Search

Network approaches have revolutionized the study of ecological interactions. Social, movement and ecological networks have all been integral to studying infectious disease ecology. However, conventional (dyadic) network approaches are limited in their ability to capture higher-order interactions. We present simplicial sets as a tool that addresses this limitation. First, we explain what simplicial sets are. Second, we explain why their use would be beneficial in different subject areas. Third, we detail where these areas are: social, transmission, movement/spatial and ecological networks and when using them would help most in each context. To demonstrate their application, we develop a novel approach to identify how pathogens persist within a host population. Fourth, we provide an overview of how to use simplicial sets, highlighting specific metrics, generative models and software. Finally, we synthesize key research questions simplicial sets will help us answer and draw attention to methodological developments that will facilitate this.

Anthropogenic disturbances are widely recognized for their far-reaching consequences on the survival and reproduction of wildlife, but we understand comparatively little about their effects on the social lives of group-living animals. Here we examined these short-term changes in affiliative behavior as part of a long-term study on a human-tolerant and socially flexible population of California ground squirrels (Otospermophilus beecheyi). We used social network analysis to examine short-term changes in affiliative behavior and individual consistency in response to disturbances by humans, domestic dogs, or a natural predator (the coyote). Overall, juveniles were more involved than adults in affiliative interactions, but the short-term directional effects of these acute disturbances on social cohesion varied by disturbance type. Human and dog presence reduced aboveground connectivity, particularly for juveniles, whereas disturbances by coyotes generally promoted it. Beyond these effects, we also detected non-random responses to disturbances, though individuals were not very consistent in their directional response to different disturbance types. Our results demonstrate the flexible changes in social behavior triggered by short-term disturbances imposed by humans and other threats. More generally, our findings elucidate the underappreciated sensitivity of animal social interactions to short-term ecological disturbances, raising key questions about their consequences on the social lives of animals.

When selection is imposed by both social and ecological environments, the costs and benefits of social relationships can depend on life-history strategy. We argue that the formation and maintenance of differentiated social relationships will prevail in species and individuals with slow life histories. Social behaviours that benefit survival can promote slower life histories. Meanwhile, longer lifespan promotes the development of strong and stable social bonds by allowing fitness payoffs to be postponed. Differentiated social behaviours should be favoured for fast life histories only when they promote the rate of reproduction. Finally, associations between life-history strategies and other traits (e.g., personality) provide a mechanism to drive inter-individual variation in social relationships, making life-history important for sociality across taxonomic scales.

Agencies reporting on disease outbreaks face many choices about what to report and the scale of its dissemination. Reporting impacts an epidemic by influencing individual decisions directly, and the social network in which they are made. We simulated a dynamic multiplex network model-with coupled infection and communication layers-to examine behavioral impacts from the nature and scale of epidemiological information reporting. We explored how adherence to protective behaviors (social distancing) can be facilitated through epidemiological reporting, social construction of perceived risk, and local monitoring of direct connections, but eroded via social reassurance. We varied reported information (total active cases, daily new cases, hospitalizations, hospital capacity exceeded, or deaths) at one of two scales (population level or community level). Total active and new case reporting at the population level were the most effective approaches, relative to the other reporting approaches. Case reporting, which synergizes with test-trace-and-isolate and vaccination policies, should remain a priority throughout an epidemic.

