Distributions, Niches and Tolerances

A small coastal plant, planted about 60 km beyond the northern end of its range, more than tripled in number over the next 12 to 14 years. If it can live there, why had it never got there? Answering that means separating three things that are easy to blur: the conditions a species can live in, the other species it meets, and the places it has been able to reach.

The idea: a niche is a set of conditions, not a place

A niche is the set of conditions and resources in which a population can persist without immigrants. G. Evelyn Hutchinson gave it this form in 1957. Treat each variable that matters as an axis: temperature, soil moisture, salt, the size of food items. The niche is the region of that space where the population’s births at least balance its deaths. With n axes it is an n-dimensional hypervolume. Axes can be biotic as well as abiotic, such as the density of a predator.

Hutchinson did not coin the word. Joseph Grinnell used it in 1917 for a bird’s habitat and habits, and Charles Elton in 1927 for an animal’s place in a food web. What Hutchinson added was a version you can measure.

The key move is to keep two spaces apart. Niche space is the space of conditions. Geographic space is the map. A thermostat analogy helps: the niche is a setting, say 15 to 25 degrees C with moist soil, and the map shows every place where that setting happens to occur. As you saw in climate and biomes, conditions vary with latitude, altitude and microclimate, so one patch of niche space can map onto scattered places, and one place can hold several microclimates.

Two consequences follow. A species can be absent from a suitable place because it has not reached it. And it can be present in an unsuitable one: a sink population is one where deaths exceed births, kept going only by immigrants from elsewhere. Neither presence nor absence on a map, by itself, tells you where the niche ends.

The persistence test uses lambda, the finite rate of increase: the factor by which a population multiplies in one time step, here one year. Lambda of 1 or more means the population replaces itself. It is built from survival, growth and reproduction across the whole life cycle, the vital rates from life histories.

Fundamental and realized niches

Hutchinson split the niche in two. The fundamental niche is the set of conditions in which a species can persist before interactions with other species are counted. The realized niche is the part it occupies when those interactions are present.

The classic field case comes from the rocky shore at Millport, on the Isle of Cumbrae in Scotland, published by Joseph Connell in 1961. Adults of the barnacle Chthamalus stellatus (now classified as C. montagui) lived in a band high on the rocks, above the larger Balanus balanoides (now Semibalanus balanoides). Yet Chthamalus larvae settled lower. When Connell kept Balanus off portions of lower rock, Chthamalus survived there; where Balanus grew, it caused heavy Chthamalus mortality. On that shore, the lower part of the Chthamalus fundamental niche was unoccupied because of a competitor. Connell attributed the upper limits mainly to heat and drying, with Chthamalus the more tolerant. The experiment itself belongs to competition; here it shows the difference between the two niches.

Competitors and predators usually shrink the realized niche. Helpers can do the opposite. A shrub that shades seedlings, or a mussel bed that holds water at low tide, can let a species persist in conditions it cannot tolerate alone. If the fundamental niche is defined before any interaction, facilitation can make the realized niche larger than the fundamental one, a way of putting it that some ecologists dispute (Hutchinson’s own wording concerned competition). So “the realized niche is always smaller” is a rule with exceptions.

Field experiments measure interactions directly. Welk and colleagues grew 14 perennial grassland species in nine German botanical gardens, with competitors present or removed and molluscs excluded or not, in a fully crossed design. In their model, competition reduced aboveground biomass by 40.8% overall. Survival also declined as precipitation fell across the gardens, but climate varied among gardens rather than being assigned at random, so that part is an association.

A caution for later lessons. Species distribution models, fitted to the climates of places where a species has been recorded, describe occupied conditions plus accessibility. They estimate something closer to the realized niche filtered by dispersal, not the fundamental niche.

Tolerance and performance curves

To measure one axis of the fundamental niche, rear individuals at several constant temperatures and record a named response. The result is a thermal performance curve. It rises from zero at Tmin, peaks at Topt and falls to zero at Tmax. The fall at high temperature has several causes, among them proteins losing their folded, active shape, as in enzyme action.

Every value belongs to its response. Tmin and Tmax are where that rate reaches zero, not lethal temperatures. And different responses can peak in different places. In a 2024 synthesis of 46 laboratory data sets from 26 species of Lepidoptera, Mariana Abarca and colleagues found that development rate generally limited performance at the cold end and survival at the hot end. They combined the two:

P = D x S (D = development rate, S = survival, each scaled to its own maximum)

Worked problem. The numbers are invented for teaching. A moth is reared at three temperatures. Survival peaked at a cooler temperature not shown, so S is below 1 at all three.

