Data and code from: The effect of resource concentration on consumer population densities depends on spatial scale and diet breadth
Data files
Jun 15, 2026 version files 26.33 MB
-
README.md
6.56 KB
-
repositoryForDryad.zip
26.32 MB
Abstract
Herbivore responses to host plant concentration and plant community diversity are predicted to depend on diet breadth, yet most studies to date have focused on dietary specialists. The few that have included generalist herbivores have largely been conducted in agricultural settings or measured resources at a single spatial scale, leaving it unclear whether these predictions hold in natural systems across multiple spatial scales. Here, we examine how host plant concentration and plant community diversity shape the population densities of dietary specialist and generalist larval Lepidoptera (caterpillars) in secondary forests of New England. We measured host plant assemblages and densities at scales from 3–400 m to assess the patch size-specific effects of host concentration and plant diversity. To accurately characterize the resource landscape for herbivores that consume >1 host species, we weighted host plant concentration by electivity (a measure of each species' observed frequency of association with each host plant relative to its availability) and demonstrate that this weighting is necessary to detect responses that are otherwise obscured. We found that dietary generalists’ population densities correlated positively with their host plant concentrations at scales up to 25 m: a pattern that was only detectable when electivity was considered. In contrast, dietary specialists’ population densities correlated positively with their host plant densities at scales < 3 m and negatively at scales 25–50 m. This suggests that specialists aggregate in small, dense patches of resources, while their density saturates or declines at intermediate scales, consistent with resource dilution, not concentration. Neither generalist nor specialist caterpillar populations responded strongly to plant community diversity at any scale. These findings demonstrate that herbivore responses to resource concentration are contingent on diet breadth and spatial scale, that detecting generalist responses requires diet- breadth-appropriate measurement, and that the influence of plant diversity on herbivore density documented in agroecosystems may not generalize to structurally complex natural forests.
Dataset DOI: 10.5061/dryad.8931zcs5z
Description of the data and file structure
These data seek to find a relationship between the spatial structure of host plants and the caterpillars that consume them. It consists of two large data collection efforts - the caterpillar dataset and the vegetation dataset - and the associated analyses on resource concentration
Files and variables
The repository is divided into four main folders:
Data
The raw data to be used in the analysis. All .rdata files in this folder are created as a result of scripts in the "analysis" folder, with the exception of those found in raw/. A description of the three raw data files is found below:
CaterpillarSurveysAllYears-v1_2.csv
Caterpillar surveys conducted between 2017 and 2019 and preliminary data collected in 2015 (from a subset of sites, not used). Each line is a captured caterpillar, or, if a branch was empty, a record that a branch was sampled.
- ID: a unique id
- Page: physical data page number. A value of 'null' indicates this information was not collected.
- LineNum: physical data page line number. A value of 'null' indicates this information was not collected.
- Enterer: initials of the individual who electronically entered the data. A value of 'null' indicates this information was not collected.
- Recorder: initials of the individual who physically recorded the data. A value of 'null' indicates this information was not collected.
- Team: initials of the individual(s) who collected the data. A value of 'null' indicates this information was not collected.
- Year: year the data was collected
- SurveyDate: date the data was collected
- BlockID: geographic block of the forest patch (data was divided into 13 blocks)
- SiteID: forest patch (32 total patches)
- Plot: plot, within patch, where sampling occurred (12 plots per patch)
- PlotID: unique identifier for each plot
- HostID: 5-letter plant code (first 3 letters of genus, first 2 letters of species)
- HostAbund: total number of that species of plant within the plot
- IndNum: indicator of which number plant of that species was sampled within the plot
- BranchNum: indicator of which number branch from that individual was being sampled.
- BranchLength: length of branch being sampled (unused). Only one of BranchLength, BranchDiam, or NumLeaves was measured per-branch, and 'n/a' in this column indicates another method was used.
- BranchDiam: diameter of branch being sampled (unused). Only one of BranchLength, BranchDiam, or NumLeaves was measured per-branch, and 'n/a' in this column indicates another method was used.
- NumLeaves: number of leaves on branch being sampled (unused). Only one of BranchLength, BranchDiam, or NumLeaves was measured per-branch, and 'n/a' in this column indicates another method was used.
- CatID: 6-letter code (first four letters of genus, first 2 letters of species) for caterpillar species (if present)
- CatSpecies: full binomial of caterpillar species, if present. A value of 'n/a' indicates no caterpillar was present, and correlates with a blank CatID
- Count: number of caterpillars of that species present in sample
- BadHost: indicator of specialist feeding on non-host (yes if "bad")
- CatIDNum: internal capture ID. Unused. A value of 'n/a' indicates no capture ID was given.
- RearID: internal rearing ID
- FieldNotes: Notes taken on sample from team in field
- EntererNotes: Notes taken from electronic data enterer
- CleaningNotes: Notes taken from data cleaner
VegetationSurveys.csv
A survey of vegetation presence in the 32 forest patches in this study. Vegetation was sampled in 21 plots per forest patch. each unique plot was surveyed once. See associated paper for a diagram of vegetation patch survey design. Each line is 1 observed tree.
Many columns in this dataset were not used for this paper: please see the "more detailed information" section for details on the unused columns.
- BlockID: geographic block of the forest patch (data was divided into 13 blocks)
- SiteID: forest patch (32 total patches)
- FragSize - ScatPredict: see "more detailed information." All unused.
- PointID: unique vegetation plot identifier
- Latitude: plot latitude
- Longitude: plot longitude
- CornerFlag: internal indicator of corner flag presence. A value of 'n/a' indicates no flag was placed.
- Year: survey year
- SurveyDate: survey date
- Soil - Laurel: plot-level data on ground cover type. Unmeasured in plots surveyed in 2017. Unused. A 'null' value indicates that this was not measured.
- Tree: species observed
- TreeGenus: genus of tree observed
- TreeSpecies: species of tree observed
- TreeFamily: family of tree observed
- DBH: diameter at breast height (1.3m) of tree, if over ~1.3m tall. If tree was under 1.3m tall, a value of 'n/a' is present.
- Height: height of tree species, if under ~1.3m tall. If tree was over 1.3m tall, a value of 'n/a' is present.
- Branches1_2m: presence or absence of branches between 1 and 2m from the ground
- Dead: is the tree dead
vegplots.csv
A legacy version of vegetation.csv that contains no unique information
Analysis
An ordered list of scripts that completes the analysis. Many scripts require the sourcing of the previous scripts in the folder, even if not explicitly stated. Scripts are commented throughout to describe their function and data products created
MakeFigures
The scripts for creating the figures for the associated manuscript
More detailed information
Can be found at https://datadryad.org/dataset/doi:10.5061/dryad.k3j9kd5k8
Code/software
All analyses were conducted in R (version 4.4.1; R Core Team 2024) using RStudio. The following packages were used: lme4 (v1.1-35.5), bbmle (v1.0.25.1), glmmTMB (v1.1.9), tidyr (v1.3.1), ggthemes (v5.1.0), ggallin (v0.1.1), future (v1.67.0), pscl (v1.5.9), MASS (v7.3-60.2), boot (v1.3-30), parallel (v4.4.1), patchwork (v1.3.0), and ggplot2 (v3.5.1). All software is freely available through CRAN (https://cran.r-project.org)
Access information
Other publicly accessible locations of the data:
