Population ecology — Lotka-Volterra

Intuition [Beginner]

How fast can a population grow? If every individual produces more than one surviving offspring each generation, the population grows exponentially. A single bacterium dividing every 20 minutes produces over 1 billion descendants in 10 hours. This is exponential growth: the population multiplies by a constant factor each time step.

But exponential growth cannot continue forever. Eventually, resources run out — food, space, light, water. The population hits a ceiling called the carrying capacity ($K$), the maximum number of individuals the environment can sustain. Growth slows as the population approaches $K$, producing an S-shaped curve called logistic growth.

What happens when two species interact? The simplest case is predator and prey. When prey are abundant, predators have plenty of food and their population grows. More predators eat more prey, reducing the prey population. With fewer prey, predators starve and their population declines. With fewer predators, prey recover. This creates a cycle — populations of both species oscillate over time.

The Lotka-Volterra equations model this cycle mathematically. Alfred Lotka (1925) and Vito Volterra (1926) independently derived these equations, which predict regular oscillations in predator and prey populations. The Canadian lynx and snowshoe hare show this pattern in real data, with roughly 10-year cycles documented over 200 years of fur-trapping records from the Hudson Bay Company.

Visual [Beginner]

Imagine a graph with population size on the vertical axis and time on the horizontal axis. Two curves oscillate out of phase.

The prey peaks first, then the predator peaks follow with a time lag. In the phase plane (prey on x-axis, predator on y-axis), the populations trace a closed loop around the equilibrium point, going counterclockwise: more prey leads to more predators, which reduces prey, which reduces predators, allowing prey to recover.

Worked example [Beginner]

Consider logistic growth with $r=0.5$ per year, $K=1000$, and initial population $N_0=10$.

Step 1. The logistic equation describes growth that slows as the population nears carrying capacity: $dN/dt=rN(1-N/K)$.

Step 2. The analytical solution is $N(t)=K/[1+((K-N_0)/N_0)e^{-rt}]$. Substituting our values: $N(t)=1000/[1+99\times e^{-0.5t}]$.

Step 3. At $t=10$: $N(10)=1000/(1+99\times 0.00674)=1000/1.667=600$. At $t=20$: $N(20)=1000/(1+99\times 0.000045)=996$.

What this tells us: the population grows from 10 to about 600 in 10 time units and approaches carrying capacity by $t=20$, following the S-shaped logistic curve. The growth rate is fastest at $N=K/2=500$, the inflection point.

Check your understanding [Beginner]

Formal definition [Intermediate+]

Exponential growth

For a population with instantaneous growth rate $r$ (births minus deaths per individual per unit time):

$$\frac{dN}{dt}=rN.$$

Solution: $N(t)=N_0{e}^{rt}$. Doubling time: $t_d=\ln 2/r$.

Logistic growth

The logistic equation adds density-dependent regulation:

$$\frac{dN}{dt}=rN\left(1-\frac{N}{K}\right).$$

Solution:

$$N(t)=\frac{K}{1+\left(\frac{K-N_0}{N_0}\right)e^{-rt}}.$$

Equilibria: $N^{\ast }=0$ (unstable for $r>0$) and $N^{\ast }=K$ (globally stable). The maximum growth rate occurs at $N=K/2$ (the inflection point), where $dN/dt=rK/4$.

Lotka-Volterra predator-prey

Let $x$ = prey density, $y$ = predator density:

$$\frac{dx}{dt}=\alpha x-\beta xy=x(\alpha -\beta y),$$ $$\frac{dy}{dt}=\delta xy-\gamma y=y(\delta x-\gamma ),$$

where $\alpha$ is the prey growth rate, $\beta$ is the predation rate, $\delta$ is the predator conversion efficiency, and $\gamma$ is the predator death rate. Equilibrium at $x^{\ast }=\gamma /\delta$, $y^{\ast }=\alpha /\beta$.

The system has a conserved quantity (first integral):

$$H(x,y)=\delta x-\gamma \ln x+\beta y-\alpha \ln y,$$

so trajectories lie on level curves of $H$, forming closed orbits around the equilibrium. The populations oscillate indefinitely with amplitude determined by initial conditions.

Lotka-Volterra competition

For two competing species:

$$\frac{d{N}_1}{dt}=r_1{N}_1\left(\frac{K_1-N_1-{\alpha }_{12}{N}_2}{K_1}\right),$$ $$\frac{d{N}_2}{dt}=r_2{N}_2\left(\frac{K_2-N_2-{\alpha }_{21}{N}_1}{K_2}\right),$$

where ${\alpha }_{12}$ is the competitive effect of species 2 on species 1, and ${\alpha }_{21}$ is the reverse. Coexistence requires:

$${\alpha }_{12}<\frac{K_1}{K_2}\text{and}{\alpha }_{21}<\frac{K_2}{K_1}.$$

For equal carrying capacities ($K_1=K_2$), coexistence requires ${\alpha }_{12}{\alpha }_{21}<1$ — each species inhibits itself more than it inhibits the other.

