Measuring Biodiversity

Why measuring biodiversity matters

• You cannot manage what you cannot count. Conservation budgets, protected-area design, and international targets (the 2022 Kunming-Montreal Global Biodiversity Framework aims to protect 30% of land and sea by 2030) all rest on quantified diversity. A number lets you rank sites, detect declines, and demonstrate that an intervention worked.
• A species list hides the collapse. Communities routinely lose most of their functional diversity while every species is technically still present, because abundances crash toward dominance by a few tolerant taxa. Evenness-sensitive indices catch this; a checklist does not.
• Sampling effort distorts raw counts. Find 50 species in a rainforest plot and 20 in a desert plot and you have learned nothing until you correct for the fact that you counted 10,000 rainforest individuals and 300 desert ones. Rarefaction makes the comparison honest.
• Not all species are equal. Losing the tuatara (Sphenodon, sole survivor of an order 250 million years old) erases far more unique biology than losing one more passerine. Phylogenetic and functional metrics let conservation prioritize evolutionary and ecological distinctness, as the Zoological Society of London's EDGE programme does.
• Evenness underpins stability. The Cedar Creek grassland experiments (David Tilman, from 1994) showed that both richness and evenness independently raise productivity and buffer ecosystems against drought — the insurance hypothesis. Redundant species covering the same role are what keep function from collapsing when one is lost.
• Beta diversity decides reserve strategy. The classic SLOSS debate — a Single Large Or Several Small reserve — is really a question about beta diversity. Where compositional turnover between sites is high, several complementary reserves capture more of the regional species pool than one big one.

How biodiversity metrics work, step by step

Start with a community sample: a list of species and how many individuals of each you counted. Convert those counts into proportional abundances, p_i = n_i/N, where n_i is the count of species i and N is the total. These proportions are the raw material of nearly every index.

Richness (S) is just the number of species with p_i > 0. It ignores abundance entirely: a community with one dominant and nineteen singletons has the same richness (20) as one with twenty equally common species. Richness is intuitive but effort-dependent and blind to structure.

The Shannon index, H′ = −Σ p_i ln p_i, measures the uncertainty in predicting the species of a randomly drawn individual — high when many species are present and abundances are even. It ranges from 0 (one species) to ln S (perfect evenness). Because it takes the logarithm of small proportions, adding a rare species measurably raises H′, making Shannon relatively sensitive to rare taxa. Dividing by ln S gives Pielou's evenness, J′ = H′/ln S, a clean 0-to-1 evenness score that separates the evenness signal from richness.

The Simpson index, D = Σ p_i², is the probability that two individuals drawn at random are the same species — a dominance measure. Squaring the proportions means the most abundant species dominate the sum, so Simpson is insensitive to rare species. Ecologists report its complements so bigger means more diverse: Gini-Simpson (1 − D), the probability two random individuals differ, and inverse Simpson (1/D), which reads as an effective number of common species.

Hill numbers unify the family. Each index becomes an effective number of species — the count of equally abundant species that would yield the same value. The order q sets abundance weighting: ^0D = richness, ^1D = exp(H′), ^2D = 1/D. Plotting ^qD against q draws a diversity profile: a flat line means an even community; a steep drop from q=0 to q=2 signals strong dominance by a few species.

Partition across scales. Whittaker's alpha is within-site diversity, gamma is the whole region, and beta = gamma/alpha is turnover. With Hill numbers this decomposition is mathematically exact and the beta component is a true effective number of distinct communities, which older Shannon/Simpson partitions could not guarantee.

Correct for effort and go beyond names. Rarefaction re-samples the richer dataset down to equal individuals or equal coverage; Chao1 extrapolates true richness from singletons and doubletons. Then functional diversity weights species by traits (Rao's quadratic entropy folds trait dissimilarity into a Simpson-like sum) and phylogenetic diversity (Faith's PD) sums the branch lengths of the evolutionary tree spanning the species present.

Comparing the core diversity indices

Metric	Formula	What it captures	Sensitivity to rare species	Hill order q
Species richness (S)	count of species	How many species	Maximal (every species counts once)	q = 0
Shannon (H′)	−Σ pi ln pi	Richness + evenness, entropy	Moderate	exp(H′) = q = 1
Simpson (D)	Σ pi²	Dominance / concentration	Low (weights common species)	1/D = q = 2
Gini-Simpson	1 − Σ pi²	Prob. two individuals differ	Low	related to q = 2
Pielou's evenness (J′)	H′ / ln S	Evenness alone, 0–1	—	ratio of q=1 to q=0
Inverse Simpson	1 / Σ pi²	Effective common species	Low	q = 2

Alpha vs beta vs gamma diversity

Property	Alpha (α)	Beta (β)	Gamma (γ)
Spatial scale	Within one site / sample	Between sites	Whole region / landscape
What it measures	Local diversity	Compositional turnover	Total regional diversity
Whittaker relation (1960)	mean local value	β = γ / α	α × β
High value means	Rich local community	Sites are differentiated	Rich region overall
Driven by	Habitat quality, area	Environmental gradients, dispersal limits	Both α and β combined
Conservation use	Rank individual sites	SLOSS, reserve complementarity	Regional targets, hotspots

