Journal of Animal Ecology 2019 ★ 2019 Elton Prize

Warming can destabilise predator–prey interactions

A three-month undergraduate project showing that warming can flip a predator's functional response from Type III to Type II, and that this one change of shape is enough to drive the prey extinct in simulation.

Daugaard U., Petchey O.L. & Pennekamp F.

doi: 10.1111/1365-2656.13053 ↗

01 · The question

Does warming reshape the way predators eat their prey?

The functional response describes how much prey a single predator consumes as prey density rises. Its shape matters for stability: a Type III response accelerates at low prey density, so predation pressure eases when prey are rare, and is generally considered the stabilising one. A Type II has no such phase. We asked whether a few degrees of warming can change which shape a predator–prey pair shows.

Plot of the Type I, Type II and Type III functional response curves, prey consumed against prey abundance, beside the study's research question. — Fig. 1 The three functional response types, and the question we put to them. Only the Type III accelerates at low prey density.

02 · What we did

A ciliate pair, 24 hours, three temperatures

We paired the ciliate predator Spathidium sp. with its prey Dexiostoma campylum, two species that co-occur in nature, and followed both over 24 hours at 15, 20 and 25 °C. At each temperature we set up eight starting prey densities, each with five predators or with none at all as a control. That came to 72 one-millilitre wells per temperature, enough of a density gradient to reconstruct the functional response at each of them.

Schematic of the feeding experiment: the two ciliate species, three temperatures, eight starting prey densities, predator present or absent, 72 wells per temperature. — Fig. 2 The experiment. Eight starting prey densities at each of three temperatures, with the predator present or absent, scored after one day.

We then fitted a population-dynamic model built around a generalised functional response, whose exponent q decides the type rather than being assumed: q = 0 gives a Type II, q > 0 a Type III. Fitting it separately at each temperature returned the space clearance constant, the handling time and the exponent itself, along with the predator’s conversion efficiency.

The generalised functional response equation and the predator–prey differential equations fitted to the data, with all parameters defined. — Fig. 3 The model we fitted. A generalised functional response with a free exponent q, inside a predator–prey model with logistic prey growth.

03 · What we found

Type III at 15 and 20 °C, Type II at 25 °C

At the two cooler temperatures the fitted exponent was positive: the response accelerated at low prey density, leaving rare prey some respite. At 25 °C that acceleration was gone and the response was Type II. Every other parameter of the model shifted with temperature too, the conversion efficiency among them.

Model fits at 15, 20 and 25 °C: change in prey density over 24 hours with and without predators, and predator density after 24 hours. — Fig. 4 Model fit. Top: change in prey density over 24 h with predators (circles) and without (triangles). Bottom: predator density after 24 h, against a starting density of five per millilitre (dashed).

The destabilising consequence. Simulating the fitted models forward drives the prey extinct in 99.8% of runs at 25 °C, against 14.6% at 15 °C and none at all at 20 °C. The jump arrives with the change in type, not gradually with warming.

Left: estimated functional responses at the three temperatures, labelled Type III, Type III and Type II. Right: percentage of simulations in which the prey went extinct, and time to extinction. — Fig. 5 Left: the estimated functional responses, Type III at 15 and 20 °C and Type II at 25 °C. Right: how often the simulations drove the prey extinct, and how fast.

Conversion efficiency rose with warming as well, so it should not be assumed temperature-independent either. It is not what destabilises the system, though: pinning every parameter but the three functional response terms to its 20 °C value leaves the extinction pattern in place, if weaker.

04 · Why it matters

Shape changes are usually missing from food-web models

Models that project food webs under warming generally hold the type of the functional response fixed and let temperature scale its parameters. The type itself can change, and scaling alone will not capture what follows. How often this happens, and how it propagates through a multi-species food web, is still open.