Counterexamples to common slips

• Carrying capacity $K$ is not a fixed environmental constant. $K$ varies with resource availability, disease, predation pressure, and environmental fluctuations. A population at $K=1000$ in a good year may face $K=200$ during a drought.
• Lotka-Volterra predator-prey cycles are not stable limit cycles. The basic model produces neutrally stable (centre) oscillations: any perturbation shifts the system to a new orbit permanently. Real predator-prey systems require additional density-dependent mechanisms for true stability.
• Competition does not always lead to exclusion. The LV competition model predicts exclusion when ${\alpha }_{12}{\alpha }_{21}>1$, but coexistence is possible when each species limits itself more than it limits the other. Niche differentiation, spatial heterogeneity, and temporal variation all promote coexistence beyond the simple LV prediction.

Key theorem with proof [Intermediate+]

Theorem (Lotka-Volterra cycles are neutrally stable). The Lotka-Volterra predator-prey system has a centre (neutrally stable equilibrium) at $(x^, y^) = (\gamma/\delta, \alpha/\beta)$. All non-equilibrium trajectories are closed orbits, and the amplitude of oscillation depends on initial conditions.

Proof. The equilibrium satisfies $dx/dt=0$ and $dy/dt=0$. From $x(\alpha -\beta y)=0$: either $x=0$ or $y=\alpha /\beta$. From $y(\delta x-\gamma )=0$: either $y=0$ or $x=\gamma /\delta$. The interior equilibrium is $(\gamma /\delta ,\alpha /\beta )$.

To show closed orbits, we demonstrate the conserved quantity. Rearranging the equations:

$$\frac{dx}{x(\alpha -\beta y)}=dt=\frac{dy}{y(\delta x-\gamma )}.$$

Separating:

$$\frac{(\delta x-\gamma )}{x}dx=\frac{(\alpha -\beta y)}{y}dy.$$

Integrating both sides:

$$\delta x-\gamma \ln x=\alpha \ln y-\beta y+C,$$

which gives $H(x,y)=\delta x-\gamma \ln x+\beta y-\alpha \ln y=C$ (constant).

Since $H$ is strictly convex with a unique minimum at the equilibrium, the level curves $H=C$ are closed curves surrounding the equilibrium. Each trajectory lies on one such curve, producing a periodic orbit. $\square$

Neutral stability means the system is structurally unstable: any small perturbation to the equations (e.g., adding prey self-limitation $-d{x}^2/K$) changes the dynamics qualitatively, typically producing a stable limit cycle or a stable equilibrium instead of neutral cycles.

Bridge. The conserved-quantity proof builds toward the phase-plane analysis in 02.12.01 where centre-type equilibria and first integrals are studied systematically. The structural instability of neutral oscillations appears again in 19.10.01 pending community ecology, where multi-species extensions of Lotka-Volterra require density-dependent mechanisms to produce the stable limit cycles observed in real ecosystems. This is exactly the reason that the Rosenzweig-MacArthur model — which adds logistic prey growth and a saturating functional response — replaces the basic LV predation equations in every applied context: the foundational reason is that neutral stability is an artefact of the mass-action predation term $\beta xy$, and realistic feeding rates saturate.

Exercises [Intermediate+]

Single-species dynamics [Master]

Exponential and logistic growth revisited

The exponential model $dN/dt=rN$ with solution $N(t)=N_0{e}^{rt}$ is the simplest description of an unconstrained population. The growth rate $r=b-d$ (per-capita birth rate minus per-capita death rate) is constant, independent of population density. Every population obeys exponential growth at sufficiently low density; the question is what happens as density increases.

Verhulst (1838) introduced the logistic equation $dN/dt=rN(1-N/K)$ to describe density-dependent regulation. The term $(1-N/K)$ reduces the per-capita growth rate linearly as $N$ approaches $K$. The analytical solution

$$N(t)=\frac{K}{1+\left(\frac{K-N_0}{N_0}\right)e^{-rt}}$$

produces the characteristic sigmoid curve with three phases: near-exponential growth at $N\ll K$, an inflection point at $N=K/2$ where the growth rate is maximal ($rK/4$), and asymptotic approach to $K$.

The logistic equation has two equilibria. $N^{\ast }=0$ is unstable for $r>0$ (any small perturbation away from extinction grows). $N^{\ast }=K$ is globally asymptotically stable on $(0,\infty )$: every positive initial population converges to $K$. Global stability follows from the Lyapunov function $V(N)=N-K-K\ln (N/K)$, which satisfies $dV/dt=-r(N-K{)}^2/K\le 0$.

