# Biodiversity and Ecosystem Function

> A grassland plot's production is the sum of its species, and in experiments that assigned richness at random, more species generally raised that sum and steadied it from year to year; how much steadier depends on the species' sizes and swings and on how far they rise and fall together. This lesson follows the experiments that assigned richness (Cedar Creek in Minnesota, BIODEPTH across eight European sites), splits the extra yield into complementarity and selection with the Loreau-Hector partition, and separates temporal stability from resistance and resilience. Afterwards you can compute net, complementarity and selection effects from monoculture and mixture yields, compute stability as mean over standard deviation, and design a study that tells a diversity effect from a soil effect.

Course: Ecology and Evolution (Ecosystem processes). Level: core university (third year). About 20 minutes.
Lesson page (link this one when citing): https://stemhop.com/lessons/biodiversity-ecosystem-function/

Builds on: Energy Flow and Productivity · Diversity, Disturbance and Succession · Competition

## What you'll learn

- State the central result of the randomized biodiversity experiments at Cedar Creek and in BIODEPTH, that richer plant communities generally produce more aboveground biomass and steadier total production, explain how assigning richness at random supports that conclusion, and say how field syntheses extend the question.
- Calculate net, complementarity and selection effects from monoculture and mixture yields with the Loreau-Hector partition, and explain why the partition does not identify the biological mechanism.
- Define temporal stability as mean over standard deviation of community biomass, distinguish it from resistance and resilience, and explain statistical averaging and compensatory dynamics.
- Design a study that separates an effect of richness from environmental confounding such as soil fertility or moisture.

**In grassland experiments that assigned the number of plant species at random, richer plots generally grew more aboveground biomass, and their totals varied proportionally less from year to year.**

> **The one idea.** A plot’s production is the sum of what its species produce. A mixture yields extra when that sum beats the yield expected from its monocultures, and the sum can steady from year to year because species’ swings partly cancel, unless the species rise and fall together.
>
> *Scope:* aboveground plant biomass in communities assembled at random, within the richness tested (1 to 16 species at Cedar Creek) and the years measured. Expected yield is each species’ monoculture yield times its share of the seed; with an even sowing, that is the average monoculture, not the best. Out-of-step swings help but do not by themselves make a total steadier: that also depends on the species’ sizes and swings, and on whether one species dominates. “Steadier” means steady across years, not resisting one drought or recovering afterwards.

## The question: does richness change what an ecosystem does?

An **ecosystem function** is a rate or stock of a process: a field’s yearly plant production (primary productivity, from [energy flow](https://stemhop.com/lessons/energy-flow/)), the nitrogen it holds, how fast litter decays. **[Species richness](https://stemhop.com/methods/biodiversity-metrics/)** is the number of species present, the simplest diversity measure from [diversity and succession](https://stemhop.com/lessons/diversity-and-succession/).

Surveys of real grasslands often find more biomass where there are more species, though the strength and even the direction of that relationship vary between studies. Moisture or fertility could raise both richness and biomass (the first film’s wet patch); such a third variable is a **confounder**. So ecologists built communities instead of finding them.

## Building communities at random: Cedar Creek and BIODEPTH

At Cedar Creek, Minnesota, Tilman and colleagues sowed plots in spring 1994 with 1, 2, 4, 8 or 16 perennial grassland and savanna species chosen at random from a pool of 18, with the same seed mass in every plot (10 g per square metre in 1994 and 5 g in 1995) so rich plots did not simply get more seed. Each year they measured **aboveground biomass**, the dry mass of plant material above the soil per square metre.

*Figure (AI-generated illustration).* Each year's biomass came from clipping the aboveground vegetation in sampling strips, drying it and weighing it. Image: three steps left to right: a narrow strip of grassland being clipped at ground level with shears, the cut plants drying on a tray in an oven, and the dried plants on a laboratory balance.

*Figure (AI-generated illustration).* Assigning richness at random, with equal seed in every plot, means a consistent difference between richness levels is not expected to come from soil or from species the experimenter picked. Simplified; the real experiment had 168 plots. Image: a schematic top-down field of equal square plots in five columns holding one, two, four, eight and sixteen plant tufts, with a key tray of eighteen plant types and an arrow marking the random draw; one plot is outlined and marked nine by nine metres.

Random draws do not remove species identity: a richer draw is still more likely to include a very productive species, such as a legume. If such high-yielding species also gain the most relative yield in mixture, the next section’s partition records a positive **selection effect**.

BIODEPTH repeated the logic at eight European sites with local species, and Hector and colleagues (1999) found that, across sites, communities assembled with fewer plant species had lower average aboveground biomass.

