Quantifying recovery of British Columbia's South Selkirk grizzly bear population

Small fragmented or isolated wildlife populations typically carry an elevated risk of becoming threatened (Haddad et al. 2015). This is particularly true for large mammals, especially carnivores (Ceia-Hasse et al. 2017). Fragmentation of bear populations around the world is widespread (Proctor et al. 2021). Brown bears (Ursus arctos), often called grizzly bears in much of North America, and brown bears in Alaska and elsewhere in the world, are considered “Least Concern” across their global distribution according to the International Union for the Conservation of Nature (IUCN) Red List assessment (McLellan et al. 2017). However, a population-level assessment revealed populations along their southern distribution world-wide were fragmented into many, sometimes totally isolated, populations; 10 were considered “Critically Endangered,” 13 “Endangered,” and 8 “Vulnerable.”

Grizzly bears in North America provide an opportunity for studying and potentially reversing fragmentation patterns (Apps et al. 2014, Proctor et al. 2018, McLellan et al. 2019, Sells et al. 2023). This is mainly because extensive fragmentation of varying degrees of isolation along the southern periphery of the North American distribution has been clearly described (Proctor et al. 2012) and increases in connectivity are often measurable (Kopatz et al. 2014, 2017; Proctor et al. 2018). Remnant bear populations exist at the southern extent of their North American distribution along occupied peninsulas, one along the British Columbia (BC) coast, and the other along the greater Rocky Mountains into the Lower 48 States of the United States (Fig. 1a; McLellan 1998, Mattson and Merrill 2002, Proctor et al. 2012). Several small fragmented or isolated populations that exist along the periphery of that distribution have varying levels of conservation concern and extirpation risk. The IUCN ranked 3 as “Critically Endangered” isolated populations, 2 as “Endangered,” and 2 as “Vulnerable” (McLellan et al. 2017, Morgan et al. 2020, USFWS 2021).

Fig. 1.

(a) South Selkirk population unit in context of the North American grizzly bear (Ursus arctos) distribution (McLellan et al. 2017), and (b) South Selkirk grizzly bear population unit in trans-border population context. IUCN is International Union for Conservation of Nature.

Grizzly bears along the interior Canadian–U.S. trans-border area within their remnant peninsular distribution are fragmented into several variously sized populations separated by fractures of different intensities (Proctor et al. 2012). Most are female-fragmented, with males often mediating a nominal level of gene flow between adjacent subpopulations; however, the lack of female interchange, particularly from larger to smaller populations, presents a concern because of the lack of potential for demographic rescue (Proctor et al. 2005a, Eriksson et al. 2014). Female immigrants and their ability to bear offspring are needed to augment, or rescue, a small struggling population. This female-biased fragmentation pattern exists because females are more susceptible to fragmentation than are males (Proctor et al. 2005a, 2012). Females invest in securing the necessities of life for themselves and their offspring, having smaller home ranges, and dispersing shorter distances than those of males. Their slow shift in home range over time likely contributes to their susceptibility to fragmentation (McLellan and Hovey 2001a, Proctor et al. 2004a).

One such fragmented population is the international South Selkirk population, spanning the Canada–United States (U.S.) border in southeastern British Columbia and northern Idaho, northwestern Montana, and northeastern Washington in the United States (Figs. 1a and b; Proctor et al. 2005a, 2012). Estimated in 2005 to be approximately 80 bears (95% CI = 50–70; Proctor et al. 2007, 2012), some individuals move freely between the two countries (Fig. 1b; Kasworm et al. 2023b). Using individual-based genetic analyses, the population of origin of all sampled individuals was identified; all bears detected within the South Selkirk population originated there. This pattern was corroborated by a reduced index of genetic diversity, heterozygosity (H_E) was 0.54 compared with adjacent populations (H_E values >0.66), confirming isolation from adjacent populations (Proctor et al. 2005a). Isolated for several generations (Proctor et al. 2005a, Turnock et al. 2024), this population had an elevated extirpation risk should poor conservation trends continue (Proctor et al. 2004b).

This international population was deemed “Threatened” by the U.S. Fish and Wildlife Service (USFWS) under the U.S. Endangered Species Act (USFWS 1993) and “Vulnerable” by the IUCN (McLellan et al. 2017). The BC government considered the Canadian portion Threatened in the past (Fig. 2a; Hamilton and Austin 2004); however, BC's 2019 revised conservation ranking tool changed the status to “High Conservation Concern” (Fig. 2b; M2 on a scale of M1–M5 where M1 is Extreme; Morgan et al. 2020). Ranking occurs at the Grizzly Bear Population Unit (GBPU) level in BC. The GBPUs are not distinct biological populations, but are jurisdictional units created to partition management across the province. Some units are partially fragmented, but not isolated, and have moderate or high levels of conservation concern but lower risk of being extirpated.

Fig. 2.