The Allee effect

At very low population densities, some species experience reduced rather than increased per-capita growth — a positive relationship between density and fitness. Warder Clyde Allee documented this experimentally in the 1930s for goldfish and flour beetles: below a critical threshold, individuals struggle to find mates, cooperate inefficiently, or suffer disproportionate predation. The strong Allee effect introduces a critical threshold $N_c<K$ such that populations below $N_c$ decline to extinction:

$$\frac{dN}{dt}=rN\left(\frac{N}{N_c}-1\right)\left(1-\frac{N}{K}\right).$$

This produces three equilibria: $N^{\ast }=0$ (stable), $N^{\ast }=N_c$ (unstable — the Allee threshold), and $N^{\ast }=K$ (stable). Populations between 0 and $N_c$ are drawn to extinction; populations above $N_c$ grow toward $K$. The Allee effect has direct conservation implications: populations reduced below $N_c$ by habitat loss or harvesting cannot recover without intervention, even if suitable habitat remains. The passenger pigeon (Ectopistes migratorius), which went from billions to extinction in decades despite vast remaining habitat, is a candidate example of an Allee-driven collapse — the species' colonial breeding system required large aggregations to function.

Time delays and oscillatory dynamics

Real populations do not respond instantaneously to density changes. There is a gestation period, a maturation time, or a resource-renewal lag. Hutchinson (1948) introduced the delayed logistic equation:

$$\frac{dN(t)}{dt}=rN(t)\left(1-\frac{N(t-\tau )}{K}\right),$$

where $\tau$ is the time delay between birth and density-dependent regulation. The delay destabilises the equilibrium $N^{\ast }=K$ when $r\tau >\pi /2$ (approximately 1.57). Below this threshold, $K$ is stable and trajectories converge with damped oscillations. Above it, a Hopf bifurcation produces a stable limit cycle. The mechanism: the population overshoots $K$ because regulation acts on the density $\tau$ time units ago; the overshoot triggers an undershoot; and the cycle perpetuates if the delay is long enough relative to the response time $1/r$.

The delay model explains oscillations in species without obvious predator-prey interactions. Nicholson's blowflies (Lucilia cuprina), maintained in laboratory cultures with constant food supply, exhibited sustained oscillations with a period roughly twice the delay from egg to adult — matching Hutchinson's prediction. May (1973) showed that the delay model's stability boundary $r\tau =\pi /2$ generalises to a broad class of delayed-density-dependent models, making the product $r\tau$ the single dimensionless parameter controlling stability.

Discrete-time models and chaos

When reproduction occurs in discrete seasons rather than continuously, difference equations replace ODEs. The discrete logistic map $N_{t+1}=r{N}_t(1-N_t/K)$ and the Ricker model $N_{t+1}=N_t{e}^{r(1-N_t/K)}$ are the standard formulations. May (1976) demonstrated that these simple deterministic equations exhibit the full range of dynamical behaviour — stable equilibrium, damped oscillations, periodic cycles, and chaos — depending on $r$.

For the Ricker model, linearisation around $N^{\ast }=K$ gives eigenvalue $\lambda =1-r$. Stability requires $|\lambda |<1$, i.e., $0<r<2$. As $r$ increases past 2, a period-doubling cascade begins: period-2 at $r\approx 2$, period-4 at $r\approx 2.53$, period-8 at $r\approx 2.66$, converging to chaos at $r\approx 2.692$ through the Feigenbaum cascade with the universal ratio $\delta \approx 4.669$. In the chaotic regime ($r>2.692$), trajectories are aperiodic and exhibit sensitive dependence on initial conditions: nearby starting points diverge exponentially, with a positive Lyapunov exponent.

The biological implication is that apparently random population fluctuations need not be driven by environmental stochasticity. Deterministic nonlinearity alone can produce erratic dynamics. Distinguishing noise from chaos in real time series requires careful statistical methods (Ellner 1991, Chaos and Order in the Dynamics of Populations), but the theoretical possibility reshaped how ecologists think about population variability.

Competition and coexistence theory [Master]

Isocline analysis of Lotka-Volterra competition

The Lotka-Volterra competition equations have four phase-plane regimes determined by the relative positions of the two species' nullclines (isoclines). The nullcline for species 1 is $N_1+{\alpha }_{12}{N}_2=K_1$ (a line in the $(N_1,N_2)$ plane with intercepts $K_1$ on the $N_1$-axis and $K_1/{\alpha }_{12}$ on the $N_2$-axis); similarly, species 2's nullcline is ${\alpha }_{21}{N}_1+N_2=K_2$ with intercepts $K_2/{\alpha }_{21}$ and $K_2$.

Four cases arise depending on the relative positions of these intercepts. Case 1: $K_1/{\alpha }_{12}>K_2$ and $K_2/{\alpha }_{21}>K_1$ — the nullclines cross, producing a stable coexistence equilibrium (both species limit themselves more than the other). Case 2: $K_1>K_2/{\alpha }_{21}$ and $K_1/{\alpha }_{12}>K_2$ — species 1 wins (its nullcline lies entirely above species 2's, so species 1 excludes species 2 from every initial condition). Case 3: the reverse of Case 2 — species 2 wins. Case 4: $K_1/{\alpha }_{12}<K_2$ and $K_2/{\alpha }_{21}<K_1$ — the nullclines cross but the coexistence equilibrium is unstable, with two stable boundary equilibria; the outcome depends on initial conditions (priority effects). This fourth case is competitive exclusion with bistability.

