The Species-Area Relationship

The species-area relationship states that the number of species S found in a sample increases with the area A sampled, and that this increase follows a power function:

S = cA^z Here c is a constant (species richness at unit area, which depends on taxon, region, and productivity) and z is the scaling exponent that describes how steeply richness climbs with area. Taking logarithms of both sides converts the power law into a straight line, which is why the relationship is almost always plotted and fitted on log-log axes:

log S = log c + z log A On this transformed scale, log c is the y-intercept and z is the slope. This log-log linearity is the empirical signature ecologists look for: when it holds cleanly across many orders of magnitude of area, from square centimeters to square kilometers, it signals that a genuinely scale-invariant process, not an artifact of a particular study, is at work. Two families of species-area curves are recognized, and confusing them is a classic error. Type I curves come from nested quadrats within a single, continuous habitat: you place a small plot, then a bigger one containing the first, then a bigger one still, all in the same community, and count cumulative species. These tend to have relatively low z-values (often 0.1–0.2) because you are mostly re-sampling the same habitat and community, with diminishing returns as rarer species are added. Type II curves compare separate, discrete patches, classically islands of different sizes, and here z-values run considerably higher (often 0.25–0.45), because bigger islands differ not just in individuals sampled but in habitat diversity, isolation, and the demographic stability that lets species persist rather than go locally extinct. Distinguishing "one continuous area sampled at increasing scale" from "many separate isolated patches of differing size" is essential to interpreting any z-value correctly.

Three non-exclusive mechanisms drive the pattern, operating simultaneously but with different relative weight depending on the system.

1. Passive sampling (species-individuals relationship). A larger area simply contains more individual organisms, and if you draw more individuals from a community, you are increasingly likely to encounter additional species, especially rare ones. This is a direct, almost statistical, consequence of unequal, typically lognormal, species-abundance distributions: a few species are common, many are rare, and rare species only show up once your sample of individuals is large enough. This mechanism alone, applied to a lognormal abundance distribution, was shown by Frank Preston to generate a power-law species-area curve. It requires no habitat heterogeneity or biogeography at all, just statistics acting on uneven abundances.

2. Habitat heterogeneity. Larger areas usually encompass a greater diversity of habitat types, soil types, microclimates, elevational bands, vegetation structures, and each additional habitat type brings its own subset of specialist species. A 1-hectare plot of forest might contain only one soil type and one canopy stratum; a 1000-hectare landscape spans ridges, valleys, streamsides, and canopy gaps, each favoring different species. This mechanism predicts that z should be steeper in physically heterogeneous, topographically complex regions than in flat, homogeneous ones, a pattern confirmed repeatedly in comparative surveys.

3. Dynamic equilibrium: colonization versus extinction (island biogeography). On true islands (or habitat patches functioning like islands, mountaintops, lakes, fragments in a deforested matrix), species richness at any moment reflects a balance between the rate at which new species arrive (immigration) and the rate at which resident species die out locally (extinction). Larger islands support larger populations of each resident species; larger populations have a lower probability of stochastic extinction. Larger islands also tend to intercept more colonists (larger target for dispersing propagules) and to offer greater habitat diversity once established. The equilibrium number of species is therefore higher on large islands than small ones, and this equilibrium is dynamic: species compositions turn over continually even while total richness stays roughly constant. Step by step, the equilibrium argument works like this: 1. Colonists arrive from a mainland source pool at some rate; this rate declines as the island fills up, because fewer of the arriving propagules belong to species not already present. 2. Established species go locally extinct at some rate; this rate rises as more species are packed onto the island, because population sizes per species shrink (more species sharing a fixed resource base) and competitive exclusion becomes more likely. 3. The point where the colonization curve crosses the extinction curve defines the equilibrium species number, Ŝ. 4. Because large islands support larger populations, their extinction curve sits below that of small islands at every S; because large islands are usually easier targets to hit and offer more niches, their colonization curve often sits above; both effects push Ŝ upward for large islands. 5. Because near islands receive more colonists per unit time than far islands, the equilibrium for near islands sits at higher S than for far islands, at the same island size, giving the area effect and the distance effect their separate, additive-ish signatures.

