Data from: One-stage spatial mark-resight analysis reveals an increasing grizzly bear population with declining density near roads
Data files
Mar 17, 2025 version files 26.32 MB
-
d2_cam.csv
15.65 MB
-
detect_marked.csv
6.10 MB
-
detect_year_cam.csv
156.82 KB
-
gps_day.csv
304.28 KB
-
MountainParks_2007_YaHaTinda.sqlite
856.06 KB
-
occ_long_cam.csv
2.55 MB
-
pix_xy.csv
478.18 KB
-
pixel_lookup.csv
16.64 KB
-
README.md
4.56 KB
-
site_cam.csv
84.67 KB
-
sites_camera_2023.sqlite
77.82 KB
-
statespace_2023.sqlite
28.67 KB
-
tbl_area_parks.csv
107 B
-
tbl_DataDescription_2023-11-13.csv
5.48 KB
-
tbl_juvenile.csv
1.75 KB
-
tbl_specs.csv
123 B
Abstract
Wildlife ecologists throughout the world strive to monitor trends in population abundance to help manage wildlife populations and conserve species at risk. Spatial capture-recapture studies are the gold standard for monitoring density, yet they can be difficult to apply because researchers must be able to distinguish all detected individuals. Spatial mark-resight (SMR) models only require a subset of the population to be marked and identifiable. Recent advances in SMR models with radio-collared animals required a two-staged analysis. We developed a one-stage generalized SMR (gSMR) model that used detection histories of marked and unmarked animals in a single analysis. We used simulations to assess the performance of one- and two-stage gSMR models. We then applied the one-stage gSMR with telemetry and remote camera data to estimate grizzly bear (Ursus arctos) abundance from 2012 to 2023 within the Canadian Rocky Mountains. We estimated abundance trends for the population and reproductive females (females with cubs of the year). Simulations suggest one- and two-stage models performed equally well. One-stage models are more dependable as they use exact likelihoods whereas two-stage models have shorter computation times for large datasets. Both methods had > 95% credible interval coverage and minimal bias. Increasing the number of marked animals increased the accuracy and precision of abundance estimates and > 10 marked animals were required to obtain coefficients of variation < 20% in most scenarios. The grizzly bear population increased slightly (growth rate λmean = 1.02) to a 2023 density of 10.4 grizzly bears/1000 km2. Reproductive female abundance had high interannual variability and increased to 1.0 bears/1000 km2. Population density was highest within protected areas, within high quality habitat and far from paved roads. The density of activity centers declined near paved roads over time. Mechanisms of decline may have included direct mortality and shifting activity centers to avoid human activity. Our study demonstrates the influence of human activity on localized density and importance of protected areas for carnivore conservation. Finally, our study highlights the widespread utility of remote camera and telemetry-based spatial mark-resight models for monitoring spatiotemporal trends in abundance.
https://doi.org/10.5061/dryad.g79cnp5wv
Description of the data and file structure
tbl_DataDescription_2023-11-13.csv provides a description of all data sets and column attributes for grizzly bear gSMR analysis. Here is a summary of each data set.
DataSet | Description |
---|---|
tbl_specs.csv | table of constants used for models |
pix_xy.csv | table of pixel coordinates with spatial attributes. 4 km spacing. |
site_cam.csv | Site attributes of remote camera locations |
occ_long_cam.csv | Occurrence data for grizzly bears in long format. 14 day occasions |
detect_year_cam.csv | Detection history data summarised by year |
detect_marked.csv | Detection data for marked (radio-collared) grizzly bears |
gps_day.csv | table of GPS locations with one location per day for each animal. |
d2_cam.csv | matrix for squared euclidean distances between each pixel (row) and each camera site (column) |
pixel_lookup.csv | matrix lookup table that links x.id and y.id (pixel coordinates) to a unique pixel number in pix_xy. |
tbl_juvenile.csv | table of parturition records and survival records (when known) for radio-collared female grizzly bears |
statespace_2023.sqlite | Shapefile of state-spaced used for the joint gSMR analysis. UTM Zone 11 NAD 83 |
MountainParks_2007_YaHaTinda.sqlite | Shapefile of study area including national parks and the Ya Ha Tinda ecosystem. UTM Zone 11 NAD 83 |
sites_camera_2023.sqlite | sqlite file of remote camera locations and raw detection rates. UTM Zone 11 NAD 83 |
tbl_area_parks.csv | Area in km2 of Banff National Park, Kootenay National Park, Yoho National Park, and the Ya Ha Tinda study area |
Sharing/Access information
Data was collected by Parks Canada and the University of Montana.
Code/Software
We provide R code for the simulation study and grizzly bear gSMR analysis with the following scripts:
Simulation Study
1a. gSMR simStudy functions and nimble code.R: R script with functions to generate spatial mark-resight data and nimble code for fitting two-stage and joint gSMR models.
1b. gSMR Run SimStudy One and Two Stage gSMR.R: R script to specify scenario parameters and to run simulation study. This script calls upon 1a. gSMR simStudy functions and nimble code.R
Grizzly bear analysis
2a. nimble_efficient_functions.R: R script with nimble functions that help reduce computation time.
2b. nimble model gSMR grizzly.R: R script with the nimble code for the grizzly bear joint gSMR model.
2c. fit one stage gSMR model grizzly bears.R: R script for importing data and fitting the grizzly bear joint gSMR model. The script calls upon 2a. nimble_efficient_functions.R and 2b. nimble model gSMR grizzly.R.
Female grizzly bear with cub of year analysis
3b. nimble model gSMR grizzly_COY.R: R script with the nimble code for the female grizzly bear with cub of the year joint gSMR model.
3c. fit one stage gSMR model grizzly_COY.R: R script for importing data and fitting the grizzly bear joint gSMR model. The script calls upon 2a. nimble_efficient_functions.R and 3b. nimble model gSMR grizzly_COY.R.
We used R version4.4.0 and the following packages for the analysis:
stringr_1.5.1 nimble_1.1.0 sf_1.0-16 ggplot2_3.5.1 lubridate_1.9.3 tibble_3.2.1 tidyr_1.3.1 dplyr_1.1.4
The gSMR models combine remote camera detections of marked animals, unmarked animals, and telemetry data to estimate the baseline detection rate, home range scale parameter, and spatially explicit estimates of density. Our study area encompassed 15,483 km2 and included Banff, Kootenay, and Yoho Nation Parks and the Ya Ha Tinda ecosystem within the Rocky Mountains of Canada. The remote camera data contains detection histories from 25 marked, radio-collared grizzly bears and detections of unmarked grizzly bears recorded at 625 remote cameras from 2012 to 2021. Telemetry data contains daily global positioning system (GPS) locations from fifteen female and ten male grizzly bears. We provide source code to estimate spatial and temporal trends in grizzly bear density as well as the density of female grizzly bears with cubs of the year. We describe each data set and associated attributes in tbl_DataDescription_2023-11-13.csv.