**Predict.** Going from 16 species to 8 halves the species number; so does going from 2 to 1. Which step cost more biomass?

> **Check your prediction.** On average across sites, about the same. The decline was **log-linear**: plotted against the logarithm of species number, biomass falls along a straight line, so each halving of species number lowered biomass by a similar step (see [log scales](https://stemhop.com/math-toolkit/#log-scales)).

Composition still mattered: at a fixed species number, communities with fewer functional groups (such as grasses, legumes and other herbs) were less productive.

## Splitting the extra yield (worked problem)

Picture a shared garden. A group can out-harvest its members working alone because they grow different crops, deep-rooted and shallow (**complementarity**), or because a random group is more likely to include the gardener who is best alone, and that gardener gains the most from the group (**selection**). In plants, complementarity covers resource partitioning (the niche differences from [competition](https://stemhop.com/lessons/competition/)) and help between species, such as legumes adding nitrogen. A positive selection effect is what the older sampling-effect hypothesis predicted.

Loreau and Hector (2001) measured both from one number per species, its **relative yield** (RY): its yield in mixture, Y, divided by its yield alone in monoculture, M. In the first film’s mixture, A makes 300 g per square metre against 400 alone, so RYA = 0.75. Expected RY is its sown proportion (the first film’s dashed bar), here 0.5. **ΔRY** (“delta R Y”) is observed RY minus expected: A’s is 0.25, better than its monoculture predicts. The **net effect**, observed mixture yield minus expected, then splits exactly:

> **net effect = complementarity + selection**

> **Equation card: How much of the extra yield comes from species doing better on average in mixture.**
>
>     complementarity = N × mean(ΔRY) × mean(M)
>
> In words: Complementarity is the number of species, times the average change in relative yield, times the average monoculture yield.
>
> - N: the number of species in the mixture
> - ΔRY: each species’ change in relative yield: observed minus expected
> - M: each species’ yield in monoculture (g per square metre)
>
> Sanity check: If every species did exactly as expected, every ΔRY is 0 and so is complementarity; it is positive when species do better on average in mixture.
> Try one number If A and B (monocultures 400 and 100) each had ΔRY = 0.25: complementarity = 2 × 0.25 × 250 = 125 g per square metre.

> **Equation card: Whether the gains in mixture go mainly to the high-yielding species or to the low-yielding ones.**
>
>     selection = N × cov(M, ΔRY)
>
> In words: Selection is the number of species times the covariance between monoculture yield and change in relative yield.
>
> - N: the number of species in the mixture
> - cov: do the big species gain most? For each species, multiply (M minus mean M) by (ΔRY minus mean ΔRY), then average the products, dividing by N (the population covariance). With N − 1 the two terms no longer add up to the net effect.
> - M: each species’ monoculture yield (g per square metre)
> - ΔRY: each species’ change in relative yield
>
> Sanity check: If every species gains the same ΔRY, nothing varies with M and selection is 0. It is positive when the species with big monocultures gain most, negative when the small ones gain most.
> Try one number With equal gains (both ΔRY = 0.25), no ΔRY deviates from the mean, so selection = 0 and all 125 is complementarity.

**Worked problem** (the first film’s numbers, constructed; plots of the same area, same year). Alone, A yields MA = 400 and B yields MB = 100 g of dry biomass per square metre. In the mixture, YA = 300 and YB = 100.

1. *Net effect.* With an even sowing the expected yield is the average monoculture, 250, so net effect = 400 − 250 = **+150 g per square metre**.
2. *Relative yields.* ΔRYA = 0.25 (above). RYB = 100/100 = 1.00, so ΔRYB = 0.50. Mean ΔRY = 0.375; mean M = 250.
3. *Complementarity* = 2 × 0.375 × 250 = **+187.5 g per square metre**.
4. *Selection.* Deviations from the means: A has M − 250 = +150 and ΔRY − 0.375 = −0.125; B has −150 and +0.125. Both products are −18.75, and their mean (divide by 2) is −18.75. Selection = 2 × −18.75 = **−37.5 g per square metre**.
5. *Check.* 187.5 − 37.5 = 150, the net effect.

Both species did better than expected, so complementarity is positive; the low-yielding B gained more, so selection is negative. Mean ΔRY above zero does not guarantee extra yield: had A made 120 and B 100, complementarity would be +75 but selection −105, and the mixture’s 220 would miss the expected 250.

**Predict.** Species A makes 300 of the mixture’s 400 g. Does a negative selection effect mean A failed to dominate?

> **Check your prediction.** No. A dominates the biomass, but selection asks whether the relative-yield gains went to the species big in monoculture. They went to B, so selection is negative though A is the largest species present.

Beating the best monoculture, not just the average, is **transgressive overyielding**; the first film’s mixture only matches A alone, so it shows none.