Preston (1962): the lognormal link. Frank Preston, analyzing bird and moth census data, showed that if species abundances within a community follow a lognormal distribution (a few very common species, many progressively rarer ones, a pattern seen almost universally when large numbers of individuals are censused), then simply sampling larger and larger areas will pull in progressively rarer species at a predictable, decelerating rate. Preston derived that this sampling process alone generates a species-area curve well approximated by a power law with z typically between 0.2 and 0.4, matching what field surveys already showed. This was the first rigorous demonstration that the SAR need not invoke habitat heterogeneity or biogeography at all — it can emerge purely from the statistics of sampling an unevenly abundant community. Crucially, Preston also showed the curve should flatten ("Preston's veil line") once area is large enough to sample essentially the whole regional species pool, explaining why z tends to decline at very large spatial scales.

MacArthur and Wilson (1967), The Theory of Island Biogeography. Robert MacArthur and E. O. Wilson formalized the colonization-extinction equilibrium model described above, deriving explicit predictions: richness should increase with island area (larger population sizes, lower extinction), and decrease with distance from the source of colonists (lower immigration rate). Their book made falsifiable, quantitative predictions and explicitly framed island richness as a dynamic equilibrium (species turning over) rather than a static ceiling, a genuinely novel claim at the time.

Simberloff and Wilson's mangrove island experiment (1969). The turnover claim needed a direct experimental test, and Daniel Simberloff, then a graduate student under E. O. Wilson, provided one of ecology's most elegant field experiments. In the Florida Keys, small mangrove islands (a few meters across, isolated by open water, each a natural "island" for arthropods) were surveyed for their resident arthropod faunas, then completely defaunated using methyl bromide fumigation, killing every insect, spider, and other arthropod on the island while leaving the mangrove trees themselves alive. The islands were then monitored as they were recolonized from scratch. The results were striking:

• Within about a year, each island had returned to approximately its original species number, even though the identity of individual species differed from the pre-fumigation community — direct proof of dynamic turnover, not a fixed guest list.
• Recolonization followed the sequence predicted by the model: fast initial colonization by good dispersers, followed by a slower approach to equilibrium richness as extinction rates rose with crowding.
• Larger islands (surveyed at varying sizes) reached higher equilibrium richness, and islands closer to the mangrove-rich mainland recolonized faster, than distant, smaller islands, confirming both the area effect and the distance effect predicted by MacArthur and Wilson.

This experiment is the closest thing island biogeography has to a controlled test with replicates and a true experimental manipulation (deliberate extinction, then observed recovery), and it is why the theory moved from an elegant curve-fitting exercise to a mechanistically supported, falsifiable model.

Modern confirmation across taxa and scales. Subsequent compilations (for beetles, land snails, reptiles, vascular plants, freshwater fish in lakes, and habitat "islands" such as mountaintop forest patches or fragmented rainforest) have repeatedly recovered log-log linear SARs with z typically in the 0.1–0.5 range, cementing the pattern's generality while also revealing systematic departures (steeper z on remote oceanic archipelagos, shallower z within continuous mainland habitat) that map neatly onto the sampling/heterogeneity/equilibrium mechanisms above.

Example 1 — Fitting z and c from island survey data, then predicting a new island's richness. Suppose a biogeographic survey of an island chain yields the following bird species counts:

Take logs (base 10):

The points fall almost perfectly on a line with slope

z ≈ (1.81 − 0.70) / (3.00 − 0.00) = 1.11 / 3.00 ≈ 0.37 and intercept log c ≈ 0.70, so c ≈ 10^0.70 ≈ 5. The fitted model is S ≈ 5·A^0.37. Now predict richness for a fifth island of area 5,000 km²: S = 5 × 5000^0.37 log S = log 5 + 0.37 × log 5000 = 0.70 + 0.37 × 3.70 = 0.70 + 1.37 = 2.07 S ≈ 10^2.07 ≈ 117 species This z of 0.37 is right in the typical range for oceanic island archipelagos (0.25–0.45), consistent with strong isolation effects layered on top of habitat heterogeneity and sampling.

