Data from: Quantifying impacts of recreation on elk (Cervus canadensis) using novel modeling approaches
Data files
Mar 18, 2024 version files 952.84 KB
-
Diel_activity_analysis.zip
849.53 KB
-
Habitat_use_analysis.zip
75.88 KB
-
README.md
27.44 KB
Abstract
Recreation is known to impact wildlife by displacing and sometimes extirpating sensitive species, underscoring a need for policies that balance wildlife and recreation. This is especially pressing when Indigenous rights necessitate ecological integrity and sustainable populations of wildlife throughout traditional territories. In the Cascade Mountain Range of Washington, USA, Indigenous harvest of elk (Cervus canadensis) is declining, concurrent with increases in recreation. Yet, the nature and magnitude of the effects of recreation on elk is unknown, which prevents land managers from developing informed policies regarding recreation and wildlife management. Here, we use camera traps alongside visitation models that incorporate geolocated social media to quantify impacts of recreation on elk in western Washington. Random forest models show elk detection rates are relatively constant at low levels of recreation (0 – 11 people per day), but decrease by over 41% when recreation increases from 12 to 22 people per day. Activity overlap analysis also revealed a shift towards increased evening activity by elk at cameras with higher-than-average recreation (∆ = 0.70, 95% confidence interval = 0.61 - 0.88; χ2 = 7.79, p = 0.02). Generalized additive modeling confirms that elk are more crepuscular or nocturnal at camera locations with more than 10 hiker detections per day. We compare methods for estimating recreation, showing model-based estimates are more informative than camera-based estimates. This indicates that recreational intensity along and in the immediate vicinity of trails may be a better predictor of impacts on wildlife than camera-based estimates that quantify recreational intensity at specific locations along trails. We stress the importance of examining impacts of recreation on wildlife across multiple spatiotemporal scales, and underscore how novel approaches can provide land managers valuable tools to develop management strategies that balance recreation and wildlife. We hope that our work can also serve as a strong example of collaboration between universities, state agencies, and sovereign Indigenous nations towards the broader goal of mitigating negative impacts of recreation on wildlife.
Description
The attached dataset includes detection data for wildlife species captured on camera traps in western Washington, as well as numerous variables representing environmental conditions of the camera locations during the time that the image was taken. Data was collected from June 2022 - Sept. 2022 by researchers at the University of Washington, working in collaboration with Washington Department of Natural Resources (DNR) and The Tulalip Tribes of Washington. The goal of this project was to assess the impacts of recreation on elk throughout the western Cascades. Data therefore comprise a collection of variables anticipated to influence elk habitat use (and therefore elk detection rates), in addition to detailed information about the date and time during which these images were taken.
https://doi.org/10.5061/dryad.jdfn2z3j4
Data and file structure
The data are structured as such:
Habitat use analysis (folder)
-
AllVarsDataFrom_Prockoetal2024.csv\
Spreadsheet of data used in 1.pre-analysis.R (see below for code descriptions) which constructs a series of univariate random forest models that each regress elk detection rates (“CECA.avg”) against one of these variables.- “year_week”, year and week combination (categorical variable representing weeks as YYYY-WW)\
“stat”, camera station name (categorical variable, camera name)\
“stat_year_week”, unique combination of year, week, and camera station name (categorical variable, combination of camera name and “year_week”)\
“HIKER.n”, number of hikers detected on camera during that year-week at that camera (# hikers/year-week)\
“BIKER.n”, number of mountain bikers detected on camera during that year-week at that camera (# mountain bikers/year-week)\
“MOTOR.n”, number of motorized vehicles detected on camera during that year-week at that camera (# motorized vehicles/year-week)\
“HORSE.n”, number of horseback riders detected on camera during that year-week at that camera (# horseback riders/year-week)\
“DOG.n”, number of domestic dogs detected on camera during that year-week at that camera (# domestic dogs/year-week)\
“CECA.n”, number of elk detected on camera during that year-week at that camera (# elk/year-week)\