(a) Past British Columbia government map of threatened grizzly bear (Ursus arctos) population units in British Columbia, Canada (Hamilton and Austin 2004). This represents the understanding and policy when our efforts began in 2004. (b) Current map of conservation ranking of grizzly bear population units in British Columbia (Morgan et al. 2020).

In 2004, we, the Trans-border Grizzly Bear Project (TBGBP), began a 15-year program in partnership with the USFWS grizzly bear recovery team to research, inform, and implement conservation solutions in an effort to reconnect and recover the South Selkirk grizzly bear population (Kasworm et al. 2023b). Using both DNA-based and Global Positioning System (GPS) telemetry methods, our cumulative research efforts included estimating population size, female survival and reproduction, population trend, habitat use, inter-population movements, and fragmentation; identifying corridors, family pedigrees, and gene flow; modelling major food resources, and density surfaces, and more (Proctor et al. 2012, 2015, 2018, 2023; Kasworm et al. 2023b).

Despite significant focus on research and conservation of small, isolated, and threatened populations of species, including bears, there has been limited quantitative monitoring to rigorously assess efficacy of recovery (Kopatz et al. 2014, Karamanlidis et al. 2015, Martinez-Cano et al. 2016, Lamb et al. 2018, McLellan et al. 2019, Vanape et al. 2022). Our goal was to reverse the fragmentation of this population to recover it from its Threatened status, establishing a model for recovery of other small grizzly–brown bear populations of conservation concern that could be locally adapted.

Cumulative research findings and knowledge about the status, threats, and suite of possible conservation solutions informed a Canadian Recovery Management Plan (CRMP; MacHutchon and Proctor 2016). This plan described the main threats—habitat loss and degradation, excessive human-caused mortality, and fragmentation and isolation—and identified specific targets for recovery (Fig. 3). These targets included restoring connectivity to address the population's fragmented status (Proctor et al. 2005a), and focus on improving female survival and reproductive rates, the most sensitive parameters in population vitality (McLellan 1989, Eberhardt et al. 1994, Garshelis et al. 2005, Harris et al. 2006, Mace et al. 2012). Recovery success of the population was to be measured using ongoing monitoring and a DNA-based population survey to determine whether targets within the CRMP were met.

Fig. 3.

A summary of conservation targets detailed within MacHutchon and Proctor (2016), threats to, and management actions implemented within, the Canadian South Selkirk grizzly bear (Ursus arctos) population between 2005 and 2020 in southeastern British Columbia, Canada. The top row of green cells contains the target values. The bottom row of green cells contains the measured value of the targets after our conservation management. ¹ The abundance target in MacHutchon and Proctor (2016) was adjusted from an open estimate of 80 to a closed estimate of 60 using telemetry data. ² Detecting evidence of reproduction through detecting a mother–father–offspring triad will be the most challenging target and may not be totally met.

Guided by the CRMP, between 2005 and 2020, we implemented and conducted ongoing evaluation of various management actions to mitigate the threats to the Canadian South Selkirk population. For example, to mitigate the threats of habitat loss and degradation, we recommended and informed access management that was applied by the Nature Conservancy of Canada (NCC) on their 750-km² Darkwoods property (Fig. 4). Research has shown a clear link between excessive road densities and degradation of high-quality habitat, resulting in lower bear densities and reproduction (Proctor et al. 2023). Others have associated low road densities with improved female survival (Mace et al. 1996, Boulanger and Stenhouse 2014, Schwartz et al. 2010), density (Boulanger et al. 2018), and population recovery (Lamb et al. 2018).

Fig. 4.

(a) South Selkirk study area (British Columbia, Canada) with human footprint, land ownership, and designations. NCC is Nature Conservancy Canada, and Darkwoods is land privately owned by the NCC. (b) South Selkirk study area with mortality risk model. Adapted from Proctor et al. (2018).

To reduce the threat of excessive human-caused mortality, we implemented nonlethal conflict response in cooperation with BC Conservation Officers and conflict reduction activities, such as the use of electric fences to secure attractants. Human-caused mortality is a ubiquitous factor in regulating bear populations in the region (McLellan et al. 1999), and applied conservation management has been shown to reduce those mortalities (Gunther 1994). Summarizing past mortality patterns for the South Selkirk grizzly bear population, Proctor et al. (2005b) found that attractant-related conflicts were a major cause of reported grizzly bear mortality. Furthermore, reducing conflict-related mortality through securing attractants and applying nonlethal conflict response have been cornerstones of successful recovery efforts of threatened grizzly bear populations in the Lower 48 States (Mace 2004) and worldwide (Krofel et al. 2020).

To reverse the threat and/or impacts of fragmentation and population isolation we instigated strategic land-trust purchases of private lands within identified corridors (Proctor et al. 2015), where we also applied attractant securement and nonlethal conflict response (conflict reduction programs). Others have found rural environments to be population sinks for grizzly bears (Lamb et al. 2017), and have recommended land trust purchases to reduce human density and, thus, human-caused bear mortality in threatened grizzly bear populations (Schwartz et al. 2010, 2012).

