPR w comm
CPR w/o comm −−
Scale fit
Scale misfit
++
(a) (14 )
(b) (7 )
(c) (20 )
(e) (6 )
(f) (18)
(g) (21 )
−−
++
(d) (34 )
.
(h) (137)
Agricultural landscape
Small-scale fishery
Figure 4. Frequency of social-ecological building blocks in the studied social-ecological networks: (a-d) agricultural landscape and (e-h) small-scale fishery. Black bars are drawn from the real social-ecological networks, whereas grey bars come from 10,000 randomly generated networks. If the number of building blocks in the real social-ecological network significantly deviates from the distribution of the random networks (e.g., in [a]), then one or 2 plus or minus signs (5% and 1% significance levels, respectively) are next to the number of building blocks above each graph. Marginal significance,
Conservation Success as a Function of Good Alignment of Social and Ecological Structures and Processes ¨ ORJAN BODIN,∗ § BEATRICE CRONA,∗ ‡ MATILDA THYRESSON,∗ ANNA-LEA GOLZ,† ¨∗ AND MARIA TENGO ∗
Stockholm Resilience Centre, Stockholm University, 10691 Stockholm, Sweden †Department of Ecology, Environment and Plant Science, Stockholm University, 10691 Stockholm, Sweden ‡Global Economic Dynamics and the Biosphere, Royal Swedish Academy of Sciences, P.O. Box 50005, SE-104 05, Stockholm, Sweden
Abstract: How to create and adjust governing institutions so that they align (fit) with complex ecosystem processes and structures across scales is an issue of increasing concern in conservation. It is argued that lack of such social-ecological fit makes governance and conservation difficult, yet progress in explicitly defining and rigorously testing what constitutes a good fit has been limited. We used a novel modeling approach and data from case studies of fishery and forest conservation to empirically test presumed relationships between conservation outcomes and certain patterns of alignment of social-ecological interdependences. Our approach made it possible to analyze conservation outcome on a systems level while also providing information on how individual actors are positioned in the complex web of social-ecological interdependencies. We found that when actors who shared resources were also socially linked, conservation at the level of the whole socialecological system was positively affected. When the scales at which individual actors used resources and the scale at which ecological resources were interconnected to other ecological resources were aligned through tightened feedback loops, conservation outcome was better than when they were not aligned. The analysis of individual actors’ positions in the web of social-ecological interdependencies was helpful in understanding why a system has a certain level of social-ecological fit. Results of analysis of positions showed that different actors contributed in very different ways to achieve a certain fit and revealed some underlying difference between the actors, for example in terms of actors’ varying rights to access and use different ecological resources. Keywords: common pool resource management, environmental governance, scale mismatch, social-ecological fit, social-ecological systems El E´xito de la Conservaci´ on como Funci´ on de una Buena Alineaci´ on de Estructuras y Procesos Sociales y Ecol´ ogicos Bodin et al.
Resumen: C´omo crear y ajustar a las instituciones gobernantes para que puedan alinearse (adecuarse) con los procesos complejos y estructuras de los ecosistemas es una cuesti´ on de preocupaci´ on creciente en la conservaci´ on. Se discute que la falta de dicha adecuaci´ on socio-ecol´ ogica hace que la gobernanza y la conservaci´ on sean dif´ıciles, sin embargo el progreso en la definici´ on expl´ıcita y en probar rigurosamente qu´e constituye una buena adecuaci´ on ha sido limitado. Usamos un m´etodo de modelado novedoso y datos de casos de estudio de la conservaci´ on pesquera y de bosques para probar emp´ıricamente las supuestas relaciones entre los resultados de la conservaci´ on y ciertos patrones de alineaci´ on de interdependencias socio-ecol´ ogicas. Encontramos que cuando los participantes que compart´ıan recursos tambi´en estaban conectados socialmente, la conservaci´ on en el nivel de todo el sistema socio-ecol´ ogico era afectada positivamente. Cuando las escalas en las cuales los participantes individuales usaban recursos y la escala en la cual los recursos ecol´ ogicos estaban interconectados con otros recursos ecol´ ogicos se alineaban por medio de ciclos de retroalimentaci´ on
§ email [email protected] Paper submitted September 12, 2013; revised manuscript accepted January 20, 2014.