“ODHE.n”, number of black-tailed deer detected on camera during that year-week at that camera (# black-tailed deer/year-week)\
“URAM.n”, number of black bears detected on camera during that year-week at that camera (# black bears/year-week)\
“PUCO.n”, number of cougars detected on camera during that year-week at that camera (# cougars/year-week)\
“CALA.n”, number of coyotes detected on camera during that year-week at that camera (# coyotes/year-week)\
“LYRU.n”, number of bobcats detected on camera during that year-week at that camera (# bobcats/year-week)\
“active”, a binary indication of whether the camera was active during a given year-week (0,1; 0 = not active, 1 = active)\
“days.active”, how long the camera was active (days)\
“Height”, the height of the camera (m)\
“TrailWidth”, the width of the trail or road on which the camera was placed (m)\
“DistFeat”, the distance from the lens of the camera to the anticipated path of the target–generally the middle of the trail (m)\
“elev_point”, elevation at the point of the camera (m)\
“canopy_point”, canopy cover at the point of the camera (0-100%)\
“frst_age_point”, forest age at the point of the camera (years)\
“dist_water”, distance from the camera to the nearest water source (m)\
“DHI_NDVI_1”, dynamic habitat indices, normalized difference vegetation index raster value (band 1) from the year 2015, 1 km resolution\
“DHI_NDVI_2”, dynamic habitat indices, normalized difference vegetation index raster value (band 2) from the year 2015, 1 km resolution\
“DHI_NDVI_3”, dynamic habitat indices, normalized difference vegetation index raster value (band 3) from the year 2015, 1 km resolution\
“elev_250m”, average elevation within a 250m buffer around the camera location (m)\
“frst_age_250m”, average forest age within a 250m buffer around the camera location (years)\
“canopy_250m”, average canopy cover within a 250m buffer around the camera location (0-100%) \
“pct_suit_250m”, percent of the top 40% suitable habitat that falls within a 250m buffer around the camera location (0-100%)\
“suit_mean_250m”, the average habitat suitability value within a 250m buffer around the camera location (relative unit “suitability”)\
“pct_water_250m”, the percent of land cover that is composed of water in a 250m buffer around the camera location (0-100%)\
“pct_dev_250m”, the percent of land cover that is composed of developed land in a 250m buffer around the camera location (0-100%)\
“pct_barren_250m”, the percent of land cover that is composed of barren land in a 250m buffer around the camera location (0-100%)\
“pct_frst_250m”, the percent of land cover that is composed of forested land in a 250m buffer around the camera location (0-100%)\
“pct_shrub_250m”, the percent of land cover that is composed of shrublands in a 250m buffer around the camera location (0-100%)\
“pct_herb_250m”, the percent of land cover that is composed of herbaceous land in a 250m buffer around the camera location (0-100%)\
“pct_ag_250m”, the percent of land cover that is composed of agricultural land in a 250m buffer around the camera location (0-100%)\
“pct_wet_250m”, the percent of land cover that is composed of wetlands in a 250m buffer around the camera location (0-100%)\
“elev_500m”, average elevation within a 500m buffer around the camera location (m)\
“frst_age_500m”, average forest age within a 500m buffer around the camera location (years)\
“canopy_500m”, average canopy cover within a 500m buffer around the camera location (0-100%)\
“pct_suit_500m”, percent of the top 40% suitable habitat that falls within a 500m buffer around the camera location (0-100%)\
“suit_mean_500m”, the average habitat suitability value within a 500m buffer around the camera location (relative unit, “suitability”)\
“pct_water_500m”, the percent of land cover that is composed of water in a 500m buffer around the camera location (0-100%)\
“pct_dev_500m”, the percent of land cover that is composed of developed land in a 500m buffer around the camera location (0-100%)\
“pct_barren_500m”, the percent of land cover that is composed of barren land in a 500m buffer around the camera location (0-100%)\
“pct_frst_500m”, the percent of land cover that is composed of forested land in a 500m buffer around the camera location (0-100%)\
“pct_shrub_500m”, the percent of land cover that is composed of shrublands in a 500m buffer around the camera location (0-100%)\
“pct_herb_500m”, the percent of land cover that is composed of herbaceous land in a 500m buffer around the camera location (0-100%)\
“pct_ag_500m”, the percent of land cover that is composed of agricultural land in a 500m buffer around the camera location (0-100%)\
“pct_wet_500m”, the percent of land cover that is composed of wetlands in a 500m buffer around the camera location (0-100%)\