Common misconceptions

• "More species always means more diverse." Two communities with identical richness can have wildly different diversity if one is dominated by a single species. Diversity indices exist precisely because richness alone is misleading — evenness is half the story.
• "The Shannon index is measured in species." H′ is an entropy in nats (or bits if you use log base 2), not a species count. Only its exponential, exp(H′), is an effective number of species. Reporting a bare Shannon value of "2.3" without converting to Hill units makes cross-study comparison treacherous, because 2.3 nats is not "2.3 species."
• "A higher Simpson's D means higher diversity." Raw Simpson's D is a dominance/concentration measure — it goes up as diversity goes down. You must report 1 − D or 1/D for the "bigger is more diverse" intuition to hold. This sign confusion is one of the most common errors in the literature.
• "You can compare raw richness across samples." Not without equalizing effort. Because richness rises with the number of individuals counted, comparing an intensively sampled site with a lightly sampled one always favors the former. Rarefaction or coverage standardization is mandatory.
• "Shannon and Simpson give the same ranking." They can disagree, and the disagreement is informative: Shannon weights rare species more, Simpson weights common ones. If two communities swap rank between the two indices, the difference lies in their rare-species tails — exactly what a Hill diversity profile reveals.
• "Species diversity captures everything worth protecting." Ten grass species are less functionally and evolutionarily diverse than a grass, a fern, a beetle, a fungus, and a bird. Functional and phylogenetic diversity capture ecological role and evolutionary history that a species count and even Shannon and Simpson miss entirely.

Species-abundance distributions and why evenness matters

Every diversity index is a summary of the underlying species-abundance distribution (SAD) — the full list of how many individuals each species has. Plotted as a rank-abundance (Whittaker) curve, abundance on a log axis against rank, the SAD reveals structure a single number hides. A flat curve is highly even; a steep curve signals dominance. Classic models describe these shapes: the geometric series (very uneven, niche-preemption communities), Frank Preston's 1948 lognormal (the typical hump-shaped pattern of large, well-sampled communities), and R. A. Fisher's 1943 log-series, whose parameter α (Fisher's alpha) is itself a widely used, sample-size-robust diversity index. Stephen Hubbell's 2001 neutral theory predicts a zero-sum multinomial distribution from birth, death, immigration, and speciation alone, and fits many tropical forest SADs surprisingly well.

Evenness matters because ecosystem function is delivered by populations, not by names on a list. A pollinator present as a single individual pollinates almost nothing; a decomposer at one cell per liter decomposes almost nothing. Eutrophication, overfishing, and nitrogen deposition characteristically erode evenness first — a few tolerant species bloom while the rest dwindle — so richness can stay flat for years while Shannon and Simpson diversity fall. That decoupling is why an evenness-blind checklist can pronounce a degrading ecosystem "healthy" right up until species start disappearing.

Conservation prioritization: from indices to decisions

Metrics become policy when they rank what to save. Beta diversity and complementarity drive reserve-selection algorithms (Marxan, Zonation) that assemble a network capturing the most species for the least cost — the answer to the SLOSS problem depends directly on turnover between candidate sites. Phylogenetic diversity underlies ZSL's EDGE of Existence programme, which ranks species by Evolutionary Distinctness and Global Endangerment, so a lineage with no close relatives and high extinction risk (the aye-aye, the Chinese giant salamander) is prioritized over an equally threatened but phylogenetically redundant one. Functional diversity and redundancy flag ecosystems where several species share a role (resilient) versus those where a single species performs a role alone (fragile) — the loss of that lone functional actor, a keystone, can trigger a trophic cascade. Norman Myers' 1988 biodiversity hotspots and Conservation International's later refinement combined high endemism (a beta/gamma signal) with habitat loss to direct global funding; the 36 recognized hotspots cover about 2.4% of Earth's land yet hold more than half of all endemic plant species.

A worked example and its history

• The worked case. Consider two woodlots, each with four species and 100 trees. Woodlot A is perfectly even (25 each): H′ = −4 × (0.25 × ln 0.25) = 1.386 nats, exp(H′) = 4 effective species, D = 4 × 0.25² = 0.25, and 1/D = 4. Woodlot B is dominated (97, 1, 1, 1): H′ ≈ 0.17 nats, exp(H′) ≈ 1.18, D ≈ 0.941, 1/D ≈ 1.06. Same richness, radically different diversity — the numbers make the collapse visible.
• Fisher, Corbet, and Williams (1943). From light-trap moth catches at Rothamsted, they derived the log-series distribution and its parameter α — the first widely adopted diversity index, prized because it is nearly independent of sample size.
• Edward H. Simpson (1949). Drawing on statistical work he did at Bletchley Park during the war and publishing in Nature, Simpson defined the concentration index D = Σ p_i² as the probability that two random draws match — one of the most cited short papers in ecology.
• Robert Whittaker (1960, 1972). Studying plant communities across the Siskiyou and Great Smoky Mountains gradients, Whittaker introduced alpha, beta, and gamma diversity and the rank-abundance curve, formalizing how diversity scales across space.
• Sanders and Hurlbert (1968, 1971). Sanders proposed rarefaction to compare deep-sea benthic diversity across unequal samples; Hurlbert corrected his math to the hypergeometric formulation still used today.
• Daniel Faith (1992). Faith defined phylogenetic diversity as the summed branch length of the minimum spanning tree of a set of species, giving conservation a way to value evolutionary history directly. Anne Chao, Lou Jost, and Mark Hill later unified richness, Shannon, Simpson, functional, and phylogenetic diversity under the Hill-number framework that now dominates the field.