The coexistence condition $K_1/{\alpha }_{12}>K_2$ and $K_2/{\alpha }_{21}>K_1$ can be rewritten as ${\alpha }_{12}<K_1/K_2$ and ${\alpha }_{21}<K_2/K_1$. For equal carrying capacities, the product condition ${\alpha }_{12}{\alpha }_{21}<1$ captures the requirement that intraspecific competition exceeds interspecific competition. This is the competitive exclusion principle (Gause 1934): two species cannot coexist on the same limiting resource unless they differ in their resource use such that each limits itself more than it limits the other.

Tilman's R* rule

Tilman (1982) reframed competition theory in terms of the resource rather than the competitor. For a single limiting resource $R$ consumed by species $i$ at rate $c_i$ with mortality $m_i$ and resource-dependent growth $g_i(R)$, the equilibrium resource level in monoculture is $R_i^{\ast }$, the solution of $g_i(R_i^{\ast })=m_i$. Tilman's R rule* states: the species that depresses the resource to the lowest level (${\min }_i{R}_i^{\ast }$) excludes all others. At equilibrium, the resource sits at $R^{\ast }={\min }_i{R}_i^{\ast }$, and any species whose growth requirement exceeds this level cannot sustain itself.

The R* rule provides a mechanistic basis for the phenomenological LV competition coefficients. The competitive effect ${\alpha }_{ij}$ is not an arbitrary constant but derives from the relative ability of each species to exploit and deplete the shared resource. Species with lower $R^{\ast }$ are better competitors; coexistence requires either multiple resources (each species being the best competitor for a different resource, producing the resource-ratio hypothesis) or fluctuating conditions that prevent any single species from maintaining the lowest $R^{\ast }$ at all times.

Modern coexistence theory

Chesson (2000, Annu. Rev. Ecol. Syst. 31, 343-366) unified coexistence mechanisms into two categories. Stabilizing mechanisms generate negative frequency dependence: each species has an advantage when rare, causing populations to recover from low density. Niche differentiation is the canonical stabilizing mechanism — species use different resources, are favoured in different habitats, or are limited by different predators. Equalizing mechanisms reduce fitness differences between species, slowing competitive exclusion without preventing it. Environmental fluctuations can act as equalizers if they favour different species at different times.

The quantitative framework defines the stabilizing effect as $\overline{\Delta E}/\overline{\Delta k}$ — the average niche difference divided by the average fitness difference — and coexistence requires this ratio to exceed 1. The storage effect, relative nonlinearity, and spatial-temporal covariance are specific stabilizing mechanisms that operate through environmental variability.

The storage effect (Chesson & Warner 1981, Am. Nat. 117, 923-943) arises when species have a life stage resistant to unfavourable conditions (e.g., long-lived adults, seed banks, diapausing eggs). During favourable years, a species builds up a "stored" population; during unfavourable years, it draws on this storage. The interaction between environmental response and competition generates negative frequency dependence: a rare species experiences less competition during its favourable years because the common species' favourable years are different.

Relative nonlinearity operates when species respond differently to fluctuating resource levels. If one species has a concave growth response to resources (fast at high $R$, slow at low $R$) while the other has a convex response, fluctuations in $R$ benefit the concave species on average (by Jensen's inequality). This mechanism can stabilise coexistence when fluctuations are endogenous (predator-prey cycles) or exogenous (seasonal variation).

Predator-prey dynamics [Master]

Functional responses

The Lotka-Volterra predation term $\beta xy$ assumes that each predator consumes prey at a rate proportional to prey density — a type I functional response (Holling 1959, Can. Entomol. 91, 385-398). This is unrealistic at high prey densities because predators saturate: handling and digesting each prey item takes time, setting an upper limit on consumption rate.

The type II functional response (Holling's disc equation) models this saturation:

$$f(N)=\frac{aN}{1+ahN},$$

where $a$ is the attack rate and $h$ is the handling time per prey item. At low $N$, $f(N)\approx aN$ (linear, equivalent to type I). At high $N$, $f(N)\to 1/h$ (saturated at the maximum processing rate). Type II responses are empirically documented in invertebrate predators (e.g., mantids, ladybird beetles) and are the standard model in theoretical ecology.

The type III functional response has $f(N)=a{N}^2/(1+ah{N}^2)$, sigmoidal with an accelerating phase at low $N$ and saturation at high $N$. This arises when predators switch to a prey species only when it becomes sufficiently abundant, or when prey learn to evade predators at low density (increasing search difficulty). Type III responses produce a prey refuge at low density: predation rate drops disproportionately when prey are rare, allowing prey populations to recover. This stabilising property means that type III responses promote coexistence more readily than type II.