1 Conservation Biology, Volume 00, No. 0, 1–9 C 2014 Society for Conservation Biology DOI: 10.1111/cobi.12306
Social-Ecological Fit and Conservation
2
compactos, los resultados de la conservaci´ on eran mejores que cuando no se alineaban. El an´ alisis de las posiciones de los participantes individuales en la red de interdependencias socio-ecol´ ogicas fue u ´ til en el entendimiento de por qu´e un sistema tiene cierto nivel de adecuaci´ on socio-ecol´ ogica. Los resultados del an´ alisis de las posiciones mostraron que los diferentes participantes contribuyen en diferentes formas para obtener cierta adecuaci´ on y revelaron algunas diferencias subyacentes entre los participantes, por ejemplo en t´erminos de los derechos variantes de los participantes de acceder y usar diferentes recursos ecol´ ogicos.
Palabras Clave: adecuaci´on socio-ecol´ogica, desigualdad de escala, gobernanza ambiental, manejo com´un de recursos acumulados, sistemas socio-ecol´ ogicos
Introduction Contemporary environmental problems abound on all spatial and institutional scales, are tightly intertwined with numerous anthropocentric processes, and therefore challenge the traditional separation of socioeconomic and ecological sciences (e.g., Rockstr¨ om et al. 2009; Stafford et al. 2009). Hence, research on conservation and sustainability is increasingly focused on transcending these boundaries and embracing a more integrative systems perspective that acknowledges the links between the socioeconomic and ecological domains (e.g., Berkes et al. 2003). A growing interest emerging through this endeavor is focused on the fit between ecosystem processes, ecosystem services, and institutions in place to govern them and the importance of misfit among these for explaining degradation of natural resources (e.g., Young 2002; Folke et al. 2007; Galaz et al. 2008). The problem of fit is multifaceted and has been characterized along several dimensions (e.g., functional, spatial, and institutional fit [e.g., Cash et al. 2006]), and several challenges in achieving fit have been identified (Galaz et al. 2008). Despite the increasing interest in the topic, limited work has been done to move from theoretical constructs of fit toward methodologies that can be used to analyze empirically both social and ecological data to assess social-ecological fit and its potential implications. In an attempt to address this gap, we devised a method for integrative social-ecological analysis of fit and used it to empirically study 2 broad and substantive concerns that are by nature interdisciplinary and that we argue can be conceptualized and analyzed from the perspective of social-ecological fit. These were stakeholder interaction and how it relates to successful management of common pool resources (CPR) and conservation in the face of ecological externalities and geographic displacement. The first issue relates to how well patterns of stakeholder interactions align with the ways in which stakeholders use different common ecological resources (Ostrom 1990) and is of particular importance in cases where formally sanctioned and enforced institutions are weak or lacking. The second issue concerns the level of fit between the scales at which key ecological processes occur and the scales of the patterns of resource management, use, and extraction. Misalignments of these processes can, for example, decouple costs and benefits of resource use, and Conservation Biology Volume 00, No. 0, 2014
thus reduce incentives for sustainable management (e.g. Prugh et al. 1999; Cumming et al. 2006). Although research addressing each of these 2 challenges abounds, little has been done to examine them from an explicit and integrated perspective of socialecological fit. The research approach we used builds on a recently proposed social-ecological modeling approach (Bodin & Teng¨ o 2012). The approach draws from the growing interdisciplinary field of network science and models and analyzes a coupled social-ecological system (SES) as a social-ecological network in which actors and ecological resources are conceptualized as nodes and interdependences among these entities are conceptualized as links. Using this approach, we disassembled 2 realworld SESs with contrasting conservation outcomes into separate sets of small and precisely defined SES building blocks. We then show how this deconstruction makes it possible to theoretically and empirically link patterns of social and ecological interdependencies within and between the social and ecological domains of an SES to the 2 broad and substantive concerns outlined above. The 2 study systems involve local-level and small-scale resource use, and in both cases there is in practice an absence of external third-party rules of enforcement. The first case is a coastal fisheries system in decline in Kenya (Crona & Bodin 2006) and the second is a locally based system for conservation of forest patches in an agricultural landscape in southern Madagascar which has been successful at maintaining critical ecological functions (Bodin & Teng¨ o 2012). By comparing the results from the disassembling processes of the 2 SESs, we examined how our approach can make it possible to test hypothesized relationships between governance and conservation outcomes and different patterns of social-ecological interdependencies. We also investigated how a novel positional perspective (i.e., focusing on what specific positions different actors occupy in the SES building blocks) can be used to explore the contribution of different actors to conservation outcomes.
