Skip to main content
Dryad logo

Habitat features and performance interact to determine the outcomes of terrestrial predator-prey pursuits

Citation

Wheatley, Rebecca; Pavlic, Theodore; Levy, Ofir; Wilson, Robbie (2020), Habitat features and performance interact to determine the outcomes of terrestrial predator-prey pursuits, Dryad, Dataset, https://doi.org/10.5061/dryad.wm37pvmkm

Abstract

1. Animals are responsive to predation risk, often seeking safer habitats at the cost of foraging rewards. Although previous research has examined how habitat features affect detection by predators, little is known about how the interaction of habitat features, sensory cues, and physical performance capabilities affect prey escape performance once detected.

2. To investigate how specific habitat features affect predation risk, we developed an individual-based model of terrestrial predator–prey pursuits in habitats with programmable features.

3. We ran simulations varying the relative performance capabilities of predator and prey as well as the availability and abundance of refuges and obstacles in the habitat.

4. Prey were more likely to avoid detection in complex habitats containing a higher abundance of obstacles; however, if detected, prey escape probability was dependent on both the abundance of refuges and obstacles and the predator’s relative performance capabilities. Our model accurately predicted the relative escape success for impala escaping from cheetah in open savanna versus acacia thicket habitat, though escape success was consistently underestimated.

5. Our model provides a mechanistic explanation for the differential effects of habitat on survival for different predator–prey pairs. Its flexible nature means that our model can be refined to simulate specific systems and could have applications toward management programs for species threatened by habitat loss and predation.

Methods

All data (contained in .csv files) were generated using the NetLogo model code files. The data for the cheetah vs impala experiments were generated using the predator-prey-model-with_manual_performance_input.nlogo version of the code (which allows the performance capabilities of predator and prey to be set manually); all other csv files were generated using the predator-prey-model-with_limb_length_input.nlogo version of the code (which sets performance capabilities via scaling relationships with limb length). The NetLogo code files are compatible with NetLogo v6.1. NetLogo is free to run and can be downloaded from: https://ccl.northwestern.edu/netlogo/download.shtml.

The data is broadly divided into three categories (corresponding to the three results sub-sections in the paper):

  1. Sensitivity analysis
  2. Effect of relative performance, obstacles, and refuges on detection and escape
  3. Case study: cheetah and impala in open savanna and acacia thicket

Below is an explanation for each data category, describing any processing and analysis that took place.

Sensitivity analysis

For the global sensitivity analysis, we used a reduced factorial design where each parameter was varied across its range with every other parameter in pairs (a total of 8,558 parameter combinations). We ran 100 simulation replications per parameter combination, resulting in 855,800 simulations in total. The excel file Global sensitivity analysis-parameter sampling design.xlsx contains the sampling design for the simulations described above, and gives the code (PS_XXX) corresponding to each parameter pair. The 171 files named in the form predator-prey-model-global_sensitivity_analysis-PS_XXX.csv contain the raw data from these simulations.

These raw data files were processed using the R code file global sensitivity analysis 1-calculate mean and median responses.R to generate the summarised data file predator-prey-model-global_sensitivity_analysis-parameter_sets_and_results-summarised.csv that contains the mean and median values of the simulation results for each setting combination for each parameter pair. This summarised data file was analysed for the global sensitivity analysis using the R code file global sensitivity analysis 2-multisensi sensitivity analysis.R. Based on the results from the sensitivity analysis, the effect of varying each of the four most influential parameters per model response was evaluated using the R code file global sensitivity analysis 3-examine how variation in individual parameters affects model responses.R and the data file containing all the raw simulation data, predator-prey-model-global_sensitivity_analysis-parameter_sets_and_results.csv.

Effect of relative performance, obstacles, and refuges on detection and escape

predator-prey-model-performance_vs_obstacles.csv contains the raw data from simulations where the predator and prey's relative performances were varied with the proportion of obstacles in the habitat. This data was analysed using the experiments-obstacles and refuges.R file.

predator-prey-model-performance_vs_refuges.csv contains the raw data from simulations where the predator and prey's relative performances were varied with the number of refuges in the habitat. This data was analysed using the experiments-obstacles and refuges.R file.

Case study: cheetah and impala in open savanna and acacia thicket

predator-prey-model-cheetah_vs_impala-acacia_thicket.csv contains the raw data for the cheetah vs impala in acacia thicket simulations. This data was analysed using the experiments-cheetah vs impala.R file.

predator-prey-model-cheetah_vs_impala-open_savanna.csv contains the raw data for the cheetah vs impala in open savanna simulations. This data was analysed using the experiments-cheetah vs impala.R file.

Funding

Australian Research Council, Award: DP180103134

University of Queensland, Award: Graduate Student International Travel Award