Rosenzweig-MacArthur model

Rosenzweig and MacArthur (1963, Am. Nat. 97, 209-223) combined logistic prey growth with a type II functional response:

$$\frac{dx}{dt}=rx\left(1-\frac{x}{K}\right)-\frac{\beta xy}{1+\beta hx},$$ $$\frac{dy}{dt}=\frac{\delta \beta xy}{1+\beta hx}-\gamma y.$$

The prey nullcline is hump-shaped: predator abundance first increases then decreases with prey density. The predator nullcline is a vertical line at the prey density where predator growth is zero.

Paradox of enrichment

Rosenzweig (1971, Science 171, 385-387) demonstrated that increasing the prey carrying capacity $K$ — enriching the environment — destabilises the predator-prey equilibrium. At low $K$, the predator nullcline intersects the prey nullcline on its rising (left) branch, and the equilibrium is a stable node or spiral. As $K$ increases, the intersection moves to the hump's descending (right) branch, the equilibrium becomes unstable, and a stable limit cycle appears via a Hopf bifurcation. Further enrichment increases the cycle's amplitude until prey minima during cycles fall to dangerously low levels, risking stochastic extinction.

This paradox of enrichment contradicts the intuition that more resources should make ecosystems more stable. The resolution is that enrichment benefits prey disproportionately at high density, creating larger prey oscillations that drag predator populations through boom-and-bust cycles. Empirical support comes from microcosm experiments (Luckinbill 1973, Ecology 54, 1320-1327) showing that enriched Paramecium-Didinium systems collapse faster than nutrient-poor ones, and that spatial structure (which provides prey refugia) stabilises the interaction.

Limit cycles and bifurcation structure

The Hopf bifurcation in the Rosenzweig-MacArthur model is typically supercritical: a stable limit cycle is born at the critical $K_H$ and grows continuously as $K$ increases past $K_H$. The cycle period is approximately the prey's response time ($1/r$) modified by the predator's numerical response lag. Near the bifurcation, the cycle amplitude grows as $\sqrt{K-K_H}$.

For systems with type III functional responses, the prey nullcline develops a sigmoidal shape that can produce two Hopf bifurcations as $K$ increases: first destabilising (supercritical), then restabilising (subcritical) at higher $K$. Between the two bifurcations, limit cycles coexist with a stable equilibrium, producing bistability between small-amplitude cycles near the equilibrium and large-amplitude cycles that visit very low prey densities. This structure has been documented in plankton systems where nutrient enrichment triggers destructive grazing cycles (Scheffer et al. 1997, Trends Ecol. Evol. 12, 354-359).

Prey refugia and spatial dynamics

Real predator-prey systems rarely operate in well-mixed environments. Prey refugia — physical spaces where prey are safe from predators — stabilise the interaction by maintaining a prey reservoir that can repopulate after predator-driven declines. Huffaker's 1958 mite experiments (Hilgardia 27, 343-383) demonstrated this principle: the predatory mite Typhlodromus occidentalis and its prey Eotetranychus sexmaculatus could coexist in spatially complex environments (oranges connected by petroleum jelly barriers) but not in simple ones, where the predator invariably consumed all prey and then starved. Spatial heterogeneity allowed persistent three-patch metapopulation dynamics where local extinctions were balanced by recolonisation.

Age-structured and spatial population models [Master]

Leslie matrix models

When age-specific birth and death rates vary — which they do in every real species — the simple differential-equation models must be replaced by age-structured formulations. Leslie (1945, Biometrika 33, 183-212) introduced the Leslie matrix $\mathbf{L}$, a square matrix with age-specific fertilities $f_i$ in the first row and age-specific survival probabilities $s_i$ on the subdiagonal:

$$\mathbf{L}=\left(\begin{array}{ccccc}{f}_1& f_2& f_3& \cdots & f_{\omega }\\ s_1& 0& 0& \cdots & 0\\ 0& s_2& 0& \cdots & 0\\ ⋮& & \ddots & & ⋮\\ 0& \cdots & 0& s_{\omega -1}& 0\end{array}\right).$$

The population at time $t+1$ is ${\mathbf{n}}_{t+1}=\mathbf{L}{\mathbf{n}}_t$. By the Perron-Frobenius theorem, if $\mathbf{L}$ is primitive (some power has all positive entries, which occurs when two adjacent fertilities are positive), the dominant eigenvalue ${\lambda }_1$ is real and positive, and the population asymptotically grows (or declines) by the factor ${\lambda }_1$ per time step, with the age distribution converging to the stable age distribution given by the right eigenvector of ${\lambda }_1$.

The quantity ${\lambda }_1$ is the asymptotic population growth rate. If ${\lambda }_1>1$, the population grows; if ${\lambda }_1<1$, it declines; if ${\lambda }_1=1$, it is stable. The corresponding left eigenvector gives the reproductive value $v_i$ of each age class — the expected future reproductive contribution of an individual of age $i$, discounted by the population's growth rate. Reproductive value typically peaks at the age of first reproduction and declines thereafter.

