This archive contains the data and analysis code for a multi-season robust-design mark-recapture study of coyotes (Canis latrans) across seven barrier islands on North Carolina Outer Banks. Individuals were identified from noninvasive fecal DNA. A Bayesian multistate Jolly-Seber model, fit in JAGS via the R package jagsUI, estimates island-specific abundance, apparent survival, per-capita recruitment, inter-island movement, and derived population metrics across three primary periods (Winter 2024, Summer 2024, and Winter 2025).

Contents

• OuterBanksCoyote_MarkRecapture_Script_for_dryad.R - analysis script in R
• obx_coyote_jags_output.rds - fitted model object with full posterior chains

Data

• obx_coyote_capture_histories.csv - individual capture histories
• obx_coyote_island_key.csv - per island survey effort by primary period
• obx_coyote_island_covariates.csv - per island management and detection covariates

Running the script also writes the JAGS model definition to RobustDesign_OuterBanksCoyotes_PerCapitaRecruitment.txt in the working directory.

Island Coding

Islands are numbered 1-7 throughout the data and code:

Primary periods: w1 = Winter 2024, s1 = Summer 2024, w2 = Winter 2025. Each primary period consisted of four secondary sampling occasions.

File Descriptions

OuterBanksCoyote_MarkRecapture_Script_for_dryad.R: This file contains the complete analysis. Sections: (1) loads data in; (2) defines indices and summarizes; (3) builds and writes the JAGS model; (4) formats the data, sets initial values, and MCMC settings; (5) runs JAGS; (6) summarizes the posterior, checks convergence, and saves outputs. Expects the three input files in a data/ subdirectory of the working directory.

data/obx_coyote_capture_histories.csv: This file contains one row per uniquely genotyped individual (62 rows) plus a header with 13 columns:

• recapture_id: unique individual identifier (genotype-based)
• w11-w14, s11-s14, w21-w24: the 12 secondary occasions, four within each of the three primary periods (w1, s1, w2). Each cell provides an island code (1-7) detailing which island the individual was detected on during that occasion, or 0 if the individual was not detected on that occasion.

One individual (recapture_id = “HH”) was detected on both North Core (5) and Ocracoke (4) within a single primary period. Her off-island (Ocracoke) detection was removed before the analysis so that the model's assumption of no inter-island movement within a primary period holds. So in this file HH appears on North Core only.

data/obx_coyote_island_key.csv: One row per island (7 rows) plus a header with five columns:

• island: island code (1-7)
• w1, s1, w2: survey-effort indicator for each primary period; 1 if the island was surveyed during that primary period, 0 if it was not. Bodie (w1 = 0) was not surveyed in the first primary period.
• island_name: island name.

data/obx_coyote_island_covariates.csv: One row per island (7 rows) plus a header with four columns:

• island: island code (1-7)
• t1, t2: predator-management covariate on apparent survival for the two survival intervals (interval 1 = Winter 2024 to Summer 2024; interval 2 = Summer 2024 to Winter 2025). 1 = island managers engaged in lethal removals of coyotes during that interval, 0 = not.
• roads_det: detection covariate; 1 if road-based scat surveys were conducted on that island, 0 if not.

Software

R version: 4.3.2
jagsUI version: 1.6.2
JAGS version: 4.3.1
Operating system: Windows 10 (64-bit)

Running the analysis

1. Place the three CSV files in a data subdirectory of the working directory.
2. Install JAGS and the R package jagsUI.
3. Run OuterBanksCoyote_MarkRecapture_Script_for_dryad.R

The script sets set.seed(20240724) and assigns a distinct reproducible RNG seed to each MCMC chain, so a run can reproduce our same results given the same JAGS and jagsUI versions. The full run uses 3 chains, 40,000 adaptation iterations, 200,000 burn-in iterations, and 1,000,000 post-burn-in iterations per chain, thinned by 10. This is computationally intensive and is set to run on a multi-core machine, as parallel is set to TRUE.

Outputs produced when the script is run

• obx_coyote_jags_output.rds: the fitted model object with full posterior chains.
• obx_coyote_jags_summary.csv: posterior summary (mean, SD, quantiles, Rhat, and effective sample size) for all monitored parameters.
• obx_coyote_jags_output.RData: the same fitted model object saved in R's native.RData format