Correspondence: Graeme S. Cumming, Percy FitzPatrick Institute, DST/NRF Centre of Excellence, University of Cape Town, Rondebosch, Cape Town 7701, South Africa. E-mail: email@example.com
Aims To highlight the potential value of network analysis for conservation biogeography and to focus attention on some of the challenges that lie ahead in applying it to conservation problems.
Methods We briefly review existing literature and then focus on five important challenges for the further development of network-based approaches in the field.
Results Our five challenges include (i) understanding cross-scale and cross-level linkages in ecological systems (top–down and bottom–up effects, such as trophic cascades, have been demonstrated in food webs but are poorly understood in nested hierarchies such as reserve networks and stream catchments), (ii) capturing dynamic aspects of ecological systems and networks (with a few exceptions we have little grasp of how important whole-network attributes change as the composition of nodes and links changes), (iii) integrating ecological aspects of network theory with metacommunity frameworks and multiple node functions and roles (can we link the spatial patterns of habitat patches in fragmented landscapes, the parallel networks of interacting species using those patches and community-level interactions as defined by metacommunity theory in a single framework?), (iv) integrating the analysis of social and ecological networks (particularly, can they be analysed as a single interacting system?) and (v) laying an empirical foundation for network analysis in conservation biogeography (this will require a larger data bank of well-studied networks from diverse habitats and systems).
Main conclusions Recent research has identified a variety of approaches that we expect to contribute to progress in each of our five challenge areas. We anticipate that some of the most exciting outcomes of attempts to meet these challenges will be frameworks that unite areas of research, such as food web analysis and metacommunity theory, that have developed independently.
Conservation biogeography, as distinct from bodies of theory like island biogeography or vicariance biogeography, is focused on providing the theory and tools that are needed to successfully undertake the sustainable, conservation-oriented management of geographical patterns and processes in ecosystems. Although its primary focus is ‘the distributional dynamics of taxa individually and collectively’ (Whittaker et al., 2005), the distributions of most species are strongly influenced by people. While the roots of conservation biogeography lie in ecology and geography, the spatial nature and magnitude of anthropogenic influences make it essential that conservation biogeography considers the spatial dynamics of social and economic systems as well as ecosystems.
Social dynamics in conservation contexts are commonly considered in terms of institutional structures, tenure rights and collaborative processes (e.g. Adger & Jordan, 2009; Ostrom, 2009). The differences between social and ecological approaches to conservation biogeography, and particularly the common tendency to analyse ecosystems as spatial entities and social systems as non-spatial entities, create problems of incommensurability when the key questions in a conservation situation revolve around spatial feedbacks between social and ecological systems. For example, in the case of a protected area network, a single event at one node (such as an out of control fire or a disease outbreak) can influence management (and hence, the biophysical environment) at other nodes in the reserve network. This influence travels via the social network, not the biophysical network, and physical distance may be irrelevant to its speed of travel.
In both social and ecological systems, system elements function as part of a system because they are connected with one another. Connections may be physical (e.g. infrastructure and corridors), through input–output dependencies (e.g. nutrient cycles and manufacturing chains), or via other mutualistic or antagonistic interactions (e.g. predation, competition, pollination or commerce). A critical level of connectivity is often vitally important for a variety of systemic functions, such as the coordination of broader-scale responses, information processing and decision-making (Norberg & Cumming, 2008). In a wide variety of conservation biogeographic contexts, including such classical problems as the management of migratory species (Berger, 2004), the control of invasive species (Roura-Pascual et al., 2009), conservation planning (Rouget et al., 2003) and the conservation of metapopulations of threatened species (e.g. Ball et al., 2003; Bergman & Kindvall, 2004), there is a pressing need to develop a more unified approach to analysing the spatial dynamics and connectivities of social–ecological systems.
Network analysis offers one way to unify social and ecological analyses (Webb & Bodin, 2008). Since the basic form of mathematical representation of a network is the same for both social and ecological systems, despite differences in the nature of the nodes and the connections, network analysis can function as a common currency for quantifying and analysing similarities and differences between relational patterns in social and ecological systems, and for understanding linkages and feedbacks within social–ecological systems (Janssen et al., 2006). In what follows, we first present some relevant background on network analysis and then focus on current challenges in its application to conservation biogeography.