Euler-Lotka equation

For continuous-time age-structured populations, the connection between the life table ($l_x$, $m_x$) and the population growth rate $r$ is the Euler-Lotka equation:

$${\int }_0^{\infty }{e}^{-ra}l(a)m(a)da=1.$$

This implicit equation for $r$ has a unique real root when $R_0={\int }_0^{\infty }l(a)m(a)da>0$ and $l(a)$ is monotonically decreasing. For $r=0$, the left side equals $R_0$; solving for $r$ therefore determines whether the population grows ($r>0$, requiring $R_0>1$) or declines. The Euler-Lotka equation is the continuous-time analogue of the Leslie matrix eigenvalue equation, and both encode the same information: the population growth rate is determined by the schedule of survival and reproduction.

Metapopulation dynamics

Levins (1969, Bull. Entomol. Soc. Am. 15, 237-240) introduced the metapopulation concept: a population of populations inhabiting discrete habitat patches, connected by occasional dispersal. The Levins model tracks only the fraction $p$ of occupied patches:

$$\frac{dp}{dt}=cp(1-p)-ep,$$

where $c$ is the colonisation rate and $e$ is the local extinction rate. The equilibrium $p^{\ast }=1-e/c$ exists only if $c>e$. If the extinction rate exceeds the colonisation rate, $p^{\ast }<0$ and the metapopulation collapses to regional extinction despite some patches being habitable.

Metapopulation theory has transformed conservation biology. The classic application is the northern spotted owl (Strix occidentalis caurina) in the Pacific Northwest: logging fragmented old-growth forest into isolated patches too small to sustain individual owl pairs, but the metapopulation could persist if patches were connected by dispersal corridors. The design of nature reserves as networks of connected patches rather than single large preserves follows from metapopulation theory.

Source-sink dynamics

Pulliam (1988, Am. Nat. 132, 652-661) extended the metapopulation framework to distinguish source habitats (where local reproduction exceeds mortality, ${\lambda }_1>1$) from sink habitats (where local reproduction is insufficient, ${\lambda }_1<1$). Sink populations persist only through immigration from sources. This distinction is critical for conservation: protecting sink habitats without maintaining source populations is futile, because sinks are population sinks in the mathematical sense — they absorb individuals without contributing to long-term persistence.

The rescue effect (Brown & Kodric-Brown 1977, Ecology 58, 445-449) describes how immigration reduces extinction probability in sink patches: frequent immigration maintains genetic diversity and demographic stability. In the modified Levins model with rescue effect, the extinction rate becomes a decreasing function of $p$, producing a stronger persistence threshold and faster colonisation.

Full proof set [Master]

Proposition 1 (Global stability of logistic equilibrium). For the logistic equation $dN/dt=rN(1-N/K)$ with $r>0$ and $K>0$, the equilibrium $N^ = K$isgloballyasymptoticallystableon$(0, \infty)$.*

Proof. Define $V(N)=N-K-K\ln (N/K)$ for $N>0$. Then $V(K)=0$ and $V(N)>0$ for $N\ne K$ (since $x-1>\ln x$ for $x\ne 1$ by convexity of $\ln$). Compute:

$$\frac{dV}{dt}=\frac{dV}{dN}\frac{dN}{dt}=\left(1-\frac{K}{N}\right)\cdot rN\left(1-\frac{N}{K}\right)=r\frac{(N-K{)}^2}{K}\cdot \left(-\frac{1}{K}\right)\cdot (-K)=-\frac{r(N-K{)}^2}{K}\le 0.$$

Wait — let me redo this more carefully. $dV/dN=1-K/N$. So:

$$\frac{dV}{dt}=\left(1-\frac{K}{N}\right)\cdot rN\left(1-\frac{N}{K}\right)=r(N-K)\left(1-\frac{N}{K}\right)=-\frac{r(N-K{)}^2}{K}.$$

Since $dV/dt\le 0$ with equality only at $N=K$, and $V$ is proper on $(0,\infty )$, LaSalle's invariance principle gives global asymptotic stability of $N^{\ast }=K$. $\square$

Proposition 2 (Hutchinson delay stability boundary). The equilibrium $N^ = K$ofthedelayedlogisticequation$dN/dt = rN(t)(1 - N(t-\tau)/K)$isasymptoticallystableifandonlyif$0 < r\tau < \pi/2$.*

Proof. Linearise around $N^{\ast }=K$ by setting $x(t)=(N(t)-K)/K$:

$$\frac{dx}{dt}=-rx(t-\tau ).$$

Seek solutions of the form $x(t)=e^{\lambda t}$. Substituting: $\lambda =-r{e}^{-\lambda \tau }$, or $\lambda e^{\lambda \tau }=-r$. Write $\lambda =\mu +i\omega$. At the stability boundary, $\mu =0$, so $i\omega e^{i\omega \tau }=-r$, giving $i\omega (\cos \omega \tau +i\sin \omega \tau )=-r$. Separating real and imaginary parts: $-\omega \sin \omega \tau =-r$ and $\omega \cos \omega \tau =0$. The second equation gives $\omega \tau =\pi /2$ (the first positive solution with $\omega >0$), and substituting into the first: $\omega =r$, so $r\tau =\pi /2$. For $r\tau <\pi /2$, all eigenvalues have negative real part and the equilibrium is stable. For $r\tau >\pi /2$, at least one eigenvalue has crossed into the right half-plane, producing a Hopf bifurcation. $\square$

