Wiley

Ecology Letters

Published by Wiley and Centre National de la Recherche Scientifique
Online ISSN: 1461-0248
Print ISSN: 1461-023X
Discipline: Ecology
Learn more about this page
Aims and scope

Ecology Letters is a forum for the very rapid publication of the most novel research in ecology. Manuscripts relating to the ecology of all taxa, in any biome and geographic area will be considered, and priority will be given to those papers exploring or testing clearly stated hypotheses. The journal publishes concise papers that merit urgent publication by virtue of their originality, general interest and their contribution to new developments in ecology. We discourage purely descriptive papers and those merely confirming or extending results of previous work.

 

Editors

Recent publications
Dispersal plasticity (a) and fitness sensitivity to the environment (b) across species and environmental gradients. Reaction norms are provided on the left to illustrate the meaning of dispersal plasticity (a) and fitness sensitivity (b) values along the y‐axes. (a) Dispersal plasticity estimates close to zero indicate flat reaction norms, while positive or negative values respectively denote increase and decrease of dispersal rate along environmental gradients. (b) Fitness sensitivity close to zero indicates low effects of environmental variation on fitness, while high values denote a high impact of environmental variation on fitness. Variability of dispersal plasticity and fitness sensitivity to environment between species are shown along the gradients of temperature, resources and salinity (blocs of five coloured boxplots from left to right). Boxplots show the distribution of dispersal plasticity and fitness sensitivity of the five genotypes (grey dots) of each species.
Dispersal plasticity depends on fitness sensitivity to the environment. Coloured lines and grey shaded areas represent linear regressions of dispersal plasticity function of fitness sensitivity along each environmental gradient (red: temperature; green: resources; blue: salinity). As in Figure 1, reaction norms are provided along the y and x‐axes to illustrate the meaning of dispersal plasticity and fitness sensitivity. Detailed relationships per species are provided in Figure S3.
Relationship between dispersal rate and expected fitness across species and environmental gradients. Reaction norms along the y‐axis illustrate the meaning of dispersal along fitness: estimates close to zero indicate flat reaction norms, while positive or negative values respectively denote increase and decrease of dispersal rate along expected fitness. Variability of dispersal along fitness between species are showed along the gradients of temperature, resources and salinity (blocs of five coloured boxplots from left to right). Boxplots show the distribution of dispersal along fitness of the five genotypes (grey dots) of each species.
Dispersal plasticity, when organisms adjust their dispersal decisions depending on their environment, can play a major role in ecological and evolutionary dynamics, but how it relates to fitness remains scarcely explored. Theory predicts that high dispersal plasticity should evolve when environmental gradients have a strong impact on fitness. Using microcosms, we tested in five species of the genus Tetrahymena whether dispersal plasticity relates to differences in fitness sensitivity along three environmental gradients. Dispersal plasticity was species‐ and environment‐dependent. As expected, dispersal plasticity was generally related to fitness sensitivity, with higher dispersal plasticity when fitness is more affected by environmental gradients. Individuals often preferentially disperse out of low fitness environments, but leaving environments that should yield high fitness was also commonly observed. We provide empirical support for a fundamental, but largely untested, assumption in dispersal theory: the extent of dispersal plasticity correlates with fitness sensitivity to the environment.
 
Ecological setting. (a) The study region is in Monterey Bay National Marine Sanctuary (MBNMS) along the eastern margin of the North Pacific. Hydrophones are operated through the Monterey Accelerated Research System (MARS) cabled observatory (black line and circle; main node at 36.713°N, 122.186°W, 891 m depth). The white circle marks the GPS location of a tagged blue whale, acquired 10 min after a series of blue whale calls recorded on the tag were matched to calls received at MARS, and the green‐shaded sector defines the span of MARS acoustically estimated bearings to calls originating from that tagged whale (see Results). (b) An infrared sea surface temperature (SST) image from 10 November 2020 14:44 UTC represents synoptic patterns of coastal upwelling plumes (cool SST labelled P) at a time when blue whale behavioural response to upwelling was observed. Point Año Nuevo and Point Sur are the primary locations in the Monterey Bay region where coastal upwelling plumes originate (Rosenfeld et al., 1994). The black contour defines the approximate domain within which bearings to calling blue whales can be reliably estimated (see ‘Materials and methods’). The diamond marks the location of a mooring where temperature and salinity were monitored. (c) This 6‐year time‐series from MARS passive acoustic monitoring represents seasonal and interannual variation in blue whale call activity. Shown are the monthly interquartile range and median values of the daily blue whale B‐call index (see ‘Materials and methods’). The two periods of blue whale call activity for which we examine directional acoustic vector sensor data are indicated by the numbered boxes. Whale artist: Larry Foster.
Directional acoustic vector sensor analysis methods. (a) Geometric mean spectrum levels for July 2019 through January 2020 illustrate a peak caused by the third harmonic of the blue whale B call. The frequency band for peak localization is shaded in grey; the frequency bands used for background are indicated by the thick black lines along the frequency axis, above and below the peak band. The call index is the ratio of maximum power spectral density (psd) within the frequency band of the call to average psd in the background bands (see ‘Materials and methods’). (b) Example of data processing to obtain bearing estimates for blue whale B calls (direction of call origin relative to MARS). The top panel is a spectrogram representing the third harmonic of B calls (sample rate = 8 kHz, nfft = 16,000, hanning window, 50% overlap). For each second for which the call index exceeds 1.25, the middle panel shows the frequencies at which maximum energy was detected within the peak band (note the tracking of peak energy in the downswept B calls), and the bottom panel shows estimated bearings from MARS to the caller, extracted at the frequency of peak call energy. The dashed line indicates a steadily changing bearing presumed to be from a whale moving relative to the hydrophone. (c) The domain over which peak:background ratios are sufficient to reliably estimate the direction of call origin is approximated by modelling of acoustic transmission loss. This domain is defined by received level (RL) > 1.25 times mean spectrum levels within a frequency band below the peak band (panel a, 80 dB re 1 μPa²/Hz). The black contour is the outer limit of RL > 100, smoothed with a moving mean window of 7° bearing.
Validation of acoustic bearings from MARS to the origin of received blue whale calls. (a) Observations from MARS and an animal‐borne tag characterize the period when sufficient tag GPS positions and call activity from the tagged whale enabled effective matchup with signals received at MARS. The percent of time for which the call index exceeded 1.25 at MARS is represented in 5‐min bins. Behavioural events indicated in the animal depth profile were derived from multiple sensors on the tag (Cade, Gough, et al., 2021). (b) Detailed view of data from the animal‐borne tag and MARS during the focal period of matchup (shaded in a), when calls made by the tagged whale were definitively matched to calls received at MARS. The depth profile overlaid on the spectrogram from the animal‐borne hydrophone defines the whale's surfacing (breathing) intervals. The starts of calls from the tagged whale are indicated by vertical dashed lines in all panels. To align signals from the spatially offset platforms, the time axis of the MARS data was shifted by an appropriate acoustic propagation delay between source and receiver. The period highlighted in green along the top axis defines the period from which acoustically estimated call bearings, one average bearing per call, were compared with bearings based on GPS location (Figure 1a, the green sector spans the range of acoustically estimated bearings; the white circle is the GPS position).
Overview of whale‐call bearings relative to coastal upwelling centres. (a) Upwelling centres are associated with coastal land points north and south of Monterey Bay, as illustrated by plumes of cold water originating at Point Año Nuevo and Point Sur in this sea surface temperature (SST) image from 10 November 2020 14:44 UTC. The dashed lines mark the bearings from MARS to the coastal origins of the upwelling plumes. (b) Total hours of call index values exceeding 1.25 during annual periods of song presence (numbered in Figure 1c) within 15° directional bins. Dashed lines indicate bearings from MARS to the coastal upwelling centres (as in (a)). (c) Daily representation of (b) for each annual period. For 2020–2021 (lower panel), the grey diamond indicates the time of the SST image in panel (a), and the white box defines the focal period examined in Figure 5. (d) Hourly representation of directional call activity during periods when a consistent pattern of blue whale behavioural response to upwelling was observed (grey shaded periods below panel (c)).
Ephemeral wind‐driven resource tracking by blue whales. (a) Reference map. Octant radials (dashed black lines) within the northern hemisphere of the call detection domain define ecologically significant features for interpreting blue whale movement (see the text). Red lines define recommended tracks for vessels 300 gross tons and heavier. (b) MARS call detection rate summed for the northern hemisphere (N) around MARS and the northeast quadrant (NE, the two octants northeast of MARS in panel (a)). (c) Percentage of call detection time from the northeast quadrant (NE/N). (d) Water column density measured at mooring M1 (location in panel (a)). (e) The coastal upwelling transport index (CUTI) for 37°N, at the north end of Monterey Bay (bars) and area scattering (sa) within patches in the upper 200 m above MARS (points), representing aggregation of forage species. Daily CUTI values were smoothed with a 3‐day running mean. Periods highlighted in yellow across panels ((b)–(e)) indicate persistently elevated percentages of calling from the NE quadrant (panel (c)), coincident with the presence of recently upwelled water in Monterey Bay (outcropping of isopycnals to the surface in panel (d)). (f) Percentages of call time originating from each octant of the northern hemisphere of the call reception domain are labelled and coloured according to magnitude during alternating periods of wind relaxation and upwelling (as identified in panels (b)–(e)).
Trophic transfer of energy through marine food webs is strongly influenced by prey aggregation and its exploitation by predators. Rapid aggregation of some marine fish and crustacean forage species during wind‐driven coastal upwelling has recently been discovered, motivating the hypothesis that predators of these forage species track the upwelling circulation in which prey aggregation occurs. We examine this hypothesis in the central California Current Ecosystem using integrative observations of upwelling dynamics, forage species' aggregation, and blue whale movement. Directional origins of blue whale calls repeatedly tracked upwelling plume circulation when wind‐driven upwelling intensified and aggregation of forage species was heightened. Our findings illustrate a resource tracking strategy by which blue whales may maximize energy gain amid ephemeral foraging opportunities. These findings have implications for the ecology and conservation of diverse predators that are sustained by forage populations whose behaviour is responsive to episodic environmental dynamics.
 
Left hand panels show maps of study landscapes (triangles) along the elevational gradients on both islands. Dashed line on Hispaniola separates the north paleoisland from the south paleoisland—areas that have had separate evolutionary histories during periods of higher sea levels, and which have different species pools. Note that maps of Jamaica and Hispaniola are not on the same scale. Right hand panel illustrates the total area (y‐axis on log scale) in 200 m elevational bands on both islands. Note the steep drop‐off in area on Jamaica above 700 m.
Schematic illustrating hypothesised influence of limited evolutionary opportunity in the highland sub‐region on species‐level composition within and between communities. On the island to the left, both lowland and highland sub‐regions have sufficient space for speciation. As a result, diversification has filled available niches (boxes) with species (represented by different coloured lizards) in single examples of sub‐regional communities (sets of joined boxes) for both the lowlands and highlands. On the right island, however, the limited highland area has prevented in situ speciation in that sub‐region, resulting in (i) the evolution of only a single highland endemic (blue lizard), (ii) highland communities that include lowland‐associated species (yellow lizard), and (iii) communities undersaturated at the species level (indicated by empty boxes). In the lowlands of both islands, sufficient evolutionary opportunity leads to convergence in niche‐filling and the species diversity held within communities.
Anolis abundance and species richness across survey plots on both islands. (a) Total Anolis abundance declines monotonically with elevation and is statistically indistinguishable between islands. The depicted model includes linear and quadratic effects of elevation and full interactions with island identity, though only the effect of elevation is statistically significant after backwards model selection. (b) Species richness patterns along elevational gradients deviate between islands, ranging from rough equivalency in the lowlands to greater species richness in the Hispaniolan fauna upslope. Figure depicts model best‐fit lines on Jamaica (green) and Hispaniola (yellow), with shaded regions representing standard error. Significant differences between islands are emphasised with black borders around best‐fit lines and standard error regions in panel (b). Points represent posterior mean abundance and species richness at study plots derived from mark‐resight models (NJamaica = 20, NHispaniola = 35).
Greater macroevolutionary diversity on Hispaniola generates a distinctive high‐elevation fauna. (a) Pairwise plot‐level community dissimilarity based on species identity (balancing component) against the differences in elevation between plots. (b and c) Heat maps showing abundance of species in all plots along island elevational gradients. Plots are organised along the y‐axis so that the lowest‐elevation plots are on the bottom and the highest are on top. Tick marks along the y‐axis show 500 m increments. Note that y‐axes are not scaled uniformly within islands, because low elevation habitats are more common and thus better represented, and are not identical across islands, as sampling sites were not precisely matched by elevation. Species along the x‐axis are ordered by mean elevation at which the species occurs. Note (i) the distinct highland species on Hispaniola, and (ii) differences between islands in elevation at which no anoles occur. (d) Pairwise phylogenetic dissimilarity of plots as in panel (a) highlighting the phylogenetic turnover with difference in elevation on Hispaniola. (e) Phylogeny of all Anolis species observed, with branches coloured according to island. Blue underlay highlights highland species (species not observed at sea‐level); illustrated timescale obtained by setting the crown age of Jamaican Anolis in the ultrametric tree of Mahler et al., 2013 to 15 Ma, following Buskirk, 1985. (f) Pairwise morphological dissimilarity between plots as in panel (a) highlighting greater morphological dissimilarity between elevation zones on Hispaniola. (g). Location in morphospace of all observed species (transparent squares) along first two morphological principal component axes from 11 morphometric measurements. PC1 corresponds to limb length whereas PC2 corresponds to overall body size. Square size represents overall abundance of the species across all plots. Solid circles depict the abundance‐weighted morphological means of all plots in the lowlands (<700 m asl) and triangles represent the highlands (>700 m asl). In all plots (except b and c) yellow indicates plots or species on Hispaniola whereas green indicates those on Jamaica.
Species abundance and richness across elevation, excluding highland endemics from analysis. (a) Abundance of all lizard individuals in plots across elevations, evaluating only lowland species (those that have maximum abundances <700 m). When discounting the highland fauna, Hispaniola has fewer individuals at middle elevations, though total abundance is still equivalent between islands near sea level. Compare to Figure 3a, which analyses all species and shows equivalent abundance profiles by elevation across islands. (b) Species richness in plots, evaluating only lowland species. Removing the highland species yields remarkably similar alpha diversity trajectories between islands, suggesting that differences in species richness trajectories across islands observed in Figure 3b can be attributable to the presence of a diverse highland fauna on Hispaniola. Definitions of lines, shaded regions and points are the same as in Figure 3. Note that both panels depict models with linear and quadratic effects of elevation and full interactions with island identity, though only the effect of elevation was significant after backwards model selection for lowland species richness.
Ecological community structure ultimately depends on the production of community members by speciation. To understand how macroevolution shapes communities, we surveyed Anolis lizard assemblages across elevations on Jamaica and Hispaniola, neighbouring Caribbean islands similar in environment, but contrasting in the richness of their endemic evolutionary radiations. The impact of diversification on local communities depends on available spatial opportunities for speciation within or between ecologically distinct sub‐regions. In the spatially expansive lowlands of both islands, communities converge in species richness and average morphology. But communities diverge in the highlands. On Jamaica, where limited highland area restricted diversification, communities remain depauperate and consist largely of elevational generalists. In contrast, a unique fauna of high‐elevation specialists evolved in the vast Hispaniolan highlands, augmenting highland richness and driving islandwide turnover in community composition. Accounting for disparate evolutionary opportunities may illuminate when regional diversity will enhance local diversity and help predict when communities should converge in structure.
 
An illustration of the storage effect in the lottery model. Panel (a): For the common red species, a good environment (high per capita larval production) is undermined by the competition (total larvae per empty site) that it brings about. Panel (b): The blue species recovers from rarity because a good environment does not lead to high competition.
A causal diagram shows how the EC covariance arises generally, with panels showing how the EC covariance arises in the annual plant model. The black arrows show the direction of causation, e.g. increased population density N causes increased competition C. The blue dashed arrow indicates a set‐subset relationship: the per capita growth rate r is a function of the effective regulating factor Ej−E¯jC−C¯$$ \left({E}_j-{\overline{E}}_j\right)\left(C-\overline{C}\right) $$ (which becomes CovE,C$$ \mathrm{Cov}\ \left(E,C\right) $$ when averaged over time). The negative sign indicates the negative EC interaction effect typical of competition models. The environment E has a direct effect on r. However, it takes some time (i.e., two causal arrows) for the indirect effects of E on r to be mediated through C. Therefore, for E to covary with C, there must be some mechanism for carrying the effects of the environment through time. In the annual plant model, this mechanism is the germinant life stage. Panel (a) Precipitation causes a high probability of germination. Panel (b) The germinants compete for a limited supply of soil nitrogen. Panel (c) A good environment (i.e., high germination probability) is undermined by the high competition that it brings about, manifesting as few seeds per capita (grey plants are dead germinants). Panel (d) The seeds disperse and join the seed bank (off‐season seed mortality not shown).
The covariance between environment and competition can be generated by causal chains of environmental variables. Solid arrows denote direct causal relationships. The dotted arrow denotes an indirect relationship. The causal relationship between the exogenous density‐independent factors—precipitation and pollinators—prevents the effects of the environment from changing too quickly, thus satisfying ingredient 3B. The demographic parameters are correlated because both are causally affected by precipitation on different time lags.
The storage effect is a general explanation for coexistence in a variable environment. Unfortunately, the storage effect is poorly understood, in part because the generality of the storage effect precludes an interpretation that is simultaneously simple, intuitive and correct. Here, we explicate the storage effect by dividing one of its key conditions—covariance between environment and competition—into two pieces, namely that there must be a strong causal relationship between environment and competition, and that the effects of the environment do not change too quickly. This finer‐grained definition can explain a number of previous results, including (1) that the storage effect promotes annual plant coexistence when the germination rate fluctuates, but not when the seed yield fluctuates, (2) that the storage effect is more likely to be induced by resource competition than the apparent competition, and (3) why the storage effect arises readily in models with either stage structure or environmental autocorrelation. Additionally, our expanded definition suggests two novel mechanisms by which the temporal storage effect can arise—transgenerational plasticity and causal chains of environmental variables—thus suggesting that the storage effect is a more common phenomenon than previously thought.
 