The basic building blocks for modelling a system as a network are nodes and links (also referred to as vertices and edges, and in the social sciences as actors and ties). A node is a bounded entity defined by the analyst (e.g. a species, a habitat patch, a manager and an organization); links are the relations between nodes (e.g. dispersal and communication); attributes are characteristics of the nodes or links that do not directly originate from their relations to other nodes or links (e.g. size, age and transmission speed); and boundary conditions determine which entities to include or exclude from the network.
The focus in network analysis is on patterns of relations (i.e. topology) within the system and the question of how these patterns relate to the characteristics of the system (Webb & Bodin, 2008). Although network analysis has recently been given increasing prominence in a range of disciplines, it has a history of applications in both the natural and social sciences. In the 1950s, the Odum brothers modelled ecosystems as sets of components (e.g. species, farms or wetlands) connected by flows of energy (Odum & Odum, 1953). Food web ecologists have also used network representations of trophic interactions for many decades (e.g. Lindeman, 1942) and network-related questions, such as that of whether more diverse food webs are more stable, have received considerable prominence in ecology (e.g. May, 1972, 2006; McCann, 2000; Montoya et al., 2006).
Some of the most powerful applications of network analysis are those that result from its combination with other methods. For example, analyses of the prisoner’s dilemma and other games in a lattice-based framework (i.e. where each actor is located within a single cell in a regularly spaced grid) have suggested that sustainable cooperation is more likely when interactions occur only with immediate neighbours rather than with any other individual in the community (Nowak & May, 1992; Nowak et al., 1994; Nowak, 2006). The use of a network-based system description in this instance removes the constraints of relying on a four- or eight-neighbour rule (for example) and hence allows greater flexibility in exploring how alterations in network properties may affect cooperative behaviours.
Network analysis has a longer history in the social sciences than in ecology. Social network analysis is the study of how localized interaction between individuals, organizations or other social entities gives rise to larger-scale patterns – or structures – that both facilitate and constrain individual actors while giving rise to the properties of the network (Wasserman & Faust, 1994; Diani & McAdam, 2003; Borgatti et al., 2009). In general, social networks have been shown to exhibit ‘small-world’ properties (Milgram, 1967; Watts & Strogatz, 1998), in which high levels of clustering and homophily (i.e. most people interact with only a few others) are paired with short average network distances (through ‘weak ties’ between groups; Granovetter, 1973). Through investigating this uneven topology, social network analysis has contributed to fields as diverse as group problem solving, collective action, social movements, social support, exchange and power, economic market formation and the spread of diseases (Wasserman & Faust, 1994).
Geographers pioneered many of the more explicitly spatial applications of network analysis, including studies of the spatial diffusion of farmer’s practices (Hägerstrand, 1967) and the ways in which physical distances between firms and actors influence the structure and evolution of interorganizational knowledge flows (e.g. Glückler, 2007; Ter Wal & Boschma, 2009). More recently, sociologists have developed models of how spatial proximity influences the formation of social network structures (Wong et al., 2006). Social network analysis has now been introduced to the interdisciplinary arena as a viable route to explore problems in natural resource management (Bodin et al., 2006a; Bodin & Crona, 2009), including such issues as the generation of social norms and communities of knowledge (through ‘strong ties’), the transfer of innovations and ideas (through ‘weak ties’), the coordination of collective action and power relations. Empirical studies on conservation-related fields include such topics as how patterns of social relations influence local fisheries management (Crona & Bodin, 2006; Ramirez-Sanchez & Pinkerton, 2009), the abilities of agencies to create arenas that successfully bring diverse sets of stakeholders closer to each other (Schneider et al., 2003), and how collaboration networks among civil-society organizations can explain the protection and management of urban green areas (Ernstson et al., 2008).
