What determines the structure of a food web?

This blog post is provided by Tomás I. Marina and Leonardo A. Saravia and tells the #StoryBehindThePaper for the paper “Ecological network assembly: how the regional metaweb influences local food webs”, which was recently published in the Journal of Animal Ecology.
Leonardo is a senior researcher and a professor at Universidad Nacional de General Sarmiento (UNGS), near Buenos Aires city, though in the coming weeks he will move to Ushuaia, at the end of the world, to start his new position as full researcher at Centro Austral de Investigaciones Científicas (CADIC-CONICET). Tomás earns a Ph.D. from UNGS; he is an early-career researcher in the Laboratory of Biological Oceanography at CADIC-CONICET.

In nature species interact among each other in many ways. The most common and studied interaction is that between a predator and its prey. A food web can be defined as a network of interactions that describes who eats whom in a certain ecosystem and at a certain time of the year. In any ecosystem, in a terrestrial, freshwater or marine environment, the so-called food web is comprised by lots of species (hundreds) and interactions (thousands), showing a complex network of predator-prey interactions.

In a recent paper in the Journal of Animal Ecology, we studied the assembly process of food webs, that means how the complex network of interactions is formed and which are the principal factors that drive such processes. To put it in simple words, our research question was the following: what determines the structure of a food web?

Florida Keys Mangrove Island. CC BY-SA 3.0, https://en.wikipedia.org/w/index.php?curid=11746362

Due to previous investigations of colleagues we now know that the structure of a food web is the result of community assembly, which is a repeated process of species arrival, colonisation and local extinction. This implies that there are two major components that determine the structure of a food web: 1) the composition of species from where individuals migrate to new habitats, referred to as the regional pool or metaweb; and 2) a selective process that determines which species can arrive and persist in the new habitat. We also know that the first to colonise new habitats are species with a broad diet (i.e. great variety of prey). Thus, predator-prey interactions are established during the assembly process that begins the formation of a network of interactions: the food web “in progress”. Added to this, the characteristics of the habitat (substrate, climatic conditions) and the dispersal capability of the species are important factors that also control the assembly process of the food web, determining which species might be able to persist in time.

Long Pond Lake at Adirondacks Region. By Original uploader was Mwanner at en.wikipediaLater version(s) were uploaded by XcepticZP at en.wikipedia. – Transferred from en.wikipedia, CC BY-SA 3.0, https://commons.wikimedia.org/w/index.php?curid=4044745

The structure of a food web is ultimately constrained by the species and potential interactions that exist in the regional pool, that is, the metaweb. Such metaweb is shaped by evolutionary and biogeographical processes that imply large spatial and temporal scales, containing many habitats and communities. Each of the local communities can have different food web structures, in terms of species and interactions among them. Some theorists conceive of assembly as a non-Darwinian selection process, whereby species and particular structures that destabilise the food web will be lost and stabilising structures will persist. Therefore, we should expect those stabilising structures to be over-represented in local food webs compared to the metaweb.

Tabular Icebergs at Northern Weddell Sea. By NASA Goddard Space Flight Center from Greenbelt, MD, USA – CC BY 2.0, https://commons.wikimedia.org/w/index.php?curid=68986936

To test the hypothesis that the selective processes are responsible for food web structure, we created an assembly model, where species migrate from a regional pool and persist in a theoretical local food web given at least one prey item available. The model considers colonisation-extinction and secondary extinctions (i.e. if a species arrives and none of its prey are present) events constrained by network structure that is represented by predator-prey interactions. Thus, the model does not include local selective processes and habitat constraints that are thought to control the assembly process. Then we compared the theoretical local food webs created from the assembly model with real food webs. We hypothesised that, if local processes like stability determine the food web structure, then we should observe a consistent difference in the network structural characteristics between real food webs and those assembled from the model. To test our hypothesis we compiled 58 real food webs from a variety of regions and ecosystems: 2 marine food webs from Antarctica, 50 freshwater ones from the Adirondacks region and 6 from an Arthropod community in six islands of the Florida Keys.

