Fencing is a major anthropogenic feature affecting human relationships, ecological processes, and wildlife distributions and movements, but its impacts are difficult to quantify due to a widespread lack of spatial data. We created a fence model and compared outputs to a fence mapping approach using satellite imagery in two counties in southwest Montana, USA to advance fence data development for use in research and management. The model incorporated road, land cover, ownership, and grazing boundary spatial layers to predict fence locations. We validated the model using data collected on randomized road transects (n = 330). The model predicted 34,706.4 km of fences with a mean fence density of 0.93 km/km² and a maximum density of 14.9 km/km². We also digitized fences using Google Earth Pro in a random subset of our study area in survey townships (n = 50). The Google Earth approach showed greater agreement (K = 0.76) with known samples than the fence model (K = 0.56) yet was unable to map fences in forests and was significantly more time intensive. We also compared fence attributes by land ownership and land cover variables to assess factors that may influence fence specifications (e.g., wire heights) and types (e.g., number of barbed wires). Private lands were more likely to have fences with lower bottom wires and higher top wires than those on public lands with sample means at 22 cm and 26.4 cm, and 115.2 cm and 110.97, respectively. Both bottom wire means were well below recommended heights for ungulates navigating underneath fencing (≥ 46 cm), while top wire means were closer to the 107 cm maximum fence height recommendation. We found that both fence type and land ownership were correlated (χ² = 45.52, df = 5, p = 0.001) as well as fence type and land cover type (χ² = 140.73, df = 15, p = 0.001). We provide tools for estimating fence locations, and our novel fence type assessment demonstrates an opportunity for updated policy to encourage the adoption of “wildlife-friendlier” fencing standards to facilitate wildlife movement in the western U.S. while supporting rural livelihoods.

For the fence model and fence density layers, the data was adapted from publicly available spatial layers informed by local expert opinion in Beaverhead and Madison Counties, MT. Data used included Montana Department of Transportation road layers, land ownership data from Montana State Library cadastral database, land cover data from the 2019 Montana Department of Revenue Final Land Unit (FLU), and railroad data from the Montana State Library. The data was processed in ArcMap 10.6.1 to form a hierarchical predictive fence location and density GIS model.

For the Google Earth mapped fences, data was collected by examining satellite imagery and tracing visible fence lines in Google Earth Pro version 7.3.3 (Google 2020) within the bounds of 50 random survey township polygons in Beaverhead and Madison Counties.

The data can be opened in GIS programs such as Esri ArcGIS, QGIS, or GRASS GIS (or others) or in statistical programming software such as R (or others).