Community ecology — interactions and food webs

Intuition [Beginner]

A community is all the species that live together in the same place at the same time. A forest community includes trees, shrubs, insects, birds, mammals, fungi, and soil bacteria. These species do not live in isolation — they interact constantly.

Species interactions come in four basic types, defined by their effects on the two species involved:

Competition ($-/-$): both species are harmed. Two bird species competing for the same nesting sites each have fewer sites available than they would alone.

Predation ($+/-$): one benefits, the other is harmed. A wolf eats a deer. Parasitism is a special case where the "predator" lives on or in the host, usually without killing it immediately.

Mutualism ($+/+$): both species benefit. Bees get nectar from flowers; flowers get pollinated by bees.

Commensalism ($+/0$): one benefits, the other is unaffected. Barnacles growing on a whale get a place to live; the whale is neither helped nor harmed.

A niche is the role a species plays in the community — the resources it uses, the habitat it occupies, and its interactions with other species. The competitive exclusion principle states that two species cannot coexist if they occupy exactly the same niche. One will always outcompete the other. In nature, coexisting species always differ in at least some aspect of their niche.

Community boundaries are not always sharp. A forest edge grades into a meadow; a tide pool connects to the open coast through water exchange. Ecologists define communities pragmatically: a community is the set of species found in a specified area at a specified time. The spatial scale matters enormously. A single rotting log hosts a community of beetles, fungi, and bacteria. A hundred-square-kilometre forest contains thousands of such micro-communities nested within the larger forest community. The interactions within each scale — competition for space on the log, predator-prey dynamics in the canopy, nutrient cycling across the whole forest — all contribute to the patterns community ecology seeks to explain.

Visual [Beginner]

A food web shows who eats whom in a community. Arrows point from prey to predator.

A trophic cascade occurs when a change at one level ripples through the food web. If the top predator is removed, its prey increases, which then depletes the next level down, which then allows the level below that to flourish. The effects cascade down the food chain. Cascades can also flow upward: adding nutrients to a lake increases algae, which feeds more zooplankton, which supports more fish. Whether the cascade is top-down or bottom-up, the principle is the same — species at one trophic level indirectly affect species two or more links away.

Worked example [Beginner]

Robert Paine's classic experiment (1966) demonstrated the concept of a keystone species. On the rocky coast of Washington State, Paine manually removed the starfish Pisaster ochraceus from experimental plots while leaving control plots undisturbed.

Before removal, the intertidal community contained about 15 species of invertebrates and algae. Pisaster is a top predator that feeds preferentially on mussels (Mytilus), which are dominant competitors for space on the rocks.

After Pisaster removal, mussels were freed from predation and rapidly expanded, covering the rock surface. They outcompeted barnacles, limpets, snails, and algae for space. After 2-3 years, the community had declined from 15 species to about 8 species. The mussels had competitively excluded most other inhabitants.

Pisaster is a keystone species: its impact on the community is disproportionately large relative to its abundance. Removing it transforms the community from diverse to dominated by a single species. Keystone species maintain diversity by preventing competitive exclusion — in this case, by eating the dominant competitor and creating open space for other species.

Check your understanding [Beginner]

Formal definition [Intermediate+]

Interaction types and sign matrices

A community interaction matrix $\mathbf{A}$ encodes pairwise effects:

$$A_{ij}=\{\begin{array}{ll}+& \text{if species }j\text{ benefits from species }i\\ -& \text{if species }j\text{ is harmed by species }i\\ 0& \text{if no direct effect}\end{array}$$

Interaction	$A_{ij}$	$A_{ji}$	Example
Competition	$-$	$-$	Two plants for light
Predation/Parasitism	$+$	$-$	Wolf-deer
Mutualism	$+$	$+$	Bee-flower
Commensalism	$+$	$0$	Barnacle-whale
Amensalism	$-$	$0$	Antibiotic-producing fungus

Competitive exclusion and coexistence

The Lotka-Volterra competition model (from 19.09.01) gives conditions for coexistence. More generally, Tilman's resource competition theory (1982) shows that at equilibrium, each species is limited by the resource for which it requires the lowest level to survive ($R^{\ast }$ rule). The species with the lowest $R^{\ast }$ for each resource wins competition for that resource. For $n$ species to coexist, there must be at least $n$ limiting resources or niches.

The competitive exclusion principle formalised: if $n$ species compete for fewer than $n$ limiting resources, at least one will be excluded at equilibrium.