In nearly all cases, as networks become larger, the relational pattern of interactions rapidly gets too complex for subjective interpretation. In such situations, formal analyses of relational patterns within the network can be used to extract meaningful measures of ecosystem or societal properties. In recent years, an increasing number of studies have explicitly modelled and analysed ecological and social systems as networks. For thorough reviews of network analysis and its applications, we refer the reader to (amongst others) Bascompte & Solé (1998); Barabasi (2002); Newman (2003); Watts (2004); Newman et al. (2006); Solé & Bascompte, 2006 and Pascual & Dunne (2006).
Challenges for the application of network analysis in conservation biogeography
Applications of network analysis in ecology and conservation biogeography are rapidly increasing, but some important areas remain neglected. We have identified five challenges (summarized in Table 1) that we regard as being particularly important for the development and application of network theory to conservation biogeography and related research fields over the next decade. We should stress that this list is not intended to be exhaustive and the order in which the challenges are presented is not intended to be a reflection of either their importance or their difficulty.
Table 1. Summary of the five challenges for network analysis in conservation biogeography that are highlighted in this article. Each of these challenges is discussed in more detail in the text.
Conservation biogeography examples
Challenge 1: Cross-scale and cross-level linkages in ecological systems
Most ecological systems are hierarchical. Top–down and bottom–up effects, such as trophic cascades and critical depensation, have been demonstrated between nodes in food webs but are poorly understood in nested hierarchies such as reserve networks and stream catchments.
How do changes in connectivity between scales or levels influence overall system dynamics? How can we distinguish cross-scale effects from same-scale effects? Do cross-scale effects generate thresholds and tipping points in biogeographic networks?
How does the creation of habitat corridors influence the overall resilience of a reserve network? How do anthropogenic influences on food webs (e.g. predator removals or introductions) influence spatial patterns of disease transmission?
Challenge 2: Dynamic aspects of ecological systems and networks
Ecological and social networks change through time, both growing and collapsing. Aside from studies on node removal, we have little grasp of how important whole-network attributes will change as the composition of nodes and links changes.
To what extent can networks adapt to node and link loss? As a network loses nodes, do strong links rebuild themselves, or are they simply lost? Under what conditions are networks capable of adaptation?
As a given landscape fragments, how will organisms of interest modify their movement patterns? In functional networks, such as plant–pollinator mutualisms, how do annual fluctuations in plant abundance influence overall pollination success?
Challenge 3: Integrating ecological aspects of network theory with metacommunity frameworks and multiple node functions and roles
Interactions among different species are at the forefront of many conservation problems. Is it possible to reach a synthesis that links the spatial patterns of habitat patches in fragmented landscapes, the parallel networks of interacting species using those patches, and community-level interactions as defined by metacommunity theory?
Can overlapping models of dispersal networks, competitive interactions and trophic networks (food webs) provide ways of quantifying metacommunity dynamics? How do covarying changes in parallel networks influence ecosystem function?
To what extent can we infer dispersal abilities from taxon-based surveys of habitat fragments? Can metacommunity theory, applied in a network context, be used to guide landscape planning for ecosystem service provision? Do domestic cats create thresholds in the biodiversity of bird communities in urban green spaces?
Challenge 4: Integrating the analysis of social and ecological networks
Social and ecological networks interact and overlap in space and time. Is it possible to analyse them as a single interacting system?
Can node attributes be used to link social and ecological aspects of the same system? Does information theory offer a way of uniting social and ecological nodes in the same network, or must this kind of analysis be carried out using overlapping networks?
Can we deliberately design adaptive management and monitoring networks that maximize learning opportunities while not stifling innovation? Should interagency workshops lead to more effective natural resource management? And under what conditions can they be expected to succeed or fail?
Challenge 5: Laying an empirical foundation
Testing and refining ideas about networks will require a large data bank of well-documented, well-studied networks from diverse habitats and systems. Studying network dynamics requires long time series. The collection of such data sets is an important priority in this field.
To what extent do empirical generalities emerge? Are generalizations about diversity and stability supported? How does network structure determine function? How does the distribution of strong links change as a real network collapses?