Schematic of predictions and interpretation of BayesTraits discrete‐dependent models for the correlated evolution of egg attendance and reproductive ecology. If the dependent model is the better fitting model (Supplementary Methods), we examine the posterior distributions of its eight transition rates between the four possible combinations of character states to identify evolutionary pathways supporting the predictions of the alternative hypotheses. As an example, here we illustrate the case of aquatic/terrestrial eggs and presence/absence of egg attendance. The arrows indicate the direction of change and their thickness the magnitude of the associated transition rates (the qij) between the four states (aquatic eggs and no egg attendance, aquatic eggs and egg attendance, terrestrial eggs and no egg attendance, terrestrial eggs and egg attendance). (a) Reproductive ecology drives the origin care (blue arrows): If the presence of terrestrial eggs promotes the evolution of egg attendance, the transition rate for the gain of egg attendance is higher with terrestrial eggs than aquatic eggs (q34 > q12) and/or terrestrial eggs are gained more rapidly from the absence of both traits than egg attendance (q13 > q12) and this is followed by a rapid gain of egg attendance (q34 > 0). (b) Care drives changes in reproductive ecology (red arrows); if egg attendance promotes the acquisition of terrestrial eggs, the transition rate for the gain of terrestrial eggs is faster with egg attendance than without it (q24 > q13) and/or egg attendance evolves faster than terrestrial eggs from the absence of both traits (q12 > q13) and promotes the rapid gain of terrestrial eggs (q24 > 0). (c) Care and reproductive ecology adaptations are alternatives (red, blue and grey arrows); the combination of terrestrial eggs and egg attendance is lost more rapidly than it is gained (q43 > q34; q42 > q24) indicating that it is evolutionarily unstable, while terrestrial eggs without attendance and/or attendance of aquatic eggs are evolutionary stable (q12 > q21 and q42 > q24; q13 > q31 and q43 > q34). See also Table 1 for further details.
(a) Distribution of the data on the phylogeny in amphibians. From the inner circle to the external circle: Male egg attendance, female egg attendance, terrestrial eggs, hidden eggs and direct development (n = 1202 species for all traits). (b, c) posterior density distributions of the parameter estimates from the probit models for the association between sex‐specific egg attendance and reproductive ecology predictors. The posterior distributions of the parameter estimates (β) of the predictors (i.e. terrestrial eggs, direct development, hidden eggs and egg attendance by the opposite sex) are shifted to the right of 0 (dotted vertical line) with positive effects, and left with negative effects. We report the percentage of the posterior distribution that crosses 0 (Px) and consider evidence of significance Px < 0.05. Full model details can be found in table S2.
The correlated evolution of terrestrial eggs and egg attendance. Terrestrial eggs (blue) and female egg attendance (red) in (a) and male egg attendance (red) in (d) plotted on the phylogeny using stochastic character mapping with the R package phytools to visualise their evolutionary history. In (b) and (e), summary diagrams of the transition rates across the four combinations of character states from the RJ discrete‐dependent model of evolution (care first pathway highlighted by the red arrows, ecology first pathway by the blue arrows; pathways in grey indicate reversals with losses of traits). The sample sizes for each combination of character states are reported. Within each summary diagram, the arrows are scaled to reflect the magnitude of the reported mean transition rates from the posterior distribution. Arrows are dashed when a parameter is estimated to be equal to zero in over 25% of models of the posterior distribution. In (c) and (f), the posterior distributions of the transition rates from the RJ discrete‐dependent model are shown as box plots for comparison and as posterior density plots for each transition rate alone. The central black dot in the box plots indicates the median, the box the upper and lower quartiles, the vertical lines the 95% credible intervals of the posterior distributions and the filled dots beyond the lines indicate outlier estimates (Table S5 for additional transition rate summaries).
The correlated evolution of hidden eggs and egg attendance. Hidden eggs (blue) and female egg attendance (red) in (a) and male egg attendance (red) in (d) plotted on the phylogeny using stochastic character mapping with the R package phytools to visualise their evolutionary history. In (b) and (e), summary diagrams of the transition rates across the four combinations of character states from the RJ discrete‐dependent model of evolution (care first pathway highlighted by the red arrows, ecology first pathway by the blue arrows; pathways in grey indicate reversals with losses of traits). The sample sizes for each combination of character states are reported. Within each summary diagram, the arrows are scaled to reflect the magnitude of the reported mean transition rates from the posterior distribution. Arrows are dashed when a parameter is estimated to be equal to zero in over 25% of models of the posterior distribution. In (c) and (f), the posterior distributions of the transition rates from the RJ discrete‐dependent model of evolution are shown as box plots for comparison and as posterior density plots for each transition rate alone. The central black dot in the box plots indicates the median, the box the upper and lower quartiles, the vertical lines the 95% credible intervals of the posterior distributions, and the filled dots beyond the lines indicate outlier estimates (Table S6 for additional transition rate summaries).
The correlated evolution of direct development and female egg attendance. In (a) direct development (blue) and female egg attendance (red) plotted on the phylogeny using stochastic character mapping with the R package phytools to visualise their evolutionary history. In (b), summary diagrams of the transition rates across the four combinations of character states from the RJ discrete‐dependent model of evolution (care first pathway highlighted by the red arrows, ecology first pathway by the blue arrows; pathways in grey highlight reversals with losses of traits). The sample sizes for each combination of character states are reported; the arrows are scaled to reflect the magnitude of the reported mean transition rates from the posterior distribution, with the mean value also indicated. Arrows are dashed when a parameter is estimated to be equal to zero in over 25% of models of the posterior distribution. In (c), the posterior distributions of the transition rates from the RJ discrete‐dependent model of evolution are shown as box plots for comparison and as posterior density plots for each transition rate alone. The central black dot in the box plots indicates the median, the box the upper and lower quartiles, the vertical lines the 95% credible intervals of the posterior distributions and the filled dots beyond the lines indicate outlier estimates (Table S7 for additional transition rate summaries).
Parental care is extremely diverse but, despite much research, why parental care evolves is poorly understood. Here we address this outstanding question using egg attendance, the simplest and most common care form in many taxa. We demonstrate that, in amphibians, terrestrial egg deposition, laying eggs in hidden locations and direct development promote the evolution of female egg attendance. Male egg attendance follows the evolution of hidden eggs and is associated with terrestrial egg deposition but not with direct development. We conclude that egg attendance, particularly by females, evolves following changes in reproductive ecology that are likely to increase egg survival, select for small clutches of large eggs and/or expose eggs to new environmental challenges. While our results resolve a long‐standing question on whether reproductive ecology traits are drivers, consequences or alternative solutions to caring, they also unravel important, yet previously unappreciated, differences between the sexes.
 
Typical network configuration and degree distribution in each one of the topologies considered (top panel), empirical examples from the dataset, geographical distribution and human pressure (mid panel) and food web classification (bottom panel): (a) pure scale‐free network (network and degree distribution) topology; (b) pure random network topology; (c) example of a network close to pure scale‐free topology (Kelleway et al., 2010); (d) example of a network close to pure random topology (Khan & Panikkar, 2009); (e) global distribution of the food webs in the dataset with the terrestrial impact metric (human footprint [Venter et al., 2016a, 2016b]) and the marine impact metric (cumulative impact to marine ecosystems (Halpern et al., 2015)) scaled from 0 to 1 for comparability; (f) categorisation of the food webs in the dataset where the signalled corners are those in which the food webs are closer to each of the two pure topologies (x axis: Correlation coefficient to gaussian (RG); y axis: Correlation coefficient to power‐law (RPL); see methods). The pattern observed in panel (f), whereby some values fall along the y axis (RG = 0) and almost all points fall below an imagined diagonal line (defined RG = RPL), is a consequence of Equation 2. According to this equation, the computation of RG is conducted only for food webs to which the degree distribution is such that the left portion of the curve is present (μ + σ/3 > 0), and the maximum degree (Kmax) is big enough to also allow the right portion of the curve to be present (kmax > μ + σ) (Figure S5). Food webs with degree distributions not conforming to these conditions would have an RG = 0, which is the case with most that would otherwise occupy the upper triangle of the plot.
Location of an hypothetical food web in the (RG, RPL) plane (grey circle), with the distances to pure gaussian (random; D2PG) and pure power‐law (D2PL) characterising its topology.
Plot depicting the relation between distance to pure gaussian (D2PG) and the number of nodes in each food web (nodes). Notice that distances to pure scale‐free (D2SF) are symmetrical with D2PG so that high values of D2PG corresponded to low values of D2SF and vice versa.
Relationship between human pressure and the distance to each of the three pure topologies in each ecosystem. Human pressure was evaluated as the cumulative impact on the world's oceans (Halpern et al., 2015) on coastal and marine ecosystems and the human footprint (Venter et al., 2016a, 2016b) impact on terrestrial and freshwater ecosystems. The relation between distance to pure topologies and the human pressure was characterised by the best fitting linear relationship. Note that values in the x‐axis represent the distance to pure topologies (D2PG and D2PL) to each food web (as shown in Figure 2). As such, values greater than 1 are possible, only the RG and RPL are constrained to vary between 0 and 1. Blue circles represent the raw data (scaled by food web size) and the blue dashed line represents the regression on these data. The black points and the black regression line represent the binned data. Error bars express the variance of the averaged values in each bin. The regression equation represented refers to the regression on the binned data.
Location of the food webs in the parameter space defined by the parameters (b) and (c) in Equation 5. The gradient, from red to blue, refers to the increasing distance to pure scale‐free). The parameters (b) and (c) in Equation 5 determine the shape of the R50 curve. The smaller plots (those in food webs GW330, GW185, EB444 and EB145) depict the robustness (R50) curve as the intentionality increases (the R50 plots for all the food webs are available as Appendix S2). Within each R50 plot: The grey points represent the average of R50 with increasing intentionality. The blue line refers to the cumulative Weibull function used to fit the R50 values.
Networks describe nodes connected by links, with numbers of links per node, the degree, forming a range of distributions including random and scale‐free. How network topologies emerge in natural systems still puzzles scientists. Based on previous theoretical simulations, we predict that scale‐free food webs are favourably selected by random disturbances while random food webs are selected by targeted disturbances. We assume that lower human pressures are more likely associated with random disturbances, whereas higher pressures are associated with targeted ones. We examine these predictions using 351 empirical food webs, generally confirming our predictions. Should the topology of food webs respond to changes in the magnitude of disturbances in a predictable fashion, consistently across ecosystems and scales of organisation, it would provide a baseline expectation to understand and predict the consequences of human pressures on ecosystem dynamics.
 
Competition for limited resources is a major force in structuring ecological communities. Species minimum resource requirements (R*s) can predict competitive outcomes and evolve under selection in simple communities under controlled conditions. However, whether R*s predict competitive outcomes or demonstrate adaptive evolution in naturally complex communities is unknown. We subjected natural phytoplankton communities to three types of resource limitation (nitrogen, phosphorus, light) in outdoor mesocosms over 10 weeks. We examined the community composition weekly and isolated 21 phytoplankton strains from seven species to quantify responses to the selection of R* for these resources. We investigated the evolutionary change in R*s in the dominant species, Desmodesmus armatus. R*s were good predictors of species changes in relative abundance, though this was largely driven by the success of D. armatus across several treatments. This species also demonstrated an evolutionary change in R*s under resource limitation, supporting the potential for adaptive trait change to modify competitive outcomes in natural communities.
 
Examples of predicted responses from the multiplicative, additive and dominance models when stressor pairs (SA = stressor a, SB = stressor B) have positive (a), contrasting (b) or negative (c) effects relative to the control, which varies in the different scenarios (ABS = absolute). Responses include plant or animal survival, growth/size, condition, biomass, abundance, diversity and decomposition measured from experimental treatments. An overview of our empirical results (d) shows the best performing null models across all (296) responses and the percentage of observations that fell within or beyond the range predicted by all three null models. Undefined responses are those where multiple null models were indistinguishable in their performance.
Percentage of observed responses to warming paired with another stressor that fall within or beyond the range predicted by all three null models grouped by stressor type (a) and organisation level (b) for the entire dataset (n = 296). Directions are on an absolute scale (e.g. observed responses larger than predicted were either more negative or more positive).
Asymmetric stressors. Frequency of responses best predicted by additive (blue), dominance (orange) and multiplicative (green) models (a) and their mean effect sizes (Hedges' d) showing standardised differences between observed effects and those predicted for each null model (with 95% confidence intervals; (b) across types of secondary stressors combined with temperature, when stressors were not strongly asymmetric or when either temperature, habitat alteration, nutrient levels or contamination had prevailing independent effects (effect >50% higher than the other). Mean effect sizes of zero indicate no difference between observed responses (Xo$$ {X}_o $$) and null model predictions (Xp$$ {X}_p $$). Mean observed effects are considered statistically indistinguishable from null model predictions when confidence intervals cross zero. Significant negative effect sizes indicate that observed responses were smaller (antagonistic; either less negative or less positive) than predicted. Although not seen here, significant positive effect sizes would indicate that observed responses were larger (more negative or more positive) than predicted. Panel (c) shows the absolute effect size for each null model with increasing asymmetry (% difference in independent effects) where either temperature or the second stressor are the prevailing independent stressor fitted with GLMs. Increasing (decreasing) absolute effect sizes indicate increasing (decreasing) deviation from null model predictions.
Stressor pairs. Frequency of responses best predicted by additive (blue), dominance (orange) and multiplicative (green) models (a) and mean effect sizes (Hedges' d) showing standardised differences between observed effects and those predicted for each null model (with 95% confidence intervals; (b) across types of secondary stressors combined with temperature, T. mean effect sizes of zero indicate no difference between observed responses (Xo$$ {X}_o $$) and null model predictions (Xp$$ {X}_p $$). Mean observed effects are considered statistically indistinguishable from null model predictions when confidence intervals cross zero. Significant negative effect sizes indicate that observed responses were smaller (either less negative or less positive) than null model predictions. Although not seen here, significant positive effect sizes would indicate that observed responses were larger (more negative or more positive) than predicted. Samples sizes may differ from Figure 1 because outliers were removed from mean effect calculations (see Methods).
Biological organisation. Frequency of responses best predicted by additive (blue), dominance (orange) and multiplicative (green) models (a) and their mean effect sizes (Hedges' d) showing standardised differences between observed effects and those predicted for each null model (with 95% confidence intervals; (b) across levels of biological organisation. Mean effect sizes of zero indicate no difference between observed responses (Xo$$ {X}_o $$) and null model predictions (Xp$$ {X}_p $$). Mean observed effects are considered statistically indistinguishable from null model predictions when confidence intervals cross zero. Significant negative effect sizes indicate that observed responses were smaller (either less negative or less positive) than null model predictions. Although not seen here, significant positive effect sizes would indicate that observed responses were larger (more negative or more positive) than predicted.
Climate warming is a ubiquitous stressor in freshwater ecosystems, yet its interactive effects with other stressors are poorly understood. We address this knowledge gap by testing the ability of three contrasting null models to predict the joint impacts of warming and a range of other aquatic stressors using a new database of 296 experimental combinations. Despite concerns that stressors will interact to cause synergisms, we found that net impacts were usually best explained by the effect of the stronger stressor alone (the dominance null model), especially if this stressor was a local disturbance associated with human land use. Prediction accuracy depended on stressor identity and how asymmetric stressors were in the magnitude of their effects. These findings suggest we can effectively predict the impacts of multiple stressors by focusing on the stronger stressor, as habitat alteration, nutrients and contamination often override the biological consequences of higher temperatures in freshwater ecosystems.
 
Cities pose a major ecological challenge for wildlife worldwide. Phenotypic variation, which can result from underlying genetic variation or plasticity, is an important metric to understand eco‐evolutionary responses to environmental change. Recent work suggests that urban populations might have higher levels of phenotypic variation than non‐urban counterparts. This prediction, however, has never been tested across species nor over a broad geographical range. Here, we conducted a meta‐analysis of the avian literature to compare urban versus non‐urban means and variation in phenology (i.e. lay date) and reproductive effort (i.e. clutch size, number of fledglings). First, we show that urban populations reproduce earlier and have smaller broods than non‐urban conspecifics. Second, we show that urban populations have higher phenotypic variation in laying date than non‐urban populations. This result arises from differences between populations within breeding seasons, conceivably due to higher landscape heterogeneity in urban habitats. These findings reveal a novel effect of urbanisation on animal life histories with potential implications for species adaptation to urban environments (which will require further investigation). The higher variation in phenology in birds subjected to urban disturbance could result from plastic responses to a heterogeneous environment, or from higher genetic variation in phenology, possibly linked to higher evolutionary potential.
 
Hypothetical structural equation models (SEMs) representing hypothesised direct and indirect effects of abiotic conditions and land use (i.e. land‐use type, land‐use intensity, ecological focus areas (EFAs) and landscape heterogeneity) on farmland γ‐diversity (i.e. multitrophic species richness (black), dispersal‐limited plants (green), food‐specialised butterflies (red) and migration‐limited birds (yellow)). Grey arrows indicate indirect pathways which do not differ for the individual components of farmland γ‐diversity. See supplementary 1 for the description of the individual expectations on direct and indirect relations.
The location of the study area in Europe and the 123 1‐km² investigation squares within the study area (a). Sampling of farmland within an exemplary investigation square in terms of (b) vegetation types and plant species on a sampling grid, (c) butterfly occurrences along transects and (d) bird territories across the entire area.
Structural equation models (SEMs) representing direct and indirect effects of abiotic conditions and land use (i.e. land‐use type, land‐use intensity, ecological focus areas (EFAs) and landscape heterogeneity), on farmland γ‐diversity (i.e. multitrophic species richness (black), dispersal‐limited plants (green), food‐specialised butterflies (red) and migration‐limited birds (yellow)). Shown are the final SEMs, and only variables that were retained in the final SEMs. Line thickness of paths is proportional to the values of the standardised coefficients (if variables are presented as groups, the largest standardised coefficient is used). The direction of the standardised coefficients is indicated with a ‘+’ or ‘−’ behind the explanatory variable. The amount of explained variance is indicated by ‘R²’. The exact values of standardised coefficients, standardised errors and significance values are given in Table S5.
To stop the ongoing decline of farmland biodiversity there are increasing claims for a paradigm shift in agriculture, namely from conserving and restoring farmland biodiversity at field scale (α‐diversity) to doing it at landscape scale (γ‐diversity). However, knowledge on factors driving farmland γ‐diversity is currently limited. Here, we quantified farmland γ‐diversity in 123 landscapes and analysed direct and indirect effects of abiotic and land‐use factors shaping it using structural equation models. The direction and strength of effects of factors shaping γ‐diversity were only partially consistent with what is known about factors shaping α‐diversity, and indirect effects were often stronger than direct effects or even opposite. Thus, relationships between factors shaping α‐diversity cannot simply be up‐scaled to γ‐diversity, and also indirect effects should no longer be neglected. Finally, we show that local mitigation measures benefit farmland γ‐diversity at landscape scale and are therefore a useful tool for designing biodiversity‐friendly landscapes. To stop the ongoing decline of farmland biodiversity there are increasing claims for a paradigm shift in agriculture, namely from conserving and restoring farmland biodiversity at the field scale (α‐diversity) to doing it at the landscape scale (γ‐diversity). By analysing 123 landscapes with structural equation models, this study demonstrates that the relationships between factors shaping α‐diversity cannot simply be up‐scaled to γ‐diversity, and that also indirect effects should taken into account when designing biodiversity‐friendly landscapes.
 
Microbial thermal adaptation is considered to be one of the core mechanisms affecting soil carbon cycling. However, the role of microbial community composition in controlling thermal adaptation is poorly understood. Using microbial communities from the rhizosphere and bulk soils in an 8‐year warming experiment as a model, we experimentally demonstrate that respiratory thermal adaptation was much stronger in microbial K‐strategist‐dominated bulk soils than in microbial r‐strategist‐dominated rhizosphere soils. Soil carbon availability exerted strong selection on the dominant ecological strategy of the microbial community, indirectly influencing respiratory thermal adaptation. Our findings shed light on the linchpin of the dominant ecological strategy exhibited by the microbial community in determining its respiratory thermal adaptation, with implications for understanding soil carbon losses under warming. Microbial respiratory thermal adaptation is considered to be one of the core mechanisms affecting soil carbon cycling. We found that respiratory thermal adaptation was much stronger in the K‐strategist‐dominant microbial community than in the r‐strategist‐dominant microbial community, and emphasise the importance of the dominant ecological strategy exhibited by the microbial community in determining its respiratory thermal adaptation, with implications for understanding soil carbon losses under warming.
 