Food web metrics

Key metrics characterise food web structure:

• Connectance $C=L/S^2$, where $L$ is the number of trophic links and $S$ is species richness.
• Mean food chain length: average number of links from a basal species to a top predator.
• Omnivory: fraction of species that feed at more than one trophic level.

Empirical food webs show consistent patterns: connectance decreases with species richness ($C\propto S^{-1}$ approximately), chain lengths are typically 3-5 links, and omnivory is common.

Counterexamples to common slips

• A food web is not a random graph. Real food webs have non-random structure: predators consume prey within a constrained size range, specialists feed on subsets of the prey consumed by generalists, and many potential links are never realised. Drawing a food web as a random directed graph with the same $S$ and $L$ would produce a network with fundamentally different stability properties.

• "Who eats whom" is not the only interaction. Food webs depict trophic links, but communities include non-trophic interactions: competition for shared resources, habitat modification (ecosystem engineers like beavers), and indirect effects such as apparent competition where two prey species share a predator. A complete interaction web includes all pairwise effects, not just consumption.

• Connectance $C=L/S^2$ uses $S^2$, not $S(S-1)/2$. The denominator counts all possible directed links (every species could potentially eat every species, including cannibalism). Some authors use $L/[S(S-1)]$ excluding self-links, but the $S^2$ form is the standard in the May-Pimm tradition.

• High diversity does not automatically mean high stability. May's (1972) result showed that, for randomly assembled communities, increasing species richness destabilises the system. The stability of real diverse communities arises from their non-random structure, not from diversity per se.

Biodiversity indices

The Shannon diversity index:

$$H^{\prime }=-\sum _{i=1}^S{p}_i\ln p_i,$$

where $p_i$ is the proportional abundance of species $i$ and $S$ is species richness. $H^{\prime }$ captures both richness and evenness: a community with many equally abundant species has higher $H^{\prime }$ than one dominated by a single species.

The Simpson index:

$$D=1-\sum _{i=1}^S{p}_i^2,$$

which gives the probability that two randomly sampled individuals belong to different species.

Key theorem with proof [Intermediate+]

Theorem (Competitive exclusion from the Lotka-Volterra model). In the two-species Lotka-Volterra competition model, species 1 excludes species 2 if and only if $K_1>K_2/{\alpha }_{21}$ and $K_2<K_1/{\alpha }_{12}$. Coexistence occurs if and only if $K_1>K_2/{\alpha }_{21}$ and $K_2>K_1/{\alpha }_{12}$ simultaneously (which simplifies to ${\alpha }_{12}{\alpha }_{21}<1$ when $K_1=K_2$).

Proof. The zero-growth isoclines are: for species 1, $N_1=K_1-{\alpha }_{12}{N}_2$; for species 2, $N_2=K_2-{\alpha }_{21}{N}_1$. These are straight lines in the $(N_1,N_2)$ phase plane.

Case 1: Species 1 wins. The species 1 isocline lies entirely above the species 2 isocline. This occurs when $K_1>K_2/{\alpha }_{21}$ (the x-intercept of the species 1 isocline exceeds that of species 2) and $K_1/{\alpha }_{12}>K_2$ (the y-intercept of species 1 projected exceeds $K_2$). Then ${\dot{N}}_1>0$ everywhere that $N_2$ is positive, and species 1 drives species 2 to extinction.

Case 2: Coexistence. The isoclines cross in the positive quadrant. This requires: the species 1 isocline has a higher x-intercept ($K_1>K_2/{\alpha }_{21}$) and the species 2 isocline has a higher y-intercept ($K_2>K_1/{\alpha }_{12}$). At the intersection, both ${\dot{N}}_1=0$ and ${\dot{N}}_2=0$, and the equilibrium is stable (both eigenvalues of the Jacobian are negative) because intraspecific competition exceeds interspecific competition. $\square$

The condition ${\alpha }_{12}{\alpha }_{21}<1$ (for equal $K$) has a biological interpretation: each species must inhibit itself (through intraspecific competition) more than it inhibits the other species. Niche differentiation ensures this — when species use different resources, competition between them is weaker than competition within each species for its own specialised resource.