*Interactive on the lesson page:* Overyielding — complementarity and selection

Applied to BIODEPTH, the complementarity effect was positive overall, while the selection effect averaged zero and ranged from negative to positive among sites.

A positive complementarity term does not prove resource partitioning: the partition is exact algebra on yields, and legume nitrogen or reduced pathogen pressure could produce it too. It needs species-level yields in both mixtures and monocultures.

At Cedar Creek the same plots were harvested every year for a decade, so the split could be calculated year by year. Did it stay put as the plots aged, and how steady was each plot’s total?

**Film transcript: Ten Years at Cedar Creek**

### Film transcript: Ten Years at Cedar Creek

The first film showed a field as a sum of its species. Now follow one real field for a decade. How much of its extra yield is complementarity, and how much is selection? Does the answer change as the plots age? The field is Cedar Creek, in Minnesota. In spring nineteen ninety-four, David Tilman and colleagues sowed a hundred and sixty-eight plots there. Each plot was nine metres on a side. Each plot got one, two, four, eight or sixteen species. They were drawn at random from a pool of eighteen grassland and savanna species. Every plot got the same seed mass. Because the draw was random, nobody chose strong species for the rich plots. Better soil could line up with richness only by chance.

Every year the plants were clipped, dried and weighed. Year after year, plots with more species generally grew more aboveground biomass. Joseph Fargione and colleagues then split the extra yield in each year, using every species' yield in mixture and alone. The green bar is complementarity. The violet bar is selection. In the early years, selection was positive. Rich plots outyielded poor ones partly because they more often held a productive species. Then the balance shifted toward complementarity. Selection shrank, crossed zero and ended strongly negative. Now the biggest gains went to species that yield little alone. The extra biomass came to be associated with more nitrogen held in the plants.

So the same plots gave a different answer depending on which years you looked at. Each year's split is a snapshot, and the snapshots changed. Now zoom into the sixteen-species plots and follow their species. On average, each species swings proportionally more from year to year than species in poorer plots. So the parts became less steady, on average. Will the plot's total be more steady, or less steady, than a monoculture's?

More steady. Tilman, Peter Reich and Johannes Knops reported this in two thousand six. The total varied proportionally less, even though its species, on average, varied more. Did opposite swings grow stronger in richer plots? Their analysis found no evidence of that. Averaging, together with a higher mean, explained the steadier total. A higher mean helps only if the swings grow less than in proportion. Assigned at random and watched for a decade, richer plots grew more and grew steadier. Their species, on average, grew less steady.

From the film

- At Cedar Creek, 168 plots were sown in 1994 with 1, 2, 4, 8 or 16 species drawn at random from a pool of 18, with the same seed mass in every plot.
- Because richness and composition were assigned at random, soil differences and the experimenter's choices are not expected to explain the consistent rise in biomass with richness.
- Partitioned year by year over 1996 to 2005, the extra yield shifted from selection toward complementarity, and by the end selection was strongly negative.
- In richer plots species varied proportionally more from year to year on average, yet the plot's total varied proportionally less.
- The analysis found no evidence of a covariance effect; statistical averaging and a higher mean explained the steadier total.

## Stability: how to score it, and when cancelling fails

Scoring the second film’s steadier totals starts with a warning: “stable” has several meanings, and they need not agree:

- **Temporal stability** is the mean of total community biomass across years, μ (“mu”), divided by its standard deviation, σ (“sigma”).
- **Resistance** is how little a function changes during a disturbance such as a drought.
- **Resilience** is how fully or quickly it returns afterwards.

> **Equation card: How steady a community’s total biomass is from year to year.**
>
>     temporal stability = μ / σ
>
> In words: Temporal stability is the mean total biomass divided by its standard deviation.
>
> - μ: the mean of total community biomass across years (g per square metre)
> - σ: the standard deviation of that total across years: how much it typically swings (g per square metre)
>
> Sanity check: The units cancel, so stability has none. A constant positive total has σ = 0; the ratio is then undefined, and is read as unbounded, perfect stability. A larger mean raises μ/σ as long as the standard deviation grows less than in proportion, which is why overyielding alone can raise stability.
> Try one number The first film’s single species: a mean of 50 g per square metre and a standard deviation of 25, so μ/σ = 50/25 = 2.

The first film’s total of four independent species swung only twice as much because standard deviations do not add; **variances** (squared standard deviations) do, when the species are uncorrelated. Four variances of 25² sum to 2,500, whose square root is 50, so μ/σ = 200/50 = 4. Stability grows with the square root of species number (16 species give 8), with no species any steadier.