Method SESs as Sets of Simple Building Blocks At the core of our approach is the concept of minimal SES building blocks (i.e., small and precisely defined SES
Bodin et al.
configurations or motifs [Bodin & Teng¨ o 2012]). It is often difficult or impossible to, in a theoretically informed way, to infer direct relationships between patterns of interdependences and governance and conservation outcomes in a full-scale SES consisting of myriads of social and ecological components and interdependencies. A simple building block capturing some well-defined pattern of social and ecological interdependencies can, by applying appropriate social- and ecological theories, however, more easily be linked to specific governance challenges and associated conservation outcomes via hypothesized and theoretically informed causal relationships. We used this social-ecological network approach to formulate hypotheses related to the outlined governance challenges by constructing a set of precisely defined SES building blocks. CPR Management and Stakeholder Interaction CPR management is a central concept in studies of SES (Ostrom 2009). A CPR is, essentially, a resource that is extracted by several actors, and extraction by one actor means there is less of the resource available for others. Thus, in a social-ecological network context, 2 social actors sharing one ecological resource implies competition that can lead to overharvesting (Ostrom 1990). In the absence of clearly defined regulation that is officially sanctioned and enforced by formal authorities, these 2 actors would, in order to avoid overharvesting, in effect be reliant upon mutual agreement if they were to regulate resource extraction themselves. Such agreements would be easier to reach if the actors were communicating. In other words, we hypothesize that social connectivity is a prerequisite for the establishment of mutual agreement on resource regulations and that such mutual agreements are beneficial for conservation in the absence of other institutions regulating resource extraction. These assumptions can be captured by 2 different SES building blocks consisting of 2 social actors and one ecological resource as outlined in Figs. 1(a) and (b), and yield the following hypotheses: (1a) the SES building block in which there are 2 socially connected actors that share a resource (CPR with communication in Fig. 1a) is more prevalent than by chance in cases with sustainable conservation of ecological resources; (1b) the SES building block in which 2 socially unconnected actors share a resource (CPR without communication in Fig. 1b) is more prevalent than by chance in cases without sustainable conservation of ecological resources. Externalities and Displacement It is commonly argued that social and ecological processes need to be aligned in space and time; otherwise, such scale mismatches will make governance inherently difficult because management measures cannot, by def-
3
A. CPR w comm
B. CPR w/o comm
C. Scale fit
D. Scale misfit
Figure 1. Stylized configurations of minimal social-ecological system (building blocks) related to a set of governance challenges. Nodes (white circles) are actors and ecological resources (grey circles). In (a) the link between the 2 actors is realized (they are communicating [comm]); therefore, the potential for the actors to avoid overharvesting the common pool resource (CPR) is assumed to increase. In (b) the link between actors is not realized (they are not communicating), and CPR management is more difficult than in (a). In (c) the 2 links between the actor and the ecological resources are realized (both interlinked ecological resources are used by the actor); therefore, indirect ecological effect feeds back directly to the actor, creating a tight feedback loop (scale fit). In (d) the link between the actor and the resource on the right is not realized, and any indirect effect caused by using the ecological resource on the left, which affects the resource on the right, is not feedback directly to the actor (i.e., scale misfit). inition, embrace the full extent of the managed entity (i.e., the ecosystem). This in turn can lead to ineffective management (Cumming et al. 2006; Guerrero et al. 2013) and overall resource decline. Our concern is, therefore, to differentiate between situations where actors are linked (though use) to all or just to a subset of the constituent components of an ecosystem consisting of several interdependent ecological resources. As a social-ecological building block this would be expressed as follows. If 2 ecological resources are interdependent, and a specific actor is linked only to one of them, the other resource can, in the eyes of that actor, be perceived as external (Fig. 1d). Hence, the incentive to use the first resource in such a way as to minimize potential negative side effects on the second resource is reduced. In economics this would be referred to as externality (Prugh et al. 1999), and it creates a misalignment between how costs and benefits related to resource use are distributed and the extent of the underlying ecosystem that provides for the resources. The misalignment could have a negative effect on the actor using one of the resources, but this feedback could be long delayed (this is an example of a temporal misfit). However, if the actor in this example uses both resources simultaneously, the incentive to manage both resources as interdependent
Conservation Biology Volume 00, No. 0, 2014
Social-Ecological Fit and Conservation
4
