Plant interaction networks reveal the limits of our understanding of diversity maintenance.

Species interactions are key drivers of biodiversity and ecosystem stability. Current theoretical frameworks for understanding the role of interactions make many assumptions which unfortunately, do not always hold in natural, diverse communities. This mismatch extends to annual plants, a common model system for studying coexistence, where interactions are typically averaged across environmental conditions and transitive competitive hierarchies are assumed to dominate. We quantify interaction networks for a community of annual wildflowers in Western Australia across a natural shade gradient at local scales. Whilst competition dominated, intraspecific and interspecific facilitation were widespread in all shade categories. Interaction strengths and directions varied substantially despite close spatial proximity and similar levels of local species richness, with most species interacting in different ways under different environmental conditions. Contrary to expectations, all networks were predominantly intransitive. These findings encourage us to rethink how we conceive of and categorize the mechanisms driving biodiversity in plant systems.

Global and regional ecological boundaries explain abrupt spatial discontinuities in avian frugivory interactions.

Species interactions can propagate disturbances across space via direct and indirect effects, potentially connecting species at a global scale. However, ecological and biogeographic boundaries may mitigate this spread by demarcating the limits of ecological networks. We tested whether large-scale ecological boundaries (ecoregions and biomes) and human disturbance gradients increase dissimilarity among plant-frugivore networks, while accounting for background spatial and elevational gradients and differences in network sampling. We assessed network dissimilarity patterns over a broad spatial scale, using 196 quantitative avian frugivory networks (encompassing 1496 plant and 1004 bird species) distributed across 67 ecoregions, 11 biomes, and 6 continents. We show that dissimilarities in species and interaction composition, but not network structure, are greater across ecoregion and biome boundaries and along different levels of human disturbance. Our findings indicate that biogeographic boundaries delineate the world's biodiversity of interactions and likely contribute to mitigating the propagation of disturbances at large spatial scales.

Exploiting node metadata to predict interactions in bipartite networks using graph embedding and neural networks.

Networks are increasingly used in various fields to represent systems with the aim of understanding the underlying rules governing observed interactions, and hence predict how the system is likely to behave in the future. Recent developments in network science highlight that accounting for node metadata improves both our understanding of how nodes interact with one another, and the accuracy of link prediction. However, to predict interactions in a network within existing statistical and machine learning frameworks, we need to learn objects that rapidly grow in dimension with the number of nodes. Thus, the task becomes computationally and conceptually challenging for networks. Here, we present a new predictive procedure combining a statistical, low-rank graph embedding method with machine learning techniques which reduces substantially the complexity of the learning task and allows us to efficiently predict interactions from node metadata in bipartite networks. To illustrate its application on real-world data, we apply it to a large dataset of tourist visits across a country. We found that our procedure accurately reconstructs existing interactions and predicts new interactions in the network. Overall, both from a network science and data science perspective, our work offers a flexible and generalizable procedure for link prediction.

Niche Packing and Local Coexistence in a Megadiverse Guild of Frugivorous Birds Are Mediated by Fruit Dependence and Shifts in Interaction Frequencies.

AbstractNiche packing is one of the prevailing mechanisms underlying the increase in the number of co-occurring species and the extraordinary diversity of tropical ecosystems. However, it is not yet understood whether niche packing is facilitated by higher specialization and reduced niche overlap or, rather, by diffuse competition and increased niche overlap. We combined highly resolved bird-plant interaction networks, bird phylogenies, and plant functional traits to compare dietary niche overlap and foraging frequencies among frugivorous birds at seven sites in the tropical Andes. We quantified niche overlap on the basis of the traits of the plants used by each bird and related it to the degree of niche packing at the different sites. Niche complementarity decreased with increasing niche packing, suggesting that increasingly dense niche packing is facilitated by increased niche overlap. Pairwise niche overlap was mediated by shifts in foraging frequencies away from shared resources, and it decreased with decreasing phylogenetic relatedness and increasing dependence on fruit as resource. Our findings suggest that foraging choices are a key axis of diversification in frugivorous birds and that differences in resource use frequencies are already sufficient to reduce potential competition between ecologically similar species and facilitate niche packing, especially if species differ in their dependence on particular resources.

Predictions of biodiversity are improved by integrating trait-based competition with abiotic filtering.