Bridge. The Lotka-Volterra coexistence condition builds toward 20.05.02 pending ecosystem ecology, where multi-species competition models become the biotic engine of nutrient cycling and energy flow. The condition ${\alpha }_{12}{\alpha }_{21}<1$ appears again in 19.09.01 population ecology as the two-species instance of the general principle that intraspecific regulation must exceed interspecific effects for coexistence. The foundational reason coexistence requires niche differentiation is exactly the competitive exclusion principle: without it, one species inevitably draws the shared resource below the other's $R^{\ast }$ threshold. Putting these together with Tilman's resource competition theory generalises the two-species picture to $n$ species competing for $m$ resources, where coexistence requires at least as many limiting resources as species — the bridge is between the phenomenological $\alpha$ coefficients of Lotka-Volterra and the mechanistic $R^{\ast }$ values of resource competition.

Exercises [Intermediate+]

Stability, complexity, and May's criterion [Master]

The relationship between species richness and community stability is one of the central questions in theoretical ecology. Intuitively, diverse communities seem stable — a tropical rain forest persists for millennia despite containing hundreds of tree species and thousands of insect species. Yet Robert May's 1972 analysis showed that this intuition is misleading for randomly assembled communities, creating what has become known as the complexity-stability paradox.

May modelled a community of $S$ species near equilibrium using a linearised dynamics equation $\dot{\mathbf{x}}=\mathbf{Jx}$, where $\mathbf{x}$ is the vector of deviations from equilibrium and $\mathbf{J}$ is the Jacobian (community interaction) matrix. He assumed that the off-diagonal elements $J_{ij}$ are drawn independently from a distribution with mean zero and variance ${\sigma }^2$, that each species has self-regulation (diagonal elements $J_{ii}=-1$), and that a fraction $C$ of the off-diagonal entries are non-zero (the connectance). This is the random-matrix model of a food web.

Theorem (May 1972). In the limit of large $S$, the equilibrium of the random community $\dot{\mathbf{x}}=\mathbf{Jx}$ is almost surely stable if and only if $\sigma \sqrt{SC}<1$.

The proof uses results from random-matrix theory, specifically the circular law for the eigenvalue distribution of large random matrices. The eigenvalues of $\mathbf{J}$ lie inside a circle of radius $\sigma \sqrt{SC}$ in the complex plane, centred on the self-regulation term $-1$. Stability requires all eigenvalues to have negative real parts. The circle extends into the right half-plane when $\sigma \sqrt{SC}>1$, meaning at least one eigenvalue has a positive real part and the equilibrium is unstable.

The biological implication is immediate. For a community with $S=50$ species, connectance $C=0.3$, and interaction-strength standard deviation $\sigma =0.3$, the stability criterion gives $\sigma \sqrt{SC}=0.3\sqrt{15}\approx 1.16>1$. A random community with these parameters would almost certainly be unstable. Yet real communities with these characteristics persist. The conclusion is that real food webs are not random — their structure confers stability that random assembly would not produce.

Three structural features of real food webs resolve the paradox. First, compartmentalisation: species cluster into subgroups (trophic guilds, habitat associations) with strong interactions within groups and weak interactions between groups. Pimm and Lawton (1980) showed that compartmentalised food webs are more stable than randomly connected ones with the same $S$ and $C$, because perturbations are contained within compartments rather than propagating through the entire network. Empirical food webs from lakes, intertidal zones, and forests consistently show modular structure.

Second, interaction strength skew: most links in real food webs carry weak interactions, with a few strong ones. Paine's (1992) careful measurements in the Tatoosh Island intertidal showed that roughly 90% of trophic links transmitted less than 10% of the energy flowing through the web, while a handful of keystone links dominated energy transfer. McCann, Hastings, and Huxel (1998) demonstrated theoretically that this skew stabilises food webs: weak interactions provide indirect pathways that damp oscillations generated by strong consumer-resource interactions. When a predator suppresses its preferred prey, it switches to a weakly connected alternative prey, releasing the preferred prey from predation pressure and allowing it to recover. The weak link acts as a stabilising feedback.

Third, trophic pyramids and allometric constraints: body-size ratios between predators and prey limit which interactions are physically possible. A whale cannot feed on copepods individually; a copepod cannot eat a whale. These allometric constraints restrict the interaction matrix $\mathbf{J}$ to a block structure where strong interactions occur only between species of appropriate size ratios. Cohen, Pimm, Yodzis, and others have documented that empirical body-size ratios in food webs follow regular patterns — predator mass is typically 10-1000 times prey mass — producing a matrix structure far from random.