“elev_1km”, average elevation within a 1km buffer around the camera location (m)\
“frst_age_1km”, average forest age within a 1km buffer around the camera location (years)\
“canopy_1km”, average canopy cover within a 1km buffer around the camera location (0-100%)\
“pct_suit_1km”, percent of the top 40% suitable habitat that falls within a 1km buffer around the camera location (0-100%)\
“suit_mean_1km”, the average habitat suitability value within a 1km buffer around the camera location (relative unit, “suitability”)\
“pct_water_1km”, the percent of land cover that is composed of water in a 1km buffer around the camera location (0-100%)\
“pct_dev_1km”, the percent of land cover that is composed of developed land in a 1km buffer around the camera location (0-100%)\
“pct_barren_1km”, the percent of land cover that is composed of barren land in a 1km buffer around the camera location (0-100%)\
“pct_frst_1km”, the percent of land cover that is composed of forested land in a 1km buffer around the camera location (0-100%)\
“pct_shrub_1km”, the percent of land cover that is composed of shrublands in a 1km buffer around the camera location (0-100%)\
“pct_herb_1km”, the percent of land cover that is composed of herbaceous land in a 1km buffer around the camera location (0-100%)\
“pct_ag_1km”, the percent of land cover that is composed of agricultural land in a 1km buffer around the camera location (0-100%)\
“pct_wet_1km”, the percent of land cover that is composed of wetlands in a 1km buffer around the camera location (0-100%)\
“elev_2km”, average elevation within a 2km buffer around the camera location (m)\
“frst_age_2km”, average forest age within a 2km buffer around the camera location (years)\
“canopy_2km”, average canopy cover within a 2km buffer around the camera location (0-100%)\
“pct_suit_2km”, percent of the top 40% suitable habitat that falls within a 2km buffer around the camera location (0-100%)\
“suit_mean_2km”, the average habitat suitability value within a 2km buffer around the camera location (relative unit, “suitability”)\
“pct_water_2km”, the percent of land cover that is composed of water in a 2km buffer around the camera location (0-100%)\
“pct_dev_2km”, the percent of land cover that is composed of developed land in a 2km buffer around the camera location (0-100%)\
“pct_barren_2km”, the percent of land cover that is composed of barren land in a 2km buffer around the camera location (0-100%)\
“pct_frst_2km”, the percent of land cover that is composed of forested land in a 2km buffer around the camera location (0-100%)\
“pct_shrub_2km”, the percent of land cover that is composed of shrublands in a 2km buffer around the camera location (0-100%)\
“pct_herb_2km”, the percent of land cover that is composed of herbaceous land in a 2km buffer around the camera location (0-100%)\
“pct_ag_2km”, the percent of land cover that is composed of agricultural land in a 2km buffer around the camera location (0-100%)\
“pct_wet_2km”, the percent of land cover that is composed of wetlands in a 2km buffer around the camera location (0-100%)\
“elev_4km”, average elevation within a 4km buffer around the camera location (m)\
“frst_age_4km”, average forest age within a 4km buffer around the camera location (years)\
“canopy_4km”, average canopy cover within a 4km buffer around the camera location (0-100%)\
“pct_suit_4km”, percent of the top 40% suitable habitat that falls within a 4km buffer around the camera location (0-100%)\
“suit_mean_4km”, the average habitat suitability value within a 4km buffer around the camera location (relative unit, “suitability”)\
“pct_water_4km”, the percent of land cover that is composed of water in a 4km buffer around the camera location (0-100%)\
“pct_dev_4km”, the percent of land cover that is composed of developed land in a 4km buffer around the camera location (0-100%)\
“pct_barren_4km”, the percent of land cover that is composed of barren land in a 4km buffer around the camera location (0-100%)\
“pct_frst_4km”, the percent of land cover that is composed of forested land in a 4km buffer around the camera location (0-100%)\
“pct_shrub_4km”, the percent of land cover that is composed of shrublands in a 4km buffer around the camera location (0-100%)\
“pct_herb_4km”, the percent of land cover that is composed of herbaceous land in a 4km buffer around the camera location (0-100%)\
“pct_ag_4km”, the percent of land cover that is composed of agricultural land in a 4km buffer around the camera location (0-100%)\
“pct_wet_4km”, the percent of land cover that is composed of wetlands in a 4km buffer around the camera location (0-100%)\
“elev_8km”, average elevation within a 8km buffer around the camera location (m)\