All organisms must simultaneously tolerate the environment and access limiting resources if they are to persist. Approaches to understanding abiotic filtering and competitive interactions have generally been developed independently. Consequently, integrating those factors to predict species abundances and community structure remains an unresolved challenge. We introduce a new synthetic framework that models both abiotic filtering and competition by using functional traits. First, our framework estimates species carrying capacities along abiotic gradients. Second, it estimates pairwise competitive interactions as a function of species trait differences. Applied to the study of a complex wetland community, our combined approach more than doubles the explained variance of species abundances compared to a model of abiotic tolerances alone. Trait-based integration of competitive interactions and abiotic filtering improves our ability to predict species abundances, bringing us closer to more accurate predictions of biodiversity structure in a changing world.

Non-additive biotic interactions improve predictions of tropical tree growth and impact community size structure.

Growth in individual size or biomass is a key demographic component in population models, with wide-ranging applications from quantifying species performance across abiotic or biotic conditions to assessing landscape-level dynamics under global change. In forest ecology, the responses of tree growth to biotic interactions are widely held to be crucial for understanding forest diversity, function, and structure. To date, most studies on plant-plant interactions only examine the additive competitive or facilitative interactions between species pairs; however, there is increasing evidence of non-additive, higher-order interactions (HOIs) impacting species demographic rates. When HOIs are present, the dynamics of a multispecies community cannot be fully understood or accurately predicted solely from pairwise outcomes because of how additional species "interfere" with the direct, pairwise interactions. Such HOIs should be particularly prevalent when species show non-linear functional responses to resource availability and resource-acquisition traits themselves are density dependent. With this in mind, we used data from a tropical secondary forest-a system that fulfills both of these conditions-to build an ontogenetic diameter growth model for individuals across 10 woody-plant species. We allowed both direct and indirect interactions within communities to influence the species-specific growth parameters in a generalized Lotka-Volterra model. Specifically, indirect interactions entered the model as higher-order quadratic terms, i.e., non-additive effects of conspecific and heterospecific neighbor size on the focal individual's growth. For the whole community and for four out of 10 focal species, the model that included HOIs had more statistical support than the model that included only direct interactions, despite the former containing a far greater number of parameters. HOIs had comparable effect sizes to direct interactions, and tended to further reduce the diameter growth rates of most species beyond what direct interactions had already reduced. In a simulation of successional stand dynamics, the inclusion of HOIs led to rank swaps in species' diameter hierarchies, even when community-level size distributions remained qualitatively similar. Our study highlights the implications, and discusses possible mechanisms, of non-additive density dependence in highly diverse and light-competitive tropical forests.

Linking functional traits and demography to model species-rich communities.

It has long been anticipated that relating functional traits to species demography would be a cornerstone for achieving large-scale predictability of ecological systems. If such a relationship existed, species demography could be modeled only by measuring functional traits, transforming our ability to predict states and dynamics of species-rich communities with process-based community models. Here, we introduce a new method that links empirical functional traits with the demographic parameters of a process-based model by calibrating a transfer function through inverse modeling. As a case study, we parameterize a modified Lotka-Volterra model of a high-diversity mountain grassland with static plant community and functional trait data only. The calibrated trait-demography relationships are amenable to ecological interpretation, and lead to species abundances that fit well to the observed community structure. We conclude that our new method offers a general solution to bridge the divide between trait data and process-based models in species-rich ecosystems.

The context dependency of pollinator interference: How environmental conditions and co-foraging species impact floral visitation.

Animals often change their behaviour in the presence of other species and the environmental context they experience, and these changes can substantially modify the course their populations follow. In the case of animals involved in mutualistic interactions, it is still unclear how to incorporate the effects of these behavioural changes into population dynamics. We propose a framework for using pollinator functional responses to examine the roles of pollinator-pollinator interactions and abiotic conditions in altering the times between floral visits of a focal pollinator. We then apply this framework to a unique foraging experiment with different models that allow resource availability and sublethal exposure to a neonicotinoid pesticide to modify how pollinators forage alone and with co-foragers. We found that all co-foragers interfere with the focal pollinator under at least one set of abiotic conditions; for most species, interference was strongest at higher levels of resource availability and with pesticide exposure. Overall our results highlight that density-dependent responses are often context-dependent themselves.

Identifying "Useful" Fitness Models: Balancing the Benefits of Added Complexity with Realistic Data Requirements in Models of Individual Plant Fitness.

