Diverse migration patterns and seasonal habitat use of Stone’s sheep (Ovis dalli stonei)

Study area

The Cassiar Mountains span approximately 4,301 km² of interior northwestern British Columbia, Canada (Pojar & MacKinnon, 2013) located near Cassiar, Jade City, and Good Hope Lake, within the Dease River and Kaska First Nations traditional territories. Our study area spanned 2,090 km² in the north-central ranges of the Cassiar Mountains (Fig. 3). Most of the land in the Cassiar Mountains is managed through the Dease-Liard Sustainable Resource Management Plan (Government of British Columbia Ministry of Sustainable Resource Management, 2004) developed by the Province of British Columbia and local First Nations governments.

The Cassiar Mountains are within the Boreal Cordillera ecozone comprised of three ecosystems across different elevations (Meidinger & Pojar, 1991). At the lowest elevations (650–900 m) is the montane ecosystem, which consists of black spruce (Picea glauca), white spruce (P. mariana), Engelmann spruce (P. englemannii), lodgepole pine (Pinus contorta), trembling aspen (Populus tremuloides), dwarf birch (Betula glandulosa), and willow (Salix spp.). The subalpine ecosystem ranges from 900–1,500 m and contains white spruce, willow, dwarf birch in krummholz form, and subalpine fir (Abies lasiocarpa). Above treeline (>1,500 m) is the alpine ecosystem consisting of rocky, rugged terrain and plant communities comprised of few subalpine fir in krummholz form, grasses, sedges, alpine-flowering plants, lichens, and bryophytes (Meidinger & Pojar, 1991). Summer daylight periods extend beyond 18 h, with the longest days never becoming completely dark; core winter months provide less than 8 h of daylight, with much longer periods of extended darkness. Mean regional temperature during summer (1 June to 31 August) was calculated as 12.3 °C (SD = 3.1) and during winter (1 December to 31 March) as −12.1 °C (SD = 8.5) using daily temperatures recorded from 1998 to 2022 at the nearest Environment Climate Change Canada weather station located approximately 95 km from the study site in Dease Lake, BC (Government of Canada, 2023). Average annual precipitation in the Cassiar Mountains ranged from 460–700 mm and consisted of 35–60% snowfall (Geist, 1971; Wang et al., 2012).

Predators of Stone’s sheep in the Cassiar Mountains include golden eagles (Aquila chrysaetos), grizzly bears (Ursus arctos), black bears (U. americanus), wolves (Canis lupus), coyotes (C. latrans), wolverines (Gulo gulo) and lynx (Lynx canadensis) (Geist, 1971; Nichols & Bunnell, 1999; Koizumi & Derocher, 2019). Other ungulates include moose (Alces alces), mountain goats (Oreamnos americanus), caribou (Rangifer tarandus), and some elk and mule deer (Odocoileus hemionus).

Most of the Cassiar Mountains have remained unaltered by substantial anthropogenic disturbance because of low human densities and limited access to much of the area. Anthropogenic land use in the Cassiar Mountains includes vehicle traffic on Highway 37, several small jade mine operations in Troutline Creek and other nearby valleys, the abandoned Cassiar Asbestos Mine, gold placer-mining, hunting, and recreation including hiking, camping and off-road recreational vehicle use (snowmobiles, all-terrain vehicles; Jex et al., 2016). Although relatively low human disturbance, there is increasing interest in seasonal recreation, especially off-road vehicle-use and tourism, and potential for increased mineral exploration and extraction.

Captures and GPS location data

We captured 18 adult female Stone’s sheep by helicopter net-gun in February 2018 and February and April in 2019. All individuals were collared with Iridium GPS collars (model G2110E2, Advanced Telemetry Systems, ATS, Isanti, MN, USA (Bertin, Single & Reflex, 2018) that collected GPS locations at 2 h (2018) and 1 h (2019) intervals. Capture and animal handling procedures were in accordance with BC Ministry of Forests, Lands, and Natural Resource Operations protocols, and approved by the Animal Care and Use Committee at the University of Alberta (AUP00002992).

