DDDonald DeAngelis

Donald DeAngelis

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Donald DeAngelis has a total of 203 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Faculty Opinions recommendation of Transient polymorphisms in parental care strategies drive divergence of sex roles.

    by Donald DeAngelis, Hyo Won Lee

    The authors use individual-based modeling to project how sex roles in parental care can evolve. The results challenge previous mathematical theory. The way parental care is divided among male and female parents varies among animal species. While in mammals most of the care is provided by the female, biparental care is most common in birds. In teleost fishes, male-based care is common, e.g., male smallmouth bass guard eggs at the nest. Long and Weissing review some of the previous theoretical approaches to explaining the evolution of patterns of parental care, for which there have been many theories. A recent mathematical approach (Kokko and Jennions 2008 {1}) combined a number of relevant factors into an analytical model based on a derived fitness function. As corrected later by Fromhage and Jennions (2016) {2}, the model showed that the division of care, starting from a given condition of division of care plane, converges along a straight line to an equilibrium of divided care in the female care – male care state plane. Initial conditions could be chosen arbitrarily, such that any type of division of care, female-only, male-only, egalitarian care, or anything in between, could evolve to equilibrium. The analytic model of Kokko and Jennions (2008) {1} involved simplifications needed to be solvable. In contrast, Long and Weissing propose an individual-based simulation model (IBM) with a framework similar to the analytic model, but that can employ more realistic assumptions. Their model follows males and females as individuals from birth to death, with individual traits regarding their amount of care for offspring. The total amount of care provided by the two parents determines the survival probability of their offspring, but time invested in care by the male or female reduces their number of additional mating opportunities, and thus reduces their fitness. Traits for the level of care are passed to surviving offspring, and traits that produce the most offspring over generations were most likely to dominate in the population. The model of Long and Weissing produces results far different from the analytic model. Instead of an infinite spectrum of strategies along an equilibrium curve in the female care – male care plane, any initial condition approaches only one of two equilibrium points in the plane, which are symmetrical with respect to the ratios of the care of one sex to the other. The final equilibrium depends on stochastic factors in the IBM as well as any initial bias towards more care by one. The IBM differs from the analytic model in that polymorphisms with respect to the trait for level of parental care can develop within the populations of each sex. Each sex develops a cluster of individuals with traits for low care and a cluster for high care. Selection acts to cause the bimodality to amplify in each sex, such that over evolutionary time there is a single cluster of traits for high care in one sex and that for low care in the other. In the simulations, the selection acting on the parental care of both sexes leads to an initially assumed high level of care by each sex sinking rapidly by 75% for each, next leading to an almost complete rebound in care by one of the sexes. The sex that rebounds may depend on some initial bias. But total care at equilibrium is still far below the total possible care if both parents were contributing at their maximum possible levels. The above results are specific to the case in which both species provide the same type of care, so that the care is substitutable between the sexes. The authors also consider the case where the types of care of the two sexes are complementary in part, that is, exhibit synergy. Oddly, the addition of a small degree of synergy among the parents leads to striking oscillations through evolutionary time in the dominance of parental care by one sex to dominance by the other sex. Higher levels of synergy lead to equilibria, but, surprisingly, lower than in the case without synergy. The advantages of using an IBM in exploring the evolution of parental care are evident in this study. Analytic models necessarily make simplifying assumptions, such as monomorphism of traits, whereas the dimorphism that evolved from differences in individuals in the authors’ IBM was a key evolutionary mechanism. The IBM also enabled the authors to extend it beyond the simulations noted above to include traits involved in mating, such as female preference and male ornamentation. As Long and Weissing note, although IBMs have their own disadvantages, such as the need to perform huge numbers of simulations to obtain general results, the authors clearly show that for at least some evolutionary questions IBMs may be essential.

  • Faculty Opinions recommendation of Incorporating neurological and behavioral mechanisms of sociality into predator-prey models.