“frst_age_8km”, average forest age within a 8km buffer around the camera location (years)\
“canopy_8km”, average canopy cover within a 8km buffer around the camera location (0-100%)\
“pct_suit_8km”, percent of the top 40% suitable habitat that falls within a 8km buffer around the camera location (0-100%)\
“suit_mean_8km”, the average habitat suitability value within a 8km buffer around the camera location (relative unit, “suitability”)\
“pct_water_8km”, the percent of land cover that is composed of water in a 8km buffer around the camera location (0-100%)\
“pct_dev_8km”, the percent of land cover that is composed of developed land in a 8km buffer around the camera location (0-100%)\
“pct_barren_8km”, the percent of land cover that is composed of barren land in a 8km buffer around the camera location (0-100%)\
“pct_frst_8km”, the percent of land cover that is composed of forested land in a 8km buffer around the camera location (0-100%)\
“pct_shrub_8km”, the percent of land cover that is composed of shrublands in a 8km buffer around the camera location (0-100%)\
“pct_herb_8km”, the percent of land cover that is composed of herbaceous land in a 8km buffer around the camera location (0-100%)\
“pct_ag_8km”, the percent of land cover that is composed of agricultural land in a 8km buffer around the camera location (0-100%)\
“pct_wet_8km”, the percent of land cover that is composed of wetlands in a 8km buffer around the camera location (0-100%)\
“poly_vis_week”, the model-based recreation estimate for that year-week at that camera location (# recreators/year-week)\
“NOMOTOR.n”, the camera-based recreation estimate for that year-week at that camera location (# recreators/year-week)\
“HIKER.avg”, the average daily hikers detected for that year-week at that camera location (# hikers/year-week divided by “days.active”)\
“BIKER.avg”, the average daily mountain bikers detected for that year-week at that camera location (# mountain bikers/year-week divided by “days.active”)\
“MOTOR.avg”, the average daily motorized vehicles detected for that year-week at that camera location (# motorized vehicles/year-week divided by “days.active”)\
“HORSE.avg”, the average daily horseback riders detected for that year-week at that camera location (# horseback riders/year-week divided by “days.active”)\
“DOG.avg”, the average daily domestic dogs detected for that year-week at that camera location (# domestic dogs/year-week divided by “days.active”)\
“NOMOTOR.avg”, the average daily non-motorized recreators detected for that year-week at that camera location (# non-motorized recreators/year-week divided by “days.active”)\
“CECA.avg”, the average daily elk detected for that year-week at that camera location (# elk/year-week divided by “days.active”)\
“ODHE.avg”, the average daily black-tailed deer detected for that year-week at that camera location (# black-tailed deer/year-week divided by “days.active”)\
“PUCO.avg”, the average daily cougars detected for that year-week at that camera location (# cougars/year-week divided by “days.active”)\
“URAM.avg”, the average daily black bears detected for that year-week at that camera location (# black bears/year-week divided by “days.active”)\
“CALA.avg”, the average daily coyotes detected for that year-week at that camera location (# coyotes/year-week divided by “days.active”)\
“LYRU.avg”, the average daily bobcats detected for that year-week at that camera location (# bobcats/year-week divided by “days.active”)\
“poly_vis_day”, the model-based recreation estimate for that year-week at that camera location divided by seven (# recreators/year-week divided by 7)\
“Area”, a categorical designation of the area in which the camera was deployed, based on land manager (categorical variable)
- “year_week”, year and week combination (categorical variable representing weeks as YYYY-WW)\
-
HabUseDataFrom_Prockoetal2024.csv\
Data used in 2.analysis.R which uses the top performing variables from the preliminary analyses to construct random forest regression models contrasting elk detection rates (“CECA.avg”) against the remainder of the variables listed here cumulatively.- “canopy_500m”, average canopy cover within a 500m buffer around the camera location (0-100%) \
“frst_age_250m”, average forest age within a 250m buffer around the camera location (years)\
“pct_herb_500m”, the percent of land cover that is composed of herbaceous land in a 500m buffer around the camera location (0-100%)\
“pct_shrub_250m”, the percent of land cover that is composed of shrublands in a 250m buffer around the camera location (0-100%)\
“pct_suit_8km”, percent of the top 40% suitable habitat that falls within a 8km buffer around the camera location (0-100%)\
“pct_wet_4km”, the percent of land cover that is composed of wetlands in a 4km buffer around the camera location (0-100%)\