Our main objective, post-management, was to re-assess the conservation status of the Canadian South Selkirk GBPU by measuring the success of the targets identified in the CRMP (Fig. 3). During the course of implementing the CRMP, we monitored female mortality and our work culminated in a DNA-based population survey in 2020 and 2021, used to measure other target metrics. We hypothesized that after 15 years of conservation management, significant progress toward population recovery would have been made. We predicted that the recovery targets would be met. We also assessed the variation of several targets within the GBPU to identify areas that might benefit from additional future management.

Study area

Our 4,074-km² study area is the Canadian portion of the South Selkirk grizzly bear population. It is situated in southeastern BC west of the south arm and south of the west arm of Kootenay Lake; the Columbia River exiting BC into Washington state is the western boundary and the Canada–U.S. border is the southern boundary (Figs. 2 and 4). The mountainous study area is predominantly covered by conifer forests with patches of deciduous trees throughout. It consists of mountain valleys, typically at 550–600 m above sea level (masl), upland forests, avalanche, riparian, and alpine habitats on the slopes of peaks that reach 2,500 masl. The region is relatively wet with much of the annual precipitation received (0.98 mm rain/month, Mar–Oct) as snow in winter (70 cm/month, Nov–Mar), especially at elevations above 1,000 masl. Summers (mid-Jun–Aug) can be somewhat dry (<20 mm of rain/month) and hot (mean daily highs of 23°C) particularly at lower elevations. The predominant ecosystem types are western red cedar (Thuja plicata)–western hemlock (Tsuga heterophylla) forests at lower elevations (500–1,500 masl) and Engelmann spruce (Picea engelmannii)–subalpine fir (Abies lasiocarpa) forests at higher elevations (1,500–2,200 masl; Meidinger and Pojar 1991).

The timber-extraction industry and sporadic mining operate throughout this region except for protected areas. Extraction activities have left an extensive network of backcountry roads in many areas. Mountain ranges are typically separated by valleys containing major highways and railways that connect urban centers. These valleys often support a linear assemblage of rural landowners or communities along portions of their length. Human settlement along highways varies from stretches with continuous rural settlement to stretches with very little development (Proctor et al. 2012). The West Arm Provincial Park (PP; 262 km²) and the Midge Creek Wildlife Management Area (WMA, 147 km²) occupy the northwestern portion of the study area and NCC owns the large Darkwoods conservation property (Fig. 4a). These three tracts of land have varying degrees of protection on a continuum from lesser (Midge Creek WMA) to most (West Arm PP). The Darkwoods property was previously owned by Pluto Darkwoods, a timber harvest company that logged extensively for decades, but is now managed for conservation, particularly grizzly bears, because they have deactivated or closed roads relative to good grizzly habitat (Fig. 4a).

Much of the perimeter of this population unit consists of human-settled valleys with highways and small villages and medium-sized towns. This pattern, coupled with the existence of the large Kootenay Lake, has resulted in historical grizzly bear mortality patterns such that the South Selkirk population became fragmented from neighboring populations (Figs. 4a and b; Proctor et al. 2005a). Partly this was a result of European settlement accompanied by persecution (circa 1870–1970), which subsided in the 1980–1990s (Mattson and Merrill 2002). However, the pattern of valley bottom human-caused mortality persisted into the late 20th and early 21st centuries (Fig. 4b) after a continent-wide paradigm shift where large, sometime dangerous, carnivores were valued and their conservation became a societal goal.

The types of habitats that grizzly bears use vary seasonally but occur throughout the study area and are therefore influenced by the human activities described above. These habitats include valley bottoms, riparian areas, avalanche chutes, old burns, and older clearcuts in spring and early summer. Late summer and autumn habitats primarily include berry patches. The primary food berry in the area is black huckleberry (Vaccinium membranaceum) in open higher elevation forest and subalpine areas (Proctor et al. 2023). Other berries fed on include buffaloberry, (Shepherdia canadensis), bearberry (Arctostaphylos spp.), Saskatoon (Amelanchier alnifolia), mountain ash (Sorbus aucuparia), and elderberry (Sambucus spp.). These habitats can be compromised to various degrees by roads, which deliver an increased mortality risk, not from vehicle collisions, but from people for various reasons including conflicts at backcountry camps, self-defense, misidentification, and malicious kills (Fig. 4b; Lamb et al. 2019; Proctor et al. 2019, 2023).

Methods

Study design

Our overall objective was to assess the efficacy of conservation management applied over the previous 15 years (2005–2019). We used several tools to accomplish this. In addition to monitoring reported female mortality and several bears' inter-population movements through telemetry (Proctor et al. 2015, 2023), we relied primarily on a DNA-based survey in 2020–2021.

