Skip to main content
Dryad logo

Effects of enhanced productivity of resources shared by predators in a food-web module: Comparing results of a field experiment to predictions of mathematical models of intra-guild predation

Citation

Wise, David; Farfan, Monica (2021), Effects of enhanced productivity of resources shared by predators in a food-web module: Comparing results of a field experiment to predictions of mathematical models of intra-guild predation, Dryad, Dataset, https://doi.org/10.5061/dryad.5qfttdz6k

Abstract

This dataset contains data from a field experiment described in the publication “Wise, D. H. & Farfan, M.A. (2021) Effects of enhanced productivity of resources shared by predators in a food-web module:  Comparing results of a field experiment to predictions of mathematical models of intra-guild predation. Ecology and Evolution, 00: 1-11. https://doi.org/10.1002/ece3.8375”.

The field experiment compared the response to increased input of nutrients and energy (artificial detritus) to an empirical model of intra-guild predation (IGP) to the predictions of published, simple mathematical models of asymmetric IGP (a generalist IG Predator that feeds both on a specialist IG Prey and a Resource that it shares with the IG Prey). The empirical model was a food-web module created by pooling species abundances across many families in a community of soil micro-arthropods into three response variables: IG Predator (large predatory mites), IG Prey (small predatory mites) and a shared Resource (fungivorous mites and springtails). The pattern of change over time in densities of the three response variables (IG Predator, IG Prey and Resource) was compared to the predictions of mathematical models of IGP to determine if the feeding relationships in this community of soil micro-arthropods could be abstracted into a simple IGP module. Thus, we were testing the hypothesis that IGP is a dominant organizing principle in this community.

Simple mathematical models predict that increased input of nutrients and energy to the shared Resource will increase the equilibrium density of Resource and IG Predator, but will decrease that of IG Prey. By the experiment’s end, densities of fungivores (Resource) had increased ~1.5x (ratio of pooled fungivore densities in the High treatment to plots with no addition of detritus (None treatment); and IG Predator densities had increased ~4x. Contrary to the prediction of mathematical models, IG Prey had not decreased, but instead had increased ~1.5x. We discuss possible reasons for the failure of the empirical model to agree with IGP theory.

Methods

The field experiment was conducted in 200 circular 1-m2 plots located within a 1-ha area of forest at the Morton Arboretum in Lisle, Illinois, USA. One hundred and fifty plots were fenced; the remaining 50 open plots served as a reference (REF) treatment. Three levels of detrital supplementation (None,  Low, or High) were randomly assigned to the fenced plots. Treatments were spread among five Blocks (note that this is not the classic randomized block design because each block had 10 replicates of each of the four treatments). The added detritus was flakes of fruitfly-culture medium mixed with chopped mushrooms and potatoes. Detrital enhancement was applied every two weeks from mid-July through late September in 2014 (6 applications), and from mid-April through mid-August in 2015 (9 applications).

Soil micro-arthropods were sampled about a week before detrital additions started (initial conditions; designated as sampling period “S0”) and three months later (sampling period “S1”). Animals were sampled again in the fenced plots in early April the following year, just prior to resumption of detrital supplementation in the Low and High treatments (sampling period “S2”); and were sampled in all plots at the end of the experiment, 13 months after detrital supplementation commenced (sampling period “S3”). One sample of leaf litter and the underlying organic soil horizon was taken from each plot during each of the four sampling periods. Micro-arthropods were extracted in the laboratory, counted, and identified to at least the family level.

Numbers of the three response variables in each sample from each sampling period in the three fenced treatments (None, Low and High) were then analyzed using a combination of mixed-effects and fixed-effect linear and generalized linear models (LM, LMM, GLM, GLMM) using R-software packages (R Core Team, 2020). The naturalness of the experiment was evaluated in two ways: (1) by using LM and GLM to determine if densities in the None (fenced, no detritial supplement) and REF (open, no detrital supplement) treatments differed in sampling periods S1 and S3 (a direct, explicit test of a possible effect of fencing on densities of micro-arthropods); and (2) by comparing graphs of the temporal patterns of mean densities (with 95% CI’s) of the response variables in the REF and three fenced treatments (an indirect, implicit test of a possible effect of fencing on the response of Resource, IG Pred and IG Predator to the imposed detrital enhancement).

Usage Notes

Each line of the ReadMe file contains the numbers of each of the response variables (categories of micro-arthropods extracted from each litter/soil sample (last three columns)), and the variables that define the experimental design. There are no missing values in the dataset.

Variables in bold lettering were input to the statistical modeling using R software. The names are identical to those in Appendix 3 of the Supplemental Materials linked to the published article, which contains output from R of the tests of model assumptions, ANOVA tables, and descriptions of criteria used in model selection. The three response variables are underlined (and are italicized in the “Meaning” column, where their names correspond to usage in the above-referenced publication).

The ReadMe file also contains two other variables not needed for the statistical modeling: the date and year each sample was taken.  Note that samples were taken over several days in each of the four sampling periods (levels “S0”, “S1”, “S2” and “S3” of the variable “RmTot”). Samples were divided among treatments and blocks on each date.

The derivation of the variable names is self-explanatory except for “RmTot”, in which “Rm” indicates that the four sampling periods were used in the repeated-measures LMM and GLMM analyses.

Funding

National Science Foundation, Award: DGE-0549245

University of Illinois at Chicago, Award: Elmer Hadley Research Award