Data from: Multiscale threats shape the occurrence dynamics of a threatened aquatic salamander and reveal a possible extinction debt
Data files
Mar 17, 2026 version files 105.33 KB
-
Data.csv
65.52 KB
-
model_script.R
17.17 KB
-
README.md
4.30 KB
-
turnover_equilibrium_script.R
5.31 KB
-
YearlySiteCovs.csv
13.03 KB
Abstract
Freshwater ecosystems are impacted by anthropogenic stressors, resulting in roughly one-third of freshwater fauna being threatened with extinction. The Neuse River Waterdog (Necturus lewisi) is a large aquatic salamander endemic to the Neuse and Tar-Pamlico River basins of eastern North Carolina, USA, and it was listed as threatened under the federal Endangered Species Act in 2021. Habitat degradation has been identified as the dominant threat driving N. lewisi occurrence, and its effect may be delayed. The USFWS Draft Recovery Plan classified investigation into the species’ occurrence dynamics (colonization/extinction) as a high-priority action. We hypothesized that extinction probabilities would decrease in high-quality local instream habitats, increase with high proportions of disturbed land cover in the contributing watershed, and increase in years with intense droughts. We evaluated these hypotheses within a dynamic occupancy modeling framework using five consecutive years of detection/non-detection data collected across 176 locations. We derived estimates of annual occurrence, turnover, and equilibrium occupancy (stability) to investigate if spatial responses to threats were delayed – an extinction debt. The top-ranked model supported the hypotheses on drivers of site-specific extinction probabilities, including a negative effect of habitat quality, a positive effect of developed land cover in the watershed, and a positive effect of annual drought intensity. The derived estimates broadly indicated that annual occurrence was highest in rural subpopulations (i.e., management units), turnover was highest in urban subpopulations, and equilibrium occupancy was lower than required to maintain stability in most subpopulations of the Neuse River basin. The estimated occurrence dynamics and their derived quantities suggested an extinction debt in urban subpopulations that may be accelerated by stochastic drought events. This study describes a novel use of the dynamic occupancy model framework within an extinction debt context and provides partnering conservation agencies with information important to guiding recovery of the Neuse River Waterdog.
Dataset DOI: 10.5061/dryad.63xsj3vgg
Description of the data and file structure
This folder contains four files in addition to this metadata text file:
1) Data.csv
2) YearlySiteCovs.csv
3) model_script.R
4) turnover_equilibrium_script.R
Data.csv
This .csv file contains the detection history, site-specific covariates, and visit-specific covariates of interest that are associated with each of the 176 sites in the study.
- Column A: Site #
- Columns B-U: Detection history (1= detected, 0= not detected, [BLANK]= not surveyed) of up to four consecutive visits in five consecutive years (20 total visits)
- Columns V-X: Site-specific local covariates (Bottom Substrate, Cover Score, Total). Scores taken from field-collected rabid habitat assessments.
- Columns Y-AC: Site-specific landscape covariates (devel= % developed land cover in the watershed, grass= % herbaceous/pasture land cover in the watershed, crop= % agricultural crop land cover in the watershed, wetland= % wetland land cover in the watershed, forest= % forested land cover in the watershed). Raw land cover data collected and reclassified from the NLCD 2019 data set. Percentages calculated using an inverse distance-weighted function in ArcGIS Pro using a toolkit created by Peterson and Pearse 2017.
- Column AD: MU= subpopulation (i.e., management unit)
- Columns AE-AX: Daily mean stream discharge (m^3/s) associated with each day in the detection history. Because N. lewisi captures occurred overnight, discharge values are those from the day prior to the detection history (e.g., Day1= detection history on Tuesday, Discharge1= discharge data from Monday). We interpolated discharge values for sites without USGS stream gages.
- Columns AY-BR: Bait age, in days, associated with each day in the detection history.
YearlySiteCovs.csv
This .csv file contains the covariates of daily mean stream discharge and annual drought intensity index, formatted in two columns, respectively, to match the format required of yearly dynamic covariates in the 'unmarked' framework.
model_script.R
This .R script file contains annotated code that was used in the analysis, including:
- Packages used
- Data formatting
- Candidate models fit
- Global model fit and Goodness-of-fit test
- Model selection process
- Formatting results, including bootstrapping SEs for finite-sample estimates
- Example covariates effects plot
turnover_equilibrium_script.R
This .R script file contains annotated code that was used to derive and analyze the turnover and equilibrium occupancy estimates, including:
TURNOVER
- Example turnover function from the colext (unmarked) vignette on CRAN
- Our adaptation of the turnover function, shown as an example with the Eno/Flat subpopulation
- Bootstrapping code to estimate 95% CIs of turnover probabilities
- Plotting the annual turnover of the example subpopulation
EQUILIBRIUM OCCUPANCY
- Our adaptation of the equilibrium occupancy function, shown as an example with the Eno/Flat subpopulation
- Bootstrapping code to estimate 95% CIs of equilibrium occupancy probabilities
- Plotting annual equilibrium occupancy of the example subpopulation
- Calculate the mean non-equilibrium (difference between the mean finite-sample estimate and the mean equilibrium occupancy) to estimate if the example subpopulation currently has occupancy greater than, lower than, or equal to what is required to maintain stability.
Code/software
Viewing/use of this data requires Program R and the loaded packages 'ggplot2', 'unmarked', 'corrplot', 'AICcmodavg', and 'dplyr'.
We uploaded annotated code scripts that describe the data, formatting, use, and workflow for the entire study analysis. We also uploaded a file 'metadata.txt' that succinctly describes all uploaded data and scripts.
Access information
Other publicly accessible locations of the data:
Data was derived from the following sources:
- None
