Data for: Behavioral responses of migratory caribou to semi-permeable roads in Arctic Alaska
Data files
Jul 30, 2025 version files 198.41 KB
-
README.md
7.83 KB
-
WAH_BaBA_Data_2025-05-09.csv
190.58 KB
Apr 20, 2026 version files 288.57 KB
-
BaBA_Analysis_Examples_2026-04-02.R
2.74 KB
-
README.md
10.35 KB
-
WAH_BaBA_Data_2026-04-02.csv
275.48 KB
Abstract
Migration conveys many benefits to species, ecosystems, and people, but relies upon connected landscapes. Anthropogenic development can present barriers for migrants, but many barriers are semi-permeable, allowing unhindered or delayed passage. We use a modified version of the Barrier Behavior Analysis (BaBA) to investigate movement responses and permeability of roads in northwestern Alaska by adult female Western Arctic Herd caribou (Rangifer tarandus) from 2009–2024. These data include the BaBA results that we used to reveal behavioral responses to roads by study caribou. Our analyses revealed movement responses to all five focal roads. We found that the roads were semi-permeable barriers to movement, with altered behaviors including bouncing away, moving back-and-forth, and tracing along roads. Overall, 63.1% of collared animals encountered (entered a road-specific buffer) at least one focal road. Of these, 61.5% displayed altered movements. At the scale of individual encounters with roads, we found altered movement in 27.1% of road encounters. Each row of the dataset represents one burst, indicating subdivisions of road encounters that reflect specific behavioral responses. Encounter timing is also indicated for each record. Most encounters occurred during fall migration, and caribou with altered behavior spent longer near focal roads than those with unaltered movement. We confirm prior findings of altered fall movements near the Red Dog road and demonstrate that movement behavior is also altered around other roads and in other seasons. Nonetheless, many collared caribou did not display altered movements in response to roads, emphasizing the need for further research to understand the mechanistic drivers of caribou movement responses. Given increasing pressures for infrastructure development and global challenges facing migratory species, it is critical to identify mitigation measures and inform management decisions seeking to balance responsible development with conservation of natural systems, including migratory species and the people that rely upon them.
Dataset DOI: 10.5061/dryad.8sf7m0d1n
Description of the data and file structure
We investigated movement responses to roads for caribou (Rangifer tarandus) in northwestern Alaska. We ran GPS telemetry data for 366 adult female caribou through a modified version of the Barrier Behavior Analysis (BaBA) tool, resulting in this dataset. The BaBA compares movement parameters near a potential barrier with those away from barriers to identify altered movement behavior. We heavily modified the original BaBA methodology for this analysis. The resulting dataset indicates interactions with roads at the encounter and altered burst scales, including information about timing and behavioral responses.
Files and variables
File: WAH_BaBA_Data_2026-04-02.csv
Description: Behavioral responses of adult female caribou of the Western Arctic Herd to roads in northwestern Alaska as indicated by our modified Barrier Behavior Analysis (BaBA). Each row of the dataset represents one burst (see definition below and in the manuscript) analyzed in this study. For details of how these behaviors were identified from caribou GPS telemetry data, see the Methods of this study and details in Supplementary Information 1.
Variables
animalID: Unique alphanumeric identifier for each caribou analyzed in the dataset. SomeanimalIDs are completely numeric while others also include letters.encounterID: Unique identifier for each encounter of a caribou with a road, encompassing the time from which a caribou entered a focal road buffer until it left that buffer. Encounters take theanimalIDfollowed by a numeric indicator to distinguish the potential for multiple encounters per caribou.burstID: Unique identifier of each burst analyzed by the BaBA. Bursts represent subdivisions of an encounter, identified by a focal road being crossed or, in the event of overlapping road buffers (Red Dog – Kivalina only), the animal moving from being closer to one focal road to another while remaining within the buffers. Bursts were used to identify multiple behavioral responses to a single focal road or to multiple nearby roads with overlapping buffers. TheburstIDmatches theencounterID, with a possible additional numeric indicator if there were multiple bursts for a given encounter.enc_year: The year in which a given encounter began.enc_start: The start date and time for a given encounter. This will match theburst_startof the first burst in a multi-burst encounter. Times are reported here in Universal Coordinated Time (UTC) but were converted to the local Alaskan time zone (tz = "America/Anchorage") using the lubridate package in R for assignment of seasons and for the other analyses reported in our study. All telemetry location times were rounded to the nearest hour, as this was their target capture time.enc_end: The end date and time for a given encounter. This will match theburst_endof the last burst in a multi-burst encounter. Times are reported here in Universal Coordinated Time (UTC) but were converted to the local Alaskan time zone (tz = "America/Anchorage") using the lubridate package in R for assignment of seasons and for the other analyses reported in our study. All telemetry location times were rounded to the nearest hour, as this was their target capture time.enc_season: The season of a given encounter. This will match theburst_startof the first burst in a multi-burst encounter.enc_duration: The duration of the encounter, in hours, indicating the time elapsed betweenenc_startandenc_end. Note that for multi-burst encounters this will differ from the sum ofburst_durationdue to the overlap between bursts as a result of adding a location onto either end of a given burst to enable step-based analyses.enc_road: Indication of the road with which the caribou interacted at the encounter scale. We combined the Red Dog and Kivalina road responses at the encounter scale as their overlapping buffers often led to joint reactions in proximity to both roads within a single encounter. Thus, possible values ofenc_roadinclude: Dalton, Kobuk, Nome, and Red Dog - Kivalina. Please see the Methods section of the manuscript for descriptions of each road. Note that for interactions with the Red Dog - Kivalina road this may differ fromburst_roaddue to the burst-scale analysis distinguishing between each road.enc_behavior: Behavioral response of the caribou to road(s) at the encounter scale. Behavior was reduced to a binary of unaltered or altered movement.- Unaltered movement - Movements indistinguishable from season-specific movement in the absence of focal roads or for which crossing did not appear hindered. These consisted of normal movement and quick cross behaviors (see below for definitions).