We linked threats identified in the CMRP to management actions and measured target metrics to assess conservation status (Fig. 3). We measured abundance and density using the 2020–2021 DNA-based survey. To assess trend, we compared abundance estimates from an area within the 2020 survey with the results from the exact same area surveyed in 2005 (a baseline survey in the eastern two-thirds of the Canadian South Selkirk study area, Figs. 5a and b; Proctor et al. 2007, 2023). The number and distribution of females, and evidence of their reproduction, were evaluated with the 2020–2021 DNA survey. We used numbers of detected immigrants into the South Selkirks as indicators of connectivity, as well as offspring of those immigrants as indicators of gene flow (Proctor et al. 2018). We assessed reported human-caused female mortality from our long-term monitoring of bear deaths across the study area. And finally, we assessed the variation of abundance and female distribution within the study area by developing a probability of occurrence surface. To accomplish this, we partitioned the study area into 6 subunits (Fig. 6).

Fig. 5.

(a) Grizzly bear (Ursus arctos) DNA grid and sampling sites during 2020 in the eastern two-thirds and 2021 in the western one-third of the South Selkirk Grizzly Bear Population Unit (British Columbia, Canada); and (b) The 2005 grizzly bear DNA survey carried out by the Trans-border Grizzly Bear Project (Proctor et al. 2007).

Fig. 6.

(a) Female and (b) male grizzly bear (Ursus arctos) detections in the South Selkirk Grizzly Bear Population Unit (GBPU) of southwestern British Columbia, Canada, 2020–2021.

Although most similar surveys usually occur during a single year, this DNA survey was carried out over 2 years (2020–2021) because of constraints associated with the COVID-19 pandemic. The use of recently developed spatially explicit capture–recapture (SECR) population estimate techniques have made it easier to combine multiple-year sampling efforts, without compromising abundance estimation objectives. The remaining objectives needed no special accommodation for the 2-year survey effort.

Results

Abundance and density

Between 2020 and 2021 we collected 4,356 hair samples (2020: 2,772, 2021: 1,584) at 121 sampling sites (2020: 77, 2021: 44), 497 of which were determined to be from grizzly bears (Table 1). After genetic analyses (see below), 73 individual grizzly bears (41 females, 32 males) were identified at survey sites, of which 2 females and 1 male were detected in both years (Table 2, Fig. 6). Most of the remaining samples were from American black bears. We captured 53 of 73 grizzly bears used in the population estimate more than one time (mean = 2.3, range = 1–6; Fig. 7). We had a relatively even number of detected grizzly bears across the 4 sampling occasions within each year (Table 2). In 2020 on the eastern two-thirds of the study area, we detected 50% more females than males, but the sex ratio of detections was almost equal in 2021 in the western one-third of the study area (Table 2). In addition, 8 bears were detected opportunistically and were only used in pedigree and migrant analyses.

Table 1.

Grizzly bear (Ursus arctos) detections across 2 years of field sampling (2020–2021) in the South Selkirk Grizzly Bear Population Unit of British Columbia, Canada.

Table 2.

All grizzly bear (Ursus arctos) detections by session across 2 years of field sampling in 2020, 2021, and both years combined in the South Selkirk Grizzly Bear Population Unit, British Columbia, Canada.

Fig. 7.

Grizzly bear (Ursus arctos) recaptures (detections at different sites) across years in the South Selkirk Grizzly Bear Population Unit, British Columbia, Canada. Different colors represent different individuals. Small open circles and 1 symbols are sites with no detections.

The SECR abundance estimate was 69 (95% CI = 56–86, coefficient of variation [CV] = 10.8%) bears on average in the study area at any one time (Table 3). This estimate was lower than the number of detected bears (73) because some bears moved off the study area during the sampling period (primarily into the United States). The study area was likely closed (no bear movement outside of the area) to the north and west sides and partially on the east side but was fully open across the U.S. border. Radiocollar data in the international population revealed that 24 of 65 radiocollared grizzly bears in Canada had home ranges spanning the Canada–U.S. border to varying degrees. We used the telemetry data of all bears that spent time within Canada and the United States at some point (24 bears) to estimate a closure correction factor of 0.75 (McLellan 1989, Kendall et al. 2016). This closure correction factor was used to adjust the CMRP abundance target to 60 bears, from the original 2005 open target of 80 (MacHutchon and Proctor 2016). Therefore, the abundance target was met (Fig. 3). Our estimated average density of 17.2 bears/1,000 km² (95% CI = 13.5–21.8; Table 3) also exceeded the target density of 15 bears/1,000 km².

Table 3.

Grizzly bear (Ursus arctos) abundance and density estimates across 2 years of field sampling (2020–2021) in the South Selkirk Grizzly Bear Population Unit, British Columbia, Canada. Estimates are the average number of bears in the study area. This number can be lower than the number of bears detected because some bears left the study area during the survey. The average overall abundance estimate for the 2-year survey is 69.2, in bold in Table. The study area is open, with the United States just south of the border—bears move freely between countries. “CI” is confidence interval.