Collaring efforts prioritized collaring females without a lamb-at-heal and from different bands to collect data that would represent movements across multiple bands in the Cassiar Mountains. As a result, the bands included in this study only had a subset of individuals collared.

GPS collars collected a total of 169,591 GPS locations from the time of captures until September 2020, unless an individual died before the end of the study, if their collar malfunctioned, or the collar batteries died. We removed erroneous GPS locations and locations that could have been influenced by capture effects by removing locations collected in the first week after captures (Werdel et al., 2021), locations collected with ≤ 2 satellites (251 locations), and locations with movement rates >20 km/h (2 locations) (D’Eon & Delparte, 2005; Lewis et al., 2007; Frair et al., 2010). We considered the time range of an individual’s location data to be adequate for evaluating an individual’s migration cycle if an individual’s location data extended at least 30 days before their spring migration to at least 90 days after their fall migration when they occupied their winter range. As a result, we removed data from two females that had <90 days of location data on their winter range after returning from fall migration. The remaining 16 females included in this study were dispersed across nine bands. For individuals for which data were collected for two full years in 2018 and 2019 (n = 4), and thus had two full migration cycles, we presented the findings on their migrations from 2019 only.

Timing of geographic migrations

To identify females that exhibited geographic migrations (migrated horizontally to a distinct summer range), we identified individuals that exhibited an abrupt movement away from their winter range and did not return during summer months. To identify this abrupt movement, we plotted each individual’s net squared displacement (NSD) from their winter range in Migration Mapper (University of Wyoming, 2021) and R v. 4.0.3 (R Core Team, 2020). To calculate NSD, we determined a start date of which Euclidean distances were calculated from for each subsequent GPS location. We defined March 15 as the start date because we assumed Stone’s sheep would be residing on their winter range at that time. However, for individuals collared in early April 2019, we defined their start date as April 12, which was one week after the captures in April 2019. We assumed the individuals collared in April 2019 would still be occupying their winter ranges at this time.

Next, we plotted the GPS locations of each female alongside their NSD in Migration Mapper (University of Wyoming, 2021) and confirmed whether a female travelled away from her winter range to a spatially distinct (non-overlapping) summer range for at least 30 days during the summer (Eggeman et al., 2016). If a female used a spatially distinct summer range, she was considered a geographic migrant. Vacillating migration was described by Denryter, Stephenson & Monteith (2021) as occupying a spatially distinct range for most of the summer, but returning (‘vacillating’) to the winter range ≥ 2 times. Females that used a distinct summer range, but ‘vacillated’ between ranges at least twice in the summer based on their NSD and GPS location data, were identified as vacillating migrants. We identified the number of times vacillating migrants returned to the winter range during summer. Females that remained on the same range year-round, thus did not exhibit a geographic migration to a distinct summer range and did not vacillate between winter and summer ranges, were considered geographic residents.

We assessed each migrating female’s GPS locations and NSD to define the start and end dates of their spring migration as the date an individual moved away from their winter range and arrived on their summer range, respectively (Eggeman et al., 2016). Similarly, we defined the start and end dates of each female’s fall migration as the date an individual travelled away from their summer range and returned back to their winter range for at least 30 days (Eggeman et al., 2016). The winter and summer seasons for females that migrated were defined as the length of time between the end date of fall migration and the start date of spring migration, and the end of spring migration and start of fall migration, respectively.

We could not distinguish a winter and summer season for geographic residents, because geographic residents do not exhibit clear migrations to seasonal ranges. As a result, we based the summer and winter seasons for geographic residents on the timing of the spring and fall migrations of geographic migrants. The summer season for residents was defined as the time between the 66th percentile of spring migration end dates and the 33rd percentile of fall migrations start dates of the geographic migrants. Likewise, the winter season for residents was defined as the time between the 66th percentile of fall migration end dates and the 33rd percentile of spring migration start dates of the geographic migrants (Poole & Lamb, 2020, unpublished).