Many behavioral, ecological, and evolutionary processes are closely intertwined with patterns of social interactions, such as the evolution of cooperation (Croft et al. 2006), information and disease transmission (VanderWaal et al. 2014; Aplin et al. 2015), predator–prey dynamics (Ioannou et al. 2012), and dispersal decisions (Blumstein et al. 2009). Even in species where individuals are traditionally viewed as leading a relatively solitary existence, interactions occur across diverse contexts, including territorial defense, resource competition, and courtship. Moreover, among members of a population, there is often substantial variation in terms of whom individuals interact with, how frequently they do so, and the intensity of these interactions. Quantifying these patterns and elucidating their functional and ultimate consequences is a central goal of behavioral ecology (Whitehead 2009). In recent years, these efforts have been facilitated by the widespread adoption of social network techniques imported from the physical and social sciences (Croft et al. 2008; Hasenjager and Dugatkin 2015; Krause et al. 2015). Network analysis provides a flexible framework for describing systems of interacting agents. In the context of animal populations, network nodes generally represent individuals, whereas connections between nodes (referred to as edges) quantify some form of social interaction, association, or relationship (e.g., agonistic, affiliative, proximity). Such networks are formally represented as an adjacency matrix or edge list, enabling the use of a rich set of mathematical tools for describing various aspects of a network’s structure (Whitehead 2009; Farine and Whitehead 2015). For instance, measures derived from networks can be used to characterize an individual’s influence over others (Flack et al. 2006; Rosenthal et al. 2015), the existence of subgroups within the population (Mersch et al. 2013), or how social relationships are structured according to phenotype (Aplin et al. 2013). In addition, these measures can facilitate investigation of the ecological and evolutionary consequences of social structure. For example, an individual’s position in a network can influence the speed it learns a new skill (Claidiere et al. 2013), while network structure can influence how quickly a disease spreads through a population (Otterstatter and Thomson 2007). Despite their flexibility, standard network approaches are not without limitations. Studies of animal social networks have traditionally represented social structure within a population using a network in which all edges represent the same type of relationship. For example, a network might quantify grooming interactions, spatiotemporal co-occurrences, or shared group membership. Yet animals can interact in different ways (e.g., grooming, play, aggression) and across different contexts (e.g., courtship, foraging). Considering only a single interaction type or combining multiple behaviors to produce a single aggregate measure may obscure important information about social structure (Finn et al. 2019). Furthermore, where multiple network types are considered (e.g., agonistic, affiliative), these are often analyzed independently of one another, tantamount to assuming that the patterning of each interaction type does not depend on the other(s). However, we know that this is unlikely to be the case in reality; agonistic interactions will change patterns of affiliative interactions not only among the interactants, but also their interactions with other group members and affiliative interactions among group members more widely. Social interactions can also be shaped by nonsocial forms of relationship, such as genetic relatedness or shared space-use, though incorporating such information using standard network approaches is not always straightforward (Pinter-Wollman et al. 2014). In addition, most network analyses use static network representations that provide “snapshots” of social structure at a particular point in time, whereas in reality, patterns of social interaction are dynamic, shifting in response to factors such as resource distributions, seasonal change, predation pressure, or demography (Blonder et al. 2012). Multilayer network analysis has recently been proposed as a framework that can help to address these shortcomings (Silk et al. 2018; Finn et al. 2019). In brief, a multilayer network incorporates multiple sets of relationships into the same mathematical structure, often with each layer representing a distinct form of connectedness (e.g., a layer of grooming interactions and a layer of aggressive interactions, or layers for associations in different seasons). Crucially, because a multilayer formulation includes these networks within a single structure, the interdependencies between different forms of connectedness can be explicitly modeled and investigated. For example, an individual’s social importance may only become apparent when multiple forms of interactions are simultaneously considered (De Domenico et al. 2015; Beisner et al. 2020). Furthermore, layers are not limited to simply capturing different types of social interaction, but can also represent nonsocial forms of relationship (e.g., genetic relatedness, patterns of shared space-use), include different types of entities (e.g., nodes may represent physical locations in one layer and individuals in another), or represent different time points. By enabling the construction of more nuanced representations of social structure, multilayer approaches hold great potential to advance the study of animal social behavior and its relationship to ecological and evolutionary processes (Silk et al. 2018; Finn et al. 2019; Montiglio et al. 2020; Mourier et al. 2020). For this Special Column, we have 2 primary aims. First, although a number of useful reviews have recently highlighted the potential of multilayer networks and related approaches for investigating animal behavior (Silk et al. 2018; Finn et al. 2019; Montiglio et al. 2020), there remain relatively few empirical studies that have employed these approaches thus far. The contributions to this Special Column help to fill this gap by applying multilayer network analysis to probe the causes and consequences of social structure across a diverse array of study systems. Second, as multilayer network analysis is still relatively new, there remains scant guidance on how best to employ these techniques. Multilayer networks inherit all the complexities of standard network analysis (see Farine and Whitehead 2015), while adding their own set of unique challenges (Finn et al. 2019). The contributions to this Special Column provide a wealth of practical guidance for researchers interested in employing these approaches, either serving as empirical case studies or explicitly addressing methodological questions. Here, we showcase how these contributions illustrate both the promise of multilayer networks and the challenges associated with their use.

