1. Structural complexity of habitats modifies trophic interactions by providing refuges and altering predator and prey behaviour. Nonlinear effects on trophic interaction strengths driven by these mechanisms may alter food web dynamics and community structure in response to habitat modifications. However, changes in functional response, the relationship between prey density and feeding rate, along habitat complexity (HC) gradients are little understood.

2. We quantified functional responses along a HC gradient from an entirely unstructured to highly structured habitat in a freshwater system, using dragonfly larvae (Aeshna cyanea) preying on Chaoborus obscuripes larvae. To disentangle mechanisms by which changes in HC affect functional responses, we used two different approaches – a population-level and a behavioural experiment–, applied an information theoretic approach to identify plausible links between HC and functional response parameters, and compared our results to previous studies.

3. Functional response shape did not change, but we found strong evidence for nonlinear dependence of attack rate and handling time on HC in our study. Combined results from both experiments imply that attack rate increased stepwise between the unstructured and structured habitats in line with the threshold hypothesis, because the predators gained better access to the prey. Handling time was lowest at an intermediate HC level in the population-level experiment, while the direct estimate of handling time did not vary with HC in the behavioural experiment. These differences point toward HC-driven changes in foraging activity and other predator and prey behaviour.

4. Most previous studies reported stepwise decrease in attack rate in line with the threshold hypothesis or no change with increasing HC. Moreover, changes in the handling time parameter with HC appear to be relatively common but not conforming to the threshold hypothesis. Overall, increased HC appears to respectively weaken and strengthen trophic links in 2D and 3D predator-prey interactions.

5. We conclude that detailed understanding of HC effects on food webs requires complementary experimental approaches across HC gradients that consider predator foraging strategies and predator and prey behaviour. Such studies can also help guide conservation efforts as addition of structural elements is frequently used for restoration of degraded aquatic habitats. 23-Feb-2021

The data in this repository consist of five datasets from the two experiments presented in the paper. The first dataset “Mocq-et-al_Exp1_DataSet.csv” contains the raw data from Experiment 1. Each row provides data on one predation trial in the following columns:

• “Date”: date when the experiment was conducted
• “Plants_Number(HC)”: habitat complexity level, given as the number of artificial plants in the arena
• “Duration_h”: duration of the trial (hours)
• “N_Predator”: number of predators (i.e. Aeshna cyanea) used in the trial (0 = controls)
• “N_Prey”: number of prey individuals (i.e. Chaoborus obscuripes) at the beginning of the trial
• “Predator_Length”: body length of the predator (mm)
• “Predator_Head_Width”: head width of the predator (mm)
• “Predator_Mentum_Length”: mentum length of the predator (mm)
• “Predator_WetWeight”: predator wet weight (g, calculated from length using the formula specified in the paper)
• “Predator_Final_Location”: location of the predator at the end of the trial
• “N_Prey_Survivor”: number of surviving prey at the end of the trial
• “N_Prey_Dead”: number of dead prey at the end of the trial
• “N_Prey_Eaten”: number of eaten prey at the end of the trial, inferred from the number of surviving and dead prey
• “Prey_WetWeight”: total wet weight of all prey recovered from each control.

The other four datasets summarize the following raw data from Experiment 2:

1. “Mocq-et-al_Exp2_substrate_preference.csv”: substrate preferences of the individual predators (each row represents one trial);
2. “Mocq-et-al_Exp2_handlingTime.csv”: prey handling times (each row represents one prey handling event);
3. “Mocq-et-al_Exp2_attacks.csv”: numbers of successful and failed attacks of the individual predators (each row represents one trial);
4. “Mocq-et-al_Exp2_arTimes.csv”: durations of individual foraging bouts (each row represents one event).

The following columns are used in these four files:

• “Aeshna”: predator identity (all four files)
• “obsID”: unique trial ID (all four files)
• “HC”: habitat complexity level (number of artificial plants) used in the trial (all four files)
• “”preyDens.ini”: initial number of prey individuals used in the trial (all four files)
• “P.bottom”, “P.plant”, “P.wall”: proportion of time spent respectively on the bottom, artificial plants and walls of the arena (only data on substrate preferences)
• “start_sec”, “stop_sec” and “duration_sec”: start, end and duration of prey handling (seconds; only data on handling times)
• “W_g”: estimated wet weight of the predator (grams; data on handling times, numbers of attacks and durations of individual foraging bouts)
• “ln(W^c)”: offset ln(W^c) used in the analysis of handling times, see the main text for details (only data on handling times)
• “assayDate_yymmdd”: date of assay (yy-mm-dd format, only data on handling times)
• “N.success” and “N.failed”: numbers of successful and failed attacks (only data on attacks)
• “obsTime_hr”: time of an attack event since the start of the trial (hours; only data on the duration of individual foraging bouts)
• “attack Delay”: time to the attack since last recoded event (hours; only data on the duration of individual foraging bouts, see the main text for details)
• “rightCensored”: whether the event is right-censored (1 = end of trial) or not (0 = observed attack event; only data on the duration of individual foraging bouts)
• “preyDens.scaled”: scaled prey density at the time of event (only data on the duration of individual foraging bouts)