Which theoretical models can we rely on in real-world conservation biogeographic applications? For instance, when and why do habitat corridors work? How much genetic connectivity is sufficient to maintain population viability under real-world conditions of environmental variation?
Challenge 1: Cross-scale and cross-level linkages in ecological systems
Ecosystems contain both nested and non-nested hierarchies (Allen & Starr, 1982). Most biogeographic studies deal with nested hierarchies, in which areas of interest are contained within other, larger biomes or continents. Cross-scale and cross-level linkages in a network analysis refer to the ways in which parts of the system at different levels (and/or scales) constrain, explain or influence one another. For example, in a food web, predators can limit the abundance of their prey species (a cross-level linkage); and in a classical biogeographic context, the extent of a particular vegetation type may determine food or microsite availability for an invertebrate species (a cross-scale linkage).
Cross-scale linkages are often poorly understood, even in classical ecological analyses (e.g. Levin, 1992, 2000; Schneider, 2001; Stalmans et al., 2001). Holling (1992, 2001) and Peterson et al. (1998), among others, have argued that the overall resilience of an ecosystem is influenced by the ways in which pattern–process relationships at different scales interlock with one another. If higher or lower scales provide no homeostatic mechanisms to buffer fluctuations at scales above or below themselves, systems can in theory become unstable and difficult to manage. For example, in managing the interactions between fire and vegetation, the maintenance of a mosaic of different successional stages at a broad scale buffers the potential for the loss of early- and late-successional species at a fine scale; landscape homogenization, for example through synchronization of successional stages (as might be induced by clear-cutting), can result in species loss and the introduction of invasive alien plants that may be difficult to eradicate.
Ecologists have been interested in cross-scale effects for many years (e.g. Levin, 1992; Walters & Korman, 1999; Wardwell et al., 2008), but there are relatively few quantitative methods for cross-scale analysis in the ecological toolbox. As a result, it is difficult to decide how best to obtain quantitative evidence of cross-scale effects and how to test for cross-scale effects in cases where they are hypothesized to be important (Cumming & Barnes, 2007).
Network approaches in conservation tend to be applied to questions of connectivity at a single scale and for particular species and processes in fragmented landscapes. While the focus is often on migration or other forms of long-range dispersal (Keitt et al., 1997; Urban & Keitt, 2001; Bodin & Norberg, 2007), or on genetic networks that have evolved over long time periods (e.g. Rozenfeld et al., 2008), other studies have considered species daily needs at significantly smaller spatial scales (e.g. Andersson & Bodin, 2009). Network analysis itself is not restricted to analyses at a single scale; in other disciplines, it has been applied to assessing different hierarchical levels of organization. For example, a hierarchical clustering algorithm using the concept of betweenness centrality proposed by Girvan & Newman (2002) has been applied to decomposing a fragmented landscape into different compartments (Bodin & Norberg, 2007). Network analytical approaches thus seem like a promising way of quantifying cross-scale interactions between patterns and processes within a landscape.
The quantification of cross-scale interactions deals with the analysis of nested hierarchies. It is closely related to the problem of analysing interactions between different levels in non-nested hierarchies, such as food webs. In food web theory, different areas of ecological research – such as trophic cascades and theories of food web collapse – have yet to be unified (e.g. Bascompte & Melian, 2005). Trophic cascades typically involve the regulation of lower levels by higher levels in a hierarchical system. For example, the addition of starfish (predators) to a coral reef system may result in a local reduction in numbers of sea urchins (grazers) and the colonization of the reef by algae (primary producers), which can drastically alter the overall community composition of the reef, effectively propelling it into an alternate stable state (Hughes, 1994; Hughes et al., 2003). It should be possible, if challenging, to integrate the core aspects of each of these problems – spatial scale and connectivity in reserve networks, trophic connections at different scales in food webs, and reserve and trophic dynamics – using network analysis and the right kinds of long-term data sets.