Example 2 — Predicting extinction from deforestation (the conservation calculation). A tropical forest region originally covers 100,000 km² and, from regional surveys, harbors an estimated 4,000 tree species, with a fitted z of 0.25 (typical for a continuous mainland forest block, Type I-like heterogeneity-driven scaling). Logging and conversion to agriculture reduce standing forest to 10,000 km², a 90% areal loss, leaving only 10% of the original habitat. Using S₂/S₁ = (A₂/A₁)^z: A₂/A₁ = 10,000 / 100,000 = 0.10 S₂/S₁ = 0.10^0.25 log(0.10^0.25) = 0.25 × log(0.10) = 0.25 × (−1) = −0.25 S₂/S₁ = 10^(−0.25) ≈ 0.562 So the model predicts the forest retains about 56% of its original species, meaning roughly 44% of the 4,000 species, about 1,760 species, are predicted to be lost (eventually, as extinction debt is paid off), even though 90% of the area was destroyed. This asymmetry, losing "only" 44% of species after losing 90% of area, is the single most important quantitative lesson of the SAR for conservation: because z is well below 1, species loss lags badly behind habitat loss in percentage terms, but the absolute numbers (1,760 tree species here) are still catastrophic, and the loss is not instantaneous. Species with small remaining populations may persist for decades as "the living dead" before finally going extinct, a lag called extinction debt, which is why post-hoc surveys taken too soon after fragmentation can badly underestimate the ultimate toll predicted by this calculation.

Robert H. MacArthur was a mathematically inclined ecologist trained under G. Evelyn Hutchinson at Yale, known for insisting that ecology could and should generate quantitative, falsifiable theory rather than purely descriptive natural history. Edward O. Wilson, a myrmecologist (ant specialist) at Harvard with deep field experience across tropical islands and archipelagos, brought the natural-history grounding and taxonomic breadth that kept the theory tethered to real biogeographic data. Their collaboration, beginning with a 1963 paper and culminating in the 1967 monograph The Theory of Island Biogeography, is often cited as the moment community and biogeographic ecology became a genuinely predictive, mathematical science, an ambition that also shaped MacArthur's later work on niche theory and consumer-resource dynamics.

Reserve design and the SLOSS debate. Conservation planners use SAR predictions to weigh "single large or several small" (SLOSS) reserve configurations: because z < 1, one large reserve typically retains more species than several small reserves of the same total area, especially for area-sensitive species, though the answer depends on species turnover between fragments (beta diversity) and is still actively debated case by case.

Extinction risk assessment from deforestation and land-use change. International bodies and researchers use S₂/S₁ = (A₂/A₁)^z relationships, calibrated with region- and taxon-specific z-values, to estimate the number of species likely to be lost from continuing tropical deforestation, feeding into global extinction-rate estimates and into the IUCN Red List's habitat-loss-based criteria.

Corridor and connectivity design. Because isolation lowers the effective colonization rate (shifting the equilibrium to lower Ŝ even at fixed patch area), wildlife corridors and stepping-stone habitats are designed explicitly to counteract the "distance effect" documented by MacArthur and Wilson, raising effective connectivity between fragments.

Agricultural landscape and biocontrol planning. SAR logic informs how much semi-natural habitat (hedgerows, field margins, remnant woodlots) is needed within farmland to sustain pollinator and natural-enemy diversity sufficient for pest control and crop pollination services.

Marine protected area sizing. The same power-law logic, adapted for reef area, seagrass bed area, or protected-zone area, is used to set minimum sizes for marine reserves intended to sustain fish and invertebrate diversity, though z-values and mechanisms differ somewhat from terrestrial systems due to larval dispersal dynamics.

Ask the tutor: How does the choice of z-value change if I'm predicting extinction from fragmentation (many small isolated patches) versus a single contiguous habitat loss of the same total area?