Real species often move partly together; their **correlation** runs from +1 (lockstep) through 0 (no linear relation) to −1 (one rises exactly as the other falls). In general, the variance of a total is the sum of the species’ variances plus twice the **covariance** of every pair, where a pair’s covariance is their correlation times their two standard deviations.

**Predict.** Give the four species a correlation of 0.5 with each other: they tend to move together, but not perfectly. Is μ/σ closer to 4, closer to 2, or exactly halfway?

> **Check your prediction.** Closer to 2: about 2.5. Each pair’s covariance is 0.5 × 25 × 25 = 312.5, and four species make six pairs, so the variance is 4 × 625 + 2 × 6 × 312.5 = 6,250. Its square root is about 79, and μ/σ = 200/79 ≈ 2.5. Partial synchrony removes most of the averaging gain: the one idea’s “unless the species rise and fall together” is a matter of degree.

Negative correlations cancel more: one species rising as another falls is **compensatory dynamics**. If richer plots had more of it, part of their extra steadiness would be a **covariance effect**. At Cedar Creek, Tilman, Reich and Knops found no evidence of a covariance effect. They credited statistical averaging and overyielding: over 2001 to 2005, the 16-species plots averaged 180 percent more biomass than monocultures, and a higher mean raises μ/σ (see the card’s check). Later analyses differ in how strongly they support averaging and in how long a record is needed to infer the mechanism (see Going deeper).

**Predict.** Suppose diverse plots lose a smaller share of their production in a drought. Will they also recover more fully a year later?

> **Check your prediction.** Not detectably. Across 46 grassland diversity experiments, Isbell and colleagues (2015) found that during wet or dry climate events, productivity in communities of one or two species changed by about 50 percent, against about 25 percent at 16 to 32 species. That is greater resistance. Yet by one year after each event, they detected no dependence of recovery on diversity.

Stability measures can even move in opposite directions: in 690 aquatic ciliate [microcosms](https://stemhop.com/methods/microcosm-experiments/), Pennekamp and colleagues (2018) found that richness raised temporal stability but lowered resistance to warming.

## Beyond the plots

In the BEF-China experiment, a planted subtropical forest, Huang and colleagues reported in 2018 that 16-species mixtures had accumulated more than twice the carbon of the average monoculture after eight years, and an amount similar to two commercial monocultures. The benchmark matters.

Real landscapes do not assign richness. Duffy and colleagues (2017) synthesized 67 field studies of wild systems, 133 estimates in all, and found positive biodiversity-biomass associations after statistically adjusting for environmental covariates. Adjustment is not random assignment: an unmeasured variable could still drive both. These results carry forward into [global change](https://stemhop.com/lessons/global-change/) and [conservation](https://stemhop.com/lessons/conservation/).

### Design problem

An observational survey finds that grassland plots with more plant species have more aboveground biomass. Design a study to decide whether richness causes the difference.

> **One way to do it.** Soil fertility or moisture could raise both; break the link by assigning richness yourself.
>
> 1. Divide a uniform field into blocks by soil and slope. Within each block, randomly assign plots to 1, 2, 4 or 8 species.
> 2. At each richness level, draw several independent random compositions from a fixed local pool, and include a monoculture of every species so the partition can be computed.
> 3. Keep plot area and seed mass constant, and weed every plot alike to hold its assigned composition.
> 4. Measure soil nitrogen, moisture and light before sowing, and check they do not differ by treatment.
> 5. Cross richness with a randomly assigned nitrogen or drought treatment, to see whether the effect holds in each environment.
> 6. Treat the plot as the unit, harvest species separately each year for several years, and analyse assigned richness, with block as a factor.
> 7. Report net, complementarity and selection effects alongside raw yields, compare mixtures with the best monoculture as well as the average, and predefine the stability measures (μ/σ over a fixed window, resistance and recovery for any event).

## Going deeper

### Uncorrelated is not the same as independent

Variances add whenever species’ swings are uncorrelated. Independent species are always uncorrelated, but uncorrelated species need not be independent.

### The mechanism debate

Zhao and colleagues (2022) found statistical averaging was the principal mediator in their North American grassland analysis. Luo and colleagues (2025) showed that synchrony patterns can reverse with the length of the time series, so the inferred mechanism depends on how long you watch.

### Assembled is not the same as lost

These experiments compared communities assembled with fewer species; they did not remove species. Real extinctions are not random draws from a pool, so random assembly does not mimic the order in which species are lost in nature. Species loss can still change what an ecosystem does; the experiments show what richness can do in the communities they built.

---

From STEMhop (https://stemhop.com), free lessons by Christopher Deppmann. Lesson text © the authors. Practice questions: https://stemhop.com/lessons/biodiversity-ecosystem-function/