Challenge 2: Dynamic aspects of ecological systems and networks
Ecological systems are dynamic and non-equilibrial; population sizes fluctuate continually, species enter or leave communities, and the number and nature of interspecific and intraspecific interactions changes over time. The abiotic environment and the broader landscape in which ecosystems are embedded also change; habitats may be created or lost, and seasonal and decadal variations can cause differences in community composition through time. As a result, the composition of any given ecosystem, at any given time, is the outcome of numerous dynamic processes operating at a variety of spatial and temporal scales (Brown, 1995).
Network analysis has mostly been concerned with analyses of static representations of ecological systems. Although an increasing number of published articles consider the potential for ecosystem collapses in different contexts through exploring the impacts of the selective removal of links and/or nodes (e.g. Bascompte & Sole, 1995; Urban & Keitt, 2001; Dobson et al., 2006; Dunne & Williams, 2009), these analyses focus on a specific kind of vulnerability (i.e. the potential for an ecosystem to withstand the loss of nodes, whether these are individual species in a food web or individual habitat patches in a fragmented landscape). Most such analyses start with a static representation of an ‘undisturbed’ system and then assess how much disturbance the system can absorb before the topology of the network reaches a critical point. Criticality typically occurs when large groups of nodes get separated from other large groups of nodes. In a human-modified food web, for example, this could mean that upper trophic levels become disconnected from lower trophic levels, leading to food web collapse (Dunne et al., 2004).
A loss of connectivity is not the only kind of network threshold criterion. There are other examples of criticalities, such as those that derive transmission rates and the level of congestion among the remaining nodes and links (Holme, 2002; Dodds et al., 2003). All of these examples, however, ignore at least two important and related dynamical aspects of ecological systems: the potential for adaptation and rewiring, and the question of how to differentiate between links that are redundant and those that are not. Changes in networks may alter the rules that produced the network in the first place (McMahon et al., 2001). In landscapes that have been significantly modified, for example, dispersal behaviour and predator avoidance may become increasingly more important; and as organisms respond to changes in landscape structure, new species or genes may enter networks (Proulx et al., 2006), and new interactions may arise between species that have not previously come into contact with one another (e.g. Crooks & Soulé, 1999; Buckley et al., 2006). Such changes can alter the basic framing of a conservation biogeographic problem that revolves around dispersal or trophic dynamics.
The question of redundancy relates to the characteristics of the links themselves rather than to the more general topology of the network. Nodes in ecological systems may differ in important ways, with some organisms or locations having a greater overall impact on network functionality than others. For example, nitrogen-fixing plant species may be essential for other plants, while other species are less important for the system as a whole. Network analysis currently offers limited help in predicting the details of network change from the topological properties of the network. This is particularly evident when subtle but important distinctions, such as differences in the strengths of interactions, are taken into account.
Some recent developments may help in resolving this problem. One example is that of inferring ‘node-level’ processes from the overall topology of the network (e.g. see Alon (2003) and the P* models reviewed in Wasserman & Faust (1994)). These methods make some basic assumptions about archetypal configurations of relations among a small set of nodes and then consider the entire network to infer how prevalent these different configurations are. If fine-scale changes in configuration are predictable, relationships between microlevel processes and the resulting macrolevel structure of the network can be established.
A second example is that of the application of agent-based simulation approaches at the level of individual nodes (c.f. Snijders, 2005). By defining some basic rules that determine how and when links are created and destroyed, and implementing them in a simulation model, emerging outcomes can be used to create expectations that can be tested against real-world time series data on network structures. These kinds of analysis offer some guidance in predicting how networks may change and evolve over time.
We see a high potential for linking dynamic and process-oriented network analytical approaches with studies of change in ecological systems. More research is needed to understand how individual species may act and respond to dynamic changes at different scales. A better understanding of these local-level processes and more sophisticated use of available network analytical approaches, paired with the development of new approaches specific for ecological contexts, could bring with it new and exciting insights into the dynamics of ecological systems.