- Altered movement - Periods where movement behavior changed near a focal road compared to seasonal average movement parameters outside of road buffers. These consisted of back-and-forth, bounce, and trace behaviors (see below for definitions). If any of the bursts within the encounter featured altered movement, then the encounter was classified as altered.
burst_start: The start date and time for a given burst within an encounter. Times are reported here in Universal Coordinated Time (UTC) but were converted to the local Alaskan time zone (tz = "America/Anchorage") using the lubridate package in R for assignment of seasons and for the other analyses reported in our study. All telemetry location times were rounded to the nearest hour, as this was their target capture time.burst_end: The end date and time for a given burst within an encounter. Times are reported here in Universal Coordinated Time (UTC) but were converted to the local Alaskan time zone (tz = "America/Anchorage") using the lubridate package in R for assignment of seasons and for the other analyses reported in our study. All telemetry location times were rounded to the nearest hour, as this was their target capture time.burst_season: The season of the year in which the burst took place. Seasons reflect biologically relevant periods of similar caribou behavior, following Joly and Cameron 2024. Seasons included: spring migration, calving, insect harassment, late summer, fall migration, and winter. Bursts were assigned to the season which occupied the majority of the burst locations. Note that the specificburst_seasonmay differ from the overallenc_season.burst_duration: The duration of the burst, in hours, indicating the time betweenburst_startandburst_end.burst_road: Indication of the road with which the caribou interacted in the given burst. Possible roads included: Dalton, Kivalina, Kobuk, Nome, and Red Dog. Please see the Methods section of the study for descriptions of each road. For Dalton, Kobuk, and Nome values ofburst_roadwill matchenc_road. For Kivalina and Red Dog values will differ fromenc_roadas the specific road being interacted with in a given burst is able to be identified.burst_behavior: Behavioral response of the caribou to a road at the burst scale. Each burst was assigned a movement behavior by the BaBA (details in Methods and Supplementary Information 1). Possible values included:- Normal movement - Movement in which movement parameters fell within seasonal average parameters or all locations in the burst were closest to the same point on the road (identifying situations where a caribou briefly dipped inside a buffer and then exited again rapidly).
- Quick cross - Rapid, linear movement across one or more focal roads, without apparent hindrance by the road.
- Back-and-forth - Clustered movement near a focal road, with repeated changes in movement direction leading to relatively confined space use.
- Bounce - Movement in which a caribou approached a focal road and then bounced back away, typically without crossing.
- Trace - Movement in which a caribou paralleled a road for a sustained amount of time.
burst_crossing: Binary indication of whether the caribou crossed theburst_roadduring the indicated burst (1) or not (0). This corresponds to thecross_trueparameter in Supplementary Fig. S1 as it distinguished between presumed true and false crossings by testing for locations on either side of a road. If all locations were on the same side of a road, we assumed this indicated a false crossing (i.e., when the straight line between subsequent locations might falsely indicate that a crossing occurred). Thus, crossing only takes a value of1when a crossing is indicated and locations occur on both sides of a road.burst_closest_dist_km: Numeric indicator of the distance between the closest recorded telemetry location andburst_road, in units of kilometers. Note that this is the closest observed distance but may not reflect the closest the caribou actually came in between recorded locations. For example, any caribou that crossed the road would have an actual closest distance of 0 km. This was calculated by taking the minimum distance between the telemetry locations for a given burst and the line locations ofburst_road, using the st_distance() function of the sf package in R.
