Fig 1.
The food web including all possible interactions that we allowed in the model.
Species are, from top left: lady beetle (Coccinella septempunctata); wolf spiders (Pardosa spp.); minute pirate bug (Orius majusculus); bird cherry-oat aphid (Rhopalosiphum padi); pea aphid (Acyrthosiphon pisum); and ground beetle (Bembidion spp). Arrows indicate potential feeding interactions which we then parameterize through least squares minimization (Section 3.2). Arrows point from prey to predator. Double headed arrows indicate that species could potentially eat each other and arrows beginning and ending with the same species indicate cannibalism. We removed all interactions to and from C. septempunctata except for C. septempunctata preying on aphids and O. majusculus, and assumed that the aphids did not consume any predators. This arthropod community was dependent on two species of plants; barley (Hordeum vulgare) and fava beans (Vicia faba).
Fig 2.
An overview of the predator-prey combinations used in the experiment.
Each combination was replicated six times. Figure replicated from [30].
Fig 3.
Microhabitat preferences as estimated from the proportion of time each species spent in each of the four microhabitat zones.
The final column shows the relative size of each area in the experimental cages.
Table 1.
Parameter values (±95% confidence intervals) and model fit (JLS and AIC) for models with and without microhabitat use and non-consumptive predator-predator effects.
Fig 4.
Model predictions of aphid population growth (lines) across time, compared to data of aphid counts per day (boxes) in single-aphid (rows) single-predator (columns) treatments.
Boxes show first and third quartiles and outlying points indicate data further from the hinge than 1.5 times the interquartile range. Lines show predictions of the different models. The final row shows model predictions when including a hypothetical new prey species that resides entirely on the ground and has a body size of 1mg. The closer a model trajectory is to the mean of the experimental data (horizontal line within boxes) the better a particular model performs. A model underestimates (overestimates) the effect of a predator (columns) on an aphid prey (rows) when its trajectory value for the prey is greater (smaller) than the experimentally observed data.
Fig 5.
Model predictions for an individual predator’s feeding rate (number of prey consumed per predator per day, y axis) on prey of different body sizes (x axis).
Curves show the “microhabitat-preference free” feeding rate, i.e. assuming all species use all microhabitats in proportion to their area. Vertical arrows show the difference in feeding rate when accounting for observed microhabitat preferences of the predator with aphid prey (R.padi, mass = 0.155mg and A.pisum, (mass = 0.67mg). The arrow tip shows the predicted feeding rate of the predator on a prey of that size when accounting for microhabitat preferences. A longer vertical arrow therefore means that observed microhabitat preferences have a larger effect on feeding rate. Line color corresponds to different models. Models with microhabitat use (blue and dark green lines) predict the highest “microhabitat-preference free” feeding rates, but when actual microhabitat preferences are accounted for their feeding rates drop closer to that predicted by the other models. We show the instantaneous feeding rate with a population of 250 aphids and in the absence of any other predator individuals (i.e. not accounting for non-trophic predator effects. To see their effect, compare with S1 Fig). Observe that, for Bembidion, accounting for microhabitat use of R.padi does not substantially change the predictions of models with microhabitat (i.e. the arrows sit on the curve). This is because the overlap of Bembidion with R.padi works out to be almost the same as if they used all microhabitats in proportion to the size of the microhabitat. Note the varying scales of the y-axis.
Table 2.
Handling times (hij, units = indi/day) for each predator with R. padi and A. pisum for each model.
Fig 6.
Experimental results (purple boxes) and model predictions for the proportion of each (focal) predator population surviving on the final day of the experiment when combined with different intraguild predators (i.e. the top right panel shows the proportion of the Pardosa population remaining at the end of the experiment when combined with Bembidion).
The experimental data is shown as a box plot, showing the 25th, 50th, and 75th quartiles. Model predictions depend on the focal and intraguild predator, but also on the prey treatment. The three prey treatments are shown as a point and error bars; where all three are stacked, prey treatment makes no difference to the predicted predator population. Note that C. septempuntata was not modelled dynamically and did not predate on most other predators, so we exclude it from this plot.