AbstractDirect species interactions are commonly included in individual fitness models used for coexistence and local diversity modeling. Though widely considered important for such models, direct interactions alone are often insufficient for accurately predicting fitness, coexistence, or diversity outcomes. Incorporating higher-order interactions (HOIs) can lead to more accurate individual fitness models but also adds many model terms, which can quickly result in model overfitting. We explore approaches for balancing the trade-off between tractability and model accuracy that occurs when HOIs are added to individual fitness models. To do this, we compare models parameterized with data from annual plant communities in Australia and Spain, varying in the extent of information included about the focal and neighbor species. The best-performing models for both data sets were those that grouped neighbors based on origin status and life form, a grouping approach that reduced the number of model parameters substantially while retaining important ecological information about direct interactions and HOIs. Results suggest that the specific identity of focal or neighbor species is not necessary for building well-performing fitness models that include HOIs. In fact, grouping neighbors by even basic functional information seems sufficient to maximize model accuracy, an important outcome for the practical use of HOI-inclusive fitness models.

Hidden layers of density dependence in consumer feeding rates.

Functional responses relate a consumer's feeding rates to variation in its abiotic and biotic environment, providing insight into consumer behaviour and fitness, and underpinning population and food-web dynamics. Despite their broad relevance and long-standing history, we show here that the types of density dependence found in classic resource- and consumer-dependent functional-response models equate to strong and often untenable assumptions about the independence of processes underlying feeding rates. We first demonstrate mathematically how to quantify non-independence between feeding and consumer interference and between feeding on multiple resources. We then analyse two large collections of functional-response data sets to show that non-independence is pervasive and borne out in previously hidden forms of density dependence. Our results provide a new lens through which to view variation in consumer feeding rates and disentangle the biological underpinnings of species interactions in multi-species contexts.

Tricky partners: native plants show stronger interaction preferences than their exotic counterparts.

In ecological networks, neutral predictions suggest that species' interaction frequencies are proportional to their relative abundances. Deviations from neutral predictions thus correspond to interaction preferences (when positive) or avoidances (when negative), driven by nonneutral (e.g., niche-based) processes. Exotic species interact with many partners with which they have not coevolved, and it remains unclear whether this systematically influences the strength of neutral processes on interactions, and how these interaction-level differences scale up to entire networks. To fill this gap, we compared interactions between plants and frugivorous birds at nine forest sites in New Zealand varying in the relative abundance and composition of native and exotic species, with independently sampled data on bird and plant abundances from the same sites. We tested if the strength and direction of interaction preferences differed between native and exotic species. We further evaluated whether the performance of neutral predictions at the site level was predicted by the proportion of exotic interactions in each network from both bird and plant perspectives, and the species composition in each site. We found that interactions involving native plants deviated more strongly from neutral predictions than did interactions involving exotics. This "pickiness" of native plants could be detrimental in a context of global biotic homogenization where they could be increasingly exposed to novel interactions with neutrally interacting mutualists. However, the realization of only a subset of interactions in different sites compensated for the neutrality of interactions involving exotics, so that neutral predictions for whole networks did not change systematically with the proportion of exotic species or species composition. Therefore, the neutral and niche processes that underpin individual interactions may not scale up to entire networks. This shows that seemingly simplistic neutral assumptions entail complex processes and can provide valuable understanding of community assembly or invasion dynamics.

Systematic bias in studies of consumer functional responses.

Functional responses are a cornerstone to our understanding of consumer-resource interactions, so how to best describe them using models has been actively debated. Here we focus on the consumer dependence of functional responses to evidence systematic bias in the statistical comparison of functional-response models and the estimation of their parameters. Both forms of bias are universal to nonlinear models (irrespective of consumer dependence) and are rooted in a lack of sufficient replication. Using a large compilation of published datasets, we show that - due to the prevalence of low sample size studies - neither the overall frequency by which alternative models achieve top rank nor the frequency distribution of parameter point estimates should be treated as providing insight into the general form or central tendency of consumer interference. We call for renewed clarity in the varied purposes that motivate the study of functional responses, purposes that can compete with each other in dictating the design, analysis and interpretation of functional-response experiments.

Untangling the seasonal dynamics of plant-pollinator communities.