increases, and the overall level of fit would be higher (Fig. 1c). This situation yields the following hypotheses: (2a) the SES building block in which an actor and 2 interdependent ecological resources are fully connected (scale fit, Fig. 1c), is more prevalent than by chance in cases with sustainable conservation of ecological resources; (2b) the SES building block in which an actor is only linked to one of 2 interdependent ecological resources (scale misfit, Fig. 1d), is more prevalent than by chance in cases without sustainable conservation of ecological resources. Deconstructing the Full Scale SES A real SES is naturally much more complex than any of the defined building blocks. To be able to infer relationships between conservation outcomes and the socialecological fit of a given system, we therefore needed to proceed in 2 steps. First, we disassembled the fullscale social-ecological network, encompassing all actors, ecological resources, and interdependencies of the studied SES into sets of building blocks. In other words, the full scale SES was deconstructed into sets of irreducible building blocks. Second, we zoomed in on the patterns of occurrences of each of the types of building blocks. In that way, we had a precise measure of how common the occurrence of each specific building block was in the SES. This measure should however be related to a baseline measure (i.e., the measure needs to be grounded). To that end, we compared the empirical measure with the distribution of measures acquired from a large number of randomly generated social-ecological networks. These random networks each had the same number of social and ecological nodes as the empirical network. The amount of social-to-social, social-toecological, and ecological-to-ecological link were also the same. The randomly generated social-ecological network thus provided a baseline estimate (i.e., the null model), and the degree to which the distributions of building blocks deviated from the baseline informed us as to whether these different building blocks were over- or underrepresented in the full scale social-ecological network. Conceptually, this could be seen as a sort of spectral analysis of the real social-ecological network (cf. Fourier transform). Our approach makes it possible, in general and beyond this study, to empirically test theoretically informed relationships between patterns of social-ecological interdependences and specific governance and conservation outcomes of an SES. Hence, this approach is not about searching for any type of pattern of social-ecological interdependences; it is about searching for specific patterns that could be linked, theoretically, to different governance and conservation outcomes. Also, the socialecological networks could be interpreted more explicitly from a perspective of multilevel network interdepen-
Conservation Biology Volume 00, No. 0, 2014
dency. The social and ecological networks, respectively, would from this perspective be seen as directly dependent on each other. For example, the structure of the social network could be seen to be the result of various social processes constrained both by the structure of the underlying ecological network itself and by the structure in which actors are linked to the ecological network. In other words, these structures would be seen as given, and it would only be the links between the actors in the social network that could vary. In contrast to a multilevel network interdependency perspective, we based our hypotheses on the assumption that actors can choose which ecological resources and which other actors to connect to. That does not mean that creating a link is effortless and will never involve negotiation or engaging in conflict with other actors or that there will not be other kinds of costs. Rather, it means there are no imposed constrains that would fundamentally prevent actors from succeeding in creating certain links (or to abandon certain links). Given the nature of the studied cases, we argue this assumption essentially holds true. In other cases, the actors’ degrees of freedom might be more limited, and in such cases the issue of multilevel network interdependency becomes more pronounced. Our disassembly approach is conceptually similar to recent work on multilevel exponential random graph models (MLERGM) used for analyzing social networks (Wang et al. 2013). Such models are fairly complex, and further description is beyond the scope of this work. However, MLERGM could provide, among other things, a more statistically robust way to separate different building blocks that to some extent overlap. It also explicates the perspective of network interdependencies. MLERGM currently supports quite a few different building blocks, and could be extended to support even more. Finally, our approach has some similarities to recent methods applied in studying motifs in food webs (e.g., Stouffer et al. 2012). Thus, there is high potential for methodological synergies across different research field. Positional Analysis The disassembling approach is not limited to investigating characteristics at the level of the entire socialecological network; it also allows for investigations of the extent to which individual actors and resources are engaged in certain building blocks, as well as what specific positions they tend to occupy (cf. Gould & Fernandez 1989; Stouffer et al. 2012). This provides a means for linking the patterns of interdependencies of the larger SES with the characteristics of individual actors and ecological resources. This can provide information on whether certain actors more commonly occupy positions in the SES building blocks characterized by good or bad fit, thereby suggesting that certain actors might have a
Bodin et al.