Dietary partitioning plays a central role in biological communities, yet the extent of partitioning often varies dramatically over time. Food availability may drive temporal variation in dietary partitioning, but alternative paradigms offer contrasting predictions about its effect. We compiled estimates of dietary overlap between co‐occurring vertebrates to test whether partitioning is greater during periods of high or low food abundance. We found that dietary partitioning was generally greatest when food abundance was low, suggesting that competition for limited food drives partitioning. The extent of dietary partitioning in birds and mammals was also related to seasonality in primary productivity. As seasonality increased, partitioning increased during the nonbreeding season for birds and the breeding season for mammals. Although some hypotheses invoke changes in dietary breadth to explain temporal variation in dietary partitioning, we found no association between dietary breadth and partitioning. These results have important implications for the evolution of dietary divergence. The extent of dietary partitioning between co‐occurring species often shows dramatic temporal variation. Using a meta‐analytic approach, we find that temporal variation in resource abundance is associated with variation in the extent of dietary partitioning in vertebrates. Our results indicate that competition for scarce resources is a primary cause of dietary partitioning.
 
Mean annual temperature map of sub‐Saharan Africa from GHCN‐CAMS (Fan & van den Dool, 2008) illustrating how temperature decreases with increasing latitude and elevation. Two possible mechanisms can lead to predictable variation of song frequency across space following Bergmann's or Allen's rules predictions. According to Bergmann's rule, increasing body size would result in lower frequency songs towards higher latitudes and elevation (a, c). By contrast, Allen's rule predicts that birds with longer beaks relative to body size, found at lower latitudes and elevation, sing lower frequency songs (b, d)
Map showing localities where tinkerbirds were sampled, including (a) the collection localities of museum specimens, (b) field collected samples, (c) recording localities, and (d) examples of song spectrograms for the four species, from Song ID: MW10_68_10 for P. bilineatus; 1007_44337 for P. chrysoconus; 1022_4439 for P. pusillus and AN05_009_04 for P. subsulphureus. The background map represents the digital elevation model (3‐arc second) of Africa. Tinkerbird illustrations courtesy of Lynx Edicions (del Hoyo et al., 2020)
Plots showing the model effects of absolute latitude and elevation on body size as defined by PC1 in both field (a, b) and museum (g, h) measurements respectively as well as their respective effects on beak length after controlling for body size (d, e and j, k). Raw data plots with latitude are shown in the rightmost column (c, f, i, l), with elevation represented by the colour scale legend. Grey shaded area indicates 95% confidence interval
Variation in peak frequency along latitudinal and elevational gradients in (a) P. bilineatus, (b) P. chrysoconus and P. pusillus combined and (c) P. subsulphureus. Elevation is represented by the colour scale legend. The lower panels show model effects of (d) absolute latitude, (e) elevation and (f) EVI on peak frequency for the four species modelled together. Grey shaded area indicates 95% confidence interval. Tinkerbird illustrations courtesy of Lynx Edicions (del Hoyo et al., 2020)
Physiological constraints related to atmospheric temperature pose a limit to body and appendage size in endothermic animals. This relationship has been summarised by two classical principles of biogeography: Bergmann's and Allen's rules. Body size may also constrain other phenotypic traits important in ecology, evolution and behaviour, and such effects have seldom been investigated at a continental scale. Through a multilevel‐modelling approach, we demonstrate that continent‐wide morphology of related African barbets follows predictions of Bergmann's rule, and that body size mirrors variation in song pitch, an acoustic trait important in species recognition and sexual selection. Specifically, effects on song frequency in accordance with Bergmann's rule dwarf those of acoustic adaptation at a continental scale. Our findings suggest that macroecological patterns of body size can influence phenotypic traits important in ecology and evolution, and provide a baseline for further studies on the effects of environmental change on bird song. Temperature‐related physiological constraints shape functional traits in animals at a global scale, yet the implications of continent‐wide morphological variation on acoustic communication signals is little understood. This is despite a well‐established relationship between body size and acoustic signal frequencies across the animal kingdom. In a continent‐wide study on birds, we show that song frequency mirrors predictable patterns of body size variation along a latitudinal and elevational gradient, suggesting climate change may drive predictable shifts in the frequency of acoustic signals that play a critical role in animal ecology and evolution.
 
How the (a) Preston‐plot SAD results from the transformation by (b) the TAD of (c) the regional species‐density function (tick marks indicate discrete species). Flat parts of the TAD concentrate species density in the SAD, whereas sloped parts diffuse it (indicated by the width of the beams reflected off the TAD)
Our first example (one niche, quadratic fitness function, uniform regional TAD). Equilibrium (a) TADs and (b) SADs for four dispersal rates m$$ m $$. (c–e) the effect of dispersal on (c) total abundance N̂tot$$ {\hat{N}}_{\mathsf{tot}} $$, (d) width of the Lorentzian TAD w$$ w $$ and (e) abundance of 21 evenly spaced species between x=−1$$ x=-1 $$ and x=1$$ x=1 $$. Parameters: a=1,r*=1,x*=0,γ=1,nRx=100,ρx=100$$ a=1,{r}^{\ast }=1,{x}^{\ast }=0,\gamma =1,{n}_R(x)=100,\rho (x)=100 $$ so that NRx=1$$ {N}_R(x)=1 $$.
Our second example (one niche, quadratic fitness function, gaussian species‐density function and therefore gaussian regional TAD). (a) Species‐density function, (b) local TAD and (c) unimodal SAD. Parameters as in Figure 2 except σρ=3,ρ*=100,NR=1,m=10−4$$ {\sigma}_{\rho }=3,{\rho}^{\ast }=100,{N}_R=1,m={10}^{-4} $$.
Our third example (one niche, gaussian fitness function, uniform regional TAD). (a) TAD, (b) growth/exclusion rate, (c) SAD. Parameters as in Figure 2 except σr=1,m=10−5.$$ {\sigma}_r=1,m={10}^{-5}. $$
Our fourth example (localised competition, quadratic fitness function, uniform regional TAD). (a) the evolutionarily stable community reached in the absence of dispersal and the corresponding invasion‐fitness landscape. (b) TAD with dispersal and (c) the corresponding fitness landscape. (d) Continuum‐based SAD. (e) Discrete‐species‐based SAD. Parameters as in Figure 2 except: a0=1,σa=0.4.$$ {a}_0=1,{\sigma}_a=0.4. $$
Species‐abundance distributions (SADs) describe the spectrum of commonness and rarity in a community. Beyond the universal observation that most species are rare and only a few common, more‐precise description of SAD shape is controversial. Furthermore, the mechanisms behind SADs and how they vary along environmental gradients remain unresolved. We lack a general, non‐neutral theory of SADs. Here, we develop a trait‐based framework, focusing on a local community coupled to the region by dispersal. The balance of immigration and exclusion determines abundances, which vary over orders‐of‐magnitude. The local trait‐abundance distribution (TAD) reflects a transformation of the regional TAD. The left‐tail of the SAD depends on scaling exponents of the exclusion function and the regional species pool. More‐complex local dynamics can lead to multimodal TADs and SADs. Connecting SADs with trait‐based ecological theory provides a way to generate more‐testable hypotheses on the controls over commonness and rarity in communities.
 
The growth rate hypothesis (GRH) posits that variation in organismal stoichiometry (C:P and N:P ratios) is driven by growth‐dependent allocation of P to ribosomal RNA. The GRH has found broad but not uniform support in studies across diverse biota and habitats. We synthesise information on how and why the tripartite growth‐RNA‐P relationship predicted by the GRH may be uncoupled and outline paths for both theoretical and empirical work needed to broaden the working domain of the GRH. We found strong support for growth to RNA (r² = 0.59) and RNA‐P to P (r² = 0.63) relationships across taxa, but growth to P relationships were relatively weaker (r² = 0.09). Together, the GRH was supported in ~50% of studies. Mechanisms behind GRH uncoupling were diverse but could generally be attributed to physiological (P accumulation in non‐RNA pools, inactive ribosomes, translation elongation rates and protein turnover rates), ecological (limitation by resources other than P), and evolutionary (adaptation to different nutrient supply regimes) causes. These factors should be accounted for in empirical tests of the GRH and formalised mathematically to facilitate a predictive understanding of growth.
 
Distribution of species mean values of baseline blood glucose level (a), the intensity of blood glucose stress response (b) and stress‐induced blood glucose level (c) across passerines. The grey‐scale heat map represents sample size per species (N). Taxonomic families are separated by alternate background shading.
Allometric scaling (left), elevational variation (centre) and latitudinal differences in (right) baseline blood glucose level (a–c), the intensity of blood glucose stress response (d–f) and stress‐induced blood glucose level (g–i) in temperate (blue) and tropical (yellow) passerines. The right column shows differences in blood glucose measures between temperate and tropical passerines estimated over the range of species‐specific elevations available for both zones, with negative values indicating lower values in tropics. Predicted values and their CrI95 were obtained from the presented Bayesian phylogenetic models and are controlled for all the included covariates. Body mass is plotted on the natural logarithmic scale. The points represent species mean values calculated from the raw data and the point area is proportional to species sample size.
Directed acyclic graphs depicting our candidate path models (a–h; the labels identify the models in Figure 4c). The path analysis tested whether the effects of latitude (breeding zone; BZ) and species‐specific elevation (EL) on baseline blood glucose are direct and/or mediated by fecundity (clutch size; CS), while controlling for the effects of body mass (Mb). The grey arrows are established relationships that were present in all the models. The black arrows represent relationships that were tested in phylogenetic path analysis and thus varied among the models. The link between breeding zone and elevation was included to control for the differences in species‐specific elevation between tropical and temperate species in our data set.
The results of phylogenetic path analysis testing the causal relationships between latitude (breeding zone; BZ), species‐specific elevation (EL), clutch size (CS) and baseline blood glucose level, while controlling for the effects of body mass (Mb). Blue and red arrows in the directed acyclic graph (a) represent positive and negative relationships, respectively, with arrows pointing to response variables of the underlying regression models. The arrow widths are proportional to the log‐transformed standardized effect sizes (untransformed coefficients shown as decimals above the arrows) resulting from model averaging of all the path models. Light red arrows represent model‐averaged coefficients (b) with 95% confidence intervals including zero. The sum of weights of each causality chain (d, e) consists of weights of all the path models (c) that included the complete focal chain, that is, BZ → G0 or EL → G0 for direct effects and BZ → CS → G0 or EL → CS → G0 for effects mediated by clutch size. The model labels in part (c) refer to the candidate model labels in Figure 3.
Macrophysiological research is vital to our understanding of mechanisms underpinning global life history variation and adaptation to diverse environments. Here, we examined latitudinal and elevational variation in a key substrate of energy metabolism and an emerging physiological component of pace‐of‐life syndromes, blood glucose concentration. Our data, collected from 61 European temperate and 99 Afrotropical passerine species, revealed that baseline blood glucose increases with both latitude and elevation, whereas blood glucose stress response shows divergent directions, being stronger at low latitudes and high elevations. Low baseline glucose in tropical birds, compared to their temperate counterparts, was mainly explained by their low fecundity, consistent with the slow pace‐of‐life syndrome in the tropics. In contrast, elevational variation in this trait was decoupled from fecundity, implying a unique montane pace‐of‐life syndrome combining slow‐paced life histories with fast‐paced physiology. The observed patterns suggest that pace‐of‐life syndromes do not evolve along the single fast‐slow axis. Using data collected from 61 European temperate and 99 Afrotropical passerine species, we revealed that baseline blood glucose increases with both latitude and elevation, whereas blood glucose stress response shows divergent directions, being stronger at low latitudes and high elevations. Low baseline glucose in tropical birds, compared to their temperate counterparts, was mainly explained by their low fecundity, consistent with the slow pace‐of‐life syndrome in the tropics. In contrast, elevational variation in this trait was decoupled from fecundity, implying a unique montane pace‐of‐life syndrome combining slow‐paced life histories with fast‐paced physiology.
 
(a) Typical algal bloom dynamics, generated by model (1) with a fixed growth parameter. (b) 100 Monte Carlo simulations (thin lines) and mean behaviour (bold lines) of the same algal bloom model as in (a), but with a random variable describing growth (model (2)). The maximum bloom height (Pmax and maxtEP) and asymptotic nutrient levels (N* and EN*) are shown. In (b), the mean of the peak of each Monte Carlo simulation, EPmax, is also shown for comparison with maxtEP, the peak of the mean curve. Both are ecologically meaningful, but only the former is analytically computable. In (a), the growth rate is β=1, and in (b), the growth rate is a random variable B with a uniform distribution U0.6,1.4 with a mean of EB=1. Other parameters, α=0,γ=0.8,δ=0.1,η=0,n0=0.22,p0=0.001, consistent with the closed model described in the text.
Comparison of blooms with different peak heights Pmax and asymptotic nutrient levels N*. Green and yellow curves are from model (1) with a fixed growth parameter β. Purple and blue curves show EP and EN from 10,000 Monte Carlo simulations of (2), the model with random growth parameter B. Depending on model parameters, the model with a fixed growth rate underestimates (a) or overestimates (b) the maximum bloom height. (c) The model with a fixed growth rate underestimates the nutrients remaining after the bloom, and thus overestimates the amount of algae present during the bloom. Here B is uniformly distributed with mean β and width 2ν. Curves of increasing colour intensity for both EP and EN correspond to increasing values of ν, ranging through 0.1,0.2,…,0.4. (a) Parameters: α=0,β=1,γ=0.8,δ=0.1,η=0,n0=0.16,p0=0.001. (b, c) Parameters as in (a) except n0=0.22. Parameters in (a) and (b) were chosen to satisfy the conditions of (6) and (7), respectively.
(a) Comparison of (expected) total algae present throughout the algal bloom for the model with a fixed algal growth rate, β, and with a random growth rate, B. The difference in total algae between the two models increases as the variance in B increases. Here B is uniformly distributed with mean β and width 2ν; variance values shown correspond to ν=0.1 to 0.4. Parameters in (a) and (b) correspond to those in Figure 2a,b, respectively.
Mean and standard deviation of nutrient and algae dynamics approximated using three different uncertainty quantification methods for model (2) with the growth rate B as a random parameter; (a) 10,000 Monte Carlo simulations, (b) intrusive polynomial chaos expansion methods (n=6) and (c) nonintrusive polynomial chaos expansion methods (M=7 samples). Shaded regions show ± 1 standard deviation from the mean. Parameters: α=0,γ=0.8,δ=0.1,η=0,n0=0.22,p0=0.001. The growth rate B has a uniform distribution U0.8,1.2 with a mean of EB=1.
Statistics of the algal concentration Pt obtained using nonintrustive polynomial chaos. (a) Mean and ±1 standard deviation of Pt, along with the nominal trajectory (all parameters set to their mean values), and several realisations with randomly sampled parameters. (b) Coefficient of variation (CV) and skewness of Pt. (c) Global sensitivities of Pt, indicating the relative contributions of variability in the random parameters to variability in Pt. Most sensitivities not involving B and N0 are visually close to zero for time t≳15. Parameters: α=0.00075,γ=0.8,δ=0.1,η=0.0005. B is uniformly distributed on 0.6,1.4. N0 and P0 are lognormally distributed with respective mean values of 0.22 and 0.001, each with a CV of 0.3. The number of samples is M=1331, as described in S8.
There is often considerable uncertainty in parameters in ecological models. This uncertainty can be incorporated into models by treating parameters as random variables with distributions, rather than fixed quantities. Recent advances in uncertainty quantification methods, such as polynomial chaos approaches, allow for the analysis of models with random parameters. We introduce these methods with a motivating case study of sea ice algal blooms in heterogeneous environments. We compare Monte Carlo methods with polynomial chaos techniques to help understand the dynamics of an algal bloom model with random parameters. Modelling key parameters in the algal bloom model as random variables changes the timing, intensity and overall productivity of the modelled bloom. The computational efficiency of polynomial chaos methods provides a promising avenue for the broader inclusion of parametric uncertainty in ecological models, leading to improved model predictions and synthesis between models and data.
 
Non‐linear effects of temperature on consumption and respiration rates interact with resource density to shift the optimum temperature for population growth. The colours of the lines represent consumption at different resource availabilities, and mT represents the rate of respiration. Panel a) shows a family of curves representing different resource densities (scaled relative to the half‐saturation rate) and the respiration rate of a consumer. The difference between each consumption curve and the respiration curve gives the per‐capita growth rate (panel b) and shows a decline in the temperature which maximizes per‐capita growth. As the density of resources declines, the ingestion rate decreases, leading to changes in the intercepts and optimum of the TPC; the lower threshold temperature (Tmin) and upper threshold temperature (Tmax) converge under resource limitation and at the same time the optimum shifts toward lower temperatures due to the interplay of the non‐linear functions governing ingestion and maintenance. In panel c, the filled area represents combinations of temperature and resource density that generate a positive per‐unit growth rate and the black line represents the zero‐net‐growth‐isocline of the consumer. This isocline can be separated into the lower and upper intercepts of the thermal performance curves in panel b. this isocline is nearly symmetric about minimum, reflecting the strong impact of resources on Imax. The optimum temperature for growth at each resource density is given by the dotted line in panel c. here =25, β=150, =0.01, =0.1, =0.05, and δ=0.5.
Equilibrium densities of the consumer‐resource dynamic model when resources are supplied by a temperature‐independent chemostat for an efficient (a) and prudent (b) consumer. Here solid (dotted) lines correspond to stable (unstable) equilibria. We have omitted branches some unstable branches for clarity. Transcritical bifurcations occur where the two equilibrium branches exchange stability (e.g. at approximately 10 and 30.5 °C in panel a). The U‐shaped equilibrium branch for the resource population (filled in grey) is a rescaled version of the curve shown in Box Figure 1c. Higher values of R0 correspond to more prudent consumers which require higher resource density to achieve the same saturation of the functional response. The equilibrium densities are analytically solved in Appendix A1. Parameter values are as in Box Figure 1 and S=D=1.
The dynamics of the consumer and resource when the resource's dynamics are given by the logistic model with temperature‐sensitive parameters r and K for three values of R0 (0.5, 1.0 and 2.0). Stable equilibrium branches are shown as solid lines and unstable branches as dashed lines. We have omitted some unstable branches for clarity. The exchange of stability between equilibrium branches (ξi) occurs at transcritical bifurcations. The coexistence equilibrium (ξ+,+) loses stability (and regains) stability at Hopf bifurcation points which generate limit cycles (with amplitudes shown by the shaded regions). The U‐shaped equilibrium branch for the resource population (filled in grey) is a rescaled version of the curve shown in Box Figure 1c. Higher values of R0 correspond to more prudent consumers which require higher resource density to achieve the same saturation of the functional response. Parameters are: .
The effect of thermal asymmetry on the dynamics of consumers and resources for logistic resource model with R0=1. Panel a) the shaded areas depict which of the dynamical regimes is the stable outcome for each combination of temperature and thermal asymmetry (positive values shift the resource's thermal performance to warmer temperatures). In the unshaded area, neither resource nor consumer can persist (ξ0,0); in the lightly shaded area only the resource persists (ξ+,0); in the darkest region resources and consumers persist at a stable equilibrium (ξ+,+). No limit cycles emerge for these parameter values. For negative asymmetry, resources limit Tmax whereas consumption limits Tmin; for positive asymmetry, the results are reversed. Panel b) shows the realized thermal performance curves for three values of thermal asymmetry and panel c) shows the full range of thermal performance curves that are generated as thermal asymmetry varies. Parameter values are given in Figure 3.
Panels (a) and (b) show traces of the resource equilibria across temperatures for the scenarios shown in Figure 4 for (a) positive thermal asymmetry (∆T=+9°C) and (b) negative thermal asymmetry (∆T=−18°C). See Figure 2 for details on curves. For resource densities to continue to decline as consumers approach their thermal limits the upper (lower) limit must fall to the left (right) of the minimum of the consumer's thermal niche, which is given by the shaded region. These scenarios highlight the differences in the resource dynamics when consumer persistence is limited by temperature (Figure 3) vs. when it is limited by resources. Parameter values are as in Figure 3
Recent work has demonstrated that changes in resource availability can alter a consumer's thermal performance curve (TPC). When resources decline, the optimal temperature and breadth of thermal performance also decline, leading to a greater risk of warming than predicted by static TPCs. We investigate the effect of temperature on coupled consumer‐resource dynamics, focusing on the potential for changes in the consumer TPC to alter extinction risk. Coupling consumer and resource dynamics generally reduces the potential for resource decline to exacerbate the effects of warming via changes to the TPC due to a reduction in top‐down control when consumers near the limits of their thermal performance curve. However, if resources are more sensitive to warming, consumer TPCs can be reshaped by declining resources, leading to increased extinction risk. Our work elucidates the role of top‐down and bottom‐up regulation in determining the extent to which changes in resource density alter consumer TPCs. We investigate the relationship between resource density and temperature (warming) on the persistence of a consumer population. Our work elucidates the importance of jointly considering temperature and resource limitation when assessing the thermal performance of species. We demonstrate how knowledge of the thermal performance of a resource population can be used to generate realized consumer thermal performance curves.
 