Seasonal ranges

We developed summer and winter 95% utilization distributions (UDs) to represent each female’s summer and winter ranges by calculating Brownian bridge movement models (BBMM; Horne et al., 2007) at 50 m ×50 m resolution using the GPS locations from each female during their summer and winter seasons. We used the kernel Brownian bridge home range estimation function (Calenge, 2015) in R v. 4.0.3 (R Core Team, 2020). We defined the first smoothing parameter (σ1) using the R function liker (Horne et al., 2007), and the second (σ2) as 30 m (Frair et al., 2010; Kittle et al., 2015). We quantified the area of each female’s winter and summer ranges using the Calculate Geometry tool in ArcMap (Esri, Redlands, CA, USA). Next, we calculated the area and percent of overlap of a female’s summer range with their winter range using the intersect tool in ArcMap (Esri, Redlands, CA, USA). We provided the median and ranges for these summer and winter parameters separately for geographic migrants and geographic residents.

We assessed the amount of overlap between the winter ranges of all the females to determine which females were in the same bands. We considered females to be in the same band if their winter ranges overlapped >70% and if their telemetry locations showed them frequently moving together and occupying the same areas.

We assessed whether females exhibited fidelity to winter range in subsequent years by comparing the winter range used during the first year they were collared (2018 or 2019) to the winter range used in the following year (2019 or 2020). If a female’s consecutive winter ranges had >50% overlap, we determined that they expressed fidelity to their winter range. We had GPS locations over three winters for four females, and thus had three winter ranges with which we could assess winter-range fidelity. We determined that these four females expressed fidelity to their winter range if there was >50% overlap among all three years.

Migration routes and stopover sites

We identified migration corridors and stopover sites used by migrating Stone’s sheep during spring and fall migrations by estimating BBMMs (Horne et al., 2007). Similar to the summer and winter ranges (UDs), we delineated 95% UDs for each spring and fall migration route at 50 m × 50 m resolution by estimating BBMMs (Horne et al., 2007) using the GPS locations from each female during their defined spring and fall migrations. In the BBMMs, we also included the locations collected 24 h before and after the start and end dates of the estimated migrations as suggested by Sawyer et al. (2009). As defined in the summer and winter BBMMs, we defined σ1 using the liker function (Horne et al., 2007) and σ2 as 30 m (Frair et al., 2010; Kittle et al., 2015). We estimated the distance travelled during spring and fall migrations for each female that migrated using the Calculate Geometry tool in ArcMap (Esri, Redlands, CA, USA). A combined migration corridor was generated and mapped for each band that had multiple collared migrants.

Next, we identified stopover sites located within the migration route UDs for each female. Stopover sites were defined as a spatial cluster of GPS telemetry locations collected over at least 12 h that were located along the migration route and not overlapping summer and winter ranges. We identified stopover sites as the top 10 percent of the UD from each migration corridor (University of Wyoming, 2021). We screened out stopover sites that had less than 12 h of location data per individual. If the nearest borders of adjacent stopover sites had a Euclidean distance <300 m, we combined the stopover sites into one stopover site using the reshape tool in ArcMap (Esri, Redlands, CA, USA). We presented the frequency of stopover sites used by individuals for both spring and fall migration and the population-level median and range.

Altitudinal variation

To explore changes in the elevations used by females, we first summarized the median and range of elevations used by each female and the total collared population during the entire year, and during the summer and winter seasons. Elevations were extracted to each GPS location using the ‘amt’ package (Signer, Fieberg & Avgar, 2019) in R v. 4.0.3 (R Core Team, 2020) from the Canadian Digital Elevation Model (DEM) with 20 × 20 m resolution acquired from Natural Resources Canada (2020). Next, we calculated the change in elevations used between the winter and summer for each female (median winter elevation–median summer elevation) to identify if there was a substantial change in elevations used across seasons. Lastly, we explored each female’s elevation use throughout the year, by plotting mean daily elevations calculated over a 14-day moving window. A 14-day moving window was chosen to smooth out the ‘noise’ from minor changes in elevation that occurred daily. We used the plotted mean daily elevations for each female to identify general patterns in the elevations used during winter, spring, summer, and fall. We considered females that exhibited a change in mean elevation >250 m between the summer and winter seasons as exhibiting traditional altitudinal migration. Abbreviated altitudinal migration is defined by Courtemanch et al. (2017) as consisting of a spring altitudinal migration from low-elevation spring ranges to high-elevation summer ranges, followed by a descent to low-elevations in the fall and a return to high-elevation winter ranges (Fig. 1). We considered females that exhibited seasonal changes in elevation use that coincided with abbreviated altitudinal migration (Courtemanch et al., 2017) as abbreviated altitudinal migrants. Lastly, we identified females that did not exhibit traditional or abbreviated altitudinal migration, as altitudinal residents.