Connections [Master]

• ODE phase space and vector fields 02.12.01. The phase-plane analysis of Lotka-Volterra competition isoclines, predator-prey nullclines, and the logistic flow on the line are direct applications of the ODE theory in 02.12.01. The conserved quantity $H(x,y)$ in the predator-prey system is a first integral whose level curves define the trajectories. The Hopf bifurcation in the Rosenzweig-MacArthur model is a codimension-one bifurcation of the type studied there.

• Thermodynamics and energy flow 11.01.01 pending. Population growth rates are ultimately constrained by energy budgets: the growth rate $r$ is bounded by the rate at which organisms can convert consumed energy into biomass. The predator conversion efficiency $\delta$ in the Lotka-Volterra equations reflects ecological (Lindeman) efficiency — typically 10-20% between trophic levels. 11.01.01 pending provides the thermodynamic foundation for why these efficiencies are bounded.

• Community ecology 19.10.01 pending. The two-species Lotka-Volterra models developed here are the building blocks for multi-species community theory. The competition and predator-prey modules combine into food webs whose stability properties depend on interaction strength, connectance, and network architecture. May's 1972 complexity-stability result generalises from the two-species context to the full community.

• Hardy-Weinberg equilibrium 19.02.01 pending. Population ecology and population genetics share a foundational concern with how population size changes over time. The Hardy-Weinberg equilibrium assumes infinite population size; population ecology determines how large real populations actually are and how demographic fluctuations create the conditions under which HW assumptions break down.

• Genetic drift 19.04.01. Genetic drift is inversely proportional to effective population size $N_e$, and population ecology predicts how $N_e$ varies over time through demographic fluctuations. Predator-prey cycles that periodically drive prey to low density create bottlenecks that amplify drift, reducing genetic diversity. The interaction of ecological dynamics and drift is central to conservation genetics.

• Cellular metabolism 17.04.01. Population growth rates are bounded by cellular energy metabolism: the rate at which glycolysis, the citric acid cycle, and oxidative phosphorylation convert substrate to ATP determines the maximum biomass production rate. The per-capita growth rate $r$ in logistic models is the population-level expression of the cellular metabolic rate scaled by body size and temperature (the metabolic theory of ecology, Gillooly et al. 2001, Science 293, 2248-2251).

Historical & philosophical context [Master]

Alfred Lotka, a physical chemist at Johns Hopkins, derived the predator-prey equations in Elements of Physical Biology (1925, Williams & Wilkins) as part of an ambitious programme to apply physical and mathematical principles to biological systems ^{[Lotka 1925]}. Lotka's broader vision was a "physical biology" that would treat biological communities as physical systems governed by energy flow and mass-action kinetics. Vito Volterra, a mathematician at the University of Rome, independently derived the same equations in 1926 ^{[Volterra 1926]} at the request of his son-in-law Umberto D'Ancona, who had noticed that predator fish increased in proportion to prey fish during World War I when fishing was reduced in the Adriatic. Volterra's Variazioni e fluttuazioni del numero d'individui in specie animali conviventi (Mem. Accad. Lincei Roma 2, 31-113, 1926) analysed the oscillatory dynamics and proved the existence of periodic solutions.

Gause (1934, The Struggle for Existence, Williams & Wilkins) provided the experimental validation with Paramecium species in laboratory cultures ^{[Gause 1934]}. His experiments demonstrated competitive exclusion: when two species competed for the same bacterial resource, one invariably excluded the other. Gause's principle — that complete competitors cannot coexist — became the foundation of niche theory. The competitive exclusion principle was not a mathematical theorem in Gause's formulation but an empirical generalisation supported by controlled experiments.

Huffaker (1958, Hilgardia 27, 343-383) demonstrated predator-prey coexistence in experimental mite systems, showing that spatial heterogeneity could sustain oscillations that collapsed in homogeneous environments ^{[Huffaker 1958]}. His experiments with Eotetranychus sexmaculatus (prey) and Typhlodromus occidentalis (predator) on oranges connected by petroleum jelly barriers were the first to show that spatial refugia and dispersal limitation allow long-term predator-prey persistence. The result directly motivated metapopulation theory.

Robert May's 1972 Nature paper "Will a large complex system be stable?" ^{[May 1972]} fundamentally changed community ecology by showing that randomly assembled communities become less stable as complexity (species number, connectance, interaction strength) increases. Before May, ecologists generally believed that diverse communities were more stable. May's random-matrix analysis — using eigenvalue distribution theory from physics — showed that stability requires either weak interactions, special network structure, or specific sign patterns. May (1976, Nature 261, 459-467) then demonstrated chaos in discrete-time population models, showing that simple deterministic equations produce apparently random dynamics.