Trajectories of alpine plant communities transplanted to different elevations in the Swiss Alps to simulate climate change. The communities originated from a site at 2050 m and were transplanted to sites at (a) 2000 m, (b) 1600 m, (c) 1400 m or (d) 1000 m (which correspond to approximately 0, 2.2, 3.3 and 5.5°C of warming respectively). Panels show a Principal Coordinates Analysis ordination based on Euclidean distances of all observed and simulated turfs from 2017 to 2020, including only the 11 taxa for which models adequately predicted cover changes. The arrows show the observed (black) and median predicted (grey) changes in community structure of each turf. Grey lines centred on the grey arrowheads display uncertainty in the predicted community changes, showing the 5% and 95% quantiles across 20 simulations for each principal component axis. In panel D, the thicker green arrows starting from the origin show the taxa for which the first two ordinations axes explain more than 90% of their cover variation through time. The inset plots within each panel show the mean observed (black) and predicted (grey) Shannon diversity across the ten turfs transplanted to each elevation; vertical lines show 95% Wald confidence intervals.
Trajectories of alpine plant community responses to contrasting climate change scenarios for the 21st century. (a) Depending on future greenhouse gas emissions, this region of the Alps could experience contrasting climate futures, warming from 0.5 (in RCP 2.6, solid line) to 2.3 (in RCP 4.5, dashed line) and potentially even 4.8°C (in RCP 8.5, dotted line) by the end of the century relative to the experimental period (2017–2020). (b) Increases in temperature under the contrasting climate change scenarios led to non‐monotonic changes in the diversity of the simulated alpine community. (c) Principal coordinates ordination based on Euclidean distances showing the trajectory of simulated alpine communities under the different climate change scenarios between years 2017 and 2098 (each point corresponds to a year in one scenario, in 3‐year time steps to reduce clutter). The grey arrows are vectors of taxa for which ordination axes explained more than 90% of variation in abundance through time. Panels (b) and (c) show the median of 20 simulated trajectories. The 5% and 95% quantiles are shown in panel (b) and ten runs of the simulations generating panel (c) are shown Figure S5.4.
Cover dynamics of alpine plant species under contrasting climate change scenarios RCP 2.6 (cyan), RCP 4.5 (orange) and RCP 8.5 (red). Lines show the natural logarithm of the sum of the cover of all individuals in each year of the climate change simulations. The trajectories of summer temperatures in the region under these scenarios are shown in Figure 2a. For scenario RCP 4.5, dotted lines show dynamics with a reduced influence of ecological lags, while dashed lines show dynamics following a stepwise change in climate to conditions expected by the end of the century (mean conditions between 2088 and 2098). Lines show the median of 20 simulated trajectories, and shading indicates the 5% and 95% quantiles.
Alpine community responses to gradual versus stepwise climate change. (a) Principal coordinates analysis ordination based on Euclidean distances showing the trajectory of simulated alpine communities under stepwise (dashed line) versus gradual climate change (RCP 4.5), with either full (solid line) or reduced (dotted line) influence of ecological lags. Each point corresponds to a year. The grey arrows are vectors of taxa for which ordination axes explained more than 90% of variation in abundance through time. (b) Trajectories of Shannon diversity based on taxa's total cover in simulated communities. Panels show the median of 20 simulated trajectories. Ten runs of the simulations generating panel A are shown in Figure S5.4, and 5% and 95% quantiles are shown with shading in panel (b).
Forecasting the trajectories of species assemblages in response to ongoing climate change requires quantifying the time lags in the demographic and ecological processes through which climate impacts species' abundances. Since experimental climate manipulations are typically abrupt, the observed species responses may not match their responses to gradual climate change. We addressed this problem by transplanting alpine grassland turfs to lower elevations, recording species' demographic responses to climate and competition, and using these data to parameterise community dynamics models forced by scenarios of gradual climate change. We found that shifts in community structure following an abrupt climate manipulation were not simply accelerated versions of shifts expected under gradual warming, as the former missed the transient rise of species benefiting from moderate warming. Time lags in demography and species interactions controlled the pace and trajectory of changing species' abundances under simulated 21st‐century climate change, and thereby prevented immediate diversity loss.
 
Understanding the factors affecting thermal tolerance is crucial for predicting the impact climate change will have on ectotherms. However, the role developmental plasticity plays in allowing populations to cope with thermal extremes is poorly understood. Here, we meta‐analyse how thermal tolerance is initially and persistently impacted by early (embryonic and juvenile) thermal environments by using data from 150 experimental studies on 138 ectothermic species. Thermal tolerance only increased by 0.13°C per 1°C change in developmental temperature and substantial variation in plasticity (~36%) was the result of shared evolutionary history and species ecology. Aquatic ectotherms were more than three times as plastic as terrestrial ectotherms. Notably, embryos expressed weaker but more heterogenous plasticity than older life stages, with numerous responses appearing as non‐adaptive. While developmental temperatures did not have persistent effects on thermal tolerance overall, persistent effects were vastly under‐studied, and their direction and magnitude varied with ontogeny. Embryonic stages may represent a critical window of vulnerability to changing environments and we urge researchers to consider early life stages when assessing the climate vulnerability of ectotherms. Overall, our synthesis suggests that developmental changes in thermal tolerance rarely reach levels of perfect compensation and may provide limited benefit in changing environments.
 
(a) Plant persistence was higher in restored sites, especially in late season whereas (b) pollinator persistence in the late season was higher in unrestored than in restored sites. Box plot represents median (mid line), interquartile range (box edges) and 1.5 × interquartile range (whiskers). Results shown for season‐level data
Neither nestedness (a) nor modularity (b) had an effect on plant persistence, whereas nestedness increased (c) and modularity decreased (d) pollinator persistence, and this effect was stronger in restored sites (green lines) than in unrestored sites (orange lines). Solid lines indicate the mean slope estimate from the GLMM model, and the shaded area represents the 95% confidence intervals around the estimate. Results shown for month‐level data
Species weighted closeness centrality had a clear, positive effect on mean persistence of both plants (a) and pollinators (b), and this effect was stronger in restored (green) than in unrestored (orange) sites. Weighted closeness centrality is scaled, solid line indicates the mean slope estimate from the GLMM model, and the shaded area represents the 95% confidence intervals around the estimate. Results shown for month‐level data
Past and recent studies have focused on the effects of global change drivers such as species invasions on species extinction. However, as we enter the United Nations Decade of Ecosystem Restoration the aim must switch to understanding how invasive‐species management affects the persistence of the remaining species in a community. Focusing on plant‐pollinator interactions, we test how species persistence is affected by restoration via the removal of invasive plant species. Restoration had a clear positive effect on plant persistence, whereas there was no difference between across treatments for pollinator persistence in the early season, but a clear effect in late season, with higher persistence in unrestored sites. Network structure affected only pollinator persistence, while centrality had a strong positive effect on both plants and pollinators. Our results suggest a hidden effect of invasive plants—although they may compete with native plant species, invasive plants may provide important resources for pollinators, at least in the short term.
 
The drivers of variability in species range sizes remain an outstanding enigma in ecology. The theoretical expectation of a positive dispersal‐range size relationship has received mixed empirical support, despite dispersal being one of the most prominent hypothesised predictors of range size. Here, we synthesised results from 86 studies examining the dispersal‐range size relationship for plants and animals in marine, terrestrial and freshwater realms. Overall, our meta‐analysis showed that dispersal positively affects range size, but its effect is dependent on the clade and dispersal proxy studied. Moreover, despite potential differences in habitat connectivity, we did not find an effect of realm on the dispersal‐range size relationship. Finally, the strength of the dispersal‐range size relationship was dependent on latitude, range size metric and the taxonomic breadth of the study clade. Our synthesis emphasizes the importance of developing a mechanistic understanding of the trait to dispersal to range size relationship, considering the complexity of dispersal departure, transfer and settlement, as well as evolutionary components such as time for range expansion, speciation and past geological–environmental dynamics. We, therefore, call for a more integrative view of the dispersal process and its causal relationship with range size.
 
A comparison of (a) simplicial set, (b) simplicial complex and (c) dyadic network representations of higher‐order interactions. Here, we present the example of social interactions of eight individuals (zero‐simplices). In (a) and (b), we illustrate simplices of different sizes in different layers and using different colours. For (a) the simplices present in the simplicial set are listed in the figure. (b) Represents the same set of higher‐order interactions but with additional lower‐order simplices due to a downward closure assumption, for example the presence of simplex (5,6,7,8) necessarily entails the presence of simplices (5,6,7), (5,6,8), (5,7,8), (6,7,8), (5,6), (5,7), (5,8), (6,7), (6,8), (7,8), (5), (6), (7) and (8). For (c) edge weights are the sum of the number of simplices (1—simplex or larger) that each individual is recorded in.
An illustration of how higher‐order representations can change our thinking about infection probabilities in non‐dyadic interactions. Here, we assume that each additional individual within an interaction is associated with a given infectious dose (r) with the probability of infection of a focal individual for a given dose given by the function σ(r). In a case where a single susceptible individual interacts with three infected individuals a naïve dyadic network approach would give an overall probability of infection 1‐(1‐σ(r))³ indicated by the blue line in the right‐hand figure, while the overall probability of infection if the higher‐order interaction is represented as a simplex σ(3r) is indicated by the red line. Red‐shading depicts the area in which the simplicial set approach predicts a higher probability of infection than the dyadic network approach, and blue shading the opposite. Full R code including the function for the dose–response curve σ(r) is provided in the supplementary materials.
(a) A simplicial complex built from the overlap of home ranges from simulated animal movements. Coloured polygons represent animal home ranges. Circles indicate nodes of the home range‐derived network and numbers provide an identification for the animal. Edges between nodes represent one‐simplices and (b) filled triangles represent two‐simplices as defined by the generalized Čech complex (i.e. where three home ranges overlap). Higher (than two) order simplices were not considered. (c) Displays a stochastic SIS simulation on the simplicial complex containing zero‐ and one‐simplices (red lines) and the simplicial complex containing zero‐, one‐ and two‐simplices (blue lines). The results of 25 simulations are shown for each of the red and blue lines over 50 units of time. The SIS model has three parameters: The force of infection across one‐simplices (β1 = 0.1 per time), the force of infection felt across two‐simplices (βΔ = 2 per time), and recovery rate (γ = 0.5 per time). A two‐simplex only contributes to the force of infection felt by a susceptible host if the two other individuals participating in the simplex are infected (as in Iacopini et al., 2019). Force of infection is additive across one‐ and two‐simplices (e.g. a susceptible individual participating in 2 one‐simplices with two infected hosts and a two‐simplex with the two infected hosts feels a force of infection of 2β + βΔ). In all simulations, node 2 starts as infected and all other nodes start as susceptible. (d) To test how the number of two‐simplices on the movement network affects the ability of an introduced pathogen to spread across the network, we randomly included 0 to 11 two‐simplices in a manner such that the simplicial set maintained downward closure. For example, when considering 3 two‐simplices we first identified all triangles defined by the zero‐ and one‐simplices shown in a. and then randomly selected three of these triangles to define our two‐simplices. We then simulated the SIS model on this simplicial complex for 50 time units and recorded whether the pathogen spread from node 2 to node 3. For each number of two‐simplices (a maximum of 11) we repeated this simulation 300 times and plotted the probability of spread as a function of the number of two‐simplices present in the complex. Code to generate this figure is provided in the supplementary materials.
(a) One realization of the conventional bipartite network representation of an assembly hole (left) and disassembly hole (right), compared with (b) the same topological structures illustrated as simplicial sets. The assembly hole consists of the simplices (1), (2), (3), (1,2), (1,3) and (2,3). The disassembly hole consists of the simplices (1), (2), (3) and (1,2,3). (c) and (d) Simplicial sets constructed for a hypothetical set of host individual‐microbe species interactions for (c) healthy hosts and (d) diseased hosts, with (e) and (f) comparisons of simplicial set topologies for the two types of host: (e) shows simplicial degree distributions for 0‐simplices in (c) and (d) separated into 1‐order (top), 2‐order (middle) and 3‐order (bottom) degrees; (f) shows the mean fill coefficient (Torres et al., 2021) for all simplices of order 2 or greater and the Betti numbers for the assembly and disassembly hypergraphs as defined by Angulo et al. (2021) and calculated in the Julia package CoexistenceHoles. Simplicial set c) has more 0‐simplices that feature in 2‐simplices, a higher fill coefficient and a smaller number of assembly and disassembly holes (especially 2‐dimensional holes) than simplicial set d). Full code to generate the elements of this figure are provided in the supplementary materials.
Network approaches have revolutionized the study of ecological interactions. Social, movement and ecological networks have all been integral to studying infectious disease ecology. However, conventional (dyadic) network approaches are limited in their ability to capture higher‐order interactions. We present simplicial sets as a tool that addresses this limitation. First, we explain what simplicial sets are. Second, we explain why their use would be beneficial in different subject areas. Third, we detail where these areas are: social, transmission, movement/spatial and ecological networks and when using them would help most in each context. To demonstrate their application, we develop a novel approach to identify how pathogens persist within a host population. Fourth, we provide an overview of how to use simplicial sets, highlighting specific metrics, generative models and software. Finally, we synthesize key research questions simplicial sets will help us answer and draw attention to methodological developments that will facilitate this.
 
Coexistence and evolutionary outcomes in relation to dietary overlap, the fraction of the habitat where species 2 could breed without interference from species 1, and the probability of species 1 winning interspecific fights (Pw1, based on the species difference in fighting ability). Each point represents a unique parameter set. In the top panels, symbols correspond to the observed coexistence outcomes (see legend). For example, ‘Coexist or Sp1 extinct’ means that the species coexisted to the end or species 1 went extinct, in different simulation runs. In the lower panels, symbols correspond to the evolutionary outcomes when both species persisted to the end (see legend): ‘Stasis’, neither species' mean values of z and μ changed significantly; ‘Sp2 diverged’, species 2 shifted away from species 1 and species 1 exhibited stasis; ‘divergence’, both species shifted away from the other; ‘Sp1 chased Sp2’, species 1 converged and species 2 diverged; ‘divergence/Sp2 diverged’, both outcomes occurred in different simulation runs; etc. This figure summarises a subset of the simulations in which the initial mean difference between the species in traits z and μ was 1. Figures S9–S12 summarise all of the simulations.
Probability of coexistence in relation to dietary overlap, the fraction of the habitat where species 2 could breed without interference from species 1 (colour scale), and the probability of species 1 winning interspecific fights (Pw1). The lines were generated by a non‐parametric smoothing function bounded by 0 and 1 (geom_smooth in the R package ggplot2). Here, the initial mean difference between the species in traits z and μ was 1. Figure S15 shows results for two additional levels of Pw1, and Figure S16 shows the probability of coexistence for simulations in which the initial mean difference was 2.
Probability of coexistence under high dietary overlap (≥0.75) in relation to the average probability of heterospecific males recognising each other as competitors in the final years of contact (i.e. just before one species went extinct or the simulation ended). Each point represents a unique parameter set. The colour scale indicates the fraction of the habitat where species 2 could breed without interference from species 1 (limited to the range shown in the legend). The lines were generated as in Figure 2. The initial mean difference between the species in traits z and μ was 1 or 2 in these simulations.
Theorists have identified several mechanisms through which species that compete exploitatively for resources could coexist. By contrast, under the current theory, interference competitors could coexist only in rare circumstances. Yet, some types of interference competition, such as interspecific territoriality, are common. This mismatch between theory and nature inspired us to model interference competition in an eco‐evolutionary framework. We based the model on the life cycle of territorial birds and ran simulations to examine whether natural selection could rescue a superior interference competitor from extinction without driving a superior exploitative competitor extinct. We found that coexistence between interference competitors can occur over a wide range of ecologically plausible scenarios, and up to the highest levels of resource overlap. An important caveat is that coexistence requires the species to co‐evolve. Reductions in population size and levels of genetic variation could destabilise coexistence between interference competitors, and thereby increase extinction rates over current estimates. How can species that engage in mutually costly forms of interference competition, such as territoriality, coexist? We approached this question with an eco‐evolutionary model based on territorial birds and found that territoriality can stabilize coexistence, but only if the species are able to coevolve. This is an important contribution to coexistence theory with real‐world implications.
 