Temperature	D	S
20 degrees C	0.50	0.80
25 degrees C	1.00	0.60
30 degrees C	0.80	0.20

Assume P changes in a straight line between treatments, do not extrapolate beyond 20 or 30 degrees C, and call the acceptable range P of at least 0.50 (the threshold Abarca’s group used, an operational choice).

1. Compute P. 0.50 x 0.80 = 0.40 at 20 degrees C; 1.00 x 0.60 = 0.60 at 25; 0.80 x 0.20 = 0.16 at 30. The best tested temperature is 25 degrees C. The true peak could lie between treatments.
2. Lower crossing. P must rise 0.10 of the 0.20 it gains from 20 to 25, half the way: 20 + (0.50 - 0.40)/(0.60 - 0.40) x 5 = 22.5 degrees C.
3. Upper crossing. P must fall 0.10 of the 0.44 it loses from 25 to 30: 25 + (0.60 - 0.50)/(0.60 - 0.16) x 5 = 25 + 1.14 = 26.14 degrees C.
4. Breadth. 26.14 - 22.5 = 3.64 degrees C.
5. Which rate limits each end? At 20 degrees C survival is high and development slow, so development limits the cold end. At 30 degrees C development is still fairly fast but survival has collapsed, so survival limits the hot end.

Show

P = D x S D, development S, survival

Acceptable P is at least 0.50

D x S = P at 20 / 25 / 30 °C

0.40 / 0.60 / 0.16

Lower crossing

22.50 °C

Upper crossing

26.14 °C

Breadth

3.64 °C

In this sketch: invented data for a moth at three constant temperatures, D and S each scaled to their own maximum, straight lines between treatments, no extrapolation beyond 20 and 30 °C. P leaves out fecundity, so it is not lambda; a crossing is not a lethal limit or a range boundary.

Rule to keep: the tolerance interval is where the combined response stays above a threshold you chose, found by interpolation inside the tested range, and the two ends can be limited by different rates.

Now the limits of the answer. P leaves out fecundity, so it is not lambda. The crossings are not lethal limits, and they are not a range boundary: a moth in the field can shelter in shade, and its distribution also depends on host plants, enemies and dispersal.

Testing a range edge by transplant

A range edge could be set by conditions (an abiotic limit), by other species (a biotic limit) or by the species not having arrived (a dispersal limit). The direct test is to plant the species beyond the edge and follow it.

Amy Angert and Douglas Schemske did this along an elevation gradient in California with two monkeyflowers, Mimulus cardinalis at low elevations and M. lewisii at high ones, in gardens inside and beyond each species’ elevation range. Each had zero or near-zero fitness beyond its elevation limits, which fits niche limits. An elevation gradient changes many conditions at once, so temperature cannot be separated from the other factors that vary with elevation.

Anna Hargreaves, Karen Samis and Christopher Eckert meta-analysed 11 such tests in 2014 and stressed that design limits what they can show, including the need for sites and seed from the range edge. Two further points follow from the logic of the test. First, plant the same way inside the range: if within-range plantings also fail, the cause is handling or a bad year. Second, measure the whole life cycle and estimate lambda, because a fall in one measure, such as first-year growth, does not show that a population cannot replace itself.

Regan Cross and Christopher Eckert then revisited a coastal dune plant, Camissoniopsis cheiranthifolia, 12 to 14 years after it was planted beyond its northern limit. The beyond-range population had more than tripled, and they inferred that limited dispersal had likely contributed to the edge. That is one site. It shows the edge there is not where the species fails, not that every dune farther north is suitable. Movement between populations is the subject of spatial populations, and shifting edges under warming of global change.

Design problem

A coastal dune plant’s range ends abruptly at a headland. Design a test of whether the edge is set by climate, by neighbouring plants or by dispersal.

One way to do it

Collect seed from several source populations, so that one unusual population does not decide the result. Sow it at several interior sites, at the edge and at several sites beyond it, with identical handling. At every site, cross the planting with neighbour removal: half the plots cleared of other plants, chosen at random, half left. The interior plantings are the controls for handling and for bad years.

Follow each plot for several years, recording survival, growth and seed production, and build a stage-structured model to estimate lambda for each site and treatment.

Read the results this way. If lambda is at least 1 beyond the range, with or without neighbours, the species can persist there and dispersal is the likely limit, the inference Cross and Eckert drew for Camissoniopsis at one site. If lambda falls below 1 beyond the range only where neighbours remain, a biotic limit is implicated. If it falls below 1 beyond the range in every treatment while interior controls stay at or above 1, an abiotic limit is the likeliest. If interior plots fail too, the years or the handling are to blame and the test must run longer.

Two practical issues come first. Planting beyond a range needs permits from landowners and agencies, and a species that thrives there could spread as a new invader. Choosing sites that can be monitored, and removing plants at the end if required, is part of the design.