    by Donald DeAngelis, Hyo Won Lee

    The authors note that the traits of aggression and sociality create dominance hierarchies, along with differences in foraging behaviors, within populations. They propose a way that ecological modeling can take into account such interdependence effects on functional responses in predator-prey interactions. Understanding consumer-resource dynamics and building a predictive model are key elements in managing ecosystems, and they represent important goals in community ecology. To describe consumer-resource interactions, ecologists have developed various types of functional responses. In a simple description, functional responses are mathematical expressions that show the foraging rate of consumers in relation to resource density. Since Holling's disk equation was first suggested, numerous studies have followed up by developing functional responses for various natural scenarios {1-3}. Two ways of measuring functional responses have been used. One infers these relationships from information on how consumer and resource population sizes change in time and space. The other approach uses studies of individual consumers that measure the rates at which individuals consume prey as a function of prey availability. The shape of these rates can be considered as emerging from the traits of the individuals. As Lichtenstein and Schmitz note, however, inferences from the population level and the level of individuals are not always consistent {2}. They suggest that the reason is that individual consumers do not act in isolation, and their interactions with other consumers may influence their functional responses. The authors present evidence from behavioral ecology and neuroscience for the effects of some consumers, particularly predators, in a population on other predators within the population. For example, individual fishing spiders that are stronger in attacking prey may also more readily attack conspecifics. That can have both short-term and long-term effects on the functional response of the attacked individuals, the long-term plastic effects resulting from defeated individuals becoming subordinate in status, with negative effects on their functional responses. Sociality can also lead to individuals with different foraging strategies within a population, such as producers and scroungers, with the functional response of the former influencing that of the latter {4}. The authors discuss the consequences of this intraspecific interdependence of functional responses at the individual consumer and population levels. This interdependence causes functional responses measured at the individual and population levels to differ. Increasing the density of predators increases the interactions of predators, such that the independence of the consumption rates by individuals decreases. Thus, the predation rate within the population of predators increases non-linearly rather than linearly with density. At the population level, the proportions of individuals having aggressive versus social tendencies will vary with environmental conditions so that the functional response at the population level is not stable but varies in space and time. As the authors note, various functional responses have been proposed, such as the ratio-dependent response ({5}) and related responses, to try to take into account interference among predators. However, these do not represent the plasticity of mechanisms involved in the interference, so at best, they can merely mimic some predator-prey interactions. The authors propose that a better approach would be to develop models that are analogs of two-site spatially implicit predator-prey models in which the predator or prey can forage at different sites that vary in conditions {6}. Environmental changes in conditions at one site can lead to a flux of individuals to the other site. In a similar way, populations can consist of individuals with different traits that are plastic, such that individuals can move between in response to environmental conditions. Models that are perhaps similar to the type proposed by the authors have been developed for modeling effects on food webs of inducible defenses in some species of algae, rotifers, etc., where individuals can move between defended and undefended phenotypic states, depending on perceived predation threat (e.g., {7,8}. There is a trend in the active development of multi-trophic food web models for various purposes, such as fishery management {9}. However, as Lichtenstein and Schmitz note, food web models often overlook the plasticity of foraging. Their thoughtful article highlights this crucial aspect that could potentially alter the outcomes of food web models and suggests a functional response that can describe these interdependent functional responses.

  • Faculty Opinions recommendation of The joint evolution of animal movement and competition strategies.

    by Donald DeAngelis, Hyo Won Lee

    The authors use agent-based modeling to explore how the movement and foraging strategies of foragers evolve and affect both prey density and the densities of foragers across a resource landscape. Competition is one of the main subject areas of ecology. Ecologists have mostly focused on competition between species - that is, interspecific competition. But intraspecific competition is also important in shaping ecological communities, including influencing how populations are distributed spatially. A key theoretical concept is that a consumer population distribution should be strongly correlated with the distribution of available resources. More precisely, because consumers compete for resources, both directly and through depletion of resources, it has been proposed that the ratios of consumers to resources should be approximately the same across the resource landscape. However, that idea, called the Ideal Free Distribution (IFD) {1}, has the weakness for describing real consumer distributions and IFD could only be approached if the consumers had abundant knowledge of resource distribution on the landscape. Gupte and colleagues stress that how consumers move and distribute themselves on the landscape is strongly related with how they interact with conspecific consumers, including both competitive interactions and using other consumers as cues for the presence of resources. It has long been realized that the spatial distribution of individuals and competition are related {2}. But tracking an animal’s movements and identifying its behavioral pattern in the field is difficult {3,4}, so modeling can be a useful surrogate. Gupte and colleagues argue that spatially explicit agent-based (also called individual-based) models are well-suited to capture complex outcomes resulting from spatial structure, behavioral strategies, and evolutionary dynamics. They use an agent-based model to investigate how a species' foraging strategies is shaped by joint evolution of spatial movement and behavioral interactions. Their model is fashioned after the foraging behavior of shorebirds, such as oystercatchers (Haematopus spp.), feeding on mussels, but has potentially broader applicability. In addition to exploitative competition for prey, those consumers engage in an extreme form of direct competition, kleptoparasitism, where species steal resources from other individuals {5}. Kleptoparasitism is included as the form of direct competition in the model. The spatial arena of the simulation model is a 512 x 512 grid of cells, in which resources (mussels) are distributed patchily, and where a single grid cell can hold multiple individuals. The authors tested three different scenarios. In the first scenario, there is only exploitative competition. In the second scenario, inherited kleptoparasitism also exists, while in the third scenario, individuals could forage normally or switch to kleptoparasitism, depending on local conditions that the forager observes. Individuals select patches for foraging based on a few possible cues: estimated prey density; the number of prey handling individuals; and the number of non-handling individuals. The proportion of individuals within a cell that are handling prey would be a clue to prey density in that cell, which is harder to estimate directly. Each individual had a weighting factor for each cue, and these traits were allowed to evolve over 1000 generations. In scenario 1, the strategy evolved towards foragers showing a preference to moving to cells with a high proportion of handlers, but that did not lead to a distribution close to an IFD, as foraging over the long term tended to graze down the initial prey numbers and wash out concentrations of prey. In scenario 2. kleptoparasitism evolved to be the dominant strategy (with low success rate, however, as the number of handlers was low), which enabled the prey density to recover to high values. But the distribution of foragers became negatively related to prey densities, as non-kleptoparasites tended to avoid high prey density sites, where kleoptoparasites clustered. In scenario 3, in which behavior was conditional, individuals chose kleptoparasitism when handlers were present in a cell, but kleptoparasitism as a whole declined from that of scenario 2. The correlation of foragers to prey population was positive, as in scenario 1, but again not strongly so. A key finding from the modeling is that intense foraging depletes a landscape of prey to the extent that areas of obvious prey concentrations are washed out, such that the resource landscape becomes ‘clueless’. In deciding where to move, the foragers rely on cues from other individuals, some of which may be obviously handling prey. Whether the trait of kleptoparasitism is inherited or conditional then has an effect on the proportion of foragers with the different strategies, as well as on both the resultant prey density and the degree to which foragers are correlated with prey densities across the landscape. This study demonstrates the power of application of a spatially explicit agent-based model to investigate evolution, in this case of spatial distribution and competition within a species. The authors point out that, although individual-based models have great potential in studying animal movement and behavior strategy, there are only a few studies about modeling mechanistic animal movements, and even fewer include evolutionary dynamics. This paper should inspire more modeling of this sort.