The distribution of interaction strengths also matters for trophic cascade dynamics. In a linear food chain (producer-herbivore-carnivore), removing the top predator increases herbivores, which then overgraze producers. In a complex food web, the omnivorous top predator may also feed directly on producers, creating a short-circuit that weakens or eliminates the cascade. Strong omnivory — feeding at multiple trophic levels — is a stabilising feature because it decouples predator dynamics from any single prey population. Pimm and Lawton (1978) showed that omnivory dampens population oscillations in theoretical models, a result corroborated by tank experiments with microbial communities.

Metacommunity dynamics and spatial ecology [Master]

A metacommunity is a set of local communities connected by dispersal across a landscape. The concept extends community ecology from a single patch to a spatially heterogeneous region, recognising that local species composition depends on both local interactions and regional dispersal processes. Leibold et al. (2004) organised metacommunity theory into four paradigms, each emphasising different mechanisms and making different predictions.

The patch dynamics paradigm extends Levins' metapopulation model to multiple species competing for patches in a landscape. The central mechanism is the competition-colonisation trade-off: good competitors are poor colonisers, and good colonisers are poor competitors. Tilman (1994) formalised this by assuming that species are ranked in competitive ability (species 1 excludes species 2, which excludes species 3, etc.) but that colonisation ability increases with competitive rank (species $n$ colonises empty patches at rate $c_n>c_{n-1}$). At equilibrium, the competitively dominant species occupies a fraction of patches, and each inferior species occupies a fraction of the remaining empty patches. Coexistence requires that no species can both outcompete and outcolonise all others. The prediction is that diversity increases with the total number of patches and with the disturbance rate that creates empty patches.

The species sorting paradigm emphasises environmental gradients. Different patches provide different conditions — wet versus dry, shaded versus exposed, nutrient-rich versus poor — and each species is adapted to a subset of these conditions. Dispersal is fast enough that every species can reach every patch, but local environmental filtering determines which species persist. The prediction is that species composition tracks environmental variation closely, with high beta diversity (species turnover) across environmental gradients and low beta diversity within environmentally homogeneous regions. Empirical studies of plant communities along moisture gradients, zooplankton across pH gradients in lakes, and stream invertebrates along elevation gradients support this paradigm.

The mass effects paradigm recognises that dispersal can maintain species in habitats where they would otherwise go extinct. A species that is competitively excluded from a patch may still persist there if it receives continuous immigration from a nearby source patch where it is competitively dominant. The result is a "mass effect" — the population in the sink is maintained by the inflow of individuals rather than by local reproduction. The prediction is that species richness is higher than expected from local conditions alone, particularly in patches near source habitats and in regions with high dispersal rates.

The neutral paradigm (Hubbell 2001) makes the radical assumption that all species in the metacommunity are ecologically equivalent — identical in their competitive ability, dispersal rate, and fitness. Diversity patterns arise entirely from stochastic demographic processes: birth, death, speciation, and dispersal. Hubbell's unified neutral theory predicts a specific species-abundance distribution (the zero-sum multinomial), a species-area relationship governed by the fundamental biodiversity parameter $\theta =2J_M\nu$ (where $J_M$ is the metacommunity size and $\nu$ is the speciation rate), and patterns of beta diversity determined by the dispersal-limitation parameter $m$.

Whether neutral theory describes nature or merely provides a null hypothesis remains contested. Its predictions for species-abundance distributions match empirical data surprisingly well in many tropical tree communities and coral-reef fish assemblages, but its assumption of ecological equivalence is violated by measurable niche differences in virtually every well-studied system. The current synthesis view holds that neutral and niche processes operate simultaneously: niche differentiation determines which species can persist in a given environment, and stochastic drift determines which of those capable species actually occupy any given spot.

Island biogeography provides the foundational result for metacommunity dynamics. MacArthur and Wilson (1967) modelled an island receiving immigrants from a mainland species pool of size $P$. The immigration rate decreases as the island accumulates species (fewer new species remain to colonise), while the extinction rate increases (more species means more populations at risk of local extinction). At equilibrium, the immigration rate equals the extinction rate, and the equilibrium species number $S^{\ast }$ increases with island area (larger islands support larger populations with lower extinction rates) and decreases with isolation (distant islands receive fewer immigrants). The species-area relationship $S=c{A}^z$, where $A$ is area and $z\approx 0.25$-$0.35$, is one of the most robust empirical patterns in ecology and emerges from both niche-based and neutral metacommunity models.