Challenge 3: Integrating ecological aspects of network theory with metacommunity frameworks and multiple node functions and roles
The roles of spatial location and connectivity are relatively well explored in metapopulation theory, which provides a fundamental extension to the theory of Island Biogeography and remains one of the most relevant bodies of knowledge for the analysis of population dynamics in conservation biogeography. Most conservation-oriented applications of network analysis to landscape fragmentation have focused on populations of a single species (e.g. Urban & Keitt, 2001; Saura & Rubio, 2010). Although population-level analyses are important for conservation biogeography, interactions between different species are at the forefront of many conservation problems.
Conservation science is also becoming increasingly concerned with ecosystem services, which are typically provided by a suite of species rather than a single population. While some progress has been made in connecting ideas about taxonomic and functional diversity (e.g. Díaz & Cabido, 2001; Cumming & Child, 2009), community ecology lacks a well-established framework or synthesis that rigorously connects taxonomic diversity to ecosystem services. Such a framework will inevitably have to confront spatial aspects of community composition and habitat connectivity.
One of the rapidly growing areas of theoretical development in spatial ecology is that of metacommunity theory. Metacommunity theory focuses on the interactions between competition, dispersal and spatial structuring of the environment (Leibold et al., 2004) and outlines a set of different, mechanism-based models for community assembly. Metacommunity theory offers a set of mechanisms that could provide the basis for further kinds of network analysis. An exciting challenge, and perhaps even a natural extension of ecological network analysis, is to develop a singly synthesis that links the spatial patterns of habitat patches in fragmented landscapes or dendritic aquatic systems (Campbell Grant et al., 2007), the parallel networks of interacting species using those patches, and community-level interactions as defined by metacommunity theory.
Some progress on linking network models and metacommunity theory using neutral models has already been made (Economo & Keitt, 2008), but a more complete synthesis would provide new possibilities to analyse complex structural (i.e. topological) patterns of relations characterising real ecosystems where food web dynamics and dispersal processes play out simultaneously and interdependently. Several researchers have recently started to discuss and apply an integrative network-based approach that combines food web dynamics and dispersal processes (e.g. Melian & Bascompte, 2002; Brose et al., 2005; Holland & Hastings, 2008), but we expect to see many further advances in this area. A further challenge is to connect functional and taxonomic aspects of spatially explicit ecological networks to explain and predict how the provision of ecosystem services should change following fragmentation events and/or the loss of important elements of a food- or mutualistic web and/or reserve network (Melian & Bascompte, 2002).
The integration of network approaches with metacommunity analyses and ideas about ecosystem service provision will depend heavily on the development of ways of better incorporating multifunctionality, or multiple types of relationships, into network analysis. Organisms influence one another in numerous ways via such interactions as reproduction, predation, competition, parasitism and mutualism. Many of these interactions are spatially structured; animals exhibit distinct habitat preferences, and dispersal may be hindered or facilitated by spatial constrains in real landscapes. Hence, analysing a single kind of interaction at a time may leave out important system characteristics that derive from the more complex patterns of multiple relations in ecological systems. For example, in assessing the vulnerability of food webs to species removals, the current focus is on the ways in which network fragmentation can lead to secondary extinctions. The loss of a species that is not occupying a critical position in a trophic web might thus seem innocuous. However, if the same species is an ecosystem engineer that provides spatial structure for other species (e.g. beavers creating dams for fish; or cavity-nesting birds such as barbets and woodpeckers that create cavities that are used by other birds, mammals and reptiles), its removal could have devastating consequences.
Networks in which multiple relations occur within the same system are termed ‘multiplexed’. For example, the same group of people may interact professionally and socially in two different ways. These networks have the same nodes, but the relations between nodes in each network (e.g. professional or social) may follow different patterns. Integrative analyses of multiplexed networks are not commonly deployed in ecological network analysis; it is more common to either treat each kind of relation as a separate network, which is initially analysed in isolation, or to collapse all kinds of relations into one. These approaches are only justified if one can reasonably assume that the investigated relation or relations more or less completely encompass the system characteristics pertaining to a particular research question.
There are some studies, although not necessarily within the natural sciences, that address these challenges (see, e.g. Lazega & Pattison, 1999). Although the potential seems high for integrative relational network studies in investigating important aspects of complex ecosystems, many challenges remain in developing theories and methods that can cope with multirelational interdependencies.
