Sells, S.N., and Costello, C.M., 2023, Grizzly Bear Space Use in the US Northern Rocky Mountains (ver. 3.0, July 2024): U.S. Geological Survey data release, https://doi.org/10.5066/P91EWUO8.
Introduction & Methods Grizzly bears (Ursus arctos) have been increasingly observed in central Montana’s plains in recent years. To assist with conservation planning, we sought to predict habitat use and connectivity pathways for grizzly bears east of the Northern Continental Divide Ecosystem (NCDE) and northeast of the Greater Yellowstone Ecosystem (GYE). To accomplish this goal, we used the methods described in Sells et al. (2023b), "Predicted connectivity pathways between grizzly bear ecosystems in Western Montana." Recent work by Sells et al. (2022, 2023a, 2023b) simulated grizzly bear movements using integrated step selection functions (iSSFs) developed from GPS-collared grizzly bears (F = 46, M = 19) in the NCDE. Data were collected from 2003 - 2020. Data used in the models were from the primary active season of May through November. Models were individual-based, and covariates were selected to maximize predictive power for each bear. Covariates were NDVI, terrain ruggedness, distance to and density of forest edge, density of riparian, density of buildings, and distance to secure habitat (defined by the Fish and Wildlife Service as public, state, and tribal lands at least 500m from roads). We applied the iSSFs developed in these works to simulate habitat use in a >230,000 km2 area encompassing Montana and far northern Wyoming, stretching from the lands east of the NCDE and northeast of the GYE to the eastern Montana boundary. Simulated grizzly bears could also move into the NCDE or GYE, and further south and west (i.e., they were not constrained to use our study area alone). We employed the undirected simulation methods from Sells et al. (2023b) in which simulated bears were randomly assigned start nodes on the eastern edge of the NCDE or northeastern edge of the GYE; bears had no predetermined end nodes. Simulated bears proceeded to “walk” using their movement models for a total of 20,000 steps (to ensure ample time for simulated bears to explore the simulated landscape). We repeated 100 iterations of simulations per female bear, and 242 per male bear, totaling 4,600 total simulations for females and 4,598 simulations for males. We summarized the total steps taken per raster grid cell (300 x 300 m cells) across all simulations (sexes combined) in cells outside of the NCDE and GYE boundaries, and binned results into equal-area classes of 1 (lowest relative predicted use) through 10 (highest relative predicted use). We next evaluated predictive capacity of the maps using outlier locations. In total, 50 outlier locations were verified for grizzly bears since 2010 by Montana Fish, Wildlife and Parks and collaborating agencies in the study area. Outlier locations generally involved isolated observations of presumably unmarked individuals verified with photo documentation of the bear(s) or their tracks. Observations were considered outliers if they occurred >7 km beyond the extent of the occupied range in that year, and likely involved dispersing individuals. If outlier locations corresponded with cells with no predicted pathways, they were assigned a value of 0. We measured classes predicted at outlier locations, Spearman rank correlations between classes and numbers of outliers, the percentage of outliers in the top class, and mean class at outlier locations with class 1 – 10. We also evaluated predictive capacity of the maps using data from GPS-collared bears in the study area. In total, 105,490 GPS locations were available (2004 – 2023) in the study area. We repeated the same measurements described above for Spearman rank correlations, percentages of GPS locations in the top class, and mean class at GPS locations with class 1 – 10. Results & Discussion As expected given that simulated bear movements were initiated along the edges of the NCDE and GYE, habitat use was concentrated near the NCDE and GYE. Simulated bears that moved further over the plains of central Montana primarily used river and stream courses for movements, along with isolated mountain ranges. Based on validation from available locations of grizzly bears in the area, the map has high predictive capacity. Mean iSSF classes at the 50 outlier locations was 7.1 and Spearman rank correlation was 0.84. A total of 14% of outlier locations were in the top class (10), and another 26% were in class 9; 60% of outlier locations were in the top 5 classes (6 – 10, representing 50% of the mapped area with class 1 – 10). Mean classes at at the 105,490 GPS locations was 8.2, and Spearman rank correlation was 0.99. Overall, 43.6% of GPS locations were in the top class 10, and 82.5% were in the top 5 classes. Our predictive map of potential grizzly bear movement and habitat use of lands east of the NCDE and northeast of the GYE can facilitate on-the-ground conservation. For example, it can be used to help prioritize habitat conservation, human-bear conflict mitigation, and transportation planning as grizzly bears expand their range eastward.