Schematic diagram of the assembly model: species migrate from the metaweb to a local food web carrying their potential interactions. Predators become extinct if none of its prey are present

Contrary to our expectations, we found that most structural characteristics didn’t differ between the real and modeled food webs. We particularly compared properties related to stability, such as modularity and the frequency of three-species motifs, and habitat filtering and dispersal limitation, such as topological roles that characterise how many interactions a species has within their module and/or between modules (i.e. modules are formed by a group of species that share more interactions compared to other species). This suggests that evolutionary and biogeographical processes that are established in the metaweb are more important and might drive the studied structural characteristics in a food web than local processes such as habitat filtering and dispersal limitations. In particular, our study forces us to re-think the way we approach the study of food web structure. The network properties that we as ecologists commonly use to shed light on this aspect might be a by-product of the assembly process already driven at the metaweb spatial and temporal scales.

Read the paper:

Read the full paper here: Saravia, L. A., Marina, T. I., Kristensen, N. P., De Troch, M., & Momo, F. R. (2021). Ecological network assembly: How the regional metaweb influences local food webs. Journal of Animal Ecology. https://doi.org/10.1111/1365-2656.13652

Can a generalist parasitoid act like Paine’s starfish?

This blog post is provided by Jinlin Chen and Chris Terry and tells the #StoryBehindThePaper for the paper ‘Natural enemies have inconsistent impacts on the coexistence of competing species’, which was recently published in the Journal of Animal Ecology.

A forgotten fruit in the kitchen will pretty quickly attract your attention by recruiting a swarm of fruit flies lingering around. Similar sights can be seen in the rainforest of Australian tropical mountains, where many Drosophila fruit fly species co-occur, apparently sharing the food resource. How might these species be able to coexist? To further thicken the ecological plot, there is rapid turnover in species composition across relatively small temperature gradients on these mountains1 and some species have been shown to be able to reproduce when alone outside of their observed range2. Understanding the complex patterns of co-occurrence on these mountains presents an intriguing test case for our understanding of ecological processes.

One classic answer to problems of community structure is the driving action of natural enemies – for example Paine’s seminal study3, where a starfish limits a dominant competitor (a mussel) allowing inferior competitors to co-exist. Might something similar be going on in our system?

Natural enemies of the flies certainly have a significant impact – a large proportion of  fly pupae (10%-40%) in our samples are killed by parasitoid wasps that develop inside the larvae and pupae. At some times of the year, up to 100% of flies could be parasitized in samples from other studies.

Trichopria parasitoid wasp laying eggs into a fruit-fly pupa (Drososphila sulfurigaster). The sole offspring of this parasitoid will develop inside the host, and eventually kill the host (Photo credit: Jinlin Chen)

Of course, it is not necessarily the case that natural enemies are an aid for coexistence. As generalist natural enemies may reduce the abundance of all competitors, it may be tempting to conclude that coexistence will be promoted through reduced competition for resources. However, there are a few dangerous gaps in such a chain of reasoning. For example, natural enemies may be more likely to drive the small population of the already inferior competitor to zero, thus reducing the chance of coexistence. Focusing on either the stabilizing or destabilizing aspects of the dynamics will both lead to unreliable conclusions. As a solution, modern coexistence theory allows the explicit framing of the question regarding pairwise species coexistence as a balance between niche differences promoting stabilisation and fitness differences fostering exclusion.

Paine’s experimental field approach of directly manipulating the numbers of starfish by flinging them from the rocks into the sea was not really translatable to our system. What we could do instead, taking advantage of the experimental amenability of Drosophila, was set up large laboratory trials to parameterise competition models, and this work is now published in Journal of Animal Ecology. We could then apply modern coexistence theory to ask with some precision how exactly parasitoids affected the ability for pairs of Drosophila species to coexist with each other.

The laboratory environment containing the individual vials containing food for the fruit fly maggots to compete within. The parasitoid wasps could fly within each box between the tubes to forage for their prey (Photo credit: Chris Terry)

The figure below shows an example of the results we found. Here without parasitism, it is likely that a more fecund competitor, D. sulfurigaster can exclude D. birchii (top left, in blue). However, when the competition trials are run in the presence of the parasitoid (middle, in red), the fitness difference between the pairs reduced as D. birchii was less susceptible to parasitism, and coexistence is more probable.

