Skip to main content
Dryad

Patterns of island fox habitat use in sand dune habitat on San Clemente Island

Cite this dataset

Gamblin, Holly E. L. et al. (2024). Patterns of island fox habitat use in sand dune habitat on San Clemente Island [Dataset]. Dryad. https://doi.org/10.5061/dryad.t1g1jwt9w

Abstract

On San Clemente Island (SCI), the island fox subspecies (Urocyon littoralis clementae) has been monitored annually since 1988 to track long-term population trends. Annual density estimates in most habitat types across the island range from 2–13 foxes/km2, yet unusually high estimates have repeatedly approached 50 foxes/km2 in a unique sand dune habitat area. Although sand dune habitat is restricted to one small area on the island, these estimates suggest sand dune habitat supports one of the highest population densities of any fox species in the world, and it may support > 5% of the SCI fox population. This finding prompted our investigation to determine if SCI foxes captured in the sand dunes habitat area maintained home ranges within this habitat type. Between January–July 2018, we used Global Positioning System collars to track the movements of 12 island foxes captured in the sand dune habitat area. Contrary to our initial predictions, we found that island foxes captured in the sand dune habitat area do maintain home ranges and core areas centralized in sand dune habitat. All 12 island fox home ranges estimated contained >50% sand dune habitat in either their 50% or 95% fixed kernel density estimate (KDE) home range, and island foxes were 3.14 times more likely to use active sand dune habitat when compared to the second most abundant habitat type, maritime desert scrub (Adjusted  = 3.14, 95% CI = 3.07–3.12). We also found that island foxes in sand dune habitat maintained much smaller home ranges than reported estimates in other habitat types, with an average 95% KDE home range size of 0.42 km2 (95% CI = 0.20–0.63 km2). Although sand dune habitat comprises just 2% of available habitat on SCI, our research highlights the importance of this unique habitat area for island foxes.

README: Patterns of island fox habitat use in sand dune habitat on San Clemente Island

https://doi.org/10.5061/dryad.t1g1jwt9w

We have submitted our raw data and R scripts associated with all home range and resource selection function analyses. The "DunesHRAnalysis" R file includes code for estimating Minimum Convex Polygons and Kernel Density Estimates for the 12 island foxes monitored in this study. It requires the csv file "GPSPoints_ALL_csv," which includes all GPS data points for the foxes monitored. The "RSFAnalysis" R file includes code for estimating the resource selection functions reported. It requires the csv files "all_used" and "all_rand_kde," which includes the vegetation data associated with each used and randomly generated GPS point. 

Description of the data and file structure

all_used
  • FoxID: unique identifier for each fox
  • East_X: easting for each used data point (Projected Coordinate System: NAD 1983)
  • North_Y: northing for each used data point (Projected Coordinate System: NAD 1983)
  • Alliance: alliance of associated vegetation
  • Group: group of associated vegetation
  • Macrogroup: macrogroup of associated vegetation 
all_rand_kde
  • FoxID: unique identifier for each fox
  • Alliance: alliance of associated vegetation
  • Group: group of associated vegetation
  • Macrogroup: macrogroup of associated vegetation 
GPSPoints_ALL_csv
  • Collar: unique identifier for each fox
  • East_X: easting for each used data point (Projected Coordinate System: NAD 1983)
  • North_Y: northing for each used data point (Projected Coordinate System: NAD 1983)

Sharing/Access information

Vegetation data was derived from the following sources:

Code/Software

R is required to run both scripts provided. Annotations are provided throughout the scripts through 1) library loading, 2) dataset loading and cleaning, and 3) analyses. 

Methods

We deployed GPS collars (Advanced Telemetry Systems Model W500) to track the movements of 12 captured island foxes. Each collar weighed 65 g (< 5% of the body weight of a 1.5 kg fox) and was equipped with a VHF transmitter. Collars were programmed to record 1 GPS fix location per hour between 0600–1800 each day for 188 days between January–July 2018. Data were remotely downloaded from all foxes every 7 days using a W100 Com module (Advanced Telemetry Systems) and Yagi antenna from distances up to 400 m.