Migration strategies

Classifying migration strategies used by mountain ungulates can be challenging because individuals often migrate over both horizontal (geographic migration) and vertical space (altitudinal migration). To simplify the classification of migration strategies we first classified individuals as geographic migrants or geographic residents. Geographic migrants were identified as individuals whose summer ranges overlapped <20% of their winter range for at least 30 days (Ball, Nordengren & Wallin, 2001; Eggeman et al., 2016). Geographic migrants were further classified as either long-distance migrants or short-distance migrants depending on whether the distance of their migration route was greater or less than 20 km, respectively. We chose 20 km as the cut-off for short-distance migrants because a natural break in the distribution of migration distances occurred at 20 km (Fig. S1). Any geographic migrant that exhibited vacillating migration (returned to their winter range ≥ 2 times in the summer) was reclassified as a vacillating migrant. Females that were identified as geographic residents (remaining on the same geographic range year-round) were further classified based on their seasonal changes in elevations used as a traditional altitudinal migrant, abbreviated altitudinal migrant or an altitudinal resident. Individuals that did not exhibit a geographic migration or altitudinal migration were classified as residents.

Timing of geographic migrations

Geographic migrations from winter to summer ranges were demonstrated by 12 females. The median start and end dates of the spring migration were 12 June and 17 June (range: 20 May to 05 August; Table 1), respectively. During fall migration, the median start and end dates were 30 August and 22 September (range: 21 August to 07 January; Table 1), respectively. The length of time that females migrated in the spring and fall varied among individuals and bands. Generally, females from the same band migrated together or within a few days of one another (Table S1). The median duration of spring migration was 7 days, ranging from 1 to 26 days (Table S1). Fall migrations were typically longer with a median duration of 12 days and varied more across individuals than during spring migrations, ranging from 2 to 59 days (Table S1).

Females remained on their winter range for 67 to 211 days, with a median of 99.5 days. Like winter, the duration that females used their summer ranges varied across individuals from 43 to 183 days with a median of 69 days (Table S2).

Two females (from the same band) vacillated between winter and summer ranges (Table 2). These two vacillating females returned to the winter range three times during the summer season. Figure 4 depicts the vacillating behavior exhibited by one of the vacillating females that returned to winter range three times in the summer. We observed four females (from two bands) that did not migrate geographically to distinct seasonal ranges and thus were classified to be geographic residents (Table 2). Summer and winter seasons for residents were defined to be 26 June to 18 August and 05 October to 12 June, respectively (Table 1).

Seasonal ranges

The median area of summer ranges for all geographic migrants (n = 12) was 2,829.0 ha and ranged from 1,810.2 ha to 10,196.2 ha (Table 2). Geographic migrants had much smaller winter ranges than summer ranges, where the median area of winter ranges was 630.8 ha and ranged from 233.6 ha to 2,308.0 ha. The amount of overlap between summer and winter ranges of geographic migrants was limited with a median of 0% ranging from 0% to 9.7% (Table 2).

Nine bands were identified from the 16 collared females in this analysis based on the overlap of winter ranges. The median percent of overlap of winter ranges within bands was 74% and ranged from 71% to 85% (Table 2). Most bands included two collared females. However, there were two bands with only one collared individual, and one band with three collared individuals. All bands were visually confirmed in the field to have more individuals than the number of collared individuals.