Beta diversity — the turnover in species composition between patches — quantifies the spatial structure of the metacommunity. Whittaker (1960) defined beta diversity as $\beta =\gamma /\alpha$, where $\gamma$ is regional species richness and $\alpha$ is mean local species richness. High beta diversity means different patches support different species assemblages; low beta diversity means the same species occur everywhere. Beta diversity is partitioned into a turnover component (species replacement between patches) and a nestedness component (species-poor patches contain subsets of the species in species-rich patches). Conservation strategies differ depending on which component dominates: high turnover requires protecting many distinct patches, while high nestedness allows protecting the richest patch to capture most species.

Biodiversity, ecosystem function, and disturbance [Master]

The biodiversity-ecosystem function (BEF) question asks whether species richness per se affects how ecosystems work — their productivity, nutrient retention, stability, and resilience. The question has direct conservation implications: if diverse communities function better than species-poor ones, then biodiversity loss has measurable ecological costs beyond the intrinsic value of individual species.

David Tilman's long-term grassland experiments at Cedar Creek, Minnesota, provide the most influential empirical evidence. Starting in 1982, Tilman established hundreds of plots varying in planted species richness from 1 to 24 species. Over more than two decades, plots with more species consistently produced more biomass, retained more soil nitrogen, and showed lower year-to-year variability in total biomass. The relationship between richness and productivity was asymptotic: the largest gains occurred going from 1 to ~10 species, with diminishing returns beyond ~16 species.

Two mechanisms explain the positive diversity-productivity relationship. The complementarity effect arises because different species use different resources or use the same resources at different times or in different soil layers. A legume fixes atmospheric nitrogen that a grass can subsequently use; a deep-rooted species accesses water unavailable to shallow-rooted neighbours; an early-season annual captures light before a late-season perennial has fully leafed out. Together, a diverse assemblage captures more total resources than any single species could alone. Loreau and Hector (2001) developed a partitioning method that separates the complementarity effect from the selection effect: diverse communities are more likely to include a species that is highly productive in monoculture, and that species dominates the mixture. Both effects contribute to the observed diversity-productivity pattern, but complementarity generally accounts for a larger fraction.

The insurance hypothesis extends the BEF framework to stability. Yachi and Loreau (1999) argued that biodiversity provides insurance against environmental fluctuations because different species respond differently to perturbations. In a diverse community, a drought that suppresses drought-sensitive species is buffered by drought-tolerant species that maintain ecosystem function. This asynchrony in species responses — the portfolio effect, by analogy with financial diversification — reduces temporal variability in aggregate properties like total biomass. The hypothesis predicts that stability (low temporal variability) increases with species richness, even when mean productivity saturates.

The intermediate disturbance hypothesis (Connell 1978) addresses how disturbance frequency and intensity regulate diversity. The hypothesis predicts a unimodal relationship between disturbance and diversity: at very low disturbance rates, competitive dominants exclude other species; at very high disturbance rates, only a few disturbance-tolerant species survive; at intermediate disturbance rates, the community contains a mosaic of patches at different successional stages, each supporting different species, maximising total diversity. Connell originally proposed the hypothesis to explain high diversity in tropical rain forests (where treefall gaps create a patchwork of successional stages) and coral reefs (where storm damage opens space for colonisation). Empirical support is mixed: some systems show the predicted unimodal pattern (rocky intertidal communities, some grasslands), while others show monotonic increases or decreases in diversity with disturbance, depending on the spatial scale of disturbance relative to the landscape.

Huston's (1979) dynamic equilibrium model generalises the intermediate disturbance hypothesis by considering both disturbance frequency and population growth rate. Diversity is maximised when the rate of competitive exclusion (driven by population growth rates) is roughly equal to the rate of disturbance-induced mortality. In productive environments (high growth rates), frequent disturbance is needed to prevent exclusion; in unproductive environments (low growth rates), infrequent disturbance suffices. This model predicts an interaction between productivity and disturbance in controlling diversity, a pattern documented in fertilised grassland experiments where adding nutrients increases the disturbance rate needed to maintain diversity.

The diversity-invasibility relationship — whether diverse communities resist invasion better than species-poor ones — is the mirror image of the BEF question. Elton (1958) argued that diverse communities resist invasion because fewer resources are available for colonisers. Tilman (1997) showed theoretically that increasing species richness decreases the average unused resource level, making invasion more difficult. The empirical evidence is nuanced: at small spatial scales, diverse communities do resist invasion; at large scales, regions with high native diversity also tend to have high exotic diversity, because both are driven by environmental heterogeneity and propagule pressure.