“MOTOR.avg”, the average daily motorized vehicles detected for that year-week at that camera location (# motorized vehicles/year-week divided by “days.active”)\
“NOMOTOR.avg”, the average daily non-motorized recreators detected for that year-week at that camera location (# recreators/year-week divided by “days.active”)\
“PUCO.avg”, the average daily cougars detected for that year-week at that camera location (# cougars/year-week divided by “days.active”)\
“URAM.avg”, the average daily black bears detected for that year-week at that camera location (# black bears/year-week divided by “days.active”)\
“ODHE.avg”, the average daily black-tailed deer detected for that year-week at that camera location (# black-tailed deer/year-week divided by “days.active”)\
“CALA.avg”, the average daily coyotes detected for that year-week at that camera location (# coyotes/year-week divided by “days.active”)\
“DistFeat”, the distance from the lens of the camera to the anticipated path of the target–generally the middle of the trail (m)\
“Height”, the height of the camera (m)\
“TrailWidth”, the width of the trail or road on which the camera was placed (m)\
“poly_vis_day”, the model-based recreation estimate for that year-week at that camera location divided by seven (# recreators/year-week divided by 7)\
“Area”, a categorical designation of the area in which the camera was deployed, based on land manager (categorical variable)\
“CECA.avg”, the average daily elk detected for that year-week at that camera location (# elk/year-week divided by “days.active”)
- “canopy_500m”, average canopy cover within a 500m buffer around the camera location (0-100%) \
Diel activity analysis (folder)
-
noct_GAM_data.csv\
Data used in noct_GAM_analysis.R to investigate elk diel activity responses to recreation. Variables are:- “stat_week_hour”, a unique variable representing a specific hour during a given week for a camera station (e.g., all instances of noon for one week) (categorical variable)\
“hour”, hour (time)\
“elk.week”, how many elk were detected in a given week at a given camera (# elk/camera-week)\
“nomotor.week”, how many non-motorized recreators were detected within a given week at a given camera (# recreators/camera-week)\
“stat”, the camera station identifier (categorical variable)\
“stat_week”, a combination of camera station and week (categorical variable)\
“days.active”, how long a given camera was active in a given week (days)\
“stat_year_week”, similar to “stat_week”, but with additional characters representing the year (categorical variable)\
“poly_vis_week”, the number of recreators estimated by the visitation model for a given camera location during a given week (# recreators/year-week)\
“nomotor.avg”, the number of non-motorized recreators detected on a given camera during a given week (“nomotor.week”) divided by the sampling effort of that camera (“days.active”), for a daily average detection rate (# recreators/year-week divided by “days.active”)\
“poly_vis_day”, the number of recreators estimated by the visitation model for a given camera location during a given week divided by seven to represent the daily average recreation rate (# recreators/year-week divided by 7)
- “stat_week_hour”, a unique variable representing a specific hour during a given week for a camera station (e.g., all instances of noon for one week) (categorical variable)\
-
overlap_data.csv\
Data used in overlap_analysis.R to investigate differences in elk diel activity patterns at high- vs. low-human-use camera stations. Variables are:- “File”, the filename of the image (categorical variable, unique identifier)\
“Folder”, the camera station name that corresponds to the image (categorical variable)\
“DateTime”, the date and time that the image was taken (date and time, “YYYY-mm-dd HH:MM:SS”)\
“ImageQuality”, whether the image quality was negatively impacted by lighting (categorical variable)\
“DeleteFlag”, whether the image should be deleted (categorical variable)\
“Hikers”, the number of hikers detected in a given image (# hikers)\
“MtnBikers”, the number of mountain bikers detected in a given image (# mountain bikers)\
“Horseback”, the number of horseback riders detected in a given image (# horseback riders)\
“Vehicles”, the number of motorized vehicles detected in a given image (# vehicles)\
“MuleDeer”, the number of black-tailed (mule) deer detected in a given image (# black-tailed deer)\
“Coyote”, the number of coyotes detected in a given image (# coyotes)\
“BlackBear”, the number of black bears detected in a given image (# black bears)\
“Bobcat”, the number of bobcats detected in a given image (# bobcats)\
“Rabbit”, the number of rabbits detected in a given image (# rabbits)\
