Data from: Cross or bounce? Locational and video data reveal high road encounter and altered movement in a common ungulate
Data files
Sep 23, 2026 version files 286.83 MB
-
GPS_cleaned_analysisset_125animals.csv
257.52 MB
-
README.md
6.30 KB
-
road_encounter_covariates.csv
29.30 MB
-
video_behavior_counts.csv
1.01 KB
Abstract
Roads can constrain wildlife movement and pose collision risk, but quantifying fine-scale behavioral responses to individual road encounters remains difficult, particularly for species with large home ranges. This dataset contains GPS telemetry and associated covariate data for 125 white-tailed deer (Odocoileus virginianus) collared across three sites in southern Illinois, USA (Lake Shelbyville, Touch of Nature Outdoor Education Center, and Southern Illinois University Carbondale), 2023–2025, used to classify road-encounter behavior via Barrier Behavior Analysis (BaBA). The dataset includes cleaned GPS fix data for all 125 collared individuals, individual-level capture and collar metadata, road-encounter events classified by behavioral type (quick cross or bounce) with associated road category, road density, season, and diel period, and land cover proportions within a 100 m buffer of each encounter in both raw and scaled and centered form as used in mixed-effects logistic regression modeling of encounter type. Duplicate fixes, near-capture fixes, and implausibly fast steps were removed from the GPS data, with an additional fix-quality filter applied to the classification-ready track. Road category and land cover values are derived from the publicly available USGS National Transportation Dataset and National Land Cover Database. These data may be reused to examine ungulate road-crossing behavior, habitat selection near linear infrastructure, or as a comparison dataset for other road ecology studies. Animal capture and handling were conducted under Southern Illinois University IACUC protocols 21-028 and 21-002.
Dataset DOI: 10.5061/dryad.51c59zwr4
Description of the data and file structure
We present data from 125 white-tailed deer (Odocoileus virginianus), GPS-collared across three sites in southern Illinois, USA (Lake Shelbyville, Touch of Nature Outdoor Education Center, and Southern Illinois University Carbondale) from 2023–2025, to study how deer respond behaviorally to road encounters. A subset of collars also recorded video, which we used to compare deer behavior near roads to behavior away from roads. We applied Barrier Behavior Analysis (BaBA) to the GPS tracks to classify each road encounter by behavioral type (e.g., quick cross, bounce), then used mixed-effects logistic regression to test how road category, land cover, season, and time of day predicted crossing outcome, and chi-square tests to compare video-recorded behavior across encounter types.
Files and variables
File: GPS_cleaned_analysisset_125animals.csv
Description: Cleaned GPS fixes for the 125 collared deer analyzed in the study — duplicates, near-capture fixes, and implausibly fast steps removed, plus a DOP ≤ 10 filter. The model-ready track: input to the BaBA road-encounter classification and the buffer-size sensitivity check.
Variables
animal_id: Unique animal identifier (site prefix + number, e.g. ILSV1196)
UTME: Fix easting, UTM Zone 16N (EPSG:32616)
UTMN: Fix northing, UTM Zone 16N (EPSG:32616)
t_: Fix timestamp, GMT
dop: Dilution of precision (fix quality); missing values substituted with 25 upstream, not itself a filter threshold in this file
age: Individual age class at capture
sex: Individual sex at capture
collar: Collar device ID
collar_deployment_date: Collar deployment start date
collar_end_date: Collar deployment end date
Study_area: Site (Shelbyville, SIUC, or Touch of Nature). SIUC = Southern Illinois University Carbondale
Longitude: Fix location, decimal degrees (WGS84)
Latitude: Fix location, decimal degrees (WGS84)
AnimalID: D-format animal ID (e.g. D1201); the join key to road_encounter_covariates.csv's AnimalID column
File: road_encounter_covariates.csv
Description: BaBA road-encounter events (quick cross or bounce only) with road category, land cover, road density, season, and diel period covariates used in model fitting. A merge of two files from the original analysis (OSM-based road category assignments and a scaled/centered NLCD land-cover extraction), joined on burstID.
Variables
burstID: Unique encounter/event ID ({AnimalID}_{index})
eventTYPE: BaBA classification: Quick_Cross or Bounce
road_category: OSM-based road class — L (large: primary, secondary, motorway), M (medium: tertiary, unclassified, residential), S (small: service, path, track)
season: prefawn / fawn / postfawn / rut
AnimalID: D-format animal ID (e.g. D1201); matches GPS_cleaned_analysisset_125animals.csv's AnimalID column
rdkm_sqkm: Road density (km road / km² home range)
Year: Encounter year
Collar_start: Collar deployment start date
Fate_date: Collar deployment end date
Study_area: Site
Sex: Individual sex
Age: Individual age class
Collar.ID: Collar device ID
X11_openwater: Raw proportion of a 100m buffer around the encounter classified as open water
X41_forest: Raw proportion of a 100m buffer classified as forest
X21_developedopen: Raw proportion of a 100m buffer classified as developed, open space
X22_developedlowmedhigh: Raw proportion of a 100m buffer classified as developed, low/medium/high intensity
X52_scrub: Raw proportion of a 100m buffer classified as scrub/shrub
X71_grassland: Raw proportion of a 100m buffer classified as grassland/herbaceous
X81_cropsandpasture: Raw proportion of a 100m buffer classified as crops and pasture
X90_wetlands: Raw proportion of a 100m buffer classified as wetlands
prop_roads: Raw proportion of a 100m buffer classified as road/impervious transportation surface
type: Binary outcome variable — 1 = Quick_Cross, 0 = Bounce
time.cat: Diel period — daytime, nighttime, or twilight
std_rdden: Scaled and centered road density (linear model)
stdln_rdden: Scaled and centered log-transformed road density (log model)
std_forest: Scaled and centered proportion forest (linear model)
std_agri: Scaled and centered proportion agriculture (linear model)
std_dev: Scaled and centered proportion developed (linear model)
std_lnforest: Scaled and centered log-transformed proportion forest (log model)
std_lnagri: Scaled and centered log-transformed proportion agriculture (log model)
std_lndev: Scaled and centered log-transformed proportion developed (log model)
File: video_behavior_counts.csv
Description: Counts of scored behaviors from camera collar video, summarized by BaBA encounter type. Feeds the video-behavior figure in the manuscript.
Variables
type: BaBA encounter type the video segment corresponds to (e.g., Average_movement)
behav: Scored behavior (e.g. Forage, Interaction, Resting, Scrape)
freq: Count of video segments scored with this behavior, within this encounter type
rel.freq: freq as a proportion of all scored segments within this encounter type
Code/software
Analysis code is on GitHub (https://github.com/anordseth/WTD_BaBA) and requires R to run.
Access information
Other publicly accessible locations of the data:
- None
Data was derived from the following sources:
-
Road network shapefiles (Trans_Road_Nroadsutm/Trans_Road_Sroadsutm), merged and reprojected to UTM Zone 16N from the USGS National Transportation Dataset: https://www.usgs.gov/programs/national-geospatial-program/national-transportation-dataset-ntd.
road_category in road_encounter_covariates.csv, classified from OpenStreetMap (2024): https://www.openstreetmap.org/copyright.
Land cover proportions in road_encounter_covariates.csv, derived from the USGS National Land Cover Database (Dewitz, 2024): https://www.mrlc.gov/data.