Food web assembly, network structure, and interaction strength [Master]

Food webs are not static objects. They assemble over time as species colonise a new habitat, establish interactions with residents, and either persist or go extinct. The trajectory of assembly depends on the order of arrival (priority effects), the match between species traits and available niches, and the feedbacks between resident species and the physical environment. Understanding assembly is essential for predicting how communities will respond to species invasions, extinctions, and habitat restoration.

Assembly rules describe constraints on the order and composition of species that can form a persistent community. Drake (1990) and Law and Morton (1996) showed experimentally and theoretically that assembly is historically contingent: different initial species pools or different arrival orders can produce alternative stable communities with different species compositions and different functional properties. The assembly process can be modelled as a trajectory through a high-dimensional state space where each point represents the abundance vector of all species. New species enter as "invaders" that either establish (moving the system to a new equilibrium) or fail (returning the system to its previous state). The set of species that coexist at the final equilibrium is the assembled community, and it need not be unique.

Network motifs — small subgraphs that recur more frequently than expected by chance — reveal the building blocks of food web structure. Bascompte and Melián (2005) showed that mutualistic networks (plant-pollinator, plant-frugivore) are significantly more nested than random, meaning that specialist species interact with subsets of the partners of generalist species. This nested architecture minimises competition (specialists share few partners) while maximising the number of coexisting species. Antagonistic networks (predator-prey, host-parasitoid) show modular structure instead, with groups of tightly interacting species forming semi-independent compartments. The structural difference arises because mutualisms benefit from nestedness (generalist pollinators service many plants, specialists use the most connected plants), while antagonisms benefit from compartmentalisation (constraining the spread of overexploitation).

The niche model (Williams and Martinez 2000) generates realistic food web structure from three simple rules for each species: (1) each species has a fundamental niche defined by a range on a single resource axis; (2) consumers eat all species whose niche centres fall within their range; (3) species are ordered along the niche axis, with predators generally consuming prey at lower positions (smaller body size, lower trophic level). The niche model reproduces the observed distributions of connectance, chain length, omnivory, and cannibalism across empirical food webs with remarkable accuracy using only two parameters ($S$ and $C$). Its success suggests that simple body-size and niche-overlap rules generate much of the structural regularity observed in nature.

Interaction strength in food webs is measured experimentally by removing or excluding a consumer and measuring the resulting change in prey density or community composition. Paine's (1992) synthesis of removal experiments revealed a characteristic pattern: most interactions are weak (removing the consumer produces less than a 10% change in prey density), while a few are strong (removing the consumer causes order-of-magnitude changes). This interaction strength skew is not an artefact of measurement — it reflects the biology of consumer-resource interactions. Weak interactions occur when consumers feed on abundant prey with fast population growth (the prey compensates for losses), while strong interactions occur when consumers feed on slow-growing, competitively dominant prey (the prey cannot compensate, and its release upon consumer removal transforms the community).

Trophic cascades in complex webs differ from the textbook linear chain. Hairston, Smith, and Slobodkin (HSS, 1960) proposed that terrestrial ecosystems are generically three-tiered: predators control herbivores, which would otherwise consume all vegetation. Polis and Strong (1996) challenged this view, arguing that real food webs are too complex for simple cascades — omnivory, detrital pathways, and subsidies from adjacent ecosystems blur trophic levels and dampen top-down effects. The resolution lies in distinguishing between community-level cascades (where a predator affects the entire producer community) and species-level cascades (where a predator affects specific prey species). Community-level cascades are documented primarily in aquatic systems (lakes, streams, marine kelp forests) where the food web is relatively linear; terrestrial systems with their complex omnivorous webs show weaker, species-level cascades.

Body size provides a unifying framework for food web structure. Cohen et al. (1993) showed that predator body mass is, on average, 10-1000 times prey body mass across taxa and ecosystems, and that this allometric ratio constrains which links are possible. Elton (1927) first observed that food chain length is limited by energy loss at each trophic transfer (approximately 90% of energy is lost as heat at each step), which also limits the maximum body size achievable at the top of the chain. The combination of allometric feeding constraints and energetic limits explains why food chains rarely exceed 5-6 links and why connectance decreases with species richness: in a diverse community, most potential links are physically impossible because the size ratio between consumer and resource falls outside the viable range.