Chesson (2000, Annu. Rev. Ecol. Syst. 31, 343-366) unified coexistence theory by distinguishing stabilizing mechanisms (niche differentiation, storage effect, relative nonlinearity) from equalizing mechanisms (reducing fitness differences) ^{[Chesson 2000]}. This framework quantified the conditions for coexistence as the ratio of average niche difference to average fitness difference exceeding unity, providing a common currency for comparing coexistence mechanisms across systems.

Bibliography [Master]

@book{Lotka1925,
  author = {Lotka, Alfred J.},
  title = {Elements of Physical Biology},
  publisher = {Williams \& Wilkins},
  year = {1925},
}

@article{Volterra1926,
  author = {Volterra, Vito},
  title = {Variazioni e fluttuazioni del numero d'individui in specie animali conviventi},
  journal = {Mem. Accad. Lincei Roma},
  volume = {2},
  pages = {31--113},
  year = {1926},
}

@book{Gause1934,
  author = {Gause, Georgii F.},
  title = {The Struggle for Existence},
  publisher = {Williams \& Wilkins},
  year = {1934},
}

@article{Hutchinson1948,
  author = {Hutchinson, G. Evelyn},
  title = {Circular causal systems in ecology},
  journal = {Ann. N. Y. Acad. Sci.},
  volume = {50},
  pages = {221--246},
  year = {1948},
}

@article{Leslie1945,
  author = {Leslie, P. H.},
  title = {On the use of matrices in certain population mathematics},
  journal = {Biometrika},
  volume = {33},
  pages = {183--212},
  year = {1945},
}

@article{Holling1959,
  author = {Holling, Crawford S.},
  title = {Some characteristics of simple types of predation and parasitism},
  journal = {Can. Entomol.},
  volume = {91},
  pages = {385--398},
  year = {1959},
}

@article{Huffaker1958,
  author = {Huffaker, C. B.},
  title = {Experimental studies on predation: dispersion factors and predator-prey oscillations},
  journal = {Hilgardia},
  volume = {27},
  pages = {343--383},
  year = {1958},
}

@article{RosenzweigMacArthur1963,
  author = {Rosenzweig, Michael L. and MacArthur, Robert H.},
  title = {Graphical representation and stability conditions of predator-prey interactions},
  journal = {Am. Nat.},
  volume = {97},
  pages = {209--223},
  year = {1963},
}

@article{May1972,
  author = {May, Robert M.},
  title = {Will a large complex system be stable?},
  journal = {Nature},
  volume = {238},
  pages = {413--414},
  year = {1972},
}

@article{May1976,
  author = {May, Robert M.},
  title = {Simple mathematical models with very complicated dynamics},
  journal = {Nature},
  volume = {261},
  pages = {459--467},
  year = {1976},
}

@article{Levins1969,
  author = {Levins, Richard},
  title = {Some demographic and genetic consequences of environmental heterogeneity for biological control},
  journal = {Bull. Entomol. Soc. Am.},
  volume = {15},
  pages = {237--240},
  year = {1969},
}

@article{Chesson2000,
  author = {Chesson, Peter},
  title = {Mechanisms of maintenance of species diversity},
  journal = {Annu. Rev. Ecol. Syst.},
  volume = {31},
  pages = {343--366},
  year = {2000},
}

@book{Tilman1982,
  author = {Tilman, David},
  title = {Resource Competition and Community Structure},
  publisher = {Princeton University Press},
  year = {1982},
}

@article{Rosenzweig1971,
  author = {Rosenzweig, Michael L.},
  title = {Paradox of enrichment: destabilization of exploitation ecosystems in ecological time},
  journal = {Science},
  volume = {171},
  pages = {385--387},
  year = {1971},
}

@book{Kot2001,
  author = {Kot, Mark},
  title = {Elements of Mathematical Ecology},
  publisher = {Cambridge University Press},
  year = {2001},
}

@book{Begon2006,
  author = {Begon, Michael and Harper, John L. and Townsend, Colin R.},
  title = {Ecology: Individuals, Populations and Communities},
  edition = {4th},
  publisher = {Blackwell},
  year = {2006},
}

@book{Hastings1997,
  author = {Hastings, Alan},
  title = {Population Biology: Concepts and Models},
  publisher = {Springer},
  year = {1997},
}

@article{Pulliam1988,
  author = {Pulliam, H. Ronald},
  title = {Sources, sinks, and population regulation},
  journal = {Am. Nat.},
  volume = {132},
  pages = {652--661},
  year = {1988},
}

@article{ChessonWarner1981,
  author = {Chesson, Peter L. and Warner, Robert R.},
  title = {Environmental variability promotes coexistence in lottery competitive systems},
  journal = {Am. Nat.},
  volume = {117},
  pages = {923--943},
  year = {1981},
}