Framework used to study the effects of environmental variability on fitness (stochastic growth rate lnλs). (a) Our calculations define demographic parameters as nonlinear functions of the environmental driver z (see methods), where A0 (from our selected, standardised COMADRE models, Ntot=154) defines the values of (st)age‐specific survival rates sj0 and fertilities fj0 in the mean environment (z=0). Different levels of environmental variance levels σz2 and environmental strength βz of z on the demographic parameters were considered. In the analytical approach, lnλs was calculated and decomposed into main components capturing nonlinearity and variance–covariance effects following Equation (3). The accuracy of this decomposition was tested using simulations (Supporting Information S4). (b) Two or three different link functions were considered for survival sjz and fertility fjz respectively. (c) Scenarios 1–8: Four combinations were examined including logistic functions for all parameters, loglog link functions for all parameters and two combinations of exponential fertilities fjz (log link) with logistic or loglog link function for sjz. Positive or negative covariance between survival and fertility was tested for each combination, assuming positive covariance between sjz, and between fjz. Scenarios 9–11: Scenarios of forced buffering considering demographic lability in the fertility coefficients and survival rates of the immature stages (Simmature). Scenarios 12–13: Scenarios of forced buffering assuming demographic lability in all survival rates sjz or in only the mature stages (Smature). Logistic functions were used to define lability while the other rates were held constant and fixed to the values reported in the standardised COMADRE projection matrix.
Mid panels: Stochastic growth rate (fitness) lnλs across generation time, under four scenarios of covariance and link functions of the demographic parameters. Left panels: Illustration of scenarios, with black and grey lines corresponding to the (st)age‐specific survival rates sjz and fertility coefficients fjz respectively (functions varied for each stage depending on sj0 and fj0; only one function is shown for survival and fertility here). We assumed positive covariance between survival rates of different (st)ages and between the fertilities of different (st)ages. For each scenario and for each population, positive (panels a, b) or negative (panels c, d) covariance between fjz and sjz were considered, treating fjz and sjz as logistic functions (panels a, c) or loglog link functions (panels b, d) of the environment z. Right panels: Decomposition of lnλs into main components capturing variance–covariance effects (blue triangles) and lability effects generated by nonlinear responses of fjz (red circles) and sjz (orange circles). Results for bony fish populations and populations with generation time >62 years are not shown (NMPMs=129; see Figure S14 for all MPMs).
Mid panels: Stochastic growth rate (fitness) lnλs across generation time, considering positive (panels a, b) or negative (panels c, d) covariance between (st)age‐specific survival rates sjz and fertilities fjz, treating sjz as logistic (panels a, c) or loglog (panels b, d) link functions of the environment z and fjz as log link functions. We assumed positive covariance between survival rates of different (st)ages and between the fertilities of different (st)ages. See Figure 2 for explanation of left and right panels. Results for bony fish populations and populations with generation time >62 years are not presented (NMPMs=129; see Figure S15 for all MPMs).
Results from scenarios of forced buffering assuming demographic lability only in (a) (st)age‐specific survival rates, (b) survival rates of the mature stages only, (c) (st)age‐specific fertilities and (d, e) fertilities and survival rates of the immature stages. For each scenario, the long term fitness lnλs and its main components reflecting variance–covariance effects (blue triangles) and lability effects due to nonlinearity of fjz (red circles) and sjz (orange circles) are plotted against generation time (mid and right panels; see Figure S16 for all MPMs). See Figure 2 for explanation of left panel.
Demographic buffering and lability have been identified as adaptive strategies to optimise fitness in a fluctuating environment. These are not mutually exclusive, however, we lack efficient methods to measure their relative importance for a given life history. Here, we decompose the stochastic growth rate (fitness) into components arising from nonlinear responses and variance–covariance of demographic parameters to an environmental driver, which allows studying joint effects of buffering and lability. We apply this decomposition for 154 animal matrix population models under different scenarios to explore how these main fitness components vary across life histories. Faster‐living species appear more responsive to environmental fluctuations, either positively or negatively. They have the highest potential for strong adaptive demographic lability, while demographic buffering is a main strategy in slow‐living species. Our decomposition provides a comprehensive framework to study how organisms adapt to variability through buffering and lability, and to predict species responses to climate change.
 
(a) Mean Hedge's g for plant size, damage tolerance (artificial herbivory), reproduction, herbivory (herbivore performance and herbivore damage combined) and chemistry most likely to contribute to qualitative defences for conspecific plants from native and non‐native ranges, (b) mean Hedge's g for damage done by specialist and generalist herbivores to plants and the performance of specialist and generalist herbivores for conspecific plants from native and non‐native ranges, (c) mean Hedge's g for damage done by all herbivores present naturally in field common gardens in non‐native and native ranges of plants for conspecifics from native and non‐native ranges, (d) mean Hedge's g for competitive effect (ability to suppress neighbours) and competitive response (ability to tolerate suppression by neighbours) for conspecifics from native and non‐native ranges. Bars show 1 SE and asterisks are presented for Hedge's g values that are significantly different than zero; that is when plants from one range show different responses than plants from the other range. Numbers in parentheses indicate the number of studies used and species used.
Mean Hedge's g for quantitative defences, including leaf chemical traits most related to quantitative defence (e.g. lignin, tannins) and the physical structure of leaves related to quantitative defence (primarily leaf‐specific mass; Hanley et al., 2007) for conspecifics from native and non‐native ranges. Bars show 1 SE and the asterisk is for the Hedge's g value that is significantly different than zero; that is when plants from one range show different responses than plants from the other range. Numbers in parentheses indicate the number of studies used and species used.
An important hypothesis for how plants respond to introduction to new ranges is the evolution of increased competitive ability (EICA). EICA predicts that biogeographical release from natural enemies initiates a trade‐off in which exotic species in non‐native ranges become larger and more competitive, but invest less in consumer defences, relative to populations in native ranges. This trade‐off is exceptionally complex because detecting concomitant biogeographical shifts in competitive ability and consumer defence depends upon which traits are targeted, how competition is measured, the defence chemicals quantified, whether defence chemicals do more than defend, whether ‘herbivory’ is artificial or natural, and where consumers fall on the generalist‐specialist spectrum. Previous meta‐analyses have successfully identified patterns but have yet to fully disentangle this complexity. We used meta‐analysis to reevaluate traditional metrics used to test EICA theory and then expanded on these metrics by partitioning competitive effect and competitive tolerance measures and testing Leaf‐Specific Mass in detail as a response trait. Unlike previous syntheses, our meta‐analyses detected evidence consistent with the classic trade‐off inherent to EICA. Plants from non‐native ranges imposed greater competitive effects than plants from native ranges and were less quantitatively defended than plants from native ranges. Our results for defence were not based on complex leaf chemistry, but instead were estimated from tannins, toughness traits and primarily Leaf‐Specific Mass. Species specificity occurred but did not influence the general patterns. As for all evidence for EICA‐like trade‐offs, we do not know if the biogeographical differences we found were caused by trade‐offs per se, but they are consistent with predictions derived from the overarching hypothesis. Underestimating physical leaf structure may have contributed to two decades of tepid perspectives on the trade‐offs fundamental to EICA.
 
(a) and (b) Growth rate as a function of food C:P ratio in the thermal gradient for clone US (a) and clone AR (b) of Daphnia magna. Dots are individual data, and the lines are the best fit to the Gaussian Function (Equation 1) estimated by nonlinear least squares regression, the maximum of the gaussian function is the C:P threshold elemental ratio (TERC:P). (c) and (d) TERC:P as a function of temperature and the 95% confidence intervals (CI95%) estimated by nonparametric bootstrapping (n = 1000) for clone US and clone AR, respectively. Lower‐case letters inside the graphs indicate homogeneous groups according to overlapping CI95%. Lines and dot colours represent different temperatures (see reference in figure)
Thermal response of ingestion rate (a, d, g, j, m), respiration rate (b, e, h, k, n) and growth rate (c, f, i, l, o), for Daphnia magna (AR) (a–c), Acartia tonsa (d–f), Carcinus maenas (Cadiz) (g–i), Oxyrrhis marina (j–l) and C. maenas (Helgoland) (m–o). Symbols are mean values and bars SE. In some cases, the error bars are not visible, because they are smaller than the symbols. Lower‐case letters inside the graphs indicate homogeneous groups according to the post hoc Holm‐Šídák multiple comparison test results
Thermal response of the body C:P ratios (a–e) and gross growth efficiency (GGE) of C and P (f–j) for all studied taxa. Dots are for body C:P, filled triangles are for GGEC and empty triangles for GGEP, symbols and bars are mean values and SE. In some cases, the error bars are not visible, because they are smaller than the symbols. Lower‐case letters inside the graphs indicate homogeneous groups according to the t‐test (a and f), post hoc Holm‐Šídák multiple comparison test results, or Kruskal‐Wallis ANOVA on ranks (i) for body C:P, and GGEP and upper‐case letters are for GGEC
C:P Threshold Elemental Ratio (TERC:P) as a function of temperature for (a) Daphnia magna (AR), (b) Acartia tonsa, (c) Oxyrrhis marina, (d) Carcinus maenas Cadiz and (e) C. maenas Helgoland. Symbols are mean values and bars SE. In some cases, the error bars are not visible, because they are smaller than the symbols. In (b–d), dashed vertical lines represent the mean MaxTERC:P temperature and the shaded areas the CI95% obtained through non‐linear regression fit of the thermal response of the TERC:P to a Gaussian function, with the R package nlstools (Baty et al., 2015). Dots and bars are mean values and SE. Lower‐case letters inside the graphs indicate homogeneous groups according to –‐tests (a), or the post hoc Holm‐ Šídák multiple comparison test results
Hypothesised response of the TERC:P to a broad temperature range. Light blue, green and orange areas, until the vertical dashed line, represent temperatures within the ecological environment of the species (is the experimental temperature range for A. tonsa, O. marina and C. maenas (C)). The red area represents temperatures beyond the thermal optimum (Pörtner, 2012; Pörtner & Farrell, 2008) that might be experienced by the species only in rare conditions (we suggest that this was the case of C. maenas (H) in the highest temperature in our experiments). In this hypothesised concept, the increasing demands of C relative to P when temperatures increase from cold (light blue) to intermediate (middle green) are the result of increasing respiration rates and P use efficiency, until the MaxTERC:P is reached. When temperatures increase above the normal thermal environment of the organism (from green to orange area), the increasing demands of P relative to C are the result of the decrease in P use efficiency, the higher Q10 of growth and ingestion in relation to respiration, and might prevent an excessive increase in metabolism that can result from the combination of low P diets (Ruiz et al., 2018; Ruiz et al., 2020) and increased temperatures. The increase in C demands relative to P when temperature is excessive may reflect the physiological stress that amplifies C‐demands for respiratory and catabolic processes (Schmitz, 2013) with higher Q10 of respiration than ingestion and growth
In light of ongoing climate change, it is increasingly important to know how nutritional requirements of ectotherms are affected by changing temperatures. Here, we analyse the wide thermal response of phosphorus (P) requirements via elemental gross growth efficiencies of Carbon (C) and P, and the Threshold Elemental Ratios in different aquatic invertebrate ectotherms: the freshwater model species Daphnia magna, the marine copepod Acartia tonsa, the marine heterotrophic dinoflagellate Oxyrrhis marina, and larvae of two populations of the marine crab Carcinus maenas. We show that they all share a non‐linear cubic thermal response of nutrient requirements. Phosphorus requirements decrease from low to intermediate temperatures, increase at higher temperatures and decrease again when temperature is excessive. This common thermal response of nutrient requirements is of great importance if we aim to understand or even predict how ectotherm communities will react to global warming and nutrient‐driven eutrophication.
 
Vital rates of the southern fulmar for each reproductive state and sea ice conditions (SICs). Vital rates are averaged for environments characterised by: High SICs (1979, 1998, 2001), low SICs (1986, 1987, 2000) and medium SICs (all other years), as defined by Jenouvrier et al. (2015). Colour bars refer to the 3 groups of unobserved heterogeneity (yellow: complex 1; orange: complex 2; and purple: complex 3), as well as the weighted average over the mixing distribution π = [0.14 0.67 0.19] (maroon). The panels are ordered by reproductive state at the previous breeding season (column 1: pre‐breeders (PB); column 2: successful breeders (S); column 3: failed breeders (F); and column 4: non‐breeders) and vital rates (first line: breeding probabilities β ; and second line: success probabilities given breeding γ ). Note that survival probabilities do not vary with time nor sea ice conditions, and thus are not shown here but in electronic Supplementary Material.
Individual differences in (a) breeding probabilities and (b) success probabilities given breeding across life history complexes for each set of sea ice conditions (SICs). Interindividual differences are measured by the coefficient of variation over the mixing distribution. The x‐axis indicates the reproductive state (see Figure 1 for legends) and the bar colours refer to SICs (low: Red, medium: Blue and high: Green).
Demographic outcomes of southern fulmar for each complex for each set of sea ice conditions, as if the population was permanently living in such an environment. Colour bars refer to the three groups of unobserved heterogeneity (yellow: complex 1; orange: complex 2; and purple: complex 3), as well as the weighted average over the mixing distribution π (maroon).
Percentages of time spent in each state for individuals in each complex for each set of sea ice conditions, as if the population was permanently living in such an environment: (a) high, (b) medium or (c) low. Complex 1 is shown by the left pie chart, while complex 3 is the right pie chart for each panel. Pre‐breeders are denoted: PB, successful breeders: S, failed breeders: F, and NB: non‐breeders.
Individuals differ in many ways. Most produce few offspring; a handful produce many. Some die early; others live to old age. It is tempting to attribute these differences in outcomes to differences in individual traits, and thus in the demographic rates experienced. However, there is more to individual variation than meets the eye of the biologist. Even among individuals sharing identical traits, life history outcomes (life expectancy and lifetime reproduction) will vary due to individual stochasticity, that is to chance. Quantifying the contributions of heterogeneity and chance is essential to understand natural variability. Interindividual differences vary across environmental conditions, hence heterogeneity and stochasticity depend on environmental conditions. We show that favourable conditions increase the contributions of individual stochasticity, and reduce the contributions of heterogeneity, to variance in demographic outcomes in a seabird population. The opposite is true under poor conditions. This result has important consequence for understanding the ecology and evolution of life history strategies.
 
Habitat complexity has been considered a key driver of biodiversity and other ecological phenomena for nearly a century. However, there is still no consensus over the definition of complexity or how to measure it. Up‐to‐date and clear guidance on measuring complexity is urgently needed, particularly given the rise of remote sensing and advent of technologies that allow environments to be scanned at unprecedented spatial extents and resolutions. Here we review how complexity is measured in ecology. We provide a framework for metrics of habitat complexity, and for the related concept of spatial heterogeneity. We focus on the two most commonly used complexity metrics in ecology: fractal dimension and rugosity. We discuss the pros and cons of these metrics using practical examples from our own empirical data and from simulations. Fractal dimension is particularly widely used, and we provide a critical examination of it drawing on research from other scientific fields. We also discuss informational metrics of complexity and their potential benefits. We chart a path forward for research on measuring habitat complexity by presenting, as a guide, sets of essential and desirable criteria that a metric of complexity should possess. Lastly, we discuss the applied significance of our review.
 
Climate change allows species to expand polewards, but non‐changing environmental features may limit expansions. Daylength is unaffected by climate and drives life cycle timing in many animals and plants. Because daylength varies over latitudes, poleward‐expanding populations must adapt to new daylength conditions. We studied local adaptation to daylength in the butterfly Lasiommata megera, which is expanding northwards along several routes in Europe. Using common garden laboratory experiments with controlled daylengths, we compared diapause induction between populations from the southern‐Swedish core range and recently established marginal populations from two independent expansion fronts in Sweden. Caterpillars from the northern populations entered diapause in clearly longer daylengths than those from southern populations, with the exception of caterpillars from one geographically isolated population. The northern populations have repeatedly and rapidly adapted to their local daylengths, indicating that the common use of daylength as seasonal cue need not strongly limit climate‐induced insect range expansions. Many species that expand polewards as a result of climate change need to adapt to latitudinal differences in daylength because daylength guides life history decisions. We describe two parallel northward range expansions of a butterfly in Sweden and show experimentally that range margin populations at both expansion fronts have adapted locally to daylength. This has happened despite theory suggesting constraints to local adaptation at range margins and implies that the need to evolve new responses to daylength cues need not hinder range‐expansions.
 
Definitions of eco‐evolutionary niche and competitive ability differences. (a) When rare species j (the single circle) invades a population of resident species i (squares), species i is adapted to conspecifics (the squares are black) and species j is adapted to heterospecifics (the circle is black). In the same way, when rare species i (single grey square) invades a population of resident species j (grey circles), species j is adapted to conspecifics and species i is adapted to heterospecifics. (b) The invasion growth rates need to be evaluated in the relevant evolutionary backgrounds for understanding mutual invasibility and stable coexistence. Rare species j can invade the community evolved to resident species i if αjii/αiii<1, where αji is a competition coefficient representing the per capita negative effects of species i on species j in the Lotka–Volterra competition model. The subscript after the vertical line represents the evolutionary background of the competition coefficients. (c) In this context, we can calculate the eco‐evolutionary niche difference (NDEE), −ln(ρEE), and competitive ability difference (FDEE), ln(κi/κjEE), based on the invasion growth rates for evaluating coexistence. (d) The calculated eco‐evolutionary niche and competitive ability differences can be used to predict the outcome of competition. Here, the x‐axis shows the eco‐evolutionary niche difference (NDEE), and y‐axis shows the eco‐evolutionary competitive ability difference (FDEE). Depending on the competition coefficients, there are four possible outcomes: (1) coexistence when NDEE is positive and the absolute value of FDEE is smaller than NDEE, (2) priority effect, where an initially common species excludes an initially rare species, emerging when NDEE is negative and the absolute value of FDEE is smaller than that of NDEE, (3) species i wins when FDEE is positive and the absolute value of NDEE is smaller than FDEE, and (4) species j wins when FDEE is negative and the absolute value of NDEE is smaller than that of FDEE.
Calculating eco‐evolutionary niche and competitive ability differences. (a) Eco‐evolutionary niche and competitive ability differences (NDEE and FDEE, purple point) can be calculated from the purely ecological versions of these quantities at two evolutionary end points (Appendix S1). More specifically, these metrics must be measured for competitors after the traits underlying their competition coefficients have evolved to the resident state where species i or j is dominant (red and blue points, respectively). The blue and orange lines delineate the four regions of competitive outcome. (b) Following Equations 4–5, for obtaining NDEE, we first calculate the mean value of the ND values (the vertical black line). (c) Then, if the competitive ability of species i increases when evolved to resident species j (the blue point is above the red point), half of the difference in the FD values (the vertical distance between points) is added to the mean ND. (d) For obtaining FDEE, we first calculate the mean value of the FD values (the horizontal black line). (e) Then, if the niche difference decreases when species i evolves to resident species j, half of the difference in the ND values (the horizontal distance between points) is subtracted from the mean FD.
Eco‐evolutionary dynamics in a simple haploid model with intransitive competition. (a) Intransitive competition amongst three genotypes ni, two of which belong to species 1, whilst the third belongs to species 2. n1 is competitively excluded by n3, n3 is excluded by n2, and n2 is excluded by n1 (the black arrows point to the pairwise competitive winner: Equations 2.I, 2.III). Arrows with embedded red and blue points show the competitive outcome at the two invader‐resident states shown in panel b in terms of niche and competitive ability differences. (b) Temporal shifts in the ecological niche and competitive ability differences in the simulated dynamics of the system are shown by the black curve. The coloured points connected by the grey curve indicate the ecological niche difference and competitive ability difference when species 1 has evolved to being the common resident (evolved to maximise intraspecific competition, red) or rare invader (evolved to maximise interspecific competition, blue), respectively. The right purple point EE represents the eco‐evolutionary niche difference (NDEE) and competitive ability difference (FDEE) predicting the competitive outcome. See Figure 2 and accompanying text for how to calculate the two metrices at the purple point from those at the red and blue points. Simulated time series of (c) the competing genotype densities, ni, and (d) species densities, Ni (solid lines) and allele frequency, p = n1/N1 (dashed line).
Eco‐evolutionary niche and competitive ability differences in the Vasseur et al. (2011) model of coexistence between an evolving species and non‐evolving species. (a) Based on empirical studies of plant competition mediated by allelochemicals (Lankau & Strauss, 2007), Vasseur et al. (2011) assumed adaptive evolution of one competitor's mean trait x¯, determining where it falls along a trade‐off between intraspecific and interspecific competition. Trait values θC (= 0) and θH (= 1) indicate trait values adapted to conspecifics and heterospecifics, respectively. The intraspecific competition coefficient of the non‐evolving species 1 (α11: a red line) is constant, whereas the intraspecific competition coefficient of the evolving species 2 (α¯22: a blue curve) and interspecific competition coefficient of the non‐evolving species 1 (α¯12: a magenta curve) increase and the interspecific competition coefficient of evolving species 2 (α¯21: a black curve) decreases as the mean trait value increases. See Appendix S2 for details of the model and parameters. (b) Simulated time series of competing species densities and an evolving trait in Figure 5h of Vasseur et al. (2011). Black, orange and blue lines show the evolving trait and population densities of non‐evolving species 1 and evolving species 2, respectively. The mean trait of the evolving species 2 moves to θC when dominant, which allows the invasion of the non‐evolving species 1. The mean trait evolves to θH when rare, and it favours the increase of the evolving species 2. (c) Niche difference (ND), −ln(ρ), and competitive ability difference (FD), ln(κ1/κ2), in eco‐evolutionary cycles of the two competing species (b). The simulation trajectory in panel b, shown by the black curve in panel c, connects the blue and red points, which indicate the niche difference and competitive ability difference when one species is dominant (x¯ = θC and θH, respectively).
Eco‐evolutionary niche and competitive ability differences in Mougi (2013) where both species coevolve. Mougi (2013) modelled coevolution between two competing species, again assuming a trade‐off between intraspecific and interspecific competition (mediated by allelochemicals). (a) Invasion growth rate of species 1 as a function of the traits of two coevolving species. It is maximised when species have evolved to species 1 being rare (the blue point) because at this point, species 1 evolves to heterospecifics (u1 = uH; meaning species 1 produces more allelochemicals effective against species 2) and species 2 evolves to conspecifics (u2 = uC; meaning species 2 produces less allelochemicals, which are costly and only effective against species 1). Species 1's invasion growth rate is minimised when evolved to being common (the red point) because at this point, species 1 evolves to conspecifics (u1 = uC) and species 2 evolves to heterospecifics (u2 = uH). The black point at the origin indicates the optimal traits when the two species are not interacting. The simulated coevolving trait dynamics in panel b is reproduced in top‐right. (b) Simulated time series of competing species densities and coevolving traits in Figure 1a of Mougi (2013). Solid and dashed lines represent population densities and traits, respectively. Orange and blue lines represent species 1 and 2, respectively. Here, larger trait values mean adaptation to heterospecifics (uH) and smaller trait values indicate adaptation to conspecifics (uC). See Appendix S3 for details of the model and parameters. When one species is common, it evolves to trait values (uC) that favour the invasion of the other species. Species dropping to rarity also evolve trait values (uH) that favour their recovery, causing the system to exhibit cyclic eco‐coevolutionary dynamics. (c) Niche difference (ND), −ln(ρ) and competitive ability difference (FD), ln(κ1/κ2), along the eco‐evolutionary cycles of the two competing species. The blue and red points indicate the niche difference and competitive ability difference when species 2 and 1 are common, respectively. The purple EE point indicating the eco‐evolutionary niche and competitive ability differences (based on Equations 4–5) predicts the stable coexistence observed in simulation.
Community ecology typically assumes that competitive exclusion and species coexistence are unaffected by evolution on the time scale of ecological dynamics. However, recent studies suggest that rapid evolution operating concurrently with competition may enable species coexistence. Such findings necessitate general theory that incorporates the coexistence contributions of eco‐evolutionary processes in parallel with purely ecological mechanisms and provides metrics for quantifying the role of evolution in shaping competitive outcomes in both modelling and empirical contexts. To foster the development of such theory, here we extend the interpretation of the two principal metrics of modern coexistence theory—niche and competitive ability differences—to systems where competitors evolve. We define eco‐evolutionary versions of these metrics by considering how invading and resident species adapt to conspecific and heterospecific competitors. We show that the eco‐evolutionary niche and competitive ability differences are sums of ecological and evolutionary processes, and that they accurately predict the potential for stable coexistence in previous theoretical studies of eco‐evolutionary dynamics. Finally, we show how this theory frames recent empirical assessments of rapid evolution effects on species coexistence, and how empirical work and theory on species coexistence and eco‐evolutionary dynamics can be further integrated.
 