Many organisms are experiencing changing daily light regimes due to latitudinal range shifts driven by climate change and increased artificial light at night (ALAN). Activity patterns are often driven by light cycles, which will have important consequences for species interactions. We tested whether longer photoperiods lead to higher parasitism rates by a day-active parasitoid on its host using a laboratory experiment in which we independently varied daylength and the presence of ALAN. We then tested whether reduced nighttime temperature tempers the effect of ALAN. We found that parasitism rate increased with daylength, with ALAN intensifying this effect only when the temperature was not reduced at night. The impact of ALAN was more pronounced under short daylength. Increased parasitoid activity was not compensated for by reduced life span, indicating that increased daylength leads to an increase in total parasitism effects on fitness. To test the significance of increased parasitism rate for population dynamics, we developed a host-parasitoid model. The results of the model predicted an increase in time-to-equilibrium with increased daylength and, crucially, a threshold daylength above which interactions are unstable, leading to local extinctions. Here we demonstrate that ALAN impact interacts with daylength and temperature by changing the interaction strength between a common day-active consumer species and its host in a predictable way. Our results further suggest that range expansion or ALAN-induced changes in light regimes experienced by insects and their natural enemies will result in unstable dynamics beyond key tipping points in daylength.

Social network analysis is a suite of approaches for exploring relational data. Two approaches commonly used to analyze animal social network data are permutation-based tests of significance and exponential random graph models. However, the performance of these approaches when analyzing different types of network data has not been simultaneously evaluated. Here we test both approaches to determine their performance when analyzing a range of biologically realistic simulated animal social networks. We examined the false positive and false negative error rate of an effect of a two-level explanatory variable (e.g., sex) on the number and combined strength of an individual’s network connections. We measured error rates for two types of simulated data collection methods in a range of network structures, and with/without a confounding effect and missing observations. Both methods performed consistently well in networks of dyadic interactions, and worse on networks constructed using observations of individuals in groups. Exponential random graph models had a marginally lower rate of false positives than permutations in most cases. Phenotypic assortativity had a large influence on the false positive rate, and a smaller effect on the false negative rate for both methods in all network types. Aspects of within- and between-group network structure influenced error rates, but not to the same extent. In "grouping event-based" networks, increased sampling effort marginally decreased rates of false negatives, but increased rates of false positives for both analysis methods. These results provide guidelines for biologists analyzing and interpreting their own network data using these methods.

Monitoring postrelease establishment and movement of animals is important in evaluating conservation translocations. We translocated 39 wild pine martens Martes martes (19 females, 20 males) from Scotland to Wales. We released them into forested areas with no conspecifics in 2015, followed by a second release in 2016, alongside the previously released animals. We used radio‐tracking to describe postrelease movement and habitat selection. Six martens (15%) were not re‐encountered during the tracking period, of which four undertook long‐distance dispersal. For the remaining individuals, we characterized two phases of movement, “exploration” followed by “settlement,” that differed between releases. In the first release, martens remained in exploration phase for a mean of 14.5 days (SE = 3.9 days) and settled at a mean distance of 8.7 km (SE = 1.8 km) from release sites, whereas martens released in year two, alongside resident conspecifics, traveled away from release sites at a faster rate, settling sooner, at a mean of 6.6 days (SE = 1.8 days), but further, at a mean distance of 14.0 km (SE = 1.7 km) from release sites. Animals released in year one did not exhibit habitat preferences overall but within forests they favored recently felled areas, whereas animals released in year two showed strong selection for forested habitat but did not discriminate between forest types. The presence of conspecifics appeared influential for settlement and site fidelity of translocated martens and was associated with more rapid but more distant dispersal of the later cohort. Releases of animals in close proximity appeared to promote site fidelity and rapid establishment of ranges in the recipient environment.