All 16 females (across the nine bands) returned to the same winter range that they had been collared the previous year, resulting in 100% fidelity to their winter range. Additionally, all females with three years of collar data (n = 4), returned to the same winter range in all three years.

Almost all females within the same band (n = 13 from six bands) used the same winter and summer ranges. However, one band with three collared females exhibited partial migration; where one female exhibited a long-distance migration to a distinct summer range, and the other two females were residents remaining in the same range year-round.

The median area of the winter and summer ranges for geographic residents (but later identified as altitudinal migrants) were 1,540.6 ha (range: 392.6–1,750.1 ha) and 2,593.1 ha (1,765.5–3,237.7 ha), respectively. The amount of overlap between summer and winter ranges of residents varied across individuals from 29.5% to 73.2% overlap with a median of 39.3% (Table 2).

Migration routes and stopover sites

Females (and bands) varied in the distances travelled along their migration routes between winter and summer ranges (Table 2). The median distance travelled was 16.3 km ranging from 7.6 km up to 47.4 km (Table 2). Figure 5 shows the delineated migration corridors used by geographic migrants (n = 12) from eight different bands. We observed variation across individuals and bands in their use of stopover sites while on their spring and/or fall migrations (Fig. 5). During the spring migration, eight of the 12 migrants from four bands used at least one stopover site with a median of 1.5 stopover sites used ranging from 0 to 4 stopover sites (Fig. 5; Table S1). During the fall migration, 11 of the 12 migrants from eight bands used at least 1 to 6 stopover sites with a median of 2.5 stopover sites used (Fig. 5; Table S1). Geographic migrants within the same band typically used the same stopover sites, while migrants from different bands showed differences in the number of stopover sites used during spring and fall migrations (Table S1).

Altitudinal variation

Median elevation used by females during this study was 1,679 m ranging from 702 m to 2,282 m (Table S3). During summer, the median elevation was 1,709 m and varied from 1,563 m to 1,827 m (Table S3). In the winter, the median elevation was 1,673 m and ranged from 1,478 m to 1,751 m (Table S3). Almost all individuals (n = 15) occupied similar summer and winter median elevations (Δ <150 m), and thus, we concluded that they did not exhibit traditional altitudinal migration. However, one female was an exception to this, because she used a notably lower-elevation winter range with a median winter elevation of 1,531 m and a median summer elevation of 1,789 m (Δ = 257 m; female 42704 in Table S3; Fig. 6A). Figure 6 shows this female’s mean elevation use over a 14-day moving window where she used elevations in the summer that were up to 350 m higher than those used in the winter.

Many of the collared females (n = 14) exhibited changes in elevation use that coincide with abbreviated altitudinal migration (Fig. 6B). As we expected, almost all females (n = 15) descended to lower elevations in mid-April to mid-May from their winter elevations (Δ <150 m), and then gradually moved upwards over multiple weeks where they arrived at higher-elevation summer ranges in mid-July or August (Δ >150 m). Figure 6B presents an example of a female’s altitudinal migration of approximately 300 m from mid-May to August 2018. During the fall, most females (n = 14) descended to 14-day mean elevations that were at least 100 m lower than their summer ranges. At the beginning of the winter, the same 14 females ascended at least 100 m to higher elevations and generally remained at higher elevations over the winter (Fig. 6). One of the 16 females did not exhibit clear seasonal changes in elevation-use as she remained within the same 150 m of mean elevations throughout the entire year.

Migration strategies

Based on our findings above, we classified 12 females from eight bands as geographic migrants (Table 2). Of the 12 geographic migrants, we further classified five females as long-distance migrants, five as short-distance migrants, and two as vacillating migrants (Table 2; Fig. 4). One female that was classified as a geographic migrant also exhibited elevation-use that coincided with traditional altitudinal migration (female 42704 described above; Fig. 6A). For simplicity, we classified this female as a geographic migrant only, but we acknowledge that traditional altitudinal migration was observed in this population. The remaining four females from two bands that were considered geographic residents, were classified as abbreviated altitudinal migrants (Table 2; Fig. 6B). As a result, we did not classify any females as residents (considered both a geographic and altitudinal resident).