Northern Bobwhite roost site selection influences survival during the non-breeding season
Data files
Aug 03, 2026 version files 1.19 MB
-
NOBO_roost_rsf_survival.zip
1.19 MB
-
README.md
4.31 KB
Abstract
A key assumption of the ideal free distribution theory is that animals inherently select resources based on environmental cues that ultimately benefit fitness, thus facilitating natural selection within a population. Habitat management and conservation practices are commonly based on results of resource selection models with the assumption that selection and fitness are correlated. This is an unsettling trend given this assumption has been proven to be violated in specific instances and systems where animals select resources that have adverse effects on vital rates. We studied how nocturnal habitat selection of 6 common landcover types influenced survival rates of Northern Bobwhite (Colinus virginianus) on public lands in southern Indiana, USA during the non-breeding season. We found that selection for woody cover had a positive effect on survival rates. Our results suggest that increased fitness is facilitated in this system as individuals who preferred a resource which offers predator refuge. To develop more comprehensive knowledge on habitat selection for species, studies should examine the effects that selection has on one or more vital rates.
Dataset DOI: 10.5061/dryad.wpzgmsc54
Description of the data and file structure
Approach to examine if roost site selection of northern bobwhite influences survival.
Each main folder contains data files and scripts to run specific analyses described in the manuscript.
Files and variables
Root folder: NOBO_roost_rsf_survival.zip
Resource_selection (folder)
GoosePond_roosts_2020_2021_ran20_170buf[071924].csv
Description: Data file created from NOBO_Roost_spatial_sampling[071624].R that is needed for resource selection analysis.
Columns:
- Ag_Distance: distance in meters to closest agricultural field
- Disturbance: proportion of disturbed area in a 20 meter square cell
- Grassland_Dist: distance in meters to perimeter of unburned grassland (negative values are interior distance)
- Wetland_Dist: distance in meters to wetlands
- Woody_Dist: distance in meters to woodlands
- Cell.ID: unique ID of a raster grid which is created for indexing purposes
- Band.ID: unique leg-band of radio tagged bird
- Covey.ID: unique covey ID that bird was located with
- Roost.ID: unique roost ID that the bird was at (multiple birds roosting at the same location will have the same roost ID)
- use: 1 = used site, 0 = random site
- year: 2020 = year 1, 2021 = year 2
Roost_RSF_model[032923].jags
Description: Annotated JAGS code used in conjunction with Roost_RSF_model[110624].R to estimate nocturnal habitat selection rates.
Roost_RSF_model[110624].R
Description: Annotated R script that uses GoosePond_roosts_2020_2021_ran20_170buf[071924].csv & Roost_RSF_model[032923].jags to estimate nocturnal habitat selection rates for individuals and population-level effects. Script also creates habitat selection figures used in publication and exports CSV file "GoosePond_roostRSF_2020_2021_betas[110624].csv" for individual selection coefficient estimates for individuals included in the analysis.
GoosePond_roostRSF_2020_2021_betas[110624].csv
Description: data file created from Roost_RSF_model[110624].R that is needed for survival model. Each row corresponds to a unique bird.
Columns:
- index: indexing ID used in resource selection model
- ag: mean beta coefficient estimate for effect of distance to row-crop fields on selection
- disturbed: mean beta coefficient estimate for effect of proportion of disturbed areas on selection
- grassland: mean beta coefficient estimate for effect of distance to grasslands on selection
- woody: mean beta coefficient estimate for effect of distance to woody cover on selection
- wetland: mean beta coefficient estimate for effect of distance to wetlands on selection
- woodland: mean beta coefficient estimate for effect of distance to woodlands on selection
- ID: unique band ID for specific bird. _1 and _2 correspond to year 1 and 2 of data
Survival (folder)
Fatesheet20_22.csv
Description: data file needed for survival analysis.
Irrelevant columns are: Treatment, trap, unit, sex, age, weight, freq.
Columns which are needed for analysis are as follows:
- Capture date: date of first capture and tagging
- ID: unique band ID for specific bird
- fate: censor = removed from sample, crip = crippling loss from hunting, H = harvested by hunter, hpothermia = hypothermia, Mam = mammal, UA = avian, UNK = unknown predator/cause, Vehicle = suspected or confirmed vehicle cause, blank cell = bird alive at end of period
- fate date = date of censor or death
- Year = specific year cohort the bird is in
GoosePond_roostRSF_2020_2021_betas[110624].csv
Described in Resource_selection folder.
GP_FW_20_22_survival_roostRSF.R
Description: Annoted R script that uses Fatesheet20_22.csv, GoosePond_roostRSF_2020_2021_betas[110624].csv, and GP_FW_20_22_survival_roostRSF[032023].jags to estimate effects of habitat selection on survival rates. Script also creates survival curve figure used in publication.
GP_FW_20_22_survival_roostRSF[032023].jags
Description: Annotated JAGS code used in conjunction with GP_FW_20_22_survival_roostRSF.R to estimate effects of nocturnal habitat selection on survival rates.
Code/software
RStudio and JAGS.
We collected nocturnal habitat use and survival data for two fall/winter seasons during 2020-21 and 2021-22 via radio-telemetry. Nocturnal locations of radio-tagged birds were recorded one to three times/week from a distance of ~25-30m via homing techniques. Locations were recorded between 45 minutes post-sunset and one hour pre-sunrise. Estimated locations of tracked bobwhite were recorded via GPS with high resolution (1-3 m) base maps. Spatial datasets (raster layers) were created from digitized satellite imagery. Our coarsest dataset resolution was 20m2, so we resampled all spatial layers to this resolution using the Resample tool in ArcGIS Pro 2.5 (Esri, Redlands, CA, USA). To account for our maximum estimated telemetry error of 20m, we used the Focal Statistics tool in ArcGIS Pro 2.5 (Esri, Redlands, CA, USA) with a 3x3 moving window on all spatial layers.