Many pathogens of public health and conservation concern persist in host communities. Identifying candidate maintenance and reservoir species is therefore a central component of disease management. The term maintenance species implies that if all species but the putative maintenance species were removed, then the pathogen would still persist. In the absence of field manipulations, this statement inherently requires a causal or mechanistic model to assess. However, we lack a systematic understanding of (i) how often conclusions are made about maintenance and reservoir species without reference to mechanistic models (ii) what types of biases may be associated with these conclusions and (iii) how explicitly invoking causal or mechanistic modelling can help ameliorate these biases. Filling these knowledge gaps is critical for robust inference about pathogen persistence and spillover in multihost-parasite systems, with clear implications for human and wildlife health. To address these gaps, we performed a literature review on the evidence previous studies have used to make claims regarding maintenance or reservoir species. We then developed multihost-parasite models to explore and demonstrate common biases that could arise when inferring maintenance potential from observational prevalence data. Finally, we developed new theory to show how model-driven inference of maintenance species can minimize and eliminate emergent biases. In our review, we found that 83% of studies used some form of observational prevalence data to draw conclusions on maintenance potential and only 6% of these studies combined observational data with mechanistic modelling. Using our model, we demonstrate how the community, spatial and temporal context of observational data can lead to substantial biases in inferences of maintenance potential. Importantly, our theory identifies that model-driven inference of maintenance species elucidates other streams of observational data that can be leveraged to correct these biases. Model-driven inference is an essential, yet underused, component of multidisciplinary studies that make inference about host reservoir and maintenance species. Better integration of wildlife disease surveillance and mechanistic models is necessary to improve the robustness and reproducibility of our conclusions regarding maintenance and reservoir species.

Social interactions between animals can provide many benefits, including the ability to gain useful environmental information through social learning. However, these social contacts can also facilitate the transmission of infectious diseases through a population. Animals engaging in social interactions must therefore face a trade-off between the potential informational benefits and the risk of acquiring disease. In order to understand how this trade-off can influence animal sociality, it is necessary to quantify the effects of different social structures on individuals’ likelihood of acquiring information versus infection Theoretical models have suggested that modular social networks, associated with the formation of groups or sub-groups, can slow spread of infection by trapping it within particular groups. However these social structures will not necessarily impact the spread of information in the same way if its transmission is considered as a “complex contagion”, e.g. through individuals copying the majority (conformist learning). Here we use simulation models to demonstrate that modular networks can promote the spread of information relative to the spread of infection, but only when the network is fragmented and group sizes are small. We show that the difference in transmission between information and disease is maximised for more well-connected social networks when the likelihood of transmission is intermediate. Our results have important implications for understanding the selective pressures operating on the social structure of animal societies, revealing that highly fragmented networks such as those formed in fission-fusion social groups and multilevel societies can be effective in modulating the infection-information trade-off for individuals within them.Significance statementRisk of infection is commonly regarded as one of the costs of animal social behaviours, while the potential for acquiring useful information is seen as a benefit. Balancing this risk of infection with the potential to gain useful information is one of the key trade-offs facing animals that engage in social interactions. In order to better understand this trade-off, it is necessary to quantify how different social structures can promote access to useful information while minimising risk of infection. We used simulations of disease and information spread to examine how group sizes and social network fragmentation influences both these transmission processes. Our models find that more subdivided networks slow the spread of disease far more than infection, but only group sizes are small. Our results demonstrate that showing that fragmented social structures can be more effective in balancing the infection-information trade-off for individuals within them.

Although strictly protected, populations of the hazel dormouse Muscardinus avellanarius in the UK declined by 72% from 1993 to 2014. Using National Dormouse Monitoring Programme data from 300 sites throughout England and Wales, we investigated variation in hazel dormouse population status (expressed as Indices of Abundance, Breeding, and population Trend) in relation to climate, landscape, habitat, and woodland management. Dormice were more abundant and produced more litters on sites with warmer, sunnier springs, summers, and autumns. Dormouse abundance was also higher on sites with consistently cold local climate in winter. Habitat connectivity, woodland species composition, and active site management were all correlated with greater dormouse abundance and breeding. Abundances were also higher on sites with successional habitats, whereas the abundance of early successional bramble Rubus fruticosus habitat, woodland area, and landscape connectivity were important for population stability. Diversity in the structure of woodlands in Europe has decreased over the last 100 years, and the habitats we found to be associated with more favourable dormouse status have also been in decline. The conservation status of the hazel dormouse, and that of woodland birds and butterflies, may benefit from reinstatement or increased frequency of management practices, such as coppicing and glade management, that maintain successional and diverse habitats within woodland.

Social network analysis is now used widely to study social behaviour in humans and non-human animals, and missing individuals can represent a problem for network studies. This problem is becoming especially frequent in studies using bio-logging to collect interaction data, which is an approach used particularly frequently in the construction of animal networks. This therefore represents an important audience for Smith et al. (2017) who investigate how sub-sampling from networks impacts the outcome of subsequent analysis. Here I take advantage of the progress made by this paper to outline key issues that still require addressing to understand the effect of missing individuals on social network analysis.

Summary Host social structure is fundamental to how infections spread and persist, and so the statistical modelling of static and dynamic social networks provides an invaluable tool to parameterise realistic epidemiological models. We present a practical guide to the application of network modelling frameworks for hypothesis testing related to social interactions and epidemiology, illustrating some approaches with worked examples using data from a population of wild European badgers Meles meles naturally infected with bovine tuberculosis. Different empirical network datasets generate particular statistical issues related to non-independence and sampling constraints. We therefore discuss the strengths and weaknesses of modelling approaches for different types of network data and for answering different questions relating to disease transmission. We argue that statistical modelling frameworks designed specifically for network analysis offer great potential in directly relating network structure to infection. They have the potential to be powerful tools in analysing empirical contact data used in epidemiological studies, but remain untested for use in networks of spatio-temporal associations. As a result, we argue that developments in the statistical analysis of empirical contact data are critical given the ready availability of dynamic network data from bio-logging studies. Furthermore, we encourage improved integration of statistical network approaches into epidemiological research to facilitate the generation of novel modelling frameworks and help extend our understanding of disease transmission in natural populations.

The emergence and spread of infections can contribute to the decline and extinction of populations, particularly in conjunction with anthropogenic environmental change. The importance of heterogeneity in processes of transmission, resistance and tolerance is increasingly well understood in theory, but empirical studies that consider both the demographic and behavioural implications of infection are scarce. Non-random mixing of host individuals can impact the demographic thresholds that determine the amplification or attenuation of disease prevalence. Risk assessment and management of disease in threatened wildlife populations must therefore consider not just host density, but also the social structure of host populations. Here we integrate the most recent developments in epidemiological research from a demographic and social network perspective, and synthesize the latest developments in social network modelling for wildlife disease, to explore their applications to disease management in populations in decline and at risk of extinction. We use simulated examples to support our key points and reveal how disease-management strategies can and should exploit both behavioural and demographic information to prevent or control the spread of disease. Our synthesis highlights the importance of considering the combined impacts of demographic and behavioural processes in epidemics to successful disease management in a conservation context. This article is part of the theme issue ‘Linking behaviour to dynamics of populations and communities: application of novel approaches in behavioural ecology to conservation’.

Social network analysis is pervasive in understanding animal social systems, and provides information about how individuals vary in their social strategies. Many long-term studies comprising uniquely marked individuals use social network analysis as an analytical tool. However, the assumption that it is possible to make inferences using network metrics calculated using a subset of the population has yet to be investigated in an animal social network. We use a simulation study of networks derived from social interactions in a typical fluid fission–fusion social system to determine the precision and accuracy of measures of individual social position based on incomplete knowledge. We show that individual social positions measured in partial social networks correlate strongly with positions in the full social network. This correlation typically becomes stronger as the size of the simulated population is increased and is largely not affected by network density. The choice of network metric has an important effect on the precision of partial networks only when they include a small subset of the population and therefore caution is advised using some of these measures in small partial networks. This work demonstrates that valid inferences about individual social position and strategy can be made using partial networks in a wide range of animal social networks, highlighting the value of applying these methods in large long-term study populations.