The precision of our density estimates differed between years; CVs below 20% are considered sufficient for management decisions (Pollock et al. 1990, Boulanger et al. 2002). The relatively higher CV in 2021 is related to the smaller number of bears detected in that year and area—21 in 2021 versus 55 in 2020. Our low CV of 10.8 across both years suggests we detected a high proportion of the bears in the study area overall.

Probability of occurrence surface (partitioning density)

We found that abundance and density estimates varied considerably across the 6 subunits that were identified in 2016 (MacHutchon and Proctor 2016) to assess the distribution of bears across the study area (Table 4, Fig. 8a). Overall, the northern and central portions along the highest elevations of the Selkirk range had the greatest densities of bears (Fig. 8) while the Pend Oreille subunit in the southwestern corner had very low densities (Table 4).

Table 4.

Grizzly bear (Ursus arctos) detections, mothers, offspring, density, and abundance estimates of grizzly bears in subunits of the South Selkirk Grizzly Bear Population Unit, British Columbia, Canada, in 2020–2021. Subunits were considered to identify areas with particularly low densities (e.g., Pend Orielle). Mothers and offspring are reported because these are recovery metrics in MacHutchon and Proctor (2016). “Long-term” refers to our genetic sample from 1990 to 2021. “SECR” is spatially explicit capture–recapture.

Fig. 8.

(a) Grizzly bear (Ursus arctos) both-sex probability of occurrence surface based on the most supported model in Table 5 in the South Selkirk Grizzly Bear Population Unit (GBPU) of southeastern British Columbia, Canada, 2020–2021. The number within subunits is the estimated density derived from partitioning the probability of occurrence model calibrated with the spatially explicit capture–recapture (SECR) density (abundance in parentheses); and (b) Plot of the relative importance of each variable in our top model. We reran the logistic regression of our top model and removed each variable one at a time. Variables with the greatest change in log likelihood scores had more influence within the top model. For example, alpine habitats and road density were the two most influential variables. Alpine had a positive influence while road density had a negative influence.

Table 5.

(a) Probability of occurrence model selection for grizzly bears (Ursus arctos) in the South Selkirk Grizzly Bear Population Unit, British Columbia, Canada, 2020–2021; and (b) beta coefficients for the top model in a.

The probability of occurrence model with the most support included positive correlations for alpine scaled to 8k, canopy closure 15k, and greenness 8k; and negative associations for road density 15k and terrain ruggedness 15k (Tables 5a and b, Fig. 8b). The top model's pseudo-R² score was 0.27, and the area under the curve of the receiver operator score was 0.83, which is considered good predictability (Hosmer and Lemeshow 1989). The probability of occurrence surface (Fig. 8a) predicts that the southwestern Pend Oreille subunit had the lowest abundance (4) and density (6.3 grizzly bears/1,000 km²). This contrasts with the zero bears (abundance and density) estimated by SECR, derived from detected bears within each subunit (Table 4).

Distribution of females and reproduction

Overall, 44 females were detected, with some occurring in all subunits; but their distribution varied considerably (Fig. 6a, Table 4). Similarly, the distribution of 18 detected mothers with offspring covered all subunits, with scant detections in the southwestern Pend Oreille subunit (Figs. 9a and b, Table 4). Therefore, the distribution of females target was met, while the distribution of reproductive females was 2 females short of being met (Fig. 3).

Fig. 9.

(a) Mothers and offspring detected during 2020–2021 grizzly bear (Ursus arctos) survey in the South Selkirk Grizzly Bear Population Unit (GBPU; British Columbia, Canada). Lines and arrows point from mothers to offspring. Some dots or stars represent multiple bears detected at the same location. (b) All Canadian grizzly bear mothers and offspring detected in our long-term genetic data set.

Connectivity

We detected 2 females and 1 male on both sides of Highway 6. We also detected relatively high numbers of both male and female bears in the Apex Creek area (just south of Nelson, BC), suggesting this area as a potential corridor across this highway (Fig. 7; Proctor et al. 2015). Using our long-term family-unit-triad pedigree (mother–father–offspring), we found evidence of 4 offspring on the opposite side of Highway 6 from their mother's detection, and 1 male who fathered an offspring on both east and west sides of Highway 6 (Fig. 9b). These results suggest Highway 6 was not acting as a barrier to bear movements or dispersal.

We found a considerable increase in the number of immigrants into the South Selkirk GBPU between our pre-2006 samples (Figs. 10a and b) and our more recent samples up through 2021. Using our updated long-term genetic database, telemetry data, and direct recaptures, we detected 9 immigrants into the South Selkirk population from the South Purcell or Yahk (Can)–Yaak (U.S.) populations to the east. Those immigrants were parents of 27 offspring (12 F, 15 M) that also were detected in the South Selkirk GBPU (Fig. 10b). These offspring represent gene flow from the Purcell into the South Selkirk Mountains. We also detected 3 male emigrants from the South Selkirks into the Yahk and South Purcell units. We found the expected heterozygosity of the bears detected in 2022 to be 0.56, a slight increase from the 0.54 reported for bears before 2005 in Proctor et al. (2005a). The target of achieving inter-population connectivity and gene flow recovery targets were exceeded (Fig. 3).