Challenge 4: Integrating the analysis of social and ecological networks
Given that human activities are central to the practice of conservation biogeography, it is logical to ask whether entire systems could effectively be analysed as social–ecological networks (Janssen et al., 2006). Ecological entities (e.g. species and habitat patches) and social entities (e.g. users, managers, agencies and NGOs) relate to and interact with one another (e.g. fisherman harvest fish and managers implement fire policies). Most conservation-oriented network studies have used network analysis for one part of the system but have relied on other types of methods and frameworks to reach a more integrated analysis (Tylianakis et al., 2007). Analysis of these different entities and interactions as a single network would need to be carried out in a robust and transparent manner and framed carefully to produce meaningful results. There are at least three unfolding approaches to this challenge.
The first approach is to translate each system as a network – one social and one ecological – and integrate their respective analysis. For instance, if the social network among conservation managers is clustered into disconnected subgroups but the ecological network exhibits strong connectivity, this might suggest that if ecological disturbances travel rapidly, the social management system will be slow to respond.
The second approach is to translate either the social or the ecological system into a network and add the characteristics of the second system as attributes of the first network’s nodes. For instance, an ecological network with social attributes could be captured by measuring the management and protective capacity of interlinked habitat patches, and hence determining the probability that a given patch would disappear or degrade as a consequence of social features. This approach would let social structures and dynamics enter directly into the ecological network analysis. It has been applied by Bodin et al. (2006b), who quantified protective capacity as the strength of social taboos attached to certain forest patches in Madagascar; and Ernstson et al. (2008), who explored the protective capacity of urban green areas.
Conversely, social networks with ecological or environmental attributes could be developed, for example by measuring the rate of harvest return that interacting farmers gain from farming and quantifying how information that is exchanged through the social network influences subsequent decision-making (Bodin & Norberg, 2005). The approach of using node attributes as a way of performing integrated social–ecological network analysis is in its early stages and has considerable scope for the development of ways to ‘fold’ characteristics of one system into the nodal attributes of the other.
The third approach is to translate the social–ecological system directly into a social–ecological network. To do so, the distinction between ‘social’ and ‘ecological’ nodes must disappear, such that social-to-social, ecological-to-ecological, social-to-ecological and ecological-to-social interactions are treated as the same (although the analyst could continue to distinguish between ‘social’ and ‘ecological’ by tagging nodes with these attributes). To model a system as a single social–ecological network, a common circulating currency would be needed that influences social and ecological nodes in similar ways. Two potential currencies are energy/emergy (Odum & Odum, 1953) and information, but few empirical attempts at network models using a common currency to link social and ecological systems exist; even achieving a full quantification of a network using ‘just’ ecological data, such as nutrient exchange between species through herbivory, predation and parasitism is difficult (e.g. Halfon et al., 1996).
A further challenging aspect of the integration of social and ecological network analysis is to take scale and hierarchical arrangements within networks into account. Cumming et al. (2006) and Folke et al. (2007) have argued that many social–ecological management problems arise from a mismatch between the scales of ecological processes and the scales of social processes. However, the literature on bridging organizations and institutional entrepreneurs that addresses the resolution of scale mismatch situations (e.g. Hahn et al., 2006; Olsson et al., 2007) tends not to explicitly analyse the structure of social networks and the explicit scales of relevance in the ecological system (H. Ernstson, unpublished data). An important challenge for conservation biogeography is thus to capture how the network structure of management organizations (i.e. rather than merely their individual properties) influences the development of scale-crossing management strategies (and ultimately, the effectiveness of conservation action).