Contact networks, behavioral interactions, and shared use of space can all have important implications for the spread of disease in animals. Social networks enable the quantification of complex patterns of interactions; therefore, network analysis is becoming increasingly widespread in the study of infectious disease in animals, including wildlife. We present an introductory guide to using social-network-analytical approaches in wildlife disease ecology, epidemiology, and management. We focus on providing detailed practical guidance for the use of basic descriptive network measures by suggesting the research questions to which each technique is best suited and detailing the software available for each. We also discuss how using network approaches can be used beyond the study of social contacts and across a range of spatial and temporal scales. Finally, we integrate these approaches to examine how network analysis can be used to inform the implementation and monitoring of effective disease management strategies.

The emergence of novel diseases represents a major hurdle for the recovery of endangered populations, and in some cases may even present the threat of extinction. In recent years, epizootics of infectious diseases have emerged as a major threat to marine mammal populations, particularly group-living odontocetes. However, little research has explored the potential consequences of novel pathogens in endangered cetacean populations. Here, we present the first study predicting the spread of infectious disease over the social network of an entire free-ranging cetacean population, the southern resident killer whale community (SRKW). Utilizing 5 years of detailed data on close contacts between individuals, we build a fine-scale social network describing potential transmission pathways in this population. We then simulate the spread of cetacean morbillivirus (CeMV) over this network. Our analysis suggests that the SRKW population is highly vulnerable to CeMV. The majority of simulations resulted in unusual mortality events (UMEs), with mortality rates predicted to be at least twice the recorded maximum annual mortality. We find only limited evidence that this population's social structure inhibits disease spread. Vaccination is not likely to be an efficient strategy for reducing the likelihood of UMEs, with over 40 vaccinated individuals (>50% of the population) required to reduce the likelihood of UMEs below 5%. This analysis highlights the importance of modelling efforts in designing strategies to mitigate disease, and suggests that populations with strong social preferences and distinct social units may still be highly vulnerable to disease outbreaks.

Almost all animal social groups show some form of fission–fusion dynamics, whereby group membership is not spatio-temporally stable. These dynamics have major implications at both population and individual levels, exerting an important influence on patterns of social behaviour, information transfer and epidemiology. However, fission–fusion dynamics in birds have received relatively little attention. We review the existing evidence for fission–fusion sociality in birds alongside a more general explanation of the social and ecological processes that may drive fission–fusion dynamics. Through a combination of recent methodological developments and novel technologies with well-established areas of ornithological research, avian systems offer great potential to further our understanding of fission–fusion social systems and the consequences they have at an individual and population level. In particular, investigating the interaction between social structure and environmental covariates can promote a deeper understanding of the evolution of social behaviour and the adaptive value of group living, as well as having important consequences for applied research.

Honey bee (Apis mellifera) guards discriminate nestmates from non-nestmates at the hive entrance. The acceptance threshold of guards is known to change adaptively, for example becoming less permissive when the number of intruder bees from other colonies increases. These adaptive shifts can occur within minutes. What is unknown is the mechanism behind this rapid shift. It was hypothesized that alarm pheromones released by guards may cause the adoption of a less permissive acceptance threshold. Here, we tested this hypothesis on five discriminator hives by using a behavioral assay. We used three amounts each of iso-pentyl acetate (IPA) and 2-heptanone (2H), which are the major components of the pheromones from the sting and the mandibular glands, respectively. Biologically relevant levels of chemicals were delivered to the hive entrance platform via an air pump. We found no effect of either IPA or 2H: there was no change in guard acceptance of either nestmate (on average, 91% accepted) or non-nestmate (on average, 30% accepted) under any of the pheromone treatments compared to the pentane control (98% nestmates accepted and 32% non-nestmates accepted). Therefore, we reject the hypothesis that the presence of IPA or 2H causes a rapid shift of guard acceptance threshold.