Insect herbivores modulate flower maleness across the angiosperms. (a) Relationship between flower maleness and number of insect herbivore species (log2 x + 1), (b) phylogenetic distribution of flower maleness (inner bars) and number of insect herbivore species (outer bars) across 141 hermaphrodite angiosperm species belonging to 106 genera, 44 families and 22 orders. Flower maleness was calculated as the ratio between the dry biomass of androecium by the dry biomass of the primary sexual organs (androecium plus gynoecium). The thick black regression line was generated by phylogenetic generalised least square (PGLS) to consider the phylogenetic structure of the data, while the thin grey lines represent their uncertainty (1000 bootstrap estimates).
Evolutionary relationship between flower maleness and (a) number of families of insect herbivores, (b) number of feeding guilds and (c) diversity of feeding guilds. The thick black regression lines were generated by phylogenetic generalised least square (PGLS) to consider the phylogenetic structure of the data, while the thin grey lines represent their uncertainty (1000 bootstrap estimates). All explanatory variables were log‐transformed (log2 x + 1).
Evolutionary relationship between flower maleness and (a) number of monophagous herbivores, (b) number of family specialist herbivores and (c) number of generalist herbivores. The thick black regression lines were generated by phylogenetic generalised least square (PGLS) to consider the phylogenetic structure of the data, while the thin grey lines represent their uncertainty (1000 bootstrap estimates). All explanatory variables were log‐transformed (log2 x + 1).
Conceptual model on the evolutionary processes determining the positive interspecific pattern between outcrossing effort of the hosts and their long‐term selective pressure by natural enemies. The emergence of new attack strategies and host colonisation can increase the selective pressure by natural enemies while the emergence of defences and the extinction of natural enemies can decrease the selective pressure. When the selective pressure gets higher, outcrossing investment is selected positively, while when selective pressure gets lower, outcrossing investment is selected negatively. The shaded ellipse represents an area of higher evolutionary stability (despite the intrinsic dynamic of the system), where the host outcrossing effort is tuned with the selective pressure of the natural enemies. Below and above the ellipse, there are areas of lower evolutionary stability.
Why sex has evolved and is maintained is an open question in evolutionary biology. The Red Queen hypothesis predicts that host lineages subjected to more intense parasite pressure should invest more in sexual reproduction to continuously create novel defences against their rapidly evolving natural enemies. In this comparative study across the angiosperms, we show that hermaphrodite plant species associated with higher species richness of insect herbivores evolved flowers with higher biomass allocation towards the male sex, an indication of their greater outcrossing effort. This pattern remained robust after controlling for key vegetative, reproductive and biogeographical traits, suggesting that long‐term herbivory pressure is a key factor driving the selfing–outcrossing gradient of higher plants. Although flower evolution is frequently associated with mutualistic pollinators, our findings support the Red Queen hypothesis and suggest that insect herbivores drive the sexual strategies of flowering plants and their genetic diversity.
 
The approximate positions of elevational gradients in herbivory, carnivory and parasitism (a) and strength of elevational changes in the intensity of these interactions—both overall and in relation to the thermoregulation strategies of herbivores and carnivores (b). On the map, dot colours refer to different interactions (consult panel b for explanations); each dot may include several effect sizes (ES) calculated from the same gradient. On the graph, the negative ES indicates a decrease in the interaction intensity with an increase in elevation. Horizontal lines denote 95% confidence intervals; sample sizes (numbers of ES) are shown in parentheses. For statistical analysis, see text.
The strength of elevational changes in the intensity of trophic interactions in relation to gradient location in terms of the climate zone (for all interactions combined) and in relation to the tree line (for herbivory and for carnivory combined with parasitism). For other explanations, refer to Figure 1; for statistical analysis, see text.
Meta‐regression of the strength of elevational changes in intensity of all trophic interactions to the elevation span of the respective gradients (Q = 14.8, p = 0.0001).
Sources of variations in the elevational changes in the intensity of different kinds of herbivory. For explanations, refer to Figure 1; for statistical analysis, see text.
Sources of variations in the elevational changes in the intensity of background invertebrate folivory. For explanations, refer to Figure 1; for statistical analysis, see text.
The premise that the intensity of biotic interactions decreases with increasing latitudes and elevations is broadly accepted; however, whether these geographical patterns can be explained within a common theoretical framework remains unclear. Our goal was to identify the general pattern of elevational changes in trophic interactions and to explore the sources of variation among the outcomes of individual studies. Meta‐analysis of 226 effect sizes calculated from 134 publications demonstrated a significant but interaction‐specific decrease in the intensity of herbivory, carnivory and parasitism with increasing elevation. Nevertheless, this decrease was not significant at high latitudes and for interactions involving endothermic organisms, for herbivore outbreaks or for herbivores living within plant tissues. Herbivory similarly declined with increases in latitude and elevation, whereas carnivory showed a fivefold stronger decrease with elevation than with latitude and parasitism increased with latitude but decreased with elevation. Thus, although these gradients share a general pattern and several sources of variation in trophic interaction intensity, we discovered important dissimilarities, indicating that elevational and latitudinal changes in these interactions are partly driven by different factors. We conclude that the scope of the latitudinal biotic interaction hypothesis cannot be extended to incorporate elevational gradients.
 
Our survey was conducted as a series of decision points which distinguished different experiments as they relate to field contribution (rows), and a sequence of risk effects (columns). In regards to the risk effects (columns), an increase in risk leads to a risk‐induced trait‐response of the prey, which can affect a fitness component, then population growth rate, and then abundance of prey (NCE) or indirectly that of another species such as the prey's resource (TMIE). Nested groups (a–d) correspond to the categorical divisions described in the text. Group S is shown to highlight the difference between our review and that of Sheriff et al. (2020, see text).
(a) Three‐year running average of papers that examined NCEs, TMIEs and all ecology papers shown as a proportion of the total number of papers in each of the three categories over the 29‐year period (inset provides natural logarithm of the data to facilitate visualisation of relative changes through time). (b) Composition of risk effect papers from 1990 to 2018 using all three sub‐searches (n = 3945 total papers). Each square represents ten papers. NCE (c) and TMIE (d) papers categorised by field component metrics. Each square represents one paper (except for one paper that is represented twice having a PLP study of abundance and growth rate). Though PLP studies include a field component, the field component category in the figure does not include PLP studies as to not double count.
Proportion (y‐axis) and number (indicated) of responses in PLP studies categorised by the fitness component of the prey measured in NCE studies, and the fitness component of the resource measured in TMIE studies. Multiple rates were included for studies that measured 2 (six studies) or 3 (1 study) different fitness components (fitness component categorisation described in methods).
Proportion (width of a bar) of a given prey taxon for NCE and TMIE studies, with the sum of the widths being equal across rows. Four taxa of invertebrates were combined into one group (echinoderms, rotifers, annelids and arachnids). The ‘all’ category includes PLP studies; e.g., of 29 mammal NCE studies, 12 were PLP studies. The total number of studies for each category is given.
A narrative in ecology is that prey modify traits to reduce predation risk, and the trait modification has costs large enough to cause ensuing demographic, trophic and ecosystem consequences, with implications for conservation, management and agriculture. But ecology has a long history of emphasising that quantifying the importance of an ecological process ultimately requires evidence linking a process to unmanipulated field patterns. We suspected that such process‐linked‐to‐pattern (PLP) studies were poorly represented in the predation risk literature, which conflicts with the confidence often given to the importance of risk effects. We reviewed 29 years of the ecological literature which revealed that there are well over 4000 articles on risk effects. Of those, 349 studies examined risk effects on prey fitness measures or abundance (i.e., non‐consumptive effects) of which only 26 were PLP studies, while 275 studies examined effects on other interacting species (i.e., trait‐mediated indirect effects) of which only 35 were PLP studies. PLP studies were narrowly focused taxonomically and included only three that examined unmanipulated patterns of prey abundance. Before concluding a widespread and influential role of predation‐risk effects, more attention must be given to linking the process of risk effects to unmanipulated patterns observed across diverse ecosystems.
 
Reproductive dispersion (S) versus generation time (Tc) on a log–log scale. Time unit is years. Upper panel, age‐structured animal data from COMADRE and GO; middle panel, stage‐structured animal data from COMADRE; bottom panel, stage‐structured plant data from COMPADRE. Each panel displays the fitted model and its coefficient of determination (R2) based on PGLS regression. P‐value in each panel is less than 0.001. The 95% confidence interval for the regression slope is [0.92, 1.00] for the upper panel, [1.04, 1.09] for the middle panel and [1.11, 1.14] for the bottom panel. Pagel's λ is 0.63, 0.29 and 0.00 for the upper, middle and bottom panels respectively.
Damping time (τ) versus generation time (Tc) on a log–log scale. Damping time (τ) is calculated directly from each population projection matrix. Time unit is years. Upper panel, age‐structured animal data from COMADRE and GO; middle panel, stage‐structured animal data from COMADRE; bottom panel, stage‐structured plant data from COMPADRE. Each panel displays the fitted model and its coefficient of determination (R2) based on PGLS regression. P‐value in each panel is less than 0.001. The 95% confidence interval for the regression slope of each panel is [0.74, 0.83] for the upper panel, [0.75, 0.89] for the middle panel and [0.61, 0.70] for the bottom panel. Pagel's λ is 0.00, 0.13 and 0.27 for the upper, middle and bottom panels respectively.
Phylogenetic principal component analysis (PPCA) of reproductive dispersion (S), generation time (Tc) and damping time (τ) on a log scale for age‐structured animals. Arrow length indicates the loading of each life‐history trait onto PCA axes. Points represent the position of species along the phylogenetically corrected principal component (PPC)1 and 2, and are coloured by class. Numbers in parentheses on both axes represent the proportion of variance explained by the corresponding PPC. Results for stage‐structured animals and plants can be found in Supplementary Information.
Iteroparous species may reproduce at many different ages, resulting in a reproductive dispersion that affects the damping of population perturbations, and varies among life histories. Since generation time (Tc$$ {T}_c $$) is known to capture aspects of life‐history variation, such as life‐history speed, does Tc$$ {T}_c $$ also determine reproductive dispersion (S$$ S $$) or damping time (τ$$ \tau $$)? Using phylogenetically corrected analyses on 633 species of animals and plants, we find, firstly, that reproductive dispersion S$$ S $$ scales isometrically with Tc$$ {T}_c $$. Secondly, and unexpectedly, we find that the damping time (τ$$ \tau $$) does not scale isometrically with generation time, but instead changes only as Tcb$$ {T}_c^b $$ with b<1$$ b<1 $$ (also, there is a similar scaling with S$$ S $$). This non‐isometric scaling implies a novel demographic contrast: increasing generation times correspond to a proportional increase in reproductive dispersion, but only to a slower increase in the damping time. Thus, damping times are partly decoupled from the slow‐fast continuum, and are determined by factors other than allometric constraints.
 
Predicted effects of three major factors on the performance gain conferred by bacteria to plants associated with Mutualistic Fungi (MF) or Antagonistic Fungi (AF). (a) Interaction between MF‐bacterial traits: Fungi and bacteria that confer traits promoting distinct plant functions may lead to a greater gain in plant performance than symbionts that confer traits enhancing similar/identical plant functions (i.e. functionally distinct MF‐bacteria vs. functionally equivalent MF‐bacteria). (b) Type of protection traits conferred by bacterial symbionts: bacteria that confer both pathogen resistance and tolerance traits to plants interacting with AF may alleviate to a greater degree the AF's negative effect on plant performance than bacteria that confer these trait types separately. (c) Abiotic/biotic environmental stresses: bacteria that confer stress‐protective traits to their MF‐associated plant hosts may lead to a greater gain in plant performance in the presence of abiotic/biotic stresses compared to those plants in the absence of any stress.
The effect of bacteria on the performance of plants associated with Mutualistic Fungi (MF). The overall effect of bacteria on plants associated with MF was categorised into three subgroups depending on the interaction between fungal and bacterial traits (i.e. functionally distinct group = bacteria that add traits promoting dissimilar plant functions than fungi, functionally equivalent group = bacteria that add traits promoting identical/similar plant functions than fungi, while the third group included symbiotic associations with an unknown relationship). The first two subgroups were further categorised according to the specific plant functions that bacteria and fungi promoted. Those plant functions that were exclusively promoted by bacteria are italicised. The effects of just MF and bacteria on performance of symbiont‐free plants are also shown. An effect size with a positive value (95% confidence interval (CI) not overlapping zero) indicates a positive or beneficial effect of bacteria (relative to MF‐plants or symbiont‐free plants) or MF (relative to symbiont‐free plants) on the performance of host plants. For simplicity, the 95% CI of ‘growth’ and ‘stress protection’ categories are not fully shown. We refer to plant performance as measures of fitness including biomass, survival and seed production. Values in parentheses indicate the number of studies analysed.
The effect of bacteria on the performance of plants associated with Antagonistic Fungi (AF). The overall effect of bacteria on plants associated with AF was categorised into four subgroups depending on the types of protective traits provided by bacteria to their host plant against AF (i.e. resistance and tolerance traits, only resistance traits, only tolerance traits and unknown). The effects of just AF and bacteria on performance of symbiont‐free are also shown. An effect size with a positive value (95% confidence interval (CI) not overlapping zero) indicates a positive or beneficial effect of bacteria (relative to AF‐plants or symbiont‐free plants) or AF (relative to symbiont‐free plants) on the performance of their host plants, whereas a negative value indicates the opposite. We refer to plant performance as measures of fitness including biomass and disease resistance. Values in parentheses indicate the number of studies analysed.
Relationships between performance gains conferred by bacteria to plants associated with fungi (either mutualistic or antagonistic) in the presence and absence of environmental stresses. (a) Abiotic/biotic environmental stresses: The effect of bacteria on plant performance gain was categorised into the type of stress (i.e. abiotic or biotic stress). (b) Within biotic stress: The effect of bacteria on the plant performance gain was categorised into three subgroups depending on the types of protective traits conferred by bacteria to their host plant against biotic stresses (i.e. resistance and tolerance traits, only resistance traits, only tolerance traits). Each dot represents a single study (for details see studies listed in Table S1). The discontinuous black line is a reference that indicates the 1:1 relationship between plant performance gains (the proportional effect conferred by bacteria on their plant hosts, which are associated with fungi in the presence and absence of abiotic/biotic stresses). The continuous lines represent the linear models inferred from the GLM analyses (abiotic: y = 0.43 ± 0.24 × X + 0.56 ± 0.84; biotic: y = 1.60 ± 0.25 × X + 0.31 ± 1.11; with resistance and tolerance traits: y = 0.16 ± 0.49 × X + 9.10 ± 3.79; with resistance traits: y = 1.81 ± 0.04 × X − 0.56 ± 0.43; with tolerance traits: y = 0.14 ± 0.51 × X + 4.96 ± 2.51).
Plants interacting with mutualistic fungi (MF) or antagonistic fungi (AF) can form associations with bacteria. We assessed whether the performance gain conferred by mutualistic bacteria to fungal‐associated plants is affected by the interaction between symbiont traits, type of bacterial‐protective traits against AF and abiotic/biotic stresses. Results showed that (A) performance gain conferred by bacteria to MF‐associated plants was greater when symbionts promoted distinct rather than similar plant functions, (B) bacterial‐based alleviation of the AF's negative effect on plants was independent of the type of protective trait, (C) bacteria promoted a greater performance of symbiotic plants in presence of biotic, but not abiotic, stress compared to stress‐free situations. The plant performance gain was not affected by any fungal‐bacterial trait combination but optimised when bacteria conferred resistance traits in biotic stress situations. The effects of bacteria on fungal‐associated plants were controlled by the interaction between the symbionts' functional traits and the relationship between bacterial traits and abiotic/biotic stresses.
 