Ecological communities often show changes in populations and their interactions over time. To date, however, it has been challenging to effectively untangle the mechanisms shaping such dynamics. One approach that has yet to be fully explored is to treat the varying structure of empirical communities-i.e. their network of interactions-as time series. Here, we follow this approach by applying a network-comparison technique to study the seasonal dynamics of plant-pollinator networks. We find that the structure of these networks is extremely variable, where species constantly change how they interact with each other within seasons. Most importantly, we find the holistic dynamic of plants and pollinators to be remarkably coherent across years, allowing us to reveal general rules by which species first enter, then change their roles, and finally leave the networks. Overall, our results disentangle key aspects of species' interaction turnover, phenology, and seasonal assembly/disassembly processes in empirical plant-pollinator communities.

Similar composition of functional roles in Andean seed-dispersal networks, despite high species and interaction turnover.

The species composition of local communities varies in space, and its similarity generally decreases with increasing geographic distance between communities, a phenomenon known as distance decay of similarity. It is, however, not known how changes in local species composition affect ecological processes, that is, whether they lead to differences in the local composition of species' functional roles. We studied eight seed-dispersal networks along the South American Andes and compared them with regard to their species composition and their composition of functional roles. We tested (1) if changes in bird species composition lead to changes in the composition of bird functional roles, and (2) if the similarity in species composition and functional-role composition decreased with increasing geographic distance between the networks. We also used cluster analysis to (3) identify bird species with similar roles across all networks based on the similarity in the plants they consume, (i) considering only the species identity of the plants and (ii) considering the functional traits of the plants. Despite strong changes in species composition, the networks along the Andes showed similar composition of functional roles. (1) Changes in species composition generally did not lead to changes in the composition of functional roles. (2) Similarity in species composition, but not functional-role composition, decreased with increasing geographic distance between the networks. (3) The cluster analysis considering the functional traits of plants identified bird species with similar functional roles across all networks. The similarity in functional roles despite the high species turnover suggests that the ecological process of seed dispersal is organized similarly along the Andes, with similar functional roles fulfilled locally by different sets of species. The high species turnover, relative to functional turnover, also indicates that a large number of bird species are needed to maintain the seed-dispersal process along the Andes.

Related plants tend to share pollinators and herbivores, but strength of phylogenetic signal varies among plant families.

Related plants are often hypothesized to interact with similar sets of pollinators and herbivores, but this idea has only mixed empirical support. This may be because plant families vary in their tendency to share interaction partners. We quantify overlap of interaction partners for all pairs of plants in 59 pollination and 11 herbivory networks based on the numbers of shared and unshared interaction partners (thereby capturing both proportional and absolute overlap). We test for relationships between phylogenetic distance and partner overlap within each network; whether these relationships varied with the composition of the plant community; and whether well-represented plant families showed different relationships. Across all networks, more closely related plants tended to have greater overlap. The strength of this relationship within a network was unrelated to the composition of the network's plant component, but, when considered separately, different plant families showed different relationships between phylogenetic distance and overlap of interaction partners. The variety of relationships between phylogenetic distance and partner overlap in different plant families probably reflects a comparable variety of ecological and evolutionary processes. Considering factors affecting particular species-rich groups within a community could be the key to understanding the distribution of interactions at the network level.

The marine fish food web is globally connected.

The productivity of marine ecosystems and the services they provide to humans are largely dependent on complex interactions between prey and predators. These are embedded in a diverse network of trophic interactions, resulting in a cascade of events following perturbations such as species extinction. The sheer scale of oceans, however, precludes the characterization of marine feeding networks through de novo sampling. This effort ought instead to rely on a combination of extensive data and inference. Here we investigate how the distribution of trophic interactions at the global scale shapes the marine fish food web structure. We hypothesize that the heterogeneous distribution of species ranges in biogeographic regions should concentrate interactions in the warmest areas and within species groups. We find that the inferred global metaweb of marine fish-that is, all possible potential feeding links between co-occurring species-is highly connected geographically with a low degree of spatial modularity. Metrics of network structure correlate with sea surface temperature and tend to peak towards the tropics. In contrast to open-water communities, coastal food webs have greater interaction redundancy, which may confer robustness to species extinction. Our results suggest that marine ecosystems are connected yet display some resistance to perturbations because of high robustness at most locations.

Unmasking structural patterns in incidence matrices: an application to ecological data.