Fig. 10.

(a) Grizzly bear (Ursus arctos) immigrants into the South Selkirk Grizzly Bear Population Unit (GBPU) in southeastern British Columbia, Canada, before 2006. Adapted from Proctor et al. (2018), and (b) Grizzly bear immigrants into the South Selkirk GBPU (9) and 27 offspring representing gene flow in southeastern British Columbia as of 2021.

Discussion

In an era of declining biodiversity (Ray et al. 2021), there is increased importance of developing, implementing, and evaluating recovery strategies. Recovery plans are important tools for species conservation but their effectiveness can vary significantly. Monitoring and evaluation are essential to track progress and make necessary adjustments to ensure the best possible outcomes for threatened species. Our evaluation of the targets within the Canadian Recovery Management Plan (CRMP; MacHutchon and Proctor 2016) provided quantitative evidence of improved population metrics in a small threatened grizzly bear population that previously was determined to be possibly isolated. This stemmed from 15 years of strategic conservation management, working toward 8 conservation targets. Of the 8 conservation targets that were set, we determined that ≥7 targets were met. Management actions, including nonlethal conflict response and extensive electric fencing of attractants, likely contributed to the increased number of bears and their connectivity with neighboring populations. These results are consistent with ongoing monitoring in the South Selkirk grizzly bear population, which led to our hypothesis that management activities were recovering this population (Proctor et al. 2018). Building on Proctor et al. (2018), we present results of a comprehensive DNA-based population-wide survey, providing rigorous evidence for a quantitative assessment of recovery targets.

For decades, reducing human-caused mortality has proven successful in reversing the decline in remnant threatened grizzly bear populations in the western Lower 48 States of the United States (Mace et al. 2012). However, securing or restoring habitat quality and security also plays a role. For example, Lamb et al. (2018) attributed improved conservation status of a small south-central BC grizzly bear population to improvements in security of high-quality backcountry habitats through creation of provincial parks and an access management plan. The implementation of the CRMP in our study area relied on management activities to reduce human-caused mortality and protect productive habitats. Lowered road densities and higher proportions of secured habitat (>500 m from open roads, extensive huckleberry fields) allowed females secure access to foraging patches, yielding higher female densities with higher reproductive rates (fitness; Proctor et al. 2023).

Our SECR estimate of 69 bears exceeded the adjusted abundance target of 60 bears. The abundance target adjustment considered several factors in lieu of a reliable carrying capacity estimate. It was designed to ensure there were sufficient females in the population to be sustainable as long as the population was also reconnected (inter-population movements and gene flow) to the Purcell Mountains (across Highway 3A) to the northeast. Our estimated average density of 17.2 bears/1,000 km² exceeded the target abundance density of 15 bears/1,000 km², although this is still below average compared with the average data-based density estimates of other interior BC populations (∼23 bears/ 1,000 km²; Mowat et al. 2005). Wielgus et al. (1994) measured 23.2 bears/1,000 km² for a portion of the Canadian South Selkirk unit. This estimate 30 years ago was derived from the mean annual number of bears in an 815-km² annual composite home range of radiocollared bears—considerably less that the 4,427 km² we surveyed. Wielgus et al. (1994) did not provide a map of the polygon where they estimated density, therefore a direct comparison was not possible. We suspect that they concentrated their radiocollaring effort in the area of highest chance of capturing bears—the densest areas of the Canadian portion of the population unit. If so, this was consistent with the areas of highest density areas in our study area where densities were >30 bears/1,000 km². We cannot determine whether the absolute differences in density in these good habitats were due to different methods or changes in actual bear density.

Probability of occurrence was influenced by road density. This supports previous findings within this ecosystem (Wielgus et al. 2002, Proctor et al. 2023), in the adjacent population unit to the west (Lamb et al. 2018), northern BC (Ciarniello et al. 2007), and in Alberta (Boulanger et al. 2018). Low road densities and high proportions of secure habitat (>500 m from open roads) have been shown to yield increased female survival (Schwartz et al. 2010, Boulanger and Stenhouse 2014, Lamb et al. 2020), often associated with positive population trends (McLellan 1989, Eberhardt et al. 1994, Garshelis et al. 2005, Harris et al. 2006, Mace et al. 2012). Strategic access management and stronghold areas of low road densities in our study area likely contributed to the observed increase in probability of bear detection. The NCC Darkwoods property, West Arm Provincial Park, and the Midge Creek Wildlife Management Area comprise 24% of the South Selkirk GBPU. These areas have varying degrees of habitat protections and an average road density of 0.12 km/km². This contrasts to the road density overall in the GBPU (1.5 km/km²). One recommended habitat-structured road density target for sustainable grizzly bear populations is 0.6 km/km², although this does not guarantee sustainability (Proctor et al. 2019).