  • Faculty Opinions recommendation of The power of forecasts to advance ecological theory.

    by Donald DeAngelis, Hyo Won Lee

    The authors show how ecological forecasting can be used to improve ecological theory and to help quantify predictability of ecological systems. John Platt (1964) explained the rapid progress in the development of theory in some fields of science as due to the application of 'strong inference' {1}. In this method, alternative hypotheses are posed to explain a phenomenon, then crucial experiments are designed with the aim of excluding all but one hypothesis. Further sub-hypotheses and experiments then refine and elaborate the advance in theory. The lack of use of strong inference, which the long time scales of ecological dynamics renders difficult, is often offered as an explanation for ecology's slower progress in testing theory than in, say, the molecular biology mentioned by Platt, or even disciplines closely related to ecology such as evolutionary biology. Advances in the technology of data acquisition and progress in making existing data more easily available to a wide number of users, however, may have given ecology tools that can compensate to some degree for the factors that have caused ecology to lag behind other fields of science. Thanks to automation, the paucity of data that has plagued ecology, due to the expensive and time-consuming nature of its collection, is being replaced in many areas by an almost continuous flow of data (e.g., Keitt and Abelson 2021 {2}). The increasing availability of data makes it feasible to replace the crucial experiments of strong inference by another methodology, that of ecological forecasting. The basic idea of ecological forecasting is attributed by Dietze (2017 {3}) to Clark et al. (2001 {4}), but interest has intensified recently in view of the urgent need to project the environmental changes being caused by human impacts. Lewis and colleagues define an ecological forecast as "a specific, quantitative prediction about a future ecological state, preferably including an uncertainty estimate." In emphasizing near-term, iterative predictions and testing, this methodology differs from the traditional scenario projections that attempt to project long term (multiple decade) changes in the environment. Near-term forecasting allows rapid updating of models as new data become available. Dietze et al. (2017 {3}) note that such continual testing against new data is what has allowed the accuracy of weather forecasting to steadily increase over the past 60 years. The authors stress the opportunity to use ecological modeling in a similar fashion for direct input to decision support of environmental management. As those authors point out, as in weather forecasting, a large amount of effort in ecological forecasting is in data management. Lewis and colleagues reiterate some of the same arguments for the benefits of forecasting as Dietze et al. (2018 {5}), But their emphasis is less on forecasting for decision support and more on the improvement of general ecological theory through forecasting. To their definition of forecasting cited above they add another important definition, that of forecast transferability, which they define as "the relative performance of a model, preferably including uncertainties, when applied outside of the system in which the model was developed (e.g., new location, temporal or spatial scale, biological system)." The idea is to use existing models, whether empirical or process-based, that were developed for a particular ecological system, to make forecasts by which alternative hypotheses can be tested using the data from other ecological systems. First, this procedure tests the generality of the theory represented in those models. Second, it allows biases in the existing models to be identified such that improvements can be made in the models and in their theoretical basis. Besides improving theory, forecasting can be used to quantify predictability in ecology. This can be done by testing forecasting skill of models against new data. The authors hypothesize that forecasting skill will decrease with an increase in the time horizon of the forecast, but will increase with the degree of aggregation of the temporal or spatial scales or the ecological variables such as species populations to which the forecast is applied. The authors both encourage and provide suggestions to help ecologists to add forecasting to their research toolbox. If used as the authors outline, ecological forecasting can be a methodology that perhaps can fill the role of crucial experiments in the method of strong inference and give prediction a more central goal in ecology.