Null models have become a crucial tool for understanding structure within incidence matrices across multiple biological contexts. For example, they have been widely used for the study of ecological and biogeographic questions, testing hypotheses regarding patterns of community assembly, species co-occurrence and biodiversity. However, to our knowledge we remain without a general and flexible approach to study the mechanisms explaining such structures. Here, we provide a method for generating 'correlation-informed' null models, which combine the classic concept of null models and tools from community ecology, like joint statistical modelling. Generally, this model allows us to assess whether the information encoded within any given correlation matrix is predictive for explaining structural patterns observed within an incidence matrix. To demonstrate its utility, we apply our approach to two different case studies that represent examples of common scenarios encountered in community ecology. First, we use a phylogenetically informed null model to detect a strong evolutionary fingerprint within empirically observed food webs, reflecting key differences in the impact of shared evolutionary history when shaping the interactions of predators or prey. Second, we use multiple informed null models to identify which factors determine structural patterns of species assemblages, focusing in on the study of nestedness and the influence of site size, isolation, species range and species richness. In addition to offering a versatile way to study the mechanisms shaping the structure of any incidence matrix, including those describing ecological communities, our approach can also be adapted further to test even more sophisticated hypotheses.

All ecological models are wrong, but some are useful.

In Focus: Curtsdotter, A., Banks, H. T., Banks, J. E., Jonsson, M., Jonsson, T., Laubmeier, A. N., … Bommarco, R. (2019). Ecosystem function in predator-prey food webs-Confronting dynamic models with empirical data. Journal of Animal Ecology, 88, https://doi.org/10.1111/1365-2656.12892 Species' population dynamics are influenced by a variety of abiotic and biotic factors. Curtsdotter et al. (2019) used a food web model to investigate the role of predator-prey interactions in the population dynamics of the bird cherry-oat aphid Rhopalosiphum padi. Their analysis hinged on linking the observed population dynamics to a mathematical description of the multi-species system via inverse methods-an approach less utilized in ecology but that allows one to search a wide space of possible parameterizations and identify best-fit model parameters. By scrutinizing the fit of this model to observed aphid population dynamics in 10 separate barley fields, they identified fields in which predation was the key driving force; in others, they found that accurate predictions depended on the existence of an unpredictable and unidentified extrinsic driver of aphid mortality. By scrutinizing areas where the model gave poor or biologically counterintuitive fits, their study provides a path forward to better link ecological theory to ecosystem function.

The mechanistic basis for higher-order interactions and non-additivity in competitive communities.

Motivated by both analytical tractability and empirical practicality, community ecologists have long treated the species pair as the fundamental unit of study. This notwithstanding, the challenge of understanding more complex systems has repeatedly generated interest in the role of so-called higher-order interactions (HOIs) imposed by species beyond the focal pair. Here we argue that HOIs - defined as non-additive effects of density on per capita growth - are best interpreted as emergent properties of phenomenological models (e.g. Lotka-Volterra competition) rather than as distinct 'ecological processes' in their own right. Using simulations of consumer-resource models, we explore the mechanisms and system properties that give rise to HOIs in observational data. We demonstrate that HOIs emerge under all but the most restrictive of assumptions, and that incorporating non-additivity into phenomenological models improves the quantitative and qualitative accuracy of model predictions. Notably, we also observe that HOIs derive primarily from mechanisms and system properties that apply equally to single-species or pairwise systems as they do to more diverse communities. Consequently, there exists a strong mandate for further recognition of non-additive effects in both theoretical and empirical research.

Community dynamics and sensitivity to model structure: towards a probabilistic view of process-based model predictions.

Statistical inference and mechanistic, process-based modelling represent two philosophically different streams of research whose primary goal is to make predictions. Here, we merge elements from both approaches to keep the theoretical power of process-based models while also considering their predictive uncertainty using Bayesian statistics. In environmental and biological sciences, the predictive uncertainty of process-based models is usually reduced to parametric uncertainty. Here, we propose a practical approach to tackle the added issue of structural sensitivity, the sensitivity of predictions to the choice between quantitatively close and biologically plausible models. In contrast to earlier studies that presented alternative predictions based on alternative models, we propose a probabilistic view of these predictions that include the uncertainty in model construction and the parametric uncertainty of each model. As a proof of concept, we apply this approach to a predator-prey system described by the classical Rosenzweig-MacArthur model, and we observe that parametric sensitivity is regularly overcome by structural sensitivity. In addition to tackling theoretical questions about model sensitivity, the proposed approach can also be extended to make probabilistic predictions based on more complex models in an operational context. Both perspectives represent important steps towards providing better model predictions in biology, and beyond.