The SECR density estimate for the southwestern Pend Oreille subunit was zero bears. This contrasts with the probability of occurrence extrapolated density estimate of 6.3 grizzly bears/1,000 km². The SECR estimate was based on detected bears whereas the probability of occurrence estimate was based on extrapolated habitat covariates. One interpretation of this discrepancy is that the habitat-derived probability of occurrence estimate suggests there is a mixture of habitat quality (a positive influence) and road density (a negative influence) to support more bears than we detected, which was one adult female with offspring. Some degree of access management to reduce road densities would have the potential to increase the number of bears in this subunit. Proctor et al. (2023) found that using a SECR-derived density model estimated an increase in grizzly bear numbers through simulations of decreases in road density and increases in secure habitat in their multimountain study area that included the South Selkirk GBPU.

Our estimates of varying population density are reflected in the distribution of female reproduction. Our evidence of reproduction comes from family pedigrees—triads of mother–father–offspring, where the offspring share a complementary allele from each parent over 21 genetic markers. Our robust survey was designed to maximize the detection of mothers, fathers, and offspring using a relatively intense sampling effort (36-km² cell size). We detected 18 mothers and 31 offspring of those mothers, sired by 7 different fathers. However, the distribution of reproducing females was highly variable. The northeastern subunit (West Arm), having higher security (low road density) had the most reproductive females, while the heavily roaded southwestern subunit (Pend Orielle) had only one reproductive female. This supports the findings of Proctor et al. (2023) that habitats of higher quality and higher proportions of secure habitat (low road density) yielded the highest female density and fitness values.

The CRMP's target of 20 breeding age females was derived from the estimate that approximately 25% of a population consists of breeding age females (McLellan 1989). In our DNA-based survey we were unable to estimate a full count of breeding females. However, we were able to detect 18 females with confirmed offspring (either dependent young or independent grown offspring). This may have been below our target of 20 reproductive females for 2 possible reasons: we restricted our pedigree analysis to detected triads, a challenging bar to meet; and the U.S. border is open to bear movements, exacerbating challenges in detection of family groups. Therefore, it remains a possibility that there are 20 reproductive females in this population.

Female survival, the inverse of mortality rate, has been shown to be the most sensitive parameter in grizzly bear population growth or decline (McLellan 1989, Eberhardt et al. 1994, Harris et al. 2006, Mace et al. 2012). The CRMP set a target of <1 reported human-caused female bear mortality per year, averaged over 6 years, for population sustainability. Targets are often set using the number of reported mortalities because that is what can be accurately measured. In reality, it is the total human-caused mortality that is most important. When MacHutchon and Proctor (2016) set this reported mortality target, the estimated ratio of reported to unreported bear mortality in BC was 1:1 (McLellan et al. 1999; but see below). Therefore, a total mortality target would be <2 females annually (1 reported + 1 unreported).

Estimating the ratio of reported to unreported mortalities typically is done by following the fates of radiocollared bears. Kasworm et al. (2023b) found the reported: unreported ratio of annual human-caused mortalities for the international Selkirk ecosystem to be 1:1.2 (1.2 grizzly bear is killed by people for every 1 kill that is reported). Lamb et al. (2023), working in the Canadian Rocky Mountains to the east of our study area, found a higher ratio of 1:2. Both of those results suggest that our reported female mortality target has been met.

To assess connectivity and gene flow, we primarily compared bear detections in 2 time periods. Proctor et al. (2005a) reported no immigrants of either sex in our study area, but Proctor et al. (2018) reported 3 before 2006, and here we detected 9 immigrants. In other words, numbers of immigrants into the population that we detected increased from 0 (or very few; Proctor et al. 2005a, 2018) to 9, exceeding our connectivity target. Beyond detecting immigrants, we also confirmed breeding after immigration—the 9 immigrants sired 27 offspring including 12 females. This provides evidence of not only bears moving into the South Selkirk area, but also evidence of subsequent breeding and thus gene flow. Immigrants were detected primarily through genetic methods; thus, it is sometimes not possible to exactly know when an immigration event occurred. Theoretically, a few immigration events may have occurred earlier, but we note that in the 2005 DNA survey (Proctor et al. 2007, 2023), 30 individual bears were detected, with a population estimate of 32 bears. This suggests that few individuals went undetected in that survey. This minimizes the possibility that migrants before 2005 were missed.

The IUCN Vulnerable status of this population was essentially based on 2 primary metrics: (1) smaller population size with limited females to support population growth, and (2) its isolated status, which limited an influx of immigrants to “rescue” its poor demographic potential (McLellan et al. 2017). The recent increase in connectivity with bears in the Purcell Mountains (9 immigrant males; 4 emigrant males) suggests that the South Selkirks is no longer an isolated population. As long as connectivity continues, this population will continue to exchange individuals, and ideally female immigration may also be observed. The increase of immigration we detected was primarily driven by males and they were the main driver of gene flow in this system. The future of connectivity of this population is dependent on continued management to reduce mortality of immigrants (during the dispersal process) and their resulting offspring.