The epidemiological landscape can inform classic spatially explicit disease models. The epidemiological landscape (dark grey box; (a)–(d) with select mechanisms in white boxes defined and summarised in Table 1) consists of intrinsic attributes (a–c) and emergent interactions between the environment, hosts and pathogens that shape host and pathogen movements (decomposed into pathogen canonical activity modes, or PCAMs; (d) and locations of pathogen transmission. Information garnered from those movement trajectories can be used to inform inputs to classic spatially explicit disease models (light grey box). Conventionally, spatially explicit disease models relied on summary metrics that simply described host and pathogen locations, and did not link those locations to environmental attributes (white arrows in (e)). Movement ecologists are developing environmentally informed metrics (dark grey arrows in (e); Table 2) that could be used to adapt the classic modelling structures to changing environmental contexts, bringing the epidemiological landscape framework to full reality. ƛi is the per capita rate of infection at location i and Yj is the density of infected hosts at location j.
The movement‐pathogen pace of life hypothesis and expectations about spatial patterns of transmission. (a) Which components of the epidemiological landscape dominate spatial patterns of pathogen transmission depends on the interface between movement and pathogen life history. Pathogen canonical activity movements (PCAMs) can be broken into pure‐environment and host‐as‐environment modes, and the duration and ordering of these modes determines pathogen distribution across the landscape. Duration and ordering of PCAMs are in turn determined by two pathogen life‐history traits: first passage times in the host and in the environment. Pathogen pace‐of‐life increases down the dashed diagonal line, with the fastest pathogens in the lower left‐hand corner exhibiting rapid first passage times in both the host and the pathogen. Approximate locations of several pathogens are shown for orientation (M. ovi refers to Mycoplasma ovipneumoniae, an infectious pathogen of bighorn sheep). Spatial patterns of transmission for pathogens in the upper triangle will be dominated by the locations of host–host interactions, while spatial patterns of transmission for pathogens in the lower triangle will be dominated by interactions between the host and environmental reservoirs. (b) Pathogen life‐histories can be summarised through vectors defined by host and environment first‐passage times. The further vectors point to the left, the more transmission is driven by direct contacts; vectors extending further to the right are driven by indirect contacts. Pathogens whose vectors extend further to the bottom are expected to show transmission patterns driven by short‐term contacts (mass aggregations; mass blooms), while pathogens whose vectors extend further towards the top will be driven by long‐term patterns of host space use and density.
Hypothetical workflows using the epidemiological landscape. Each workflow uses animal movement trajectories to forecast movements from environmental covariates, adjusts movement forecasts according to time lags imposed by pathogen life‐history, and integrates with disease models (Box 2) to predict spatiotemporal transmission dynamics. Auxiliary data can enter the workflows (examples indicated by steps flagged with dark green arrows), but could limit model transportability. Many of the methods already exist (checkmarks), although it is not always clear how they should be connected. Other methods have been proposed and prototyped, but are as‐yet untested on real‐word data (blue triangles). A third group remains strictly hypothetical (red stars). ‘RSF’ = resource selection function; ‘SSF’ = step selection function; ‘CTMM’ = continuous time movement model; ‘UD’ = utilisation distribution; ‘CTMC’ = continuous time Markov chain. ‘MoveSTIR’ accounts for temporal lags between pathogen deposition and acquisition (Wilber et al., 2022). Method details are in Table 2.
Pathogen transmission depends on host density, mobility and contact. These components emerge from host and pathogen movements that themselves arise through interactions with the surrounding environment. The environment, the emergent host and pathogen movements, and the subsequent patterns of density, mobility and contact form an ‘epidemiological landscape’ connecting the environment to specific locations where transmissions occur. Conventionally, the epidemiological landscape has been described in terms of the geographical coordinates where hosts or pathogens are located. We advocate for an alternative approach that relates those locations to attributes of the local environment. Environmental descriptions can strengthen epidemiological forecasts by allowing for predictions even when local geographical data are not available. Environmental predictions are more accessible than ever thanks to new tools from movement ecology, and we introduce a ‘movement‐pathogen pace of life’ heuristic to help identify aspects of movement that have the most influence on spatial epidemiology. By linking pathogen transmission directly to the environment, the epidemiological landscape offers an efficient path for using environmental information to inform models describing when and where transmission will occur.
 
Map of the study region in California, USA, showing the distribution of all persistent mainland giant kelp, where each point location represents a 500‐m segment of coastline (n = 361; not to scale). Point Conception divides central California locations (blue) from southern California locations (red). Letters correspond with representative locations plotted in Figure 2
Giant kelp canopy biomass fluctuates synchronously over time, but patterns of synchrony differ by geography and timescale (i.e. period of fluctuations). Top panel shows quarterly giant kelp canopy biomass (log scale) from 1987 through 2019 for all locations (rows) ordered based on alongshore position (see Figure 1). Bottom panel shows biomass time series for three representative locations in each region (indicated by letters in top panel and in Figure 1)
Wavelet mean fields reveal synchrony of giant kelp canopy biomass in central (left) and southern (right) California across time and timescale. Top panels (a–b) show observed synchrony; nearly all features shaded cyan, green, and yellow are highly significant in wavelet phasor mean field testing of phase synchrony (p < 0.001; Appendix S3: Figure S3). Middle panels (c–d) show synchrony predicted by wavelet models based on wave disturbance (maximum wave height), nutrient availability (mean nitrate concentration), and their interaction. Bottom panels (e–f) summarise the top and middle panels by averaging observed (solid line) and model‐predicted (dashed line) timescale‐specific synchrony across all years (i.e. the mean squared synchrony), and comparing these observations and model predictions across timescales. Vertical lines separate timescale bands for annual (<2 y period), short interannual (2–4 y period) and long interannual (4–10 y period) synchrony. Note that the x‐axis shows the timescale of synchrony on a log scale
Wave disturbance, nutrient availability, and their interaction explain synchrony in giant kelp canopy biomass across timescales and regions. The black bar shows the total average percentage of synchrony explained by wavelet models (qall) and coloured bars show the partitioned contributions from maximum wave height (qwaves), mean nitrate concentration (qnutrients), and their interaction (qint). For each panel, coloured bars sum to the black bar. Values for the main effects of waves and nutrients are always positive, but can exceed 100% when offset by negative interactions (antagonism). Such antagonistic interactions indicate that the synchronising effects of waves and nutrients counteract one another to reduce synchrony below that attributable to additive effects alone. Likewise, positive interactions indicate synergy between waves and nutrients that enhance synchrony above that attributable to additive effects alone
Synchrony of giant kelp canopy biomass is partly explained by fluctuations in the North Pacific Gyre Oscillation (NPGO), an oceanographic climate index that corresponds with large‐scale strengthening of wind‐driven upwelling. Lines and axes as in Figure 3. Numbers at the top of each timescale band show the average percentage of synchrony explained by NPGO wavelet models within a given region and timescale band (qall; analogous to black bars for total synchrony explained in Figure 4)
Spatial synchrony is a ubiquitous and important feature of population dynamics, but many aspects of this phenomenon are not well understood. In particular, it is largely unknown how multiple environmental drivers interact to determine synchrony via Moran effects, and how these impacts vary across spatial and temporal scales. Using new wavelet statistical techniques, we characterised synchrony in populations of giant kelp Macrocystis pyrifera, a widely distributed marine foundation species, and related synchrony to variation in oceanographic conditions across 33 years (1987–2019) and >900 km of coastline in California, USA. We discovered that disturbance (storm‐driven waves) and resources (seawater nutrients)—underpinned by climatic variability—act individually and interactively to produce synchrony in giant kelp across geography and timescales. Our findings demonstrate that understanding and predicting synchrony, and thus the regional stability of populations, relies on resolving the synergistic and antagonistic Moran effects of multiple environmental drivers acting on different timescales. Spatial synchrony is a ubiquitous feature of population dynamics, but it is largely unknown how multiple environmental drivers interact to determine synchrony via Moran effects, and how these impacts vary across spatial and temporal scales. Using new wavelet statistical techniques, we characterized synchrony in populations of giant kelp, a widely distributed marine foundation species, and related synchrony to variation in oceanographic conditions. We discovered that disturbance and resources—underpinned by climatic variability—act individually and interactively to produce synchrony across geography and timescales, demonstrating that predicting regional population stability relies on resolving the synergistic and antagonistic Moran effects of multiple environmental drivers acting on different timescales.
 
Understanding the role of animal behaviour in linking individuals to ecosystems is central to advancing knowledge surrounding community structure, stability and transition dynamics. Using 22 years of long-term subtidal monitoring, we show that an abrupt outbreak of purple sea urchins (Strongylocentrotus purpuratus), which occurred in 2014 in southern Monterey Bay, California, USA, was primarily driven by a behavioural shift, not by a demographic response (i.e. survival or recruitment). We then tracked the foraging behaviour of sea urchins for 3 years following the 2014 outbreak and found that behaviour is strongly associated with patch state (forest or barren) transition dynamics. Finally, in 2019, we observed a remarkable recovery of kelp forests at a deep rocky reef. We show that this recovery was associated with sea urchin movement from the deep reef to shallow water. These results demonstrate how changes in grazer behaviour can facilitate patch dynamics and dramatically restructure communities and ecosystems.
 
A diagram depicting the networks considered in this study: One‐to‐one (e.g. species A → C); one‐to‐many (e.g. species A → C and D); many‐to‐one (e.g. species A and B → C); and many‐to‐many (e.g. species A and B → C and D)
World map depicting field site locations utilised in brood parasitism studies. The size of each bubble correlates with the number of studies conducted at each field site, with larger bubbles indicating that more studies were conducted
A breakdown of the number of studies: (a) investigating the 10 most frequently studied species of brood parasites; (b) by continent for each decade between 1981 and 2020, and studies pre‐1981; (c) by system type for each decade between 1981 and 2020, and studies pre‐1981. The continent of Antarctica is omitted as no brood parasites breed in the region. One‐to‐one refers to a system with one species of brood parasite and one species of host, one‐to‐many refers to a system with one species of brood parasite and multiple species of host, many‐to‐one refers to a system with multiple species of brood parasites and one species of host, and many‐to‐many refers to a system with multiple species of brood parasites and multiple species of hosts
Plots describing the relationship between system complexity measured as the linkage density of potential brood parasite–host systems for each respective land hexagon, and three metrics of species richness: (a) number of species of brood parasites; (b) number of species of hosts; (c) number of bird species. Each dot represents one land hexagon
Heatmaps describing (a) global patterns of brood parasite species diversity; (b) host species diversity; and (c) brood parasite–host system complexity. System complexity is measured as the linkage density of potential brood parasite–host systems for each respective land hexagon. Grey land areas represent regions where either no species of brood parasites (a and c) or hosts (b) occur
The relationships between avian brood parasites and their hosts are widely recognised as model systems for studying coevolution. However, while most brood parasites are known to parasitise multiple species of host and hosts are often subject to parasitism by multiple brood parasite species, the examination of multispecies interactions remains rare. Here, we compile data on all known brood parasite–host relationships and find that complex brood parasite–host systems, where multiple species of brood parasites and hosts coexist and interact, are globally commonplace. By examining patterns of past research, we outline the disparity between patterns of network complexity and past research emphases and discuss factors that may be associated with these patterns. Drawing on insights gained from other systems that have embraced a multispecies framework, we highlight the potential benefits of considering brood parasite–host interactions as ecological networks and brood parasitism as a model system for studying multispecies interactions. Overall, our results provide new insights into the diversity of these relationships, highlight the stark mismatch between past research efforts and global patterns of network complexity, and draw attention to the opportunities that more complex arrangements offer for examining how species interactions shape global patterns of biodiversity.
 
Net changes in cover. Changes in cover of species groups at the plot level with respect to pre‐treatment values (2006) for each year (2007–2020). Points show the change in cover (i.e. ‘hits’) versus 2006 for a given plot by each species group within each year (n = 6 plots per treatment × 4 groups × 14 years). Lines reflect modelled estimates of cover change by treatment type, species group and duration of treatment (number of years since 2006) with a random effect of calendar year with 95% confidence intervals plotted around line estimates (Table 1).
Density‐independent mechanisms. Posterior parameter estimates for responses of density‐independent growth rates (ρs) to standardised (mean 0 and unit variance) environmental covariates (temp‐average summer air temperature (°C), snow depth‐mean April snow depth (cm), N dep‐average summer nitrogen deposition (g/m²/year)). Points show mean estimates and error bars show 95% Bayesian credible intervals. We set wide priors on ρ coefficients (−0.5, 0.5) to allow a 50% change (increase or decrease) in ρs in response to a 1 SD change in a given environmental covariate at each time step. Estimates here reflect posterior sampling across all time steps. Estimates are standardised by environmental covariates within treatment as models were run separately for each treatment, thus the magnitude of parameter estimates and credible intervals should be compared between species groups within a treatment but not across treatments (colours). Rare species showed weak DI responses to environmental variables (Figure S7).
Density‐dependent mechanisms. Changes in the mean competitive interactions (Δαμ) of each species group in each global change treatment versus control. Intraspecific (intra) shows the mean change in the competition of a species group on itself (i.e. self‐limitation). Interspecific (inter) shows the sum of the mean changes of all other species groups on that group. Net is the combination of intra and interspecific changes within each treatment and species group. Values to the left, right of the dotted zero line signify that competition on a species group became stronger, weaker in global change vs control conditions, respectively. Raw pairwise α and Δαμ values are shown in Tables S2, S3 and Figure S8(a‐d).
Predictive steady‐state distributions. Estimated steady‐state abundance distributions (relative plot proportions) for each species group across the observed gradients of ambient (left) and augmented (right) N deposition (zero centred ± two standard deviations) in W and NW global change plots, respectively (see Figure S9 for all treatments). Points show simulated equilibrium abundances (100 at each value of ×). Dotted lines show the best model fit from a general additive model in the geom_smooth function in ggplot2 (Wickham, 2009).
Global change is altering patterns of community assembly, with net outcomes dependent on species' responses to the abiotic environment, both directly and mediated through biotic interactions. Here, we assess alpine plant community responses in a 15‐year factorial nitrogen addition, warming and snow manipulation experiment. We used a dynamic competition model to estimate the density‐dependent and ‐independent processes underlying changes in species‐group abundances over time. Density‐dependent shifts in competitive interactions drove long‐term changes in abundance of species‐groups under global change while counteracting environmental drivers limited the growth response of the dominant species through density‐independent mechanisms. Furthermore, competitive interactions shifted with the environment, primarily with nitrogen and drove non‐linear abundance responses across environmental gradients. Our results highlight that global change can either reshuffle species hierarchies or further favour already‐dominant species; predicting which outcome will occur requires incorporating both density‐dependent and ‐independent mechanisms and how they interact across multiple global change factors.
 
Effects of decade‐long (2011–2020) nitrogen additions on soil respiration in a larch forest in Northeast China. Rs, soil respiration. Error bars represent standard errors (n = 3) and the ‘*’ represents statistically significant at p < 0.05 (Table 1).
Shifts in the effect of nitrogen addition on soil respiration over time (2011–2020) under different nitrogen addition rates in a larch forest in Northeast China. (a) A linear decrease in the effect size of soil respiration under different treatments. (b) Response of soil respiration to N addition during different stages of the fertilisation experiment. The lines in panel A represent significant linear regression with 95% confidence intervals, and the different capital letters with slopes suggest statistically significant at p < 0.05. In panel (b), the different letters for the same stages of the experiment represent significant differences among the treatments at p < 0.05, and the error bars are the standard errors (n = 3).
The relationships between soil respiration and soil ectomycorrhizal fungi richness, soil saprotrophic fungi richness and ratio of soil saprotrophic to ectomycorrhizal fungi richness during a long‐term nitrogen addition experiment. The data for 2016 are shown in the upper panel (a–c), and the data for 2018 are shown in the lower panel (d–f). We only showed significant regression lines (p < 0.10) and their 95% confidence intervals (grey bands).
Relationship between soil respiration and fine root biomass during the late stage of the long‐term nitrogen addition experiment.
Increased nitrogen (N) inputs are widely recognised to reduce soil respiration (Rs), but how N deposition affects the temporal dynamics of Rs remains unclear. Using a decade‐long fertilisation experiment in a boreal larch forest (Larix gmelini) in northeast China, we found that the effects of N additions on Rs showed a temporal shift from a positive effect in the short‐term (increased by 8% on average in the first year) to a negative effect over the longer term (decreased by 21% on average in the 11th year). The rates of decrease in Rs for the higher N levels were almost twice as high as those of the low N level. Our results suggest that the reduction in Rs in response to increased N input is accelerated by high‐level N additions, and experimental high N applications are likely to overestimate the contribution of N deposition to soil carbon sequestration in a boreal forest.
 
Marine microbial communities are extremely complex and diverse. The number of locally coexisting species often vastly exceeds the number of identifiable niches, and taxonomic composition often appears decoupled from local environmental conditions. This is contrary to the view that environmental conditions should select for a few locally well‐adapted species. Here we use an individual‐based eco‐evolutionary model to show that virtually unlimited taxonomic diversity can be supported in highly evolving assemblages, even in the absence of niche separation. With a steady stream of heritable changes to phenotype, competitive exclusion may be weakened, allowing sustained coexistence of nearly neutral phenotypes with highly divergent lineages. This behaviour is robust even to abrupt environmental perturbations that might be expected to cause strong selection pressure and an associated loss of diversity. We, therefore, suggest that rapid evolution and individual‐level variability are key drivers of species coexistence and maintenance of microbial biodiversity.
 
Continental maps show the locations of sites used in this study (a: Effective population size and genetic differentiation (FST) estimates from Schmidt et al. data compiled from raw genotypes; b: MacroPopGen genetic diversity (gene diversity) estimates; c: TetraDENSITY population density records). One species was sampled at each site. Inset maps show site level values of genetic and demographic variables for select species. The size of points denotes site distance from the nearest biogeographic region boundary.
Model coefficients for the effect of distance from biogeographic boundary on population biodiversity variables. Open circles are global coefficient estimates; narrow and thick bars represent 95% and 90% credible intervals respectively. Pale points are the species‐specific coefficient estimates that underlie the global estimate, and their diameter denotes the number of sample sites included for that species. Effective population size and genetic diversity increase moving away from region boundaries while genetic differentiation and population density are higher closer to boundaries. Select species at the tails of the distributions of species‐specific effects are shown.
Global biodiversity is organised into biogeographic regions that comprise distinct biotas. The contemporary factors maintaining differences in species composition between regions are poorly understood. Given evidence that populations with sufficient genetic variation can adapt to fill new habitats, it is surprising that more homogenisation of species assemblages across regions has not occurred. Theory suggests that expansion across biogeographic regions could be limited by reduced adaptive capacity due to demographic variation along environmental gradients, but this possibility has not been empirically explored. Using three independently curated data sets describing continental patterns of mammalian demography and population genetics, we show that populations near biogeographic boundaries have lower effective population sizes and genetic diversity, and are more genetically differentiated. These patterns are consistent with reduced adaptive capacity in areas where one biogeographic region transitions into the next. That these patterns are replicated across mammals suggests they are stable and generalisable in their contribution to long‐term limits on biodiversity homogenisation. Understanding the contemporary processes that maintain compositional differences among regional biotas is crucial for our understanding of the current and future organisation of global biodiversity.
 