  • Faculty Opinions recommendation of Mechanistic models of trophic interactions: opportunities for species richness and challenges for modern coexistence theory.

    by Donald DeAngelis, Hyo Won Lee

    The authors use a zooplankton-phytoplankton-nutrient model to show that traits regarding trophic interactions have a stronger effect than resource-related traits on species coexistence. The authors also show that multitrophic systems present challenges to modern coexistence theory. The coexistence of large numbers of plankton species, despite the seemingly relatively few distinct resource-related niches available in pelagic zones, was termed the ‘paradox of plankton’ by Hutchinson {1}. Spaak and colleagues note, however, that plankton species may differ in many traits beyond those involving competition for resources. When the possible differences in all traits are taken into account, they provide mechanisms that can contribute to the coexistence of species. The first goal of the authors was to use mechanistic modeling to compare the relative importance of mechanisms of resource- and mechanisms of predator-mediated coexistence. A second goal was to study how trade-offs among the various traits, for example, the trade-off of phytoplankton putting energy into defense against predation versus into growth, affects coexistence. To investigate these mechanisms, the authors developed a detailed zooplankton-phytoplankton-nutrient chemostat-type model. The model contains seven phytoplankton-specific traits, four zooplankton-specific traits, and two joint traits (handling time and prey selectivity for zooplankton feeding on phytoplankton). These traits were based on empirical data from which distribution functions were constructed for each trait. Allometric scaling was used to estimate correlations among pairs of traits. By sampling traits from the distributions, the authors generated 1,000 species pools, each consisting of 20 phytoplankton and 20 zooplankton species. From these pools, they assembled virtual communities through a process of sequentially adding species from these pools (phytoplankton then zooplankton) through time. They analyzed the species richness that emerged after 20 years. The authors were able to assess the importance of traits to species richness by varying the traits’ means and standard deviations. Varying the means demonstrated how strongly a given trait was related to species diversity. Increasing the standard deviation increased the differences of a given trait among species in the pools, increasing differences in fitness among the species, but possibly also creating new niches. In order to address the second goal of studying the effects of trade-offs among traits on coexistence, the authors varied the strength of trait-trait correlations. In addition, the authors were able to distinguish between niche and fitness differences among zooplankton species. To do that, they used a novel approach from Spaak and De Laender {2}. Niche difference is defined in terms of growth rate of zooplankton invading a community of competing zooplankton species at equilibrium, while fitness is defined in terms of its growth rate in invading a community consisting of members of one’s own species but at the total equilibrium density that all of its competing species. The results of the analysis are somewhat surprising. Species growth traits and phytoplankton resource use traits had little effect on zooplankton or phytoplankton species richness, but traits related to trophic interactions between zooplankton and phytoplankton had strong effects, with zooplankton species richness being more sensitive to changes in the means of traits. Regarding the second goal of studying effects on trade-offs, the strongest effects of alterations in correlations among traits were also related to phytoplankton-zooplankton interactions, affecting such trade-offs as the alternative gleaner-opportunist traits of zooplankton and the growth vs. defense traits of phytoplankton. Finally, the strength of differences in fitness among zooplankton had a greater effect on zooplankton richness than did the differences in niches. A third goal of the authors arose unexpectedly in this study. That was confronting the problem of how the so-called modern coexistence theory could be applied in cases of complex communities that include multitrophic interactions. Modern coexistence theory (e.g., {3}) usually predicts coexistence in terms of the ability of each member of a community to invade, starting from small numbers, the equilibrium community of all the other species. That condition works well for species in a single trophic level consisting of a community of competitors, which modern coexistence theory focuses on. However, mutual invasion does not easily work for most of the zooplankton-phytoplankton communities that Spaak and colleagues studied. A case in point is the situation in which two competing zooplankton are also involved in an indirect mutualism (see {4}) in which neither can invade unless the other species are present and have an indirect effect through the phytoplankton community. The authors describe current work towards overcoming that roadblock in extension of modern coexistence theory to multitrophic systems. But they note that much work needs to be done. This paper is an important contribution to understanding species coexistence by extending the community studied to multi-trophic food webs. The authors revealed the importance of trophic interactions in coexistence and showed the need to extend modern coexistence theory to deal with communities with such interactions.