Although our results provide evidence of population recovery and suggest that management efforts were successful, one must examine the outcomes through an alternative lens; could this level of population recovery have happened naturally, without any conservation management interventions? We do not believe that the population vitality we measured could have been achieved without management actions. Grizzly bears are considered a conservation-reliant species (Scott et al. 2010, Schwartz et al. 2006)—their persistence in areas with high human overlap has required constant, careful, and adaptive management in other areas. Small grizzly bear populations at the southern extent of their North American range have been shown less likely to recover with no intervening conservation-oriented management (McLellan 1998, McLellan et al. 1999, Mattson and Merrill 2002, Proctor et al. 2012). An example of a grizzly bear population that was not actively managed, and thus dwindled, is the North Cascade population in southwestern BC and northwestern Washington. This area used to contain an international grizzly bear population with serious conservation issues (North Cascades Grizzly Bear Recovery Team 2004, McLellan et al. 2017, Proctor and Morehouse 2021). Conservation attention to recover this population did not begin while there were enough bears remaining in this area. This population of grizzly bears is now considered extirpated (USFWS 2021, Proctor 2024). Human-caused mortality rapidly reduced this population through a “bottleneck” from which it never recovered (Almack et al. 1993). Recovery efforts have been proposed, but these require extensive multiyear human-assisted restoration and/or augmentation efforts (North Cascades Grizzly Bear Recovery Team 2004).

Another example of the need for active management to retain grizzly bear populations is the Cabinet Mountain population in northwestern Montana. This population has been totally isolated and was in decline in the decades before the early 1990s; it had as few as 15 bears (Kasworm and Manley 1988, Kasworm et al. 2007). Extensive management in the form of population augmentation has been keeping this population from being extirpated (Kendall et al. 2016, Kasworm et al. 2023a).

This work has been an accounting of recovery management applied within the Canadian portion of the international South Selkirk grizzly bear population. The biological population extends into the United States. Similar management efforts have been applied to that area. An evaluation of the conservation status of the entire international South Selkirk population is forthcoming because similar data for the U.S. portion of the ecosystem will be available at a later date.

Managements implications

The work done to recover the small fragmented threatened South Selkirk grizzly bear population was done in the absence of a BC provincial Species at Risk Act or government-led recovery efforts. Canada has a federal Species at Risk Act; however, it only applies on federal lands. Within BC, this applies only to approximately 1% of land area ( https://www2.gov.bc.ca/gov/content/industry/natural-resource-use). The management achieved in this study, as prescribed by MacHutchon and Proctor (2016), included partnerships and collaborations of the TBGBP, nongovernmental organizations (NGOs), land trusts, and government wildlife departments. In partnership with the TBGBP (our research team), the BC Conservation Federation delivered a long-running WildSafeBC (previously Bear Aware) public education program to reduce conflicts with bears. The BC government halted the legal grizzly bear hunt in the South Selkirk GBPU in 1995, which likely contributed to bear conservation over time; and significantly, the BC Conservation Officer Service collaborated with the TBGBP to apply nonlethal conflict response to appropriate bears, particularly females (see Proctor and Morehouse [2021] for a review of BC government regulations and actions related to Threatened populations).

Our conservation experience has highlighted the importance of certain effective science-informed management actions for continued success of bears in the South Selkirk GBPU. We recommend some degree of access management in areas (or landscape units) where the total road density exceeds 0.6 km/km². In particular, the Pend Orielle subunit appears to have excessive road densities (∼2.6 km/km²) that are likely inhibiting bear numbers. Our research also identified likely corridors of connectivity such as the Apex Wetland area in the vicinity of Nelson that should be considered as a priority for protection. Further, we recommend that the management actions mentioned in this paper be adopted and continued by the BC government and the NGO community. This includes the nonlethal efforts to reduce conflict-related mortality. One possibility is to have a qualified government employee (either a Conservation Officer or other wildlife manager) be responsible for this program in areas where conflict mortalities justify it. The other management need is a government-backed long-term fund for the cost-share electric fencing program that has been operating in this area but that will need to be maintained. This would distribute the cost of keeping grizzly bears in BC among the taxpayers, rather than placing the burden of coexistence on affected farmers and ranchers, who currently shoulder the cost. The benefit of grizzly bears to BC is a shared provincial value, codified by government and reflected by the public. The coexistence programs that foster grizzly bear conservation ensure public safety, and freedom from wildlife-related property damage, while conserving these important carnivores. Government agencies with responsibilities to oversee grizzly bear conservation (e.g., BC Grizzly Bear Science Team) could benefit from using the successful approaches outlined in our study in other populations of conservation concern in BC (Fig 2b; Morgan et al. 2020). For more details on management actions we implemented, see  Appendix II (URSUS-D-24-00012R1._Proctor_et_al._SUPPLEMENTAL_MATERIALS.docx) ( Supplemental material (URSUS-D-24-00012R1._Proctor_et_al._SUPPLEMENTAL_MATERIALS.docx)).