Niche Partitioning

From Gause to coexistence

Georgy Gause grew two species of Paramecium — P. aurelia and P. caudatum — together in a flask in 1934. They eat the same bacteria, live in the same water, occupy the same physical space. P. aurelia consistently wiped out P. caudatum within a few weeks. Run the same Paramecium against a third species, P. bursaria, that feeds slightly lower in the flask, and both persisted indefinitely. The conclusion lives on as the competitive exclusion principle: two species cannot indefinitely occupy the same ecological niche.

That immediately raised a problem. Real communities are full of species that look similar and live together. Five or six warblers forage on the same spruce tree. Twenty-plus African ungulates graze the same Serengeti plain. Eighteen finch species share the Galápagos archipelago. If Gause is right, why are they not extinct?

The answer is that looking similar is not the same as occupying the same niche. Look closely and you find subdivisions — by food size, by tree zone, by time of day, by trophic role. The community has partitioned the niche. That partitioning lowers the cross-species competition coefficient α below 1, and Lotka-Volterra theory then permits stable coexistence.

The four canonical axes

	Resource	Habitat	Temporal	Trophic
What is divided	Food type or size	Physical space	Time of activity	Position in food web
Granularity	Continuous (seed size)	Discrete or continuous (zones)	Daily, seasonal, annual	Discrete (predator vs grazer)
Iconic example	Galápagos finch beaks	MacArthur's warblers	Diurnal/nocturnal raptors	African ungulate guild
Detection method	Diet analysis (gut, isotope)	Foraging maps	Camera traps, activity logs	Body-size and stable-isotope
Evolutionary driver	Character displacement	Habitat selection	Activity-time evolution	Body-size and dentition
Time scale to evolve	10² – 10⁴ generations	10² – 10³ generations	10² – 10³ generations	10⁴ – 10⁶ generations
Conservation relevance	Predicts seed-size shifts	Predicts habitat-loss winners	Predicts light-pollution losers	Predicts predator-loss cascades

MacArthur's warblers — habitat partitioning

Five species of Dendroica warblers (now Setophaga) breed in the same northeastern spruce-fir forests of New England: bay-breasted, Cape May, black-throated green, blackburnian, and yellow-rumped (myrtle). All eat insects, all weigh 8–14 g, all migrate from the same wintering grounds. They look like Gause's prediction of certain extinction.

Robert MacArthur's 1958 study used field-glasses and a stopwatch to track where each species foraged on a single spruce tree. He found:

• Cape May: top of the canopy, outer twigs.
• Blackburnian: top, middle of the branches.
• Black-throated green: middle of the canopy, mostly outer.
• Bay-breasted: middle to lower canopy, inner branches.
• Yellow-rumped (myrtle): lower canopy and ground.

Five species, five foraging zones, almost no overlap. Each species also showed slight differences in feeding behaviour (hovering vs probing) and prey size, fine-tuning the partitioning further. The paper turned habitat partitioning from a verbal claim into a quantitative observation and is one of the most-cited papers in twentieth-century ecology.

Galápagos finches — resource partitioning

Eighteen finch species across the Galápagos archipelago descend from one ancestral pair that arrived about 2 million years ago. Beak depth — a proxy for the maximum seed size each bird can crack — varies from 6 mm in cactus finches to 23 mm in large ground finches. The ground-finch genus Geospiza shows the cleanest example.

• On Daphne Major (small island, both species present), G. fortis beak depth averages 9.4 mm and G. fuliginosa averages 7.7 mm — a 22 percent gap.
• On Los Hermanos (only G. fortis present), G. fortis beak depth averages 8.6 mm — closer to the smaller species' size.
• On Daphne, G. fortis eats predominantly large seeds and G. fuliginosa small seeds; the diet overlap is below 30 percent.

This is character displacement — divergent evolution forced by sympatric competition. Peter and Rosemary Grant's 40-year mark-recapture study on Daphne Major documented a real-time selection event during the 1977 drought: G. fortis beaks shifted by 4 percent in a single year because birds with deeper beaks could crack the only large hard seeds left.

The Lotka-Volterra coexistence condition

The competition form of Lotka-Volterra writes the two species' dynamics as:

dN₁/dt = r₁N₁(1 − (N₁ + α₁₂N₂)/K₁)
dN₂/dt = r₂N₂(1 − (N₂ + α₂₁N₁)/K₂)

Stable coexistence requires:

α₁₂ < K₁/K₂ AND α₂₁ < K₂/K₁

Multiplying the two yields α₁₂ · α₂₁ < 1. In words: each species must compete more strongly with itself than with the other. Niche partitioning is the mechanism that pushes the αs below 1. Identical species (αs = 1) lie exactly on the boundary; the slightest perturbation tips them to extinction.

Empirical α estimates for sympatric warblers, finches, and bumblebees typically fall between 0.2 and 0.7 — well within the coexistence region. For species pairs that recently invaded a community and have not yet partitioned, αs near 1 are observed and competitive exclusion is the predicted outcome.

Empirical evidence

• African ungulate guild (Serengeti). Fifteen sympatric grazers partition by grass height and migration. Wildebeest crop tall grass; zebra eat coarse stems; Thomson's gazelle clean up short regrowth.
• Anolis lizards on Caribbean islands. Four ecomorph types — trunk-ground, crown, twig, grass-bush — recur independently across Cuba, Hispaniola, Jamaica, and Puerto Rico. Convergent evolution of the same partitioning solution.
• African elephants vs hippos. Elephants browse and break trees; hippos graze grass at night. Resource and temporal partitioning let two megaherbivores share the same waterhole.
• Diurnal vs nocturnal raptors. Owls and hawks share prey species but partition by time of day. Light pollution erodes the partition and forces stronger direct competition.

Diagram sketch

• Panel A — warbler tree. A vertical spruce silhouette with five overlaid heat zones marking each warbler species's preferred foraging height and depth. Zones overlap less than 15 percent.
• Panel B — finch beak histogram. Two histograms of beak depth: one for sympatric island (well-separated bimodal), one for allopatric island (single broad mode). Character displacement made visible.
• Panel C — niche-axis schematic. Three Gaussian curves on a one-dimensional resource axis, showing low overlap. Add a second axis for two-dimensional partitioning of habitat × time.

Pitfalls

• The ghost of competition past. Joe Connell pointed out in 1980 that observed niche differences may reflect old selection, not current competition. Field experiments removing one species often produce no measurable response in the other — suggesting partitioning happened long ago and current populations no longer interact.
• Niche may not partition completely. Many real species pairs overlap by 50 percent or more on every measured axis. Possible explanations: unmeasured axes, fluctuating environments that swap competitive advantage (storage effect), or near-neutral dynamics.
• One axis is not enough. Two-dimensional niche space allows species to overlap on each axis individually but stay distinct in combination. Studies of single axes routinely overstate competition.
• Niche width is plastic. Generalists narrow their diet under competition; specialists do not. Realised niche measured in sympatry differs from fundamental niche measured alone.
• Confusing niche with habitat. Habitat is where you find a species; niche is what it does there. Two species can share a forest while occupying different niches (canopy vs ground).

Variants and modern frameworks

• Hutchinson's n-dimensional niche. The original 1957 framework — every niche axis is one dimension, the niche is a hypervolume in n-dimensional space.
• Storage effect (Chesson). Coexistence through fluctuation: species are favoured in different years and long-lived stages "store" the advantage. Allows coexistence even with αs near 1.
• Limiting similarity (MacArthur and Levins). Formula for the minimum niche separation that allows coexistence given stochasticity. Predicts species packing density.
• Modern coexistence theory (Chesson). Decomposes coexistence into stabilizing (lower α) and equalizing (closer K, r) mechanisms.