“Rodent”, the number of rodents detected in a given image (# rodents)\
“Bird”, the number of birds detected in a given image (# birds)\
“Cougars”, the number of cougars detected in a given image (# cougars)\
“Raccoon”, the number of raccoons detected in a given image (# raccoons)\
“Skunk”, the number of skunks detected in a given image (# skunks)\
“OtherSpp”, the number of other species not yet listed detected in a given image (# other species not yet listed)\
“Dogs”, the number of domestic dogs detected in a given image (# domestic dogs)\
“Elk”, the number of elk detected in a given image (# elk)\
“detDATE”, a reformatted date-time variable (date-time in POSIXct format from program R, “YYYY-mm-dd HH:MM:SS”)\
“latin_name”, a categorical variable indicating the latin name of the species in the image (categorical variable)\
“count”, how many of a given species are in a given image (# individuals per image)\
“duration”, time between the image in question and the last image taken by a given camera, used to filter down to independent detection events (seconds)
- “File”, the filename of the image (categorical variable, unique identifier)\
-
MASTER_DEPLOYMENTS.csv\
Spreadsheet detailing information about the dates and times that each camera location was active, as well as the locations of each camera and any pertinent information related to the camera deployment (e.g., camera height). In some instances “NA” values are present, indicating this data was not collected. For instance, the column “Height” shows at least two instances of “NA”, which were due to a failure to collect data on the height of the camera deployment. These data were omitted from analyses.- “Station”, the unique name of the camera station (categorical variable)\
“Status”, the status of the camera (categorical variable; Retrieved, Inactive, Stolen)\
“Lat”, latitude of the camera (decimal degrees)\
“Long”, Longitude of the camera (decimal degrees)\
“StartDate”, when the camera was initially deployed (date-time, “YYYY-mm-dd HH:MM:SS”)\
“EndDate”, the final date that the camera was active (date-time, “YYYY-mm-dd HH:MM:SS”)\
“Height”, the height of the camera (m)\
“TrailWidth”, the width of the trail on which the camera was deployed (m)\
“DistFeat”, the distance from the camera lens to the anticipated path of the target, typically the center of the trail/road (m)\
“Orientation”, the cardinal direction the camera lens faced (categorical variable, e.g., “N”, “NW”, “W”…)\
“SDID”, the identification number of the SD card associated with the camera (categorical variable)\
“KeyID”, the key number for the lock on the camera (categorical variable)\
“Notes”, any pertinent information about the camera deployment (categorical variable)\
“Area”, a categorical variable representing the management area that the camera was deployed in (categorical variable)
- “Station”, the unique name of the camera station (categorical variable)\
Code/Software
All code was written using R version 4.2.2. All packages required are specified in the beginning lines of each code. Codes associated with the habitat use analysis should be run sequentially from 1 to 3. Codes associated with the diel activity analysis are independent of one another and can be run in any order.
Habitat use analysis (folder)
- 1.pre-analysis.R\
Code for running the preliminary analysis detailed in the ‘Habitat use modeling’ section of Procko et al. 2024. This includes preliminary univariate random forest regression models that inform a variable selection process for the final models constructed in the analysis script (2.analysis.R). - 2.analysis.R\
Code for running the analysis detailed in the ‘Habitat use modeling’ section of Procko et al. 2024. This includes random forest regression models constructed with a suite of variables (informed by the preliminary analysis), calculating variable importance metrics and statistical information associated with these models, and visualizing model outputs. - 3.reviewer_requests.R\
Code for running two additional random forest regression models recommended by manuscript reviewers–one excluding the model-based recreation estimate variable (“poly_vis_day” below) and one excluding the camera-based recreation estimate variable (“NOMOTOR.avg” below).
Diel activity analysis (folder)
- noct_GAM_analysis.R\
Code for constructing generalized additive models that model elk diel activity as circular data against recreation variables (anticipated to influence elk nocturnality) by using cyclic regression splines. - overlap_analysis.R\
Code for running an overlap analysis contrasting elk diel activities at camera stations with higher-than-average human activity vs. camera stations with lower-than-average human activity.