Synthesis. The four Master-tier topics — stability, metacommunity dynamics, biodiversity-ecosystem function, and food web assembly — are not independent threads but interconnected facets of a single problem: what determines the composition, structure, and persistence of ecological communities? The foundational reason stability requires non-random structure (May 1972) is exactly the condition that real food webs satisfy through compartmentalisation and interaction-strength skew. This is the bridge to metacommunity theory: spatial structure provides one mechanism for compartmentalisation, and dispersal limitation provides one mechanism for interaction-strength reduction. Putting these together with the niche model and allometric constraints generalises the assembly problem from local coexistence conditions to landscape-scale diversity patterns. The pattern recurs throughout: local interactions generate community structure, spatial processes modulate that structure across the landscape, and the resulting metacommunity determines both the ecosystem functions measured by BEF experiments and the stability properties predicted by May's criterion. The central insight is that community ecology operates on a hierarchy of scales — from individual interactions to local communities to regional metacommunities to global biodiversity patterns — and the continuity across scales is maintained by dispersal, allometry, and energy flow.

Connections [Master]

• Population ecology 19.09.01 provides the single-species and two-species models (exponential growth, logistic growth, Lotka-Volterra competition) that community ecology extends to multi-species systems. The Lotka-Volterra coexistence condition derived there appears again in the Master-tier analysis of May's stability criterion as the two-species special case of the general $n$-species interaction matrix.

• Natural selection 19.03.01 pending drives the evolution of niche differentiation, anti-predator defences, and mutualism traits that structure communities. Evolutionary arms races between predators and prey shape the interaction strengths that May's criterion identifies as load-bearing for stability.

• Kin selection 19.03.03 pending explains cooperative behaviour within species, which can influence competitive ability and thus community structure (e.g., cooperative hunting in wolves affects prey populations, with cascading effects on vegetation).

• Speciation 19.06.01 pending adds new species to communities. The rate of speciation (versus extinction) determines regional species pools, which constrain local community diversity and set the species richness parameter $S$ in May's stability criterion.

• Ecosystem ecology 20.05.02 pending builds on community structure to analyse energy flow and nutrient cycling. Food webs link community ecology to ecosystem-level processes, and the BEF results described in this unit build toward 20.05.02 pending as the mechanistic basis for biodiversity's role in ecosystem functioning.

• Island biogeography and neutral theory 19.10.02 pending (pending) extend the metacommunity framework to large-scale biodiversity patterns. MacArthur and Wilson's equilibrium model and Hubbell's neutral theory both appear in the Master-tier treatment of metacommunity dynamics.

Historical & philosophical context [Master]

The concept of a "web of life" has ancient roots, but the formal study of communities began with Frederic Clements (1916), who viewed plant communities as superorganisms — tightly integrated units that develop through predictable succession toward a climax state ^{[Clements 1916]}. Henry Gleason (1926) challenged this view, arguing that communities are loose, accidental assemblages of individualistic species distributions ^{[Gleason 1926]}. The Clements-Gleason debate continues: modern evidence supports elements of both views. Some communities are tightly integrated (tropical fig-wasp mutualisms), while others are loosely assembled (temperate forest herb layers). The metacommunity perspective resolves the debate partially by showing that both patterns can arise from the same spatial processes operating at different scales.

Robert Paine's Pisaster experiment (1966) introduced the keystone species concept ^{[Paine 1966]}, demonstrating that individual species can have outsized effects on community structure. This work shifted ecology from cataloguing community composition to experimentally manipulating communities to test theory. Paine's later quantitative work on interaction strengths (Paine 1992) laid the groundwork for the empirical study of interaction-strength distributions that resolved the complexity-stability paradox.

Robert May's mathematical analysis (1972) brought theoretical rigour to community ecology ^{[May 1972]}. By showing that complexity destabilises random communities, May forced ecologists to explain why real diverse communities are stable, leading to research on food web structure, interaction strength distributions, and spatial dynamics. May's random-matrix approach was imported from physics (the circular law for eigenvalue distributions) and demonstrated the power of applying quantitative methods from other disciplines to ecological questions.

The neutral theory of biodiversity (Hubbell, 2001) provoked intense debate by proposing that community-level patterns (species-abundance distributions, species-area relationships) can be explained without invoking niche differences — purely through stochastic drift, speciation, and dispersal ^{[Hubbell 2001]}. The debate sharpened both niche-based and neutral theories and led to the synthesis view that both processes operate simultaneously, a conclusion that metacommunity theory formalises.

