Coyote Inventory - Population Modelling Information - Analysis Scripts for Cape Hatteras and Cape Lookout National Seashores and Pea Island National Wildlife Refuge

Sossover, Daniel

Core population modeling contains the complete robust design Jolly-Seber multi-state mark-recapture R script, density visualization script, and associated data and output subdirectories. Our population analysis uses an open-population robust design Jolly-Seber (RDJS) multi-state mark-recapture framework. Coyotes were individually identified through microsatellite genotyping of non-invasive scat samples collected across four replicate sampling rounds (survey round AKA secondary occasions) within each season (primary period). The model was implemented in R interfacing with JAGS for Bayesian estimation and simultaneously estimates apparent survival, recruitment, population size, and state-specific detection probabilities across islands and seasons. Posterior distributions from the fitted model are then passed to a separate script to calculate population growth rates (lambda) and associated credible intervals. The model was implemented in R interfacing with JAGS for Bayesian estimation. It simultaneously estimates apparent survival, recruitment, population size, and state-specific detection probabilities across islands and seasons. Posterior distributions from the fitted model are then passed to a separate script to calculate finite population growth rates (lambda) and associated credible intervals. Input Data: obx_coyote_matrix_nathan - updated.xlsx — The individual recapture history matrix. Each row is a genetically confirmed coyote (identified by recapture_id); columns represent each sampling round across Winter 2024 (w11–w14), Summer 2024 (s11–s14), and Winter 2025 (w21–w24). Cell values indicate which island the individual was detected on in a given round, with zero representing no detection. This is the primary capture history input to the RDJS model. obx_coyote_island_key - Sheet1.xlsx — Island-level sampling effort lookup table. Records which barrier islands were sampled in each broad season (w1 = Winter 2024, s1 = Summer 2024, w2 = Winter 2025) using binary indicators, and provides the mapping between numeric island codes and island names (Bodie, Pea, Hatteras, Ocracoke, etc.). Main purpose is to show the model we did not survey Bodie in Winter 2024. trapping_matrix_complete.xlsx —Island-level trapping effort matrix. Records whether each island received physical trap deployment across two inter-season trapping sessions (t1 = the interval between the end of Winter 2024 and the start of Summer 2024; t2 = the interval between Summer 2024 and Winter 2025) and whether road-based surveys were conducted (roads_det). Used to model detection probability and effort covariates in the RDJS framework. Model Outputs: jags_model_out_onemiliter_twohundburnin.RData — Saved R workspace containing the full JAGS posterior sample object from the RDJS model run (1 million iterations, 200 burn-in). This is the master checkpoint file; the growth rate and visualization scripts load this object directly, so the computationally intensive MCMC chain does not need to be re-run. jags_model_out_onmiliter_twohundburnin.csv — Tabular export of key posterior summaries (means, credible intervals) from the JAGS model run, provided in Excel format for review without requiring R. Robust_Design_Posterior_Distribution_Output_of_Abundance_to_Density.csv — Derived density estimates converted from posterior abundance estimates using island area. Provides per-island, per-season coyote density with associated credible intervals; the direct source data for the density visualization script and maps.