  • Faculty Opinions recommendation of Towards a modular theory of trophic interactions.

    by Donald DeAngelis, Hyo Won Lee

    The authors propose a general framework for describing the steps in a trophic interaction, relating them to species traits and environmental conditions, and translating them into functional and numerical responses. The functional response can be defined succinctly as the response of individual predators, in terms of their rate of consumption, to the density of their prey. Along with the predator’s numerical response, or the increase in population through reproduction in response to consumption of prey, it is a key concept in the ecology of these interactions, and of food webs made up of such interactions. These interactions are called trophic interactions in general, as they can involve any type of consumption of one organism by another, including herbivores consuming plant biomass. The mathematical modeling of a functional response usually begins with describing the sequence of steps or actions involved in a trophic interaction. Studies have been done for a huge number of particular trophic interactions, and the steps for different cases are necessarily different, or at least have different degrees of importance, for different types of interactions. For example, the trophic interaction for a lynx feeding on a snowshoe hare must take into account factors such as pursuit and subdual, which are absent from that of a hare browsing on shrub twigs, whereas, in the latter trophic interaction, the digestion step may play a larger role than the digestion step in the lynx-hare trophic interaction. In other types of trophic interactions, those of parasitoids for example, the relative importance of different steps might be still different. The question arises whether there is some way of including the steps in any trophic interaction within a single framework that would allow any functional response to be described in terms of that framework. This is the goal of Wootton and coauthors, which they call a ‘modular approach’ to describing a trophic interaction. They argue that one advantage would be that a given trophic interaction could be easily compared with any other trophic interaction. They are not the first authors with the idea, and they cite relevant previous studies. But the effort of Wootton and colleagues is possibly the most comprehensive. The authors define eight basic steps in a trophic interaction; search, detect, decide, pursue, subdue, ingest, digest, and allocate. The probability of each occurring is conditional on the success of the step preceding it. Therefore, this chain could end at any step, and few chains that are started may go all the way to the end. The steps up to and at least partially including the ingestion step (when that step interferes with searching) determine the functional response. That is, both the maximum rate at which prey can be ingested and the handling time of a functional response can be calculated from the conditional probabilities and associated times for each step leading to ingestion. Digestion and allocation (the latter being the use of digested food as energy and nutrients that go into reproduction) are essential in determining the numerical response. Describing a trophic interaction would first involve determining which of these steps are parts of any given trophic interaction. Then the conditional probability of success and duration of each of the steps can be related mathematically to organism traits (e.g., body size, which can affect pursuit, subdual, and metabolic rates) of the consumer and its resource, along with environmental context (e.g., temperature). Some of the steps, such as ‘decide’, have probably not been considered in many of the functional responses that have been formulated. However, decisions, for example, on whether to pursue a dangerous prey individual or venture into an open area to browse a detected twig, would involve weighing risks and benefits and could be important in an overall functional response. The authors provide detailed guidance on developing a functional response, including estimating parameters. They work carefully through a detailed example of ground beetles feeding on aphids. The authors have provided and described in detail a framework that can help standardize the way in which trophic interactions are described and functional and numerical responses structured using the steps in a trophic interaction in terms of underlying species traits and environmental factors. This will not only be useful for ecologists studying specific trophic interactions, but can also contribute to a general theory of trophic interactions and food web theory in general.

  • Faculty Opinions recommendation of The maximum entropy principle to predict forager spatial distributions: an alternate perspective for movement ecology.