The effect of a generalist parasitoid on the pairwise co-existence of six tropical fruit fly species (Copyright: Jinlin Chen)

But one pair isn’t all that much help in determining what is going on at the community level. In all, we extended our analysis to all 15 pairwise combinations of six representative fly species from our Australian system. This was quite an undertaking, with two of us identifying and counting over 75000 flies from nearly 2000 tubes.  

Across the set, rather than the neat pattern we described above, we saw a real mixed bag in how the parasitoids affected likelihood of coexistence: increases, decreases and cases of no change at all. Essentially, the less-susceptible fly species were not necessarily the inferior competitors. In a blow to hopes that there would be a way to shortcut the extensive competition trials, competitive ability could not be discerned just from the intrinsic fecundity – it was necessary to simultaneously take both their fecundity and intra- and inter-competitive coefficients into account.

In short, we cannot expect generalist predators to be consistent drivers of coexistence. In accordance with other empirical studies, predominantly in plant systems, the regulation of co-existence by the other trophic levels appears highly variable4,5,6. Nevertheless, our and others’ studies have demonstrated that we have the theoretical framework and analytical tools to study the seemingly messy effects in various biological systems.  

The Australian fruit fly-parasitoid system may not be the most glamorous, but it does enable the study of  ’realistically complicated’ competitive and trophic networks in both laboratory conditions and the field. Our laboratory system includes all ten of the most prevalent fruit fly species along an altitude gradient and several of their native parasitoid wasps, each with a distinctive natural history. This system can offer a uniquely powerful test for the natural enemy-mediated coexistence among insects whose interactions are naturally assembled.

Looking to the future, we are hopeful that we can  upscale the pairwise relationships to multi-species interactions in a well-controlled manner, and extend to more trophic interactions by addition of parasitoid and hyper-parasitoid species. Although there is some distance to go, we are optimistic that we are on  track to empirically understand how diversity is regulated by simultaneous interactions with different trophic levels.

  1. Terry, J., Chen, J., & Lewis, O. (2021). Natural enemies have inconsistent impacts on the coexistence of competing species. Journal of Animal Ecology.
  2. O’Brien, E. K., Higgie, M., Reynolds, A., Hoffmann, A. A., & Bridle, J. R. (2017). Testing for local adaptation and evolutionary potential along altitudinal gradients in rainforest Drosophila: beyond laboratory estimates. Global change biology, 23(5), 1847-1860.
  3. Paine, Robert T. “Food web complexity and species diversity.” The American Naturalist 100.910 (1966): 65-75.
  4. Petry, W. K., Kandlikar, G. S., Kraft, N. J., Godoy, O., & Levine, J. M. (2018). A competition–defence trade‐off both promotes and weakens coexistence in an annual plant community. Journal of Ecology, 106(5), 1806-1818.
  5. Kandlikar, G. S., Yan, X., Levine, J. M., & Kraft, N. J. (2021). Soil Microbes Generate Stronger Fitness Differences than Stabilization among California Annual Plants. The American Naturalist, 197(1), E000-E000.
  6. Mordecai, E. A. (2013). Despite spillover, a shared pathogen promotes native plant persistence in a cheatgrass‐invaded grassland. Ecology, 94(12), 2744-2753.
Read the paper

Read the full paper here: Terry, J.C.D., Chen, J. and Lewis, O.T. (2021), Natural enemies have inconsistent impacts on the coexistence of competing species. Journal of Animal Ecology. Accepted Author Manuscript. https://doi.org/10.1111/1365-2656.13534

Ecosystem restoration – insights from theory

This blog post is provided by Klementyna Gawecka and tells the #StoryBehindThePaper for the paper “Habitat restoration in spatially explicit metacommunity models“, which was recently published in Journal of Animal Ecology.