Home range estimation

We calculated home range sizes from the GPS locations using the ‘adehabitatHR’ package in RStudio (RStudio Team 2022) and ArcMap 10.3 (ESRI 2015). Utilization distributions were generated using the ad hoc method for the estimation of the smoothing parameter (Worton 1989) within ‘adehabitatHR.’ We calculated spatial use metrics including the 95% minimum convex polygon (MCP) home ranges after the removal of 5% of extreme points, 95% fixed kernel density estimate (KDE) home ranges, and 50% KDE core areas. We compared these spatial parameters among sexes using a two-sample t-test and determined where primary core areas existed on the landscape. MCP and traditional KDE methods were selected to maintain consistency with Sanchez and Hudgens's (2015) previous home range estimates for SCI foxes. We conducted exploratory analyses using autocorrelated KDE (AKDE) to determine if the fine-scale location data in our study would be impacted by autocorrelation (Calabrese et al. 2016). Traditional KDE and AKDE performed nearly identically, so we proceeded with traditional KDE to be more directly comparable to Sanchez and Hudgens (2015).

We defined a home range or a core area to be within the sand dune habitat if either the 50% or 95% KDE was composed of > 50% sand dune habitat. We evaluated the dominant habitat types within each home range by calculating the percentage of each habitat type within the home range metrics. We classified habitat type based on the vegetation communities present in the study area: sand dune habitat (active and stabilized sand dunes), maritime desert scrub, grassland, coastal unvegetated, unvegetated, California chaparral, and developed (Tierra Data Systems 2010, Uyeda et al. 2020a, Uyeda et al. 2020b).

Habitat selection analysis

We conducted a resource selection function analysis to assess third order (within the home range, Johnson 1980) habitat selection under a Design 3 study design for comparing island fox habitat use and availability (Thomas and Taylor 1990). Under a Design 3 framework, individual home ranges are identified and corresponding used and available points are measured within the individuals’ home range. Following methods described in Fieberg et al. (2021), we assessed availability by generating 10 random (i.e., available) points within the home range of each collared island fox for every 1 used point using ArcMap 10.3 (ESRI 2015). Habitat type for used and available points was assigned using a fine-scale SCI Vegetation Map with a mapping unit of 0.25 m2 (Uyeda et al. 2020a, Uyeda et al. 2020b) in ArcMap 10.3 (ESRI 2015). We used the vegetation alliances and macrogroups described in Uyeda et al. (2020a, 2020b) to reclassify habitat type based on the vegetation communities present in the study area: active sand dune habitat, stabilized sand dune habitat, maritime desert scrub, grassland, coastal unvegetated, unvegetated, California chaparral, and developed (Uyeda et al. 2020a, Uyeda et al. 2020b). Sand dune habitat was classified as either active or stabilized sand dune habitat to determine if selection differed between the two types of sand dune habitat.

We compared habitat type between used and available points using a generalized linear mixed model (GLMM) with a logit link using the ‘glmer’ function in the ‘lme4’ package in RStudio (Bates et al. 2015, RStudio Team 2022). Habitat type was a categorical variable with maritime desert scrub set as the reference level. Fox ID was included as a random effect in the model to account for non-independence of individual foxes (Fieberg et al. 2021, Johnson-Bice et al. 2023). Available points were assigned a larger weight of 1000 and used points were assigned a weight of 1 to ensure the results hold more generally (Fieberg et al. 2021, Johnson-Bice et al. 2023).

The model coefficients associated with habitat type provide a comparison of use between each categorical habitat type (i.e., active sand dune, stabilized sand dune, grassland, etc.) and the reference level maritime desert scrub, with the assumption that each habitat type is equally available within the home range. However, since each habitat type is not equally available within the home ranges, an adjusted ratio of habitat use (Adjusted ) was calculated following methods described in Fieberg et al. (2021) by multiplying the habitat type coefficient estimate,  (calculated assuming equal availability) by the ratio of overall habitat type availability relative to maritime desert scrub availability within each home range:

 

We also compared the distribution of raw used and available points for the dominant habitat types (i.e., the habitat classifications for > 95% of raw used and available data points) to assess disproportionate use of the habitat types, if any.

Funding

U.S. Navy Pacific Fleet, Award: NA, Naval Base Coronado Environmental Division