propensity to engage (or not) with others in certain ways that can disproportionally affect the overall level of fit. This propensity can be caused by individual characteristics, social norms, or other social structures. Comparison of 2 Contrasting Study Systems To test the applicability of our proposed approach, we analyzed 2 contrasting real world SESs. The study systems differed in several respects which are further described below, but most importantly they differed with regards to how well the actors in the 2 systems had managed to conserve some ecological resources on which they depended for their livelihoods. As such, the 2 cases allow for a comparative approach when evaluating the results from the social-ecological network deconstruction. Based on the theoretically informed link between each building block and conservation outcomes, we hypothesized that the study system where resources are more conservatively used would exhibit a higher frequency of building blocks hypothetically associated with a good fit (hypotheses 1a and 2a). In the case with less successful conservation building blocks associated with poor fit should be overrepresented (hypotheses 1b and 2b). The number of study systems is far too low to test the hypotheses in a statistically rigorous way; however, the comparative approach at least allowed us to make an early assessment of the potential of the approach. The first study system was a rural, coastal fishing village, approximately 50 km south of Mombasa, Kenya (Crona & Bodin 2006). The village had approximately 200 households and an estimated 1000 inhabitants. Approximately 44% of the households were directly involved in a small-scale multispecies fishery, and overharvesting and declining fish stocks characterized the fishery in the region. Despite a broad recognition of the decline, the actors involved had initiated no substantial efforts to remedy this apparently unfavorable situation. Describing this study system as a social-ecological network is a nontrivial task, and any meaningful interpretation of the revealed patterns of interdependencies is inevitably linked to how this has been done. Bodin and Teng¨ o (2012) suggest a sequential heuristic approach generally applicable to basically any SES consisting of multiple actors and resources. The approach aims to guide the conceptualization of the SES as a social-ecological network in such a way that the nodes and links are defined in a transparent way, that nodes and links are comparable in size and presumed effect, and that no possible interrelationships are defined a priori as impossible. Also, defining SES boundaries should be given careful consideration. The approach consists of a series of steps that should be reiterated with the aim to construct cross-case comparable social-ecological networks. How comparable any 2 social-ecological networks are also depends on how
5
generically the hypotheses related to the relationship between structure and function have been formulated. We formulated our hypotheses broadly; therefore, they should be broadly applicable to many different conceptualizations of SESs as social-ecological networks. The small-scale fishery study system was modeled as a social-ecological network with actors (social nodes) defined as groups of fishers using the same type of fishing gear and therefore targeting the same fish species. The way we defined the study system as a social-ecological network is described in detail in Supporting Information. Links between the fisher groups represented exchange of ecological information, and the ecological resources (ecological nodes) were the most important targeted fish species. Social-ecological links connected fisher groups to their targeted fish species, and the ecological links represented trophic interactions in the fish community. The data on trophic interactions in the fish community were based on theoretical estimates and not explicit observations, and for all fisher groups, except the ring net fishers, secondary data were used to assess the targeted fish species (Supporting Information). Because some of the data for this study system were based on assumptions, we performed a sensitivity analysis in which the number of trophic links varied and the number of targeted fish species included in the network varied systematically (Supporting Information). The second study system was a rural agricultural landscape in Androy, southern Madagascar (Bodin & Teng¨ o 2012). Forest patches, ranging in size from 3 to >90 ha, are scattered across the landscape of small fields and pastures. The forest patches are protected by taboos restricting access and use, they generate essential ecosystem services such as microclimate regulation and crop pollination, they are culturally important as ancestral burial grounds, and they serve as sites for ceremonies and as symbols of the link between people and the land (Teng¨ o & von Heland 2011). The landscape has been remarkably well preserved over a substantial amount of time despite strong pressures on land and forest resources (Teng¨ o& von Heland 2011; Bodin & Teng¨ o 2012). The social-ecological network in our study system was defined by Bodin and Teng¨ o (2012), and we used it without modification. The actors were defined as clans residing in different parts of the landscape who used or benefited from the ecosystem services (including cultural services) stemming from one or several of the forest patches (the resources). The social links were based on interclan relationships such as joint agreement on forest use and management and on kinship and intermarriages that had implications for access to forest services. The ecological links between the forest patches were based on species dispersal, which is essential for the ability of the forest patches to exist over time (Bodin & Teng¨ o 2012).