Inquiries
Inquiries regarding this data can be sent to lead author, Michael Procko, at xprockox@gmail.com
We used an elk habitat suitability model hereafter referred to as “the westside model” (Rowland et al. 2018) to characterize habitat suitability throughout the full North Rainier Elk Herd range in western Washington, identifying areas of high quality habitat for camera trap deployments. We defined high quality habitat as areas that the westside model ranked as the 40% most suitable elk habitat in the herd range (Fig. 1). The goal of this site selection criterion was to better isolate the impacts of recreation on elk detections while reducing the confounding effects of habitat quality. We then clipped these areas to land parcels managed by Washington Department of Natural Resources (WA DNR) and the United States Forest Service (USFS), as these two agencies provided permission for us to deploy cameras on their managed lands. The WA DNR and USFS properties encompassed by the North Rainier Elk Herd range span a gradient of recreational use intensity, including areas that consistently see over a thousand visitors per week in the summer and areas that are closed to the public via a 24-hour guarded security gate. We randomly selected 80 camera locations along trails or restricted access roads within these areas, with a minimum of 1 km spacing between randomized points (Fig. 1). Beginning on June 2, 2022, we navigated to each random point and deployed a Reconyx Hyperfire (HC600/PC900) camera (Reconyx, Holmen, WI) approximately 1 m from the ground. We set cameras at a 90˚ angle to the anticipated direction of travel (i.e., perpendicular to the trail) when the trail was restricted to hikers or horseback riders only, and at a 45˚ angle to the direction of travel when the trail or road allowed mountain bikers and/or vehicles. These differences in camera set orientations were undertaken to account for difficulty in detecting faster moving targets such as mountain bikes and vehicles (Miller, Leung, and Kays 2017). We also recorded the camera height (m), trail or road width (m), and the distance (m) from the camera lens to the anticipated path of the target (i.e., the nearest edge of the trail or road) for use as covariates in subsequent modeling. We set cameras to take one photo per trigger, with no delay period, on high trigger sensitivity. We performed final camera take-downs on September 18, 2022, with most cameras being deployed for approximately 6-8 weeks.
We pre-processed camera images using the artificial intelligence software MegaDetector (Beery, Morris, and Yang 2019), binning photos into “human”, “animal”, “vehicle”, or “blank” (i.e., false trigger) categories. We then blurred all human images using a publicly available photo-blurring software (WildCo Lab 2021) to protect the privacy of people detected on our cameras (Sharma et al. 2020). We used the automated tags generated by MegaDetector to sort images for rapid batch identification (Fennell, Beirne, and Burton 2022), and we manually tagged all images using the open source software Timelapse (Greenberg 2021). We classified photos of humans into three activity categories: hiking, mountain biking, and horseback riding, and we tagged all animal detections by species, including domestic dogs and cats. We then corrected any erroneous identifications provided during the automated detection process, and exported all detection data for statistical analysis in R (v. 4.2.2; R Core Team 2022).
From our raw detection data, we used an independence threshold of one minute to determine independent detection “events” for all human-related activities, and a threshold of 30 minutes for all wildlife detection events. A 30-minute independence threshold is common in most wildlife studies (Burton et al. 2015). We used the one minute threshold for human detections because numerous independent recreationists are likely to use high-traffic trails within the larger 30-minute window (Procko et al. 2022; 2023). This method may slightly under-estimate visitors, as multiple individuals are recorded as a single detection event if they all pass the camera within one minute of each other. To remedy this, recent works have recommended utilizing raw image counts instead of a time-to-independence method (Peral, Landman, and Kerley 2022), but this method would over-estimate visitation given most people are photographed multiple times when passing in front of a camera, and preliminary modeling showed worse model performance when we removed independence thresholds. We used these independent detection events to calculate detection rates (number of detection events per unit of sampling effort) of all human-related activities and all wildlife species, and we used these detection rates in subsequent modeling either as our response variables (elk detection rates) or predictor variables (detection rates of other species including humans participating in various activities). We did not include any wildlife species which were detected in fewer than 30 unique events.