Data from: Heading west: Ecology of swift foxes in a novel landscape beyond their range
Data files
Jun 10, 2025 version files 8 MB
-
ActivityPatterns.RDS
23.01 KB
-
Cams_HSM_values.RDS
1.98 KB
-
GridCentroidWithheld_data.RDS
854 B
-
IndependentWithheld_data.RDS
349 B
-
PseudoAbsence_Covariates.RDS
3.87 MB
-
RandomForest_TrainingData.RDS
4.07 MB
-
README.md
3.38 KB
-
Smithetal_VuVe_HSM_Spatiotemporal_Analyses.R
26.64 KB
Abstract
The swift fox (Vulpes velox) is generally associated with the short-grass prairie ecosystem of the North American Great Plains; a system that has declined by approximately 50% over the last century. Yet, swift fox populations seem to demonstrate regional variation in trends, with some populations declining while others appear stable to increasing. In Wyoming, USA, swift foxes have been observed and successfully reproducing over 100 km west of their historical range boundary in shrubland-dominated habitats, previously deemed unsuitable. This has created challenges in conservation, as it is unclear what factors are uniquely important to a species occupying habitats previously considered unsuitable. Therefore, we sought to investigate fundamental questions regarding swift fox ecology in this novel habitat: (1) what environmental gradients are contributing to suitable habitat?, (2) how is swift fox habitat suitability associated with competitors (American badgers [Taxidea taxus] and coyotes [Canis latrans]) in space?, and (3) how do swift foxes and competitors allocate activity in time? Between 2021 and 2023, seven swift foxes were GPS-collared, and 71 camera traps were deployed in central Wyoming. GPS-collar locations paired with environmental covariates were used to develop a habitat suitability model using Random Forest, predicting 30.28% of our study area was suitable. Important habitat resources identified were herbaceous biomass, shrub height, and sand content. We found swift foxes likely have minimal spatial and temporal separation between competitors, specifically coyotes. We hypothesize that swift foxes may successfully coexist with competitors either by being less risk-averse for the sake of resource acquisition or by relying on abundant escape cover, which reduces the need to modulate spatial or temporal activity. Our findings enhance our understanding of environmental factors contributing to suitable habitats and how swift foxes coexist in time and space with competitors in a novel shrubland environment.
The R code is divided into five sections: The first section employs Random Forest modeling to assess the environmental gradients that characterize swift fox habitat in unconventional shrublands and predict suitable habitat across the study area. The second section validates the predicted habitat suitability map using reserved GPS centroids and independent data. The third section focuses on understanding trends in swift fox habitat suitability and its covariates. The final two sections explore the association between swift fox habitat suitability and competitors (i.e., coyotes and American badgers) in space, as well as how swift foxes and their competitors allocate activity over time.
Dataset DOI: 10.5061/dryad.r2280gbq0
Description of the data and file structure
The data were collected from GPS collars deployed on swift foxes and camera stations throughout the BLM Lander field office, Wyoming, USA, as well as various remotely sensed products and airborne LiDAR. Due to the status of swift foxes in Wyoming, as classified by the Wyoming Game and Fish Department as a species of greatest conservation need (NSS4 (Cb), Tier II) and by the Wyoming Bureau of Land Management as sensitive, we excluded the GPS and camera station coordinates from the datasets (https://wyndd.org/species_list/).
The R script (VuVe_HSM_Spatiotemporal_Analyses.R) and accompanying RDS files contain all the material necessary to replicate our analyses. Below, we have provided additional context to RDS files, such as column names, column descriptions, etc.
Random Forest modeling and predictions
RandomForest_TrainingData.RDS
Column names and information:
- Id = 1 = presence; 0 = pseudo-absence
- bd_pol = soil bulk density (g cm–3)
- chm = vegetation height (m)
- clay_pol = clay content (% (kg kg–1))
- eaglenests = likelihood of golden eagle nests (index)
- herb = herbaceous biomass (kg ha-1)
- hli = heat load index (index)
- msavi = modified soil adjusted vegetation index (index)
- roads = distance from roads (m)
- sand_pol = sand content (% (kg kg–1))
- slope = slope (degrees)
Validation: Reserved GPS centroids and independent (withheld) data
Independentwithheld_data.RDS
Column information:
- ID = 60 independent swift fox observations across the BLM Lander field office
- Habitat_Values = predicted swift fox habitat suitability values (binned 1 to 10) intersected by each ID
GridCentroidwithheld_data.RDS
Column information:
- ID = 341 reserved grid centroids with two or more swift fox GPS locations
- Habitat_Values = predicted swift fox habitat suitability values (binned 1 to 10) intersected by each ID
Understanding trends in swift fox habitat suitability and covariates
PseudoAbsence_Covariates.RDS
Column information:
- Column names are the same as the RandomForest_TrainingData.RDS, except now Habitat_Values is included.
- Habitat_Values was created by intersecting all the pseudo-absence locations with the predicted swift fox habitat suitability values (binned 1 to 10).
Species occurrence models
Cams_HSM_values.RDS
Column information:
- LocationName = the name/location of camera stations
- Species = Swift fox, Coyote, and American badger
- n = 1 = a camera station with one or more detections of that species. 0 = a camera station without a detection of that species
- Habitat_Values was created by intersecting all the camera station locations with the predicted swift fox habitat suitability values (binned 1 to 10).
Activity Patterns
ActivityPatterns.RDS
Column information:
- Species = Swift fox, Coyote, and American badger
- Location = 1; all activity patterns were calculated in one study area (Lander field office)
- Time = time is in radians
- Date = timestamp of species detection at the camera station
- radsun = sun time = converted clock time (radians) to sun time