Conservation Biology Volume 00, No. 0, 2014
Social-Ecological Fit and Conservation
6
Spear gun
Ring net
Handline
Gill net Seine net
Figure 2. Estimated social-ecological network of the African small-scale fishery. Circular nodes at the top are gear-defined groups of fishers (Crona & Bodin 2006), dark grey diamonds in the middle are targeted species of predator fish, and light grey diamonds at the bottom are targeted species of prey fish. Dark grey lines with double-headed arrows are links that represent an exchange of ecological knowledge and information (i.e., communication) among groups of fishers, light grey arrows are extraction of particular fish species by certain groups of fishers, and black arrows are trophic interactions.
Results
Figure 3. Estimated social-ecological network of the Madagascar agricultural landscape (adapted from Bodin and Teng¨ o [2012]). Grey circles are nodes representing different clans residing in the landscape, polygons are nodes representing forest patches, black lines show which clans are benefiting from ecosystem services stemming from which forest patches, light grey lines show patch interconnectivity through seed dispersal, and dark grey lines are social ties between clans. The clans with a white cross reside outside the area shown.
The social-ecological networks representing the 2 study systems both exhibited nontrivial patterns of socialecological interdependencies (Figs. 2 and 3). To provide a baseline measure for the frequency analysis of the studied building blocks, we created 10,000 random social-ecological networks for each study system. The frequency analysis showed that for the CPR governance challenge, the building block corresponding to resource competition without actor communication was significantly underrepresented in the agricultural system (Fig. 4b). Resource competition paired with actor communication was, however, significantly overrepresented in the agricultural system (Fig. 4a) and was significantly underrepresented in the small-scale fishery system (Fig. 4e). For the governance challenge of scale mismatch a similar pattern was found. The building block corresponding to scale-match was significantly overrepresented in the agricultural system (Fig. 4c) and marginally significantly underrepresented in the small-scale fishery system (Fig. 4g, P = 0.053), whereas the frequency of the building block corresponding to scale mismatch was significantly underrepresented in the agricultural system (Fig. 4d).
The sensitivity analysis (Supporting Information) indicated that results for the fishery system were fairly robust in regards to trophic link uncertainties (Supporting Information). The significant patterns of underrepresentation of the building blocks associated with good fit (Fig. 4e & g) were, however, less pronounced when only the 70% most targeted fish species were accounted for. When a larger pool of species were accounted for (90%), the underrepresentation of the building blocks associated with good fit instead became more pronounced, and the building block corresponding to resource competition without actor communication became significantly overrepresented. Hence, the overall social-ecological fit decreased when more targeted species were included in the analysis. How often the different actors in each case occupied all the structurally unique positions in the studied building blocks varied greatly among actors (Table 1). The variability could, for example, range from 0% to 55% (agricultural system, scale fit, Table 1).
Conservation Biology Volume 00, No. 0, 2014
Bodin et al.
7
CPR w comm
CPR w/o comm −−
Scale fit
Scale misfit
++
(a) (14 )
(b) (7 )
(c) (20 )
(e) (6 )
(f) (18)
(g) (21 )
−−
++
(d) (34 )
.
(h) (137)
Agricultural landscape
Small-scale fishery
Figure 4. Frequency of social-ecological building blocks in the studied social-ecological networks: (a-d) agricultural landscape and (e-h) small-scale fishery. Black bars are drawn from the real social-ecological networks, whereas grey bars come from 10,000 randomly generated networks. If the number of building blocks in the real social-ecological network significantly deviates from the distribution of the random networks (e.g., in [a]), then one or 2 plus or minus signs (5% and 1% significance levels, respectively) are next to the number of building blocks above each graph. Marginal significance,