The BEF debate of the 1990s-2000s connected community ecology to conservation policy by quantifying the ecosystem consequences of biodiversity loss. Tilman's Cedar Creek experiments and the pan-European BIODEPTH experiment (Hector et al. 1999) provided the experimental foundation, while Loreau et al. (2001) supplied the theoretical partitioning framework. The debate continues in refined form: whether complementarity or selection dominates, whether functional-trait diversity matters more than species richness, and how BEF relationships scale from small plots to landscapes.

Bibliography [Master]

Primary literature — keystone species and interaction strength.

• Paine, R. T., "Food web complexity and species diversity", Am. Nat. 100 (1966), 65-75.
• Paine, R. T., "Food-web analysis through field measurement of per capita interaction strength", Nature 355 (1992), 73-75.
• Hairston, N. G., Smith, F. E. & Slobodkin, L. B., "Community structure, population control, and competition", Am. Nat. 94 (1960), 421-425.
• Polis, G. A. & Strong, D. R., "Food web complexity and community dynamics", Am. Nat. 147 (1996), 813-846.

Stability, complexity, and network structure.

• May, R. M., "Will a large complex system be stable?", Nature 238 (1972), 413-414.
• Pimm, S. L., Food Webs (Chapman & Hall, 1982).
• Pimm, S. L. & Lawton, J. H., "On feeding on more than one trophic level", Nature 275 (1978), 542-544.
• Pimm, S. L. & Lawton, J. H., "Are food webs divided into compartments?", J. Anim. Ecol. 49 (1980), 879-898.
• McCann, K., Hastings, A. & Huxel, G. R., "Weak trophic interactions and the balance of nature", Nature 395 (1998), 794-798.
• Williams, R. J. & Martinez, N. D., "Simple rules yield complex food webs", Nature 404 (2000), 180-183.
• Cohen, J. E. et al., "Improving food webs", Ecology 74 (1993), 252-258.
• Bascompte, J. & Melián, C. J., "Simple trophic modules for complex food webs", Ecology 86 (2005), 2868-2873.

Competition, coexistence, and niche theory.

• Tilman, D., Resource Competition and Community Structure (Princeton UP, 1982).
• Tilman, D., "Competition and biodiversity in spatially structured habitats", Ecology 75 (1994), 2-16.
• Chesson, P., "Mechanisms of maintenance of species diversity", Annu. Rev. Ecol. Syst. 31 (2000), 343-366.

Metacommunity and spatial dynamics.

• Leibold, M. A. et al., "The metacommunity concept: a framework for multi-scale community ecology", Ecol. Lett. 7 (2004), 601-613.
• MacArthur, R. H. & Wilson, E. O., The Theory of Island Biogeography (Princeton UP, 1967).
• Hubbell, S. P., The Unified Neutral Theory of Biodiversity and Biogeography (Princeton UP, 2001).
• Whittaker, R. H., "Vegetation of the Siskiyou Mountains, Oregon and California", Ecol. Monogr. 30 (1960), 279-338.

Biodiversity, ecosystem function, and disturbance.

• Connell, J. H., "Diversity in tropical rain forests and coral reefs", Science 199 (1978), 1302-1310.
• Huston, M. A., "A general hypothesis of species diversity", Am. Nat. 113 (1979), 81-101.
• Loreau, M. & Hector, A., "Partitioning selection and complementarity in biodiversity experiments", Nature 412 (2001), 72-76.
• Yachi, S. & Loreau, M., "Biodiversity and ecosystem productivity in a fluctuating environment: the insurance hypothesis", Proc. Natl. Acad. Sci. 96 (1999), 1463-1468.
• Hector, A. et al., "Plant diversity and productivity experiments in European grasslands", Science 286 (1999), 1123-1127.
• Elton, C. S., The Ecology of Invasions by Animals and Plants (Methuen, 1958).

Assembly and historical.

• Drake, J. A., "The mechanics of community assembly and succession", J. Theor. Biol. 147 (1990), 213-233.
• Law, R. & Morton, R. D., "Permanence and the assembly of ecological communities", Ecology 77 (1996), 762-775.
• Clements, F. E., Plant Succession: An Analysis of the Development of Vegetation (Carnegie Institution, 1916).
• Gleason, H. A., "The individualistic concept of the plant association", Bull. Torrey Bot. Club 53 (1926), 7-26.

Textbook.

• Begon, M., Harper, J. L. & Townsend, C. R., Ecology, 4th ed. (Blackwell, 2006).