Evaluating mark-resight survey design performance using simulation: a case study of endangered Steller sea lions
Amanda J. Warlick, Brian S. Fadely, Peter Mahoney, Sharon R. Melin, Tom Gelatt, Kim Raum-Suryan, Sarah J. Converse
Please contact Amanda Warlick at amanda.warlick@noaa.gov for questions about the code or data.
Secondary contact: Sarah Converse (sconver@uw.edu)
Published in Ecosphere.
Effective monitoring is fundamental to estimating wildlife population parameters with a level of accuracy and precision that is adequate to inform management decisions. However, it is important to balance the trade-offs between survey effort (and therefore costs) and the resulting data quality to ensure that monitoring surveys are both effective and efficient. As such, evaluating the expected performance of monitoring designs prior to survey implementation can help identify survey designs that require the least investment for a desired level of performance. In this study, we present a simulation framework for examining the accuracy and precision of age-specific survival estimates and the probability of detecting a change in survival within the context of mark-resight monitoring programs. We consider 90 survey designs that vary across a range of attributes, including marked cohort size, marking frequency, study duration, and resight probability. We implement this framework in the context of designing an effective and efficient monitoring program for Steller sea lions (Eumetopias jubatus), which is complicated by heterogeneity in rookery accessibility, rookery population sizes, and abundance trends across the species’ range. Our results highlight a variety of survey designs that reliably meet pre-defined precision targets, with precision and accuracy being more strongly affected by marked cohort size, marking frequency, and study duration than resight probability. We found that historical mark-resight survey effort for Steller sea lions has been sufficient to reliably achieve precision targets for younger age class survival probabilities only for survey designs characteristic of rookeries where abundance trends are stable or increasing. In contrast, the probability of achieving survival estimates with target levels of precision using survey designs characteristic of rookeries where abundance has been declining is low (< 25%) due to smaller marked cohort sizes, less frequent marking at remote sites, and fewer years of available data. Though our results indicate that the precision of survival estimates for subpopulations of conservation concern can be improved by longer-term monitoring, the constraints of monitoring small populations may limit the ability of biologists to detect changes in population dynamics on management-relevant time horizons. To identify cost-effective survey designs in the absence of actual true survey costs, we evaluated survey design performance with respect to a relative costs schema. This modeling framework can be applied in a variety of contexts to assist natural resource managers in developing monitoring programs that efficiently meet specific monitoring objectives..
- SSL_FutureProof_sim.Rmd includes simulation and estimation functions, code to run simulations and model fitting in nimble, and code to generate model performance metrics.
Code to generate figures is available upon request but generally not extensively included here.
This was primarily a simulation-based study. Contact the first author for code and data that were used to produce empirical estimates that guided data generation for the simulations.
Results files can be obtained by contacting the first author.
Hmisc_4.5-0
ggstance_0.3.5
reshape2_1.4.4
truncnorm_1.0-8
DescTools_0.99.43 c
cowplot_1.1.1
demogR_0.6.0
gtools_3.9.2
knitr_1.31
stringr_1.4.0
purrr_0.3.4
ggplot2_3.3.6
tidyverse_1.3.1
dplyr_1.0.5
readr_2.0.1
here_1.0.1
tidyr_1.1.3
lubridate_1.7.10
latex2exp_0.9.5
Warlick, AJ, Fadely, BS, Mahoney, P, Melin, SR, Gelatt, T, Raum-Suryan, K, Converse, SJ. 2025. Evaluating mark-resight survey design performance using simulation: a case study of endangered Steller sea lions. Ecosphere
The SSL_FutureProof_sim.Rmd file includes code chunks that outline functions used to designate initial values and known latent state information, nimble functions used to fit the model, model text, transition matrices and descriptions of the model parameters and simulation scenarios, and code used to run the model in parallel using nimble. Simulations were saved in batches due to computing resources.
Contact the first author for (1) data processing and mark-resight model code used to produce empirical estimates (that guided data generation for the simulation); and (2) code used to generate figures and summary statistics found in the manuscript.