    by Donald DeAngelis, Bo Zhang

    The authors use maximum entropy to derive the expected spatial distributions of foragers across a patchy resource landscape, in an approach that can include factors such as local depletions of the resources. How foraging organisms distribute themselves across a landscape consisting of patches of different resource availabilities is a basic question studied by ecologists. Clearly, foragers would seldom be expected to be distributed purely randomly among patches of different quality. In a classic paper, Fretwell and Lucas (1970) {1} proposed that foragers should distribute themselves according to an ideal free distribution (IFD), in which the foragers distribute themselves as a function of patch resource densities, such that no forager could increase fitness by moving to another patch. As Capera-Aragones and colleagues recount, however, several simplifying assumptions are made in the IFD, including that the foragers have perfect information of the resource densities across the patches, that travel costs between patches can be ignored, that resource levels remain the same despite being exploited by the foragers, that there is no direct competition between foragers, and that fear of predators on patches is not a factor in patch choice. Therefore, the question is, how might foragers distribute themselves if these assumptions are relaxed? The authors propose an approach to this question, using the maximum entropy, or MaxEnt, which is widely used in species distribution modeling (SDM). The authors’ approach differs from the common correlative SDMs by adding mechanisms that make their approach more theoretical and generic for foraging situations. The authors’ mathematics is analogous that in statistical mechanical textbooks, in which the equilibrium of a population of particles over possible energy states is found by maximizing the system’s entropy under constraints. In the present case, available resources on patches on the landscape take the place of energy states, and the particles are foragers. What is defined as the total energy of the system differs, however, from that in physics. Here, what is defined as the total energy is higher the closer the foragers are to efficiently utilizing the distribution of prey across the patches. Applying the relevant constraints with the Lagrange multiplier approach, the authors derive an analogy of the well-known Boltzmann distribution for numbers of particles in states; here numbers of foragers in different patches, each of which is a different state in terms of prey availability. One of the Lagrange multipliers, β, is a measure of the level of information the foragers have of the distribution of prey density. β is negative, which means that higher energy states tend to be more populated than lower energy states, in contrast to Boltzmann distribution for thermodynamic systems. The authors show advantages of this approach for describing forager distributions by showing that the approach can take into account forager effects on prey, i.e. the foragers reduce the prey density in each patch by a common multiple of the numbers foragers in the patch. This produces a ‘Depleted Boltzmann distribution’. The authors then compare both Boltzmann distributions for a series of values of β, with both the IFD and the distribution of foragers that maximizes total energy (max E), across the system. In this latter measure, instead of each individual forager being distributed such that none gain fitness from moving to another patch, the sum of numbers of individuals per patch times the prey in that patch is maximized. The authors show that both Boltzmann distributions, when information is very low (β small), approach random distributions across the patches. As β approaches large values, however, the Boltzmann distribution approaches the IFD, while the Depleted Boltzmann distribution approaches the max E distribution. Further, the authors add the factors left out of the IFD distribution, including direct forager competition, predation and travel costs, to the information uncertainty and prey depletion. They apply the resulting distribution function to a hypothetical spatially explicit distribution of patches with varying prey availability. In a truly remarkable figure, they show how gradients in the strengths of each of these factors affects the predicted distribution of foragers on the landscape. The authors admit limitations of their approach, such as lumping effects of several factors into a scalar energy function for the patches. But the approach has enormous potential for describing forager distributions in space, and beyond to other issues in ecology.

  • Faculty Opinions recommendation of Evolutionary emergence of alternative stable states in shallow lakes.

    by Donald DeAngelis, Bo Zhang

    The authors use adaptive dynamics theory to study the conditions under which aquatic macrophytes with alternative stable states can evolve in shallow lakes. In a classic paper, Scheffer et al. (2003) studied the competition between free-floating aquatic macrophytes and submersed macrophytes in shallow lakes {1}. In the authors’ model, the floating plants could diminish light to the submersed plants, while the submersed plants, which obtained some nutrients from its roots in the sediments, could tolerate lower water column nutrient concentration than the floating plants. Using non-linear differential equations for the two species and a limitation on total limiting nutrients in the system, the authors showed that alternative stable states, or bistability, could occur, such that there were critical transitions, or tipping points, between the two stable states. Thus, if nutrient concentration was gradually increased from a low level, a stable state of only submersed plants transitioned gradually to a stable mixture of coexisting submersed and floating plants, but then reached a tipping point, beyond which a sharp transition to floating plants alone occurred. If the nutrients were then gradually decreased, another tipping point, beyond which there was a sudden shift back to pure submersed vegetation, occurred, but at a lower nutrient concentration. Using competition experiments between floating duckweed (Lemna minor) and submersed Elodea nuttallii in small, 8-liter containers, the authors confirmed the existence of alternative stable states. Differences in initial amounts of the two plants determined which macrophyte excluded the other. While the model and experiments of Scheffer et al. showed the feasibility of alternative stable states under controlled conditions {1}, Ardichvili and coauthors ask the question of whether such alternative stable states could actually evolve in a given shallow lake. To explore this, the authors imagine such a lake in which the nutrients diffuse from the sediments, decreasing toward the surface, while light attenuates from the surface to the bottom due do turbidity. Using the same model as Scheffer et al. {1}, the authors simulate the evolution of the plant’s trait, z, for mean depth preference along the gradients in the water column. They show that, given the nutrient-light trade-off, a macrophyte with certain capabilities for light and nutrient capture will evolve towards a mean depth preference where its fitness is such that no mutant, similar except for a higher or lower depth preference, can invade. Only one plant species would exist in this case. In the language of adaptive dynamics (e.g. Geritz et al. 1998 {2}), the resident is at a depth that is an evolutionary singularity that is convergent stable – that is, if the resident were at a greater or lower depth than the singularity, a mutant could invade. In the above scenario, however, the only mechanism at work is a trade-off between nutrient concentration and light intensity. In real situations, the fact that a plant higher in the water column can shade one below it comes into play. If that is considered in the model, the macrophyte could, indeed, still evolve to the same singularity, but now the singularity becomes a branching point, such that there is positive selection for both mutants above and below the singularity, leading to coexisting macrophyte species, one closer to the surface and one deeper. In fact, up to four phenotypes could coexist for a limited set of parameters. This result only shows the possible coexistence of two to four macrophyte species, where changes in nutrients lead only to gradual changes, not sudden jumps, between the dominant species. The authors show that one further mechanism was necessary to obtain the alternative stable states described in Scheffer et al. {1}. That mechanism is the better nutrient exploitation ability by the mutant strain closer to the bottom. With the three mechanisms of nutrient-light trade-off, asymmetric competition (shading by the shallower species) and superior nutrient exploitation ability by the deeper species, the authors show that the tipping points predicted by the Scheffer et al. model can occur {1}. But they also show that the range of parameters over which the branching occurs is limited and that their model does not produce highly distinct ‘floating’ and ‘submersed’ phenotypes, but rather two macrophyte types at intermediate depths separated only by about two meters. Interestingly, possible evolution of periodic fluctuations between single species and pairs of species was also shown by the model. This work complements other studies of possible coexistence of species along opposing resource gradients of aquatic systems (e.g. Huisman and Weissing 1995 {3}). It contributes the important result that competitive asymmetries are necessary for alternative stable states to evolve. Because opposing resource gradients are ubiquitous in nature, it is important to know the circumstances under which they might have evolved, especially in view of how climate change may impact ecosystems. This paper provides such a framework.

  • Faculty Opinions recommendation of Fragmented tipping in a spatially heterogeneous world.

    by Donald DeAngelis, Bo Zhang

    Using several models of Earth systems and ecosystems, the authors show that the bifurcation diagrams showing tipping points between two alternative stable states also allow coexistence states when spatial transport and heterogeneity are taken into account. It is well known that alternative stable states occur in many types of ecological systems, from coral reefs, to savannas, deserts, cloud forests, and oceanic food webs {1}. For example, Scheffer et al. (2003) show how a simple model with alternative stable states could explain how some shallow lakes are clear with a large component of submersed aquatic vegetation (SAV), while others are dominated by floating vegetation {2}. The essentials of this bistability can be depicted on a simple state plane diagram with the non-linear zero isoclines of the two competing types of autotrophs, the exact shapes of which change as the loading of the limiting nutrient is changed. The intersection of the zero-isoclines defines an unstable saddle point that separates the two stable equilibria. If the lake is initially in the clear state, slow increases in the loading of the limiting nutrient lead at first to only small changes in the variables, but then to a large abrupt change, called a critical transition, or more familiarly, a tipping point, from an SAV-dominated to a floating vegetation-dominated lake. The SAV-dominated equilibrium disappears, and the only stable state is that dominated by floating vegetation. If the nutrient load is then gradually decreased, a critical transition can occur back to dominance by SAV, but at a lower nutrient load than the first transition. This phenomenon, called hysteresis, is represented by a bifurcation diagram with an S-shaped curve as a function of nutrient load, with the one stable state along the upper and one along the lower branch of the curve. This is now very familiar in the ecological literature. The above theory applies best to relatively small, homogeneous systems. When a system is spatially large, such that spatial transport (e.g., diffusion) can occur or that there is underlying spatial heterogeneity, one might expect that the added complexity will result in differences from the changes described in the above simple description. This paper of Bastiaansen and colleagues is one of several papers that have appeared in recent years on theoretical analysis of the differences caused by spatial extent (e.g., {3-5}). Rietkerk et al. (2021) takes as a key example the tipping point in savanna ecosystems in which, in simple models, a gradual decrease from high to low precipitation leads to continuous vegetation cover changing abruptly to bare soil {5}. However, the authors describe how diffusion in a spatially extended version of the model can allow a Turing instability to occur before the critical transition can, creating a pattern of striped or labyrinthine alternating vegetation and bare ground; that is, coexistence of the two states, separated spatially. Although the occurrence of the formation of such spatial patterning was previously thought to simply be a prelude to an imminent critical transition to homogeneous bare soil, Rietkerk et al. (2021) cite increasing evidence that the vegetation patterns are fairly resilient to further reduction in precipitation, such that further decline in vegetation is gradual rather than precipitous {5}. In the present paper, Bastiaansen and colleagues use examples of five different conceptual models to show how spatial pattern formation modifies the bifurcation diagram from the simple S-shaped curve to a more complex curve that has one or more new branches representing stable coexistence of the two states. The conceptual models are for global energy balance, oceanic circulation, atmospheric circulation, the tropical forest-grassland ecotone, and shallow lakes. For example, for the global energy balance, in the absence of spatial effects, there can only be two stable alternative states; either ‘ice-free Earth’ or ‘snowball Earth’. When spatial effects are considered, the bifurcation diagram adds a branch representing ice-free and ice-covered zone coexistence in different parts of the planet. In the tropical forest-grassland ecotone, in which the spatial heterogeneity is a precipitation gradient, a branch is added to the bifurcation diagram representing stable forest and stable grassland along different parts of the ecotone. This is also the case for the shallow lake bifurcation diagram. From the shapes of the bifurcation diagrams, one can infer the dynamics. For example, the bifurcation diagram for shallow lakes shows that, for the type of spatial heterogeneity assumed, an increase of nutrient load from low levels can still cause the direct drastic change from clear to turbid conditions known from the simpler model, but that a decrease of nutrient load from high level results in the lake passing through coexistence states rather than a direct change from turbid to clear conditions. This paper shows that simple conceptual models of many Earth system and ecosystem models, which have tipping points between two extreme alternative stable states, can also have states of coexistence in separate areas of space when spatial transport and heterogeneity are considered. This has rich implications for better understanding of changes that may occur at the global and ecosystem levels from climate change.