Ecological and evolutionary processes shape the structure of the cichlid‐Cichlidogyrus network consisting of cichlid fishes, a model system for explosive speciation research, and the parasitic flatworms belonging to Cichlidogyrus infecting the gills of cichlid and few non‐cichlid fishes. Species presented in the figure are Coptodon guineensis (Günther, 1862) and Cichlidogyrus gallus Pariselle & Euzet, 1995.
Cichlid‐Cichlidogyrus species network. (a) Whole network with unweighted links and Lake Tanganyika (LT), Lake Victoria regions (LV), and inferred meta‐communities (n ≥ 10) highlighted in colours. Circles indicate host and squares parasite species. Meta‐communities detected through the Louvain cluster algorithm include the ‘Coptodon zillii’ (CZ), ‘Oreochromis niloticus’ (ON), ‘Hemichromis’ (He) and ‘Tilapia sparrmanii’ (TS) clusters. Many small unconnected clusters belong to LT. (b) Chord diagrams of the geographically defined LT and LV clusters. (c) Inferred meta‐communities involving species of Cichlidogyrus with links weighted by number of observed infections. Unlike LT and LV, meta‐communities CZ, ON, He and TS are characterised by sampling bias towards few, economically relevant host species, for example Coptodon zillii, Oreochromis niloticus, Hemichromis fasciatus and Tilapia sparrmanii. Species names were omitted from (b) and (c) but are included in Appendix S5.
Changes of network metrics when only accounting for natural host repertoires and geographical ranges of cichlid‐Cichlidogyrus meta‐communities including Lake Victoria region (LV), ‘Oreochromis niloticus’ (ON), ‘Hemichromis’ (He) and ‘Coptodon zillii’ (CZ). Most values of the weighted nestedness based on overlap and decreasing fill (NODFw) (Almeida‐Neto & Ulrich, 2011), weighted connectance (Cw) (Bersier et al., 2002), specialisation asymmetry (SA) (Blüthgen et al., 2007), interaction evenness (Ei) (Bersier et al., 2002), and the standardised interaction diversity (H2’) (Blüthgen et al., 2006) that differed significantly from the null distributions (NM1, NM2) remained unchanged (see Appendix S2.2 for a detailed discussion).
Functional‐phylogenetic distances (FPDist) inferred from host repertoires of selected species of Cichlidogyrus calculated as mean pairwise distance (MPD) and mean nearest taxon distance (MNTD) weighted by abundancy of interactions (blue). FPDist matrices are a function of functional (FDist) and phylogenetic (PDist) distance matrices of the host species weighted by the parameter a. Shaded areas (grey) indicate 5% and 95% quantiles of 1000 null distributions resulting from taxon shuffling. If estimates fall outside the null distribution, they can be considered informative. Smaller values indicated higher functional‐phylogenetic similarities of host repertoires. A decreasing trend for FPDist estimates indicates that host communities are more phylogenetically than ecologically similar. For plots of other species of Cichlidogyrus infecting at least two host species, see Appendix S7.
Network link prediction based on host [H] and parasite [P] data in the cichlid‐Cichlidogyrus network and Lake Tanganyika (LT) and Lake Victoria regions (LV) meta‐communities including missingness map of input variables for whole networks (a), heat maps of host–parasite links (b), and bar plot of variable importances (c) predicted by the plug‐and‐play algorithm (Dallas et al., 2017). The missingness map illustrates significant gaps in the taxon coverage of phylogenetic data and host standard lengths. The heat maps show that a large proportion of cichlid‐Cichlidogyrus interactions likely remain undetected (highlighted in colour) (for taxon labels, see Appendix S7) although most interactions of the studied organisms are possibly already known for LT and LV. The variable importance graph indicates that the basins/basin types inhabited by the hosts are the most important predictor of cichlid‐Cichlidogyrus interactions, but less so for LT and LV.
Many species‐rich ecological communities emerge from adaptive radiation events. Yet the effects of adaptive radiation on community assembly remain poorly understood. Here, we explore the well‐documented radiations of African cichlid fishes and their interactions with the flatworm gill parasites Cichlidogyrus spp., including 10,529 reported infections and 477 different host–parasite combinations collected through a survey of peer‐reviewed literature. We assess how evolutionary, ecological, and morphological parameters determine host–parasite meta‐communities affected by adaptive radiation events through network metrics, host repertoire measures, and network link prediction. The hosts' evolutionary history mostly determined host repertoires of the parasites. Ecological and evolutionary parameters predicted host–parasite interactions. Generally, ecological opportunity and fitting have shaped cichlid‐Cichlidogyrus meta‐communities suggesting an invasive potential for hosts used in aquaculture. Meta‐communities affected by adaptive radiations are increasingly specialised with higher environmental stability. These trends should be verified across other systems to infer generalities in the evolution of species‐rich host–parasite networks. Many species‐rich ecological communities result from adaptive radiation events. We investigate interactions of African cichlids and their flatworm parasites belonging to Cichlidogyrus (a) through network analyses (b), host repertoire estimation, and network link prediction (heatmaps) (c). The hosts’ evolutionary history and environment determine observed host repertoires and network structure (b) but cichlid radiations in Eastern Africa have formed more specialised host‐parasite communities (c).
 
Mechanisms for pesticide resistance evolution in the insect. (A) Resistance acquisition via avoidance of the toxin, that is insecticides often fail to reach target insects under the leaf. (B) Reduce toxin penetrability through thickening of the insect cuticle. (C) Mutation in the binding site inside the target pest causes pesticide insensitivity. (D) Pesticide metabolism exploiting internal molecular machinery. I modifications may occur at the epigenetic level via DNA methylation or histone modification, leading to target gene expression alteration upon pesticide exposure. Epimutations are often heritable. II transcription factors (TFs) can modulate the expression of xenobiotic response elements, that is CncC‐Maf mediated xenobiotic response. III overexpression of phase I (i.e. Cyt P450s), phase II (i.e. GSTs), phase III (i.e. ABC transporters) enzymes can lead to detoxification or excretion of the entomotoxic pesticide molecules. (E) In‐house microbial symbionts can facilitate resistance development via detoxifying the toxic compound or facilitating the encapsulation of toxic molecules by activating the insect's immune system. (F) Single gene or multigene mutations can facilitate genetic resistance against pesticides.
Number of publications found on primary research on web of science investigating the proportion of studies on pesticide resistance with an ecology, evolution or genetic approach from 1985 to 2021. The searches were done by using the keywords “pesticide resistance” with one of the following keywords “Ecolog*”, “gene*” or “evolution*” (done in November 2021).
Scheme over pesticides' direct and indirect impacts on ecosystem functioning following cross‐resistance, from target species to non‐target species. Pesticides, the development of resistance due to their use and their potential side effects are represented in yellow. The impacts listed in the figure are not exhaustive.
Pesticide resistance development is an example of rapid contemporary evolution that poses immense challenges for agriculture. It typically evolves due to the strong directional selection that pesticide treatments exert on herbivorous arthropods. However, recent research suggests that some species are more prone to evolve pesticide resistance than others due to their evolutionary history and standing genetic variation. Generalist species might develop pesticide resistance especially rapidly due to pre-adaptation to handle a wide array of plant allelochemicals. Moreover, research has shown that adaptation to novel host plants could lead to increased pesticide resistance. Exploring such cross-resistance between host plant range evolution and pesticide resistance development from an ecological perspective is needed to understand its causes and consequences better. Much research has, however, been devoted to the molecular mechanisms underlying pesticide resistance while both the ecological contexts that could facilitate resistance evolution and the ecological consequences of cross-resistance have been understudied. Here, we take an eco-evolutionary approach and discuss circumstances that may facilitate cross-resistance in arthropods and the consequences cross-resistance may have for plant–arthropod interactions in both target and non-target species and species interactions. Furthermore, we suggest future research avenues and practical implications of an increased ecological understanding of pesticide resistance evolution.
 
Parameter inference. The raw data is an abundance time series (as obtained from direct observations or a numerical simulation of a properly parameterised model). Here this dataset of nt, nt+τ, nt+2τ, … is illustrated in panel (a). Each pair of adjacent points yields a single value Δn=nt+τ−nt. The mean value EΔn is plotted against n in panel (b). As expected, EΔn grows linearly with n, and the constant ℰ is identified from its slope. A graph of EΔn/n is shown in the inset, emphasising that this quantity is indeed n‐independent. VarΔn is plotted against n in panel (c). As expected, it takes a quadratic form, VarΔn=nVd+n2Ve, so VarΔn/n is a straight line whose slope gives Ve and the intercept with the vertical axis provides Vd. An analogous procedure leads to the parameters needed for the WKB‐based formula. The starting point is the time series zt+τ,zt+2τ,…, obtained from the time series for nt by the transformation zt=lnnt [panel (d)]. Each pair of points yields one value of Δz, and from these values one calculates EΔz [panel (e)]. Note that the parameter E0, corresponding to the mean growth rate of a rare population in coexistence theories (more precisely, Er=E0/τ), is the value of EΔz when n≫1, that is, when demographic stochasticity is negligible. For smaller values of n, EΔz=E0−Vd/2n, as demonstrated in the inset. To extract the values of EΔz, we filtered from the dataset all the cases where nt+τ=0; this yields an artificial positive bias when n is very small. Finally, the variance of Δz is shown (in panel (f)) to increase as n decreases, and the inset shows that it satisfies VarΔz=Ve+Vd/n. All the datasets were obtained for the individual‐based version of the lottery model (Chesson, 1982; Chesson & Warner, 1981) with τ=0.2, s0=0 and σe=0.25 (see Supplement S4 for model details).
Comparing the formulae. The chance of invasion, Π1→200, as obtained from Monte‐Carlo simulations of the individual‐based lottery model is plotted (blue circles, full lines) against the log amplitude of the environmental variations. These results are compared with the prediction of the diffusion approximation‐based formula, Equation (6) (red circles, dashed‐dotted lines) and the WKB‐based formula, Equation (20) (yellow diamonds, dashed lines), with n0=1, nf=200. The lottery model parameters were (a) τ=0.1, s0=0.1, N=106, and (b) τ=0.3, s0=0.2, N=106, where s0 indicates the mean fitness advantage of the invading species (see Supplement S4 for model details) and N is the total size of the whole community. The parameters required for the use of each formula were inferred from abundance time series generated using the model, following the procedure described in Figure 1. As is visible, the formula based on the diffusion approximation works well only for low levels of selection and environmental variations. In contrast, the WKB‐based formula agrees with the true chance of invasion for a much wider parameter range. Supplement S5 presents many other plots showing the same comparison for several values of τ∈0.1..0.7 and s0∈−0.2…0.4.
The chance of a single invader (n=1) to reach nf=200. This chance was obtained from numerical solution of the discrete‐time lottery model with demographic stochasticity (see Supplement S4) for various values of the model parameters τ (the dwell time) and σ (the amplitude of the temporal fitness variations). The colour of each filled circle indicates the value of σ, while its size is proportional to its τ‐value (between 0.1 and 1). In the left panel, the true chance of invasion is plotted against Er, demonstrating that Er is a poor metric: For a given value of Er, the true invasibility may have many different values, and when the model parameters τ and σ change (e.g. along the path indicated by the grey arrow), the invasibility may even decrease while Er increases. In the right panel, the true chance of invasion is plotted against the predictions of Equation (20), showing not only a data collapse but also close agreement with the numerical values. All the results were obtained for the case where there is no mean fitness advantage of the invading species, that is, s0=0 (Er is positive even in this case due to the storage effect), and with N=105. Similar graphs for nf=500 and for nf=20 are presented and discussed in Supplement S6, with the latter case providing a good example where Equation (6), based on the diffusion approximation, works better than Equation (20) based on WKB, due to higher accuracy of parameter extraction.
For the continuous‐time (Moran) version of the lottery model, the chance to reach nf=200 starting from varying n, as obtained from Monte Carlo simulations (circles), is compared, in panel (a), with the predictions of Πn→nfWKB from Equation (20) (full lines) for different values of s0, the mean selection parameter. In panel (b), the chance of invasion as found from the simulations is plotted against Er, with the colour of the points corresponding to their value of s0, and their size to the value of n. In panel (c), the same chance of invasion is plotted against Πn→nfWKB. As in Figure 3, all the observed (in simulations) data points collapse on a line when plotted against Equation (20). Other parameters are σ=0.3, τ=0.1 and N=5000. Similar graphs for different values of N are presented in Supplement S7. Even when s0 is negative, the curves in panel (a) may have the concave shape associated with a beneficial mutant. This happens because the storage effect (Chesson & Warner, 1981; Dean & Shnerb, 2020) may support invasion by an inferior species.
Chance of invasion in the Leslie–Gower forest dynamics model of Usinowicz et al. (2012) and Usinowicz et al. (2017), as described in Supplement S4. For a community size of N=10000, we calculated the chance to reach nf=1000, starting from varying n, and plotted the results against Er (panel a) and against the predictions of Equation (20) (panel b). The size of each point is proportional to the initial population n∈1..200, different colours correspond to different chances of adult tree survival d∈0.3..0.7 and different symbols represent different chances of sapling survival f∈0.5..0.9. In our simulations, we employed the empirical recruitment rates for Spondias mombin and Spondias radlkoferi, as provided in Usinowicz et al. (2012), scaled by a constant factor. A step‐by‐step description of the analysis is presented in Supplement S8.
Invasibility, the chance of a population to grow from rarity and become established, plays a fundamental role in population genetics, ecology, epidemiology and evolution. For many decades, the mean growth rate of a species when it is rare has been employed as an invasion criterion. Recent studies show that the mean growth rate fails as a quantitative metric for invasibility, with its magnitude sometimes even increasing while the invasibility decreases. Here we provide two novel formulae, based on the diffusion approximation and a large‐deviations (Wentzel–Kramers–Brillouin) approach, for the chance of invasion given the mean growth and its variance. The first formula has the virtue of simplicity, while the second one holds over a wider parameter range. The efficacy of the formulae, including their accompanying data analysis technique, is demonstrated using synthetic time series generated from canonical models and parameterised with empirical data.
 
A directed acyclic graph (DAG) representing the causal structure of a hypothetical ecological system.
Ecologists often rely on observational data to understand causal relationships. Although observational causal inference methodologies exist, predictive techniques such as model selection based on information criterion (e.g. AIC) remains a common approach used to understand ecological relationships. However, predictive approaches are not appropriate for drawing causal conclusions. Here, we highlight the distinction between predictive and causal inference and show how predictive techniques can lead to biased causal estimates. Instead, we encourage ecologists to valid causal inference methods such as the backdoor criterion, a graphical rule that can be used to determine causal relationships across observational studies.
 
Pearson correlation (r) between network size (defined by the product of the number of plant and pollinator species; i.e. rows·columns) and the three specialisation metrics of niche overlap, linkage density and mean normalised degree for 87 plant–pollinator networks.
Luna et al. (2022) concluded that the environment contributes to explaining specialisation in open plant–pollinator networks. When reproducing their study, we instead found that network size alone largely explained the variation in their specialisation metrics. Thus, we question whether empirical network specialisation is driven by the environment. Luna et al. (2022) concluded that the environment contributes to explaining specialisation in open plant–pollinator networks. When reproducing their study, we instead found that network size alone largely explained the variation in their specialisation metrics. Thus, we question whether empirical network specialisation is driven by the environment.
 
Visual representation of the various hypotheses (via simulated data), where yellow indicates high growth rates and blue low growth rates. (a) Truly asynchronous population cycles, where each population (line) cycles independently of its neighbours. (b) Partial synchrony where the neighbouring populations cycle almost simultaneously, though they are not perfectly synchronised (decomposed into subsequent models). (c) Perfectly synchronised populations where all populations cycle at precisely the same time (where rt,i$$ {r}_{t,i} $$ should best be represented as varying with time only, model N2). (d, e, g, h, j, k) All represent the spatial patterns on a given day for the various parameterisations of fitted travelling wave models, described in Table 1. (f) A purely spatial pattern (where any perceived spatio‐temporal pattern is merely spatial, model N3, as Sherratt & Smith, 2008 suggested may be the case for the apparent snowshoe hare travelling wave). (d) A single planar wave at a snapshot in time (Moss et al., 2000; Lambin et al., 1998, Bjørnstad et al., 2002, Berthier et al. 2014, model P). (e) Either an expanding or contracting single radial travelling wave (radially expanding from a central location as suggested by Johnson et al., 2006 [model RE] or contracting as suggested by Sherratt & Smith, 2008, [model RC]). (g) Two isolated planar waves separated by a physical feature, the Duero river (inferred from Sherratt & Smith, 2008, model PF). (h) Two radial waves separated by the same physical feature but may be either contracting or expanding (models RFE and RFC). (j) Dual overlapping planar waves, which additively form a single overall pattern (model PD). (k) Either dual overlapping contracting or expanding radial waves, additively forming an overall pattern (models RDE and RDC). (i) The modelling approach of an underlying cycle manifesting itself in the form of partial asynchrony in the data. The graphical pathway of analysis for the selected model (RDE) would be b →$$ \to $$ k →$$ \to $$ i, where k →$$ \to $$ i is carried out according to model RDE in Table 1.
Left: Map of the Duero basin, coloured (grey scale) according to elevation (m) with mountains in the north, east and south visible as white regions. The Bay of Biscay is visible in the north. Points represent the centroid locations and are coloured according to the population growth rate and sized according to the number of transects in each centroid. Estimated epicentre locations are noted with the black dots with white edges with their respective 95% CI profiles shown with the white polygons (the epicentre to the southeast is the activator, while inhibitor is in the north west). The Duero River is visible as the turquoise line running east to west. Elevation data were downloaded from copernicus[dot]EU (EU‐DEM v1.1) and waterway data from Ea[dot]europa[dot]eu. Top right: Time series of population growth rates for each centroid, is similarly coloured according to temporal periods. The horizontal dashed line shows a growth rate of zero (i.e. no growth). Bottom right: A histogram showing the number of transects contained in each centroid. The vertical dashed line shows the mean of 18.5.
Conditional predictions, showing the contribution of the 1st (slow) and 2nd (fast) waves, including the intercept, to mean growth rate (rit¯$$ \overline{r_{it}} $$, y‐axis) over space‐modified time (x‐axis) represented by the solid black lines, with 95% confidence intervals represented by the grey ribbons. Horizontal black dashed line indicates a growth rate of 0. The light grey points represent the partial residuals for the respective smoothing spline.
The dynamics of cyclic populations distributed in space result from the relative strength of synchronising influences and the limited dispersal of destabilising factors (activators and inhibitors), known to cause multi‐annual population cycles. However, while each of these have been well studied in isolation, there is limited empirical evidence of how the processes of synchronisation and activation–inhibition act together, largely owing to the scarcity of datasets with sufficient spatial and temporal scale and resolution. We assessed a variety of models that could be underlying the spatio‐temporal pattern, designed to capture both theoretical and empirical understandings of travelling waves using large‐scale (>35,000 km²), multi‐year (2011–2017) field monitoring data on abundances of common vole (Microtus arvalis), a cyclic agricultural rodent pest. We found most support for a pattern formed from the summation of two radial travelling waves with contrasting speeds that together describe population growth rates across the region.
 
Journal metrics
$4,800 / £3,200 / €4,000
Article Processing Charges (APC)
20%
Acceptance rate
11.274 (2021)
Journal Impact Factor™
13.6 (2021)
CiteScore™
Top-cited authors
Ingolf Steffan-Dewenter
  • University of Wuerzburg
Claire Kremen
  • University of California
Robert K Colwell
  • University of Connecticut
Alexandra Klein
  • University of Freiburg
Eric William Seabloom
  • University of Minnesota Twin Cities