How Many Species Does a Job Take?
Report — Field Note
How Many Species Does a Job Need?
Clean water, pollinated crops, carbon locked away in the sea: each is work done by living things. A synthesis of 423 studies asks how many kinds of living thing each job needs, and the answers come in three shapes.
Ecologists test the question in plots: patches of ground or tanks of water stocked with different numbers of species, with some useful output measured in each, such as how much the plants grow or how clean the water runs. They call these outputs ecosystem services and functions; here they are simply jobs. If every species does a slightly different job, each one added raises the output by about the same amount, and the line climbs straight.
If several species do the same work, the line bends. The first few carry most of the load and later ones add little, because they duplicate work already covered. Ecologists call this redundancy. A redundant species looks idle in the measurement, yet it can act as a spare: if one of the working species is lost, another can take up its share.
Emma Moffett and colleagues at Imperial College London gathered 423 studies, nearly 2,000 sets of measurements, and fitted both shapes to each. Across 23 kinds of job, the straight line fitted best for 10 and a bending curve for 13. Water quality, pollination and protection from hazards mostly ran straight, suggesting each species adds something. Carbon storage in the ocean, and the category the field calls "learning and inspiration", bent over.
Where the curve flattens depends on the habitat. Counting species, the bend came at a median of about 4 in fresh water, 14 on land and 23 in the sea. The authors suggest freshwater food webs are smaller and their members more interchangeable, while the sea's long chains of who eats whom, and partnerships such as those on coral reefs, need more species before extra ones stop counting.
Shape is one question and strength is another. The team also measured how closely each job tracked the number of species. Most rose with diversity. Carbon storage in the ocean tracked it most closely of all, by a wide margin. Air quality, and protection from floods, erosion and storms, barely tracked it at all.
Hazard protection is the instructive case. A sand dune that shields a coast from flooding is often held in place by one or two kinds of shrub. Adding other species does not make the dune hold any better, so the measurement shows little effect. The authors point out that the shrub's own survival may still depend on the community around it, a dependence that a count of species in one plot cannot see.
The size of the patch changes the answer. Across all jobs, the effect of diversity grew stronger as the area measured grew larger, from square metres toward whole landscapes. A likely reason is that different places hold different species, so a large area combines more distinct workers than any single plot. Hazard protection and plant growth went the other way, weakening at larger scales.
The evidence is lopsided. More than four in five of the datasets come from land, and only around twenty from estuaries. Most sites sit in wealthy countries, with Germany heavily represented because of a few long-running experiments. The strongest single result in the study, carbon storage in the ocean, rests on seven datasets.
A synthesis this broad buys reach at the cost of grip. For most jobs the number of species explained less than 40 per cent of the variation, so diversity is one driver among several. Many datasets are observations of wild places, where the arrow can run both ways: a productive patch may gather more species as easily as more species make a patch productive. The authors also warn that some straight lines may be straight only because the richest systems are rarely sampled, and a curve followed too short a distance cannot show its bend. The paper's forecast, that natural pest control on farms weakens most under a fossil-fuelled future and in fast-growing, poorer countries, treats species as lost at random, a limitation the authors list themselves.
The pattern underneath is about arrangement. Whether a job needs many species, a few, or one depends on how the work is divided: spread widely, duplicated, or carried by a single species that itself rests on the rest of the community. The shape of each curve records not how many species are present but how the work is divided among them, and that division decides what goes first when species are lost. The term to search is the insurance hypothesis, Shigeo Yachi and Michel Loreau's 1999 account of how species that look redundant keep a job running when conditions change.
Sourced — Moffett, Gayford, Chen et al., "Biodiversity safeguards ecosystem services and functions worldwide", Nature Ecology & Evolution (published Oct 6 2026) · Yachi & Loreau, PNAS 96:1463 (1999) · Effect-size values other than ocean carbon (1.48), air quality (−0.03) and hazards (~0) are approximate, read from the paper's Fig. 3a. Diagrams are schematic.