  • Faculty Opinions recommendation of Does deterministic coexistence theory matter in a finite world?

    by Donald DeAngelis, Bo Zhang

    The authors extend deterministic models of coexistence of annual plant species to project the effects of demographic stochasticity on how long pairs of competing species coexist. Explaining the long-term coexistence of species that occupy similar niches has been a subject of debate among ecologists for many decades. As Schreiber and coauthors note, the arguments concerning the requirements for coexistence long tended to be vague conceptualizations and have only recently been stated in rigorous fashion (e.g. Chesson, 2000 {1}). The modern theory, focusing on mechanisms guaranteeing stable coexistence, is deterministic, representing populations as continuous variables on infinite landscapes. Deterministic theory supplies methods to predict whether two species can coexist, by testing whether each species, starting from small numbers, can invade the other species at its equilibrium population size. In fact, however, populations consist of a finite number of discrete individuals. Deterministic models that predict coexistence imply that coexistence will last indefinitely, but demographic stochasticity will inevitably limit the duration of coexistence through extinction of one or both species. This raises the question of whether theory that incorporates stochasticity can predict the actual mean time that coexistence can be expected to last. To explore the effects of demographic stochasticity on species coexistence, the authors started with a deterministic discrete-time model of competition between two annual plant species that has been well-studied analytically and that has been parameterized with data on 18 plant species on California serpentine soils. From these well-studied species, 90 pairs of two-species competition could be studied. For their stochastic version of the model of two-species competition, the authors let the offspring forming a new generation be generated as a random number of discrete individuals described by a Poisson distribution. A parameter was added to prescribe the mean size of the community, which could vary from that which could fit in a few square meters of habitat to communities large enough to be described deterministically. The authors introduced a measure they called ‘intrinsic coexistence time’. This is analogous to the ‘intrinsic extinction time’ for a single species (Grimm and Wissel, 2004 {2}), but in this case it is the mean time for multiple species to coexist before one goes to extinction. With parameterization from field data, the authors could compare intrinsic coexistence time for the deterministic and stochastic models for pair-wise competing species. The authors examined the effect of several metrics from deterministic theory that could affect intrinsic coexistence time in their stochastic model. These included the deterministic coexistence time for those cases where deterministic models predicted competitive exclusion, as well as invasion growth rates, equilibrium population sizes, differences in niche overlap and fitness, and competitive interactions (increases one species being correlated with decreases in the other). To calculate intrinsic coexistence time for a pair of species, the authors required the pair to have coexisted over a long previous time period, thus exhibiting relatively stationary population dynamics. This can be described as a quasi-stationary distribution (QSD) (Méléard and Villemonais, 2012 {3}), in which there is a constant probability at each time step of a population going to the absorbing state of extinction. For large populations, solving for the intrinsic coexistence time is intractable from the QSD, but an approximation by Aldous et al. (1988) {4} makes this possible and was used by the authors. Of the 90 possible pairwise competitions, 82 were determined from the deterministic model to result in competitive exclusion, while eight resulted in coexistence. Using their stochastic simulations, the authors found that, indeed, coexistence that was predicted deterministically to persist did so stochastically for decades or millennia, and the intrinsic coexistence time increased exponentially with community size, while those predicted not to coexist lasted for only five years or less. Comparing stochastic coexistence for seven of the pairs of competitors predicted deterministically to coexist, the authors found that the invasion growth rates from the deterministic model explained much but not all of the variation in the coexistence times. Equilibrium population size explained most of the remaining variation, but with the interesting results that some cases of similar equilibrium population size differed by orders of magnitude in coexistence time. For example, the simulations showed that pairs of species with community sizes of only 10 individuals, for which standard population viability analysis would predict a high risk of imminent extinction, could coexist for times greater than 1000 years. The authors’ model also showed that coexistence time decreased with increasing amount of both niche overlap and fitness ratio for five of the seven pairs, but the behavior was more complex for the other two pairs, as equilibrium size and invasion growth rates also differed. The number of pairs was too small to elaborate much further. Answering the title’s question, the authors show that a combination of deterministic invasion growth rate and population sizes at equilibrium are very helpful in predicting coexistence time, though they are only a starting point. Since these quantities are directly obtainable from the deterministic models, the authors’ approach has practical applications beyond this paper. The analytic and computational approaches introduced here are new and certain to be widely applied.