Healthy ecosystems are vital for supporting the great diversity of life on Earth and providing us, humans, with benefits such as clean water, flood mitigation or recreation. But our activities, such as pollution, overexploitation or clearing land for agriculture are degrading ecosystems at an alarming rate. This degradation in turn accelerates the many crises, such as climate change and biodiversity loss, we now hear about every day in the news.

But there is hope. We still have a chance to restore the damaged or lost ecosystems and to replenish the life they harbour and the benefits they provide. With growing support from international organisations and governments, and improving socio-economic conditions in many parts of the world, the opportunities for restoration have never been greater. In fact, the United Nations has declared the current decade (2021-2030) as the Decade on Ecosystem Restoration

Habitat destruction is the major cause of biodiversity loss but restoration is a solution to this problem that has been gaining momentum in the recent decade (Credit: Wikimedia Commons).

The health and stability of ecosystems is maintained and regulated by the diversity of species present. This diversity in turn depends on the interactions between the species within the ecosystem. Therefore, to restore an ecosystem, we must also restore the groups of interacting species (communities), and the ecological processes that result from those interactions.

So where do we start? How do we design a successful restoration project? What strategies do we adopt so that communities recover? The aim of our recent study was to answer these questions using a theoretical approach. We built models which allowed us to disentangle and explain complex processes, and thus to provide rules of thumb for restoration projects.

In our simulations, we divided a landscape into small areas. We started by destroying this landscape a few areas at a time, and then reversed this process by restoring it, again, a few areas at a time. During this process, we allowed species to recolonise the newly restored areas. We explored two different restoration strategies: one where the areas were restored in clusters, and one where they were picked at random. In each case, we considered communities involving different interactions ranging from two-species competition, predation, and mutualism to more complex three- and four- species food webs. To compare the various scenarios, we came up with a new measure of the efficiency of restoration which corresponds to the rate of recovery of species abundance (see figure below).

Changes in species abundance as landscape is destroyed (dashed lines) and restored (solid lines). The figure on the left shows a restoration lag – a scenario where the species abundance during restoration is lower than during destruction. The figure on the right shows the opposite situation with abundance during restoration being higher than during destruction – a restoration boost. R represents the restoration efficiency.

First, we found that the more landscape we destroy, the less efficient and less successful restoration is. This means that, in highly degraded landscapes, we need to restore larger areas than we destroyed to recover the original species abundance. We refer to such scenario as a restoration lag. Moreover, there is a chance that the species will not recover at all. However, we also found that adopting the right restoration strategy can substantially improve this outcome. Restoring areas which are adjacent to undestroyed areas, where the species are still present, is more efficient than restoring areas at random. Essentially, in the former case, the landscape is restored in clusters, minimising fragmentation of the habitable landscape. This strategy can even eliminate the restoration lag, resulting in higher species abundances during restoration than in the original degraded landscape – a restoration boost.

Second, we show that species involved in different interaction types can respond differently to different restoration strategies. Let’s consider two species which compete with each other for the same resource. This could be, for example, tree species which compete for space and nutrients. In this case, restoring areas adjacent to those where one species is present, benefits the recovery of that species but is less favourable for the other species. This means that we may need to make a compromise between the recovery of the two species. As a second example, let’s consider a mutualistic interaction between two species such as a flowering plant and its pollinator. Here, restoring areas adjacent to those where one species is present, has a positive effect on the recovery of not only that species, but also its mutualistic partner. In other words, choosing the best restoration strategy is simpler than for the competitors, as both species benefit equally from restoring the landscape in clusters.

Ecosystem restoration is still a relatively new field. By considering simple communities, our study provides insights into effective design of restoration projects. However, the next step is to consider complex, real-world communities involving more species and types of interactions. Nonetheless, our study shows that the glass is still half full, and if we give it a chance through restoration, “life finds a way” as Dr Ian Malcolm from Jurassic Park used to say.

Disentangling temporal food web structure

This blog post is provided by Susanne Kortsch, Romain Frelat, Ivars Putnis, and Marie Nordström and tells the #StoryBehindThePaper for their article ‘Disentangling temporal food web dynamics facilitates understanding of ecosystem functioning’, which was recently published in the Journal of Animal Ecology.