File: BaBA_Analysis_Examples_2026-04-02.R
Description: R script providing code giving examples of how to run encounter-scale and burst-scale analyses reported in our manuscript using the WAH_BaBA_Data_2026-04-02.csv dataset. Code was developed in R version 4.4.1 and uses the tidyverse package version 2.0.0.
File: Manuscript_Updates_2026-04-02.pdf
Description: Document that reflects alterations made to the published manuscript based on updating encounter-scale duration calculations and two encounter-scale season assignments. See change log in the Files section for details. Alterations are depicted using tracked changes.
Code/software
Code for running our modified Barrier Behavior Analysis (BaBA) tool is available at https://github.com/tfullman/BaBA.
Changes after Jul 30, 2025:
1. Updated encounter-scale data
2. Added and renamed columns for clarity and consistency
3. Fixed encounter-scale duration calculations
4. Fixed two encounter-scale season assignments
5. Added a file indicating manuscript changes due to # 3 and 4
Details:
1&2) Updating encounter-scale data and adding/renaming columns
The previously posted version, WAH_BaBA_Data_2025-05-09.csv erroneously reported all statistics at the burst-scale, even though some were intended to be for encounter-scale variables. This resulted in multiple values per encounter for the encounter_scale and road columns. This was an issue when exporting the data for posting on Dryad and did not affect the results in our associated manuscript.
The newly uploaded file, WAH_BaBA_Data_2026-04-02.csv, corrects these issues and adds new columns to better clarify data used for each scale of analysis, favoring clarity over the parsimony sought in the previous version. Column names for encounter-scale data now are preceded by enc_ while those for burst-scale data are preceded by burst_. We also made slight alterations to the naming of the initial columns for consistency.
The table below lists the corresponding column names for each dataset.
| WAH_BaBA_Data_2026-04-02.csv | WAH_BaBA_Data_2025-05-09.csv |
|---|---|
| animalID | AnimalID |
| encounterID | encounter |
| burstID | burstID |
| enc_year | |
| enc_start | |
| enc_end | |
| enc_season | |
| enc_duration | |
| enc_road | road |
| enc_behavior | encounter_scale |
| burst_start | start_time |
| burst_end | end_time |
| burst_season | season |
| burst_duration | duration_hrs |
| burst_road | closest_rd |
| burst_behavior | burst_scale |
| burst_crossing | crossing |
| burst_closest_dist_km | closest_dist_km |
Note that there is still one row per burst in the dataset. Some encounters consist of multiple bursts and so encounter-scale data may be repeated. This means for encounter-scale analyses columns should first be set down and one distinct row per encounter retained. Added a BaBA_Analysis_Examples_2026-04-02.R file that gives an example of how the dataset is used for analyses at the encounter and burst scales in R.
3&5) Fixing encounter-scale durations
We also identified an issue with the encounter durations reported in our manuscript. The Barrier Behavior Analysis classification analyzes movement steps, not location points, to yield information about heading and tortuosity. Because of this the analysis adds a location onto either end of a given burst to ensure a complete step. This results in the starting step for an encounter being the step that takes the animal from outside of a barrier buffer into the buffer, and the last step being that which takes the animal out of the buffer. This also means that for multi-burst encounters a single step signifies the end of one burst and the start of the next, such that the start time of a step is used for the start of burst 2 and the end time of the step is used for the end of burst 1. This can be seen in the burst_start and burst_end columns of the updated dataset. It also means that the duration of an encounter and the sum of the durations of each burst within the encounter are not the same for multi-burst encounters. This was not indicated in the previous dataset, as encounter duration was not reported. However, WAH_BaBA_Data_2026-04-02.csv does include this information in the new enc_duration column.
Our manuscript reported encounter duration statistics based on summed burst durations, rather than accounting for overlap. This only affects encounters with multiple bursts. We corrected the encounter-scale duration measures and recalculated our summary statistics. This led to slightly altered statistics reported in the Results, Table 2, and Supplementary Information 2. However, changes were minor and did not alter any statistical test results or conclusions. Further, no burst-scale results were affected.
We have added a Supplementary file hosted via Zenodo, Manuscript Updates 2026-04-02.pdf, that uses track changes in the relevant sections to indicate where changes were made to reflect this updated information. We also are requesting a correction of the manuscript with Scientific Reports.
4&5) Fixing encounter-scale season assignments
As described in our manuscript, seasons are assigned differently at the encounter and season scales, with the burst-scale season reflecting the modal season of its individual locations and the encounter-scale season reflecting the season of its first burst. This approach was used throughout the main manuscript, however Table S8 and Fig. S8 in Supplementary Information 2 both accidentally included encounters based on modal season assignments. Correcting this had the effect of shifting two records from spring to winter, with a negligible influence on our results.
The Manuscript Updates 2026-04-02.pdf file noted above also reflects these changes, as do the data contained in WAH_BaBA_Data_2026-04-02.csv.