Challenge 5: Laying an empirical foundation
Conservation is an action-oriented discipline, and conservation biogeography must be based on solid empirical evidence if it is to have any lasting conservation impact. Many existing network models are built around simulation and a small number of empirical data sets, making it difficult to determine the true empirical generality of their results. While some kinds of interaction may be relatively inflexible, using appropriate empirical data to confront the predictions made by ecological and conservation-oriented models almost always reveals that some oversimplifications have been made in spatial models. In the real world, for example, organisms are capable of a wide range of behavioural responses to changes in their local environment (e.g. Hein et al., 2003; Hanski et al., 2004) and so-called patch dwellers frequently use matrix environments (Debinski & Holt, 2000; Prugh et al., 2008). The work of John Terborgh in Venezuela provides a fascinating example of how a rapid fragmentation event (the construction of a man-made dam that created a series of islands in a formerly continuous habitat) resulted in a series of differing and unexpected trophic outcomes (Terborgh et al., 2001). Similarly, Tylianakis et al. (2007) have shown how human modification can alter interaction networks in host–parasitoid systems, with implications for pollination.
Given concerns over the validity of theoretical conclusions, and despite the high levels of effort required, there is a clear need for additional data sets from different kinds of networks that can be used to develop and test theoretical principles. One of the greatest needs is for data that describe changes in real-world networks through time, for example during periods of trophic collapse in food webs (Dobson et al., 2006) or fragmentation events in forested landscapes (e.g. Lindenmayer et al., 2002, 2008). Deep time analyses of network dynamics are also important (Dunne et al., 2008). A parallel challenge is to find cases in which the ecological data can be associated with relevant social network data through time, to advance our understanding of coupled social–ecological systems.
As a number of other authors (e.g. Ohl et al., 2007) have pointed out, high potential exists in some existing data sets, such as those obtained by National Science Foundation-funded LTER (Long-Term Ecological Research) sites, for analysing and teasing out long-term dynamics and responses to human-driven change in food webs. It may also be possible to reconstruct some social network data, for example using data captured in public records of property rights and land-use patterns. One of the primary challenges in this area will be the development of good null models against which hypotheses of pattern–process interactions can be contrasted. For example, empirical data should clarify the question of how the distribution of strong links can be expected to decline as a network collapses; and whether a given empirical data set on food web collapse follows a random pattern or is structured by some ecologically relevant influence.
We have argued that network analysis offers a potentially valuable tool for conservation biogeography, and we have presented five important challenges for the further development of network-based approaches in the field. The challenges that we have identified include (i) understanding cross-scale and cross-level linkages in ecological systems, (ii) capturing dynamic aspects of ecological systems and networks, (iii) integrating ecological aspects of network theory with metacommunity frameworks and other multiplexing approaches, (iv) integrating the analysis of social and ecological networks and (v) laying an empirical foundation for network analysis in conservation biogeography.
One of the most fundamental challenges for conservation biogeography, and one that is touched on in different ways by other articles in this special issue, is that of how to achieve a level of analytical generality that allows us to understand and predict the ecological impacts of anthropogenic impacts on landscapes and respond in appropriate ways. Network approaches have the potential to provide a basic conceptual and empirical framework for understanding habitat fragmentation and related biogeographic questions. Consequently, they deserve further consideration within the field of conservation biogeography and ideally need to be better integrated into standard approaches and methodologies, such as conservation planning and the management of invasive species.
Network analyses also appear to offer a way to integrate knowledge between different disciplines. One of the most interesting questions in the study of networks is that of whether the structural and behavioural similarities between social and ecological networks are evidence of shared mechanisms, as suggested by McMahon et al. (2001); see also Saavedra et al. (2009). If social and ecological networks are structured by similar constraints and opportunities, the joint analysis of social and ecological networks in space may yield some fascinating and surprising results with high relevance for conservation biogeography. We look forward to further advances in this exciting and fast-moving field.
We are grateful for Jordi Bascompte and three anonymous reviewers for their useful comments on earlier versions of this manuscript.
Graeme Cumming holds the Pola Pasvolsky Chair in Conservation Biology at the University of Cape Town. He works on ecological and social–ecological systems, combining theoretical and applied approaches. Thomas Elmqvist leads the systems ecology group at the University of Stockholm; his particular interests are in ecosystem services and urban ecology. Henrik Ernstson and Örjan Bodin are interdisciplinary scientists with a shared interest in applying network analyses to solve problems of natural resource management.