Communities are organized into consumer networks, or food webs, describing who eats whom. Food webs provide the “energetic” backbones of ecosystems and are essential for energy and matter cycling. Moreover, a given food web structure can potentially buffer or accelerate human and environmental impacts such as climate change. Hence, it is important to understand how food web structure changes over time to better comprehend and anticipate ecosystem changes. In this study, we set out to uncover how food web structure changes over a 34-year period and how shifts in food web structure relate to changes in functioning.

Figure 1. The Gulf of Riga and our research vessel at sunset (photograph by Ivars Putnis) along with the Gulf of Riga food web metaweb.

One of the biggest challenges in the study of empirical temporal food webs is data availability. To study long-term variability in food web structure, one needs detailed information on species feeding relationship as well as the variation in species biomass through time. Marine biological monitoring programs that estimate species’ biomasses often focus on a single group of species (e.g. trawling for fish), but rarely sample the entire food web– ranging from primary producers (e.g. phytoplankton) to secondary consumers (e.g. zooplankton) to top predators (e.g. cod). The Gulf of Riga, a sub-basin of the Baltic Sea, is one of the few places with long-term data (>30 years of sampling) on multiple trophic groups. The Gulf of Riga presents a nice case study illustrating the importance of collecting long-term data for understanding the variability of natural systems.

Figure 2. The Gulf of Riga with our scientific research vessel to the left (photograph by Janis Gruduls). The upper right photograph shows a fourhorn sculpin – one of the few glacial relict fish species in the Baltic Sea (photograph by Kalvis Grinvalds), and the lower right photograph shows herring (photograph by Kalvis Grinvalds). The herring is the most dominant fish in the Gulf of Riga in terms of numbers (~90% of total fish abundance).

At the end of the 1980s, the Gulf of Riga underwent an ecosystem-wide structural and functional reorganization related to the disappearance of cod, a top predator in the food web, and an increasing importance of the pelagic compartment. We were curious to learn how species compositional changes alter food web structure and function. One might expect that a major compositional change will translate into a change at the food web level, but species appearing in or disappearing from a network can play similar structural and functional network roles. This means that overall food web function can be maintained despite compositional changes. We were also curious to find out how variable food web structure is over time. Although this may seem a trivial question, much of our current empirical understanding of how food webs vary over time comes from analyses on unweighted networks, which have suggested that food web macro-descriptors (e.g. connectance) are constant or invariant over time.

Figure 3. Example showing three measures of connectance: unweighted, node-weighted and link-weighted. This example demonstrates how a metric such as ‘connectance’ can display distinct and complementary temporal dynamics depending on food web approach.

To disentangle how food webs vary over time in the Gulf of Riga, we explored multiple approaches to describe food web structure and function. We compared the traditional topological approach based on species presence/absence using unweighted food webs and two weighted approaches – one node-weighted and one link-weighted – which include information on species biomasses and magnitude of feeding interactions. Node-weighted food web metrics weigh nodes by species biomass and describe the dominance of species in the food webs, whereas link-weighted metrics, based on energetic modeling, can capture changes in the magnitude of interactions. The advantage of link-weighted metrics is that they can reveal changes in energy flow and food web functioning.

Figure 4. Average link-weighted food webs for the five main periods with unique food web structure and function identified. The arrows and associated text summarize some of the main ecological changes throughout the study period.

Our results show that unweighted and weighted food web descriptors vary substantially, and distinctively, over the 34-year time series. We identified five periods with unique food web characteristics that represent distinct ecosystem structures and functions. The full extent of the temporal food web changes reported was only revealed through the complementarity between unweighted and weighted network approaches linking structure and function. Thus, our study demonstrates the benefit of using multiple methods to draw a more complete picture of temporal ecosystem dynamics. We therefore recommend using a range of descriptors from both unweighted topology-based and weighted (e.g. flux-based) food web approaches in order to characterize the dynamic and multifaceted nature of structural and functional changes in ecosystems.

If you are curious about the methods used in our study, have a look at our online tutorial: https://rfrelat.github.io/BalticFoodWeb.html