Study Area

Galiano Island lies in the rain shadow of the Olympic Mountains and the Vancouver Island Ranges, in southern coastal British Columbia, Canada, a region defined by its temperate Mediterranean-type climate, with mild, wet winters and warm, dry summers (Klassen et al. 2015). The combined effects of low precipitation, warm temperatures, and high sunshine hours result in an annual moisture deficit during summer months, which varies depending on precipitation (Moore et al. 2010). Galiano Island remains relatively intact ecologically, with about 24% of its land base conserved in protected areas (Island Trust Conservancy [ITC] 2018). Today, 78% of the island landscape remains forested, with only 9% converted for active human use, including rural development and limited small-scale agriculture (Emmings and Erickson 2004; Madrone Environmental Services Ltd [MES] 2008, 2017; ITC 2018), though forestry and cumulative land-use effects have altered habitats here, as throughout the rest of British Columbia (Shackelford et al. 2018). Approximately 60% of the forested land base comprises regenerating early seral and young forests with dense canopy structure; the remainder is in a mature to old-growth state (Emmings and Erickson 2004). Galiano Island's forests are mostly coniferous, composed primarily of Douglas-fir (Pseudotsuga menziesii (Mirb.) Franco), western redcedar (Thuja plicata Donn ex D. Don), and grand fir (Abies grandis (Douglas ex D. Don) Lindl.), with pockets of moist deciduous forests and dry woodlands, including bigleaf maple (Acer macrophyllum Pursh), red alder (Alnus rubra Bong.), arbutus (Arbutus menziesii Pursh), and Garry oak (Quercus garryana Douglas ex Hook.), species representative of British Columbia's Coastal Douglas-fir Biogeoclimatic Zone (Klassen et al. 2015).

Bumble Bee Sampling and Analysis

We sampled bumble bees as part of an ecological study investigating the impact of seasonal drought on plant-pollinator communities (Simon et al. 2021). Bumble bees were collected using blue vane traps (Stephen and Rao 2005) systematically distributed across the landscape in a 2 × 2 factorial study design contrasting dry versus wet, and disturbed versus undisturbed, site conditions. However, there was an imbalance in the study design due to difficulties in logistics of site selection and access. We selected field sites and stratified site conditions using terrestrial ecosystem mapping data (MES 2008), spanning a broad range of habitats representative of the Coastal Douglas-fir Biogeoclimatic Zone (Nuszdorfer et al. 1991), including woodlands and associated rock outcrop communities, wetlands, clearcuts, hydro-line corridors, gardens, orchards, and fields. Sites ranged in size from 0.21 to 6.3 ha and were spaced between approximately 0.5 km and 23 km apart (Figure 1). Sampling was conducted over five sample periods, from April through August 2018, across 24 field sites, with three blue vane traps allocated to each site (representing a single sample), resulting in 119 samples (120 less one sample compromised due to human interference) representing 47,896 individuals. Each sampling period lasted 11 days—the time required to concurrently estimate floral resources at each site. We modelled the abundance of each bumble bee species as a response to habitat types and other environmental variables using generalized linear mixed models (GLMMs) implemented using ‘lme4’, and ‘glmmTMB’ in cases where zero-inflation proved problematic (Bates et al. 2015, Brooks et al. 2017). Note, however, that estimates of floral resources and other aspects of the ecological study conducted in 2018 are not relevant to our analysis of historical change and are therefore not reported as results.

Analysis of Historical Change

Two sources of contemporary species occurrence data were compared with historical occurrence data to inform our analysis of historical change in these communities. These include: 1) specimens (n = 47,896) collected from Galiano Island using blue vane traps during our 2018 ecological study, as described above; and 2) iNaturalist (2022) observations of the local bumble bee fauna (n = 238), dating from 2016 to 2021. All material, including historical voucher specimens, were carefully reviewed, and species determined with reference to Williams et al. (2014).

Historical species occurrence data are based on voucher specimens (n = 278) dated from 1970 to 2010, deposited at the Beaty Biodiversity Museum in Vancouver, BC, Canada, and at the Royal British Columbia Museum (RBCM) in Victoria, BC. A total of 285 specimens were databased in collections. However, 6 specimens at the RBCM were only determined to genus and could not be located during our visit to the collection. One specimen was dated to 2017 and thus was not included among historical records. No other Galiano Island bumble bee collections exist as far as the authors are aware. Historical metadata indicate that specimens analysed in this study were collected by flight intercept traps (n = 193), aerial netting (n = 10), light traps (n = 3), pitfall traps (n = 3), window traps (n = 2), and otherwise unknown methods (n = 67). The distribution of historical and contemporary species occurrence data are presented in Figure 1. Data from 2018 blue vane samples, including catalog numbers for synoptic collections deposited at the RBCM, are summarized in Table 1. Data and R scripts for implementing analyses are available on Dryad (2023). Historical and contemporary species occurrence data were analysed by rarefaction using ‘vegan’, R package v.2.5–5 (Oksanen et al. 2019, R Core Team 2019) and the Chao2 estimator (Chao 1984) implemented using ‘fossil’ (Vavrek 2011) to estimate species richness, that is, the number of unique species occurring within historical and contemporary communities. We then compared the similarity of rarefaction curves using the R package ‘rareNMtests’ v.1.1 (Cayuela and Gotelli 2014), testing the differences in historical and contemporary species assemblages based on null models. To do this, we followed the ecological null hypothesis test procedure outlined in Cayuela et al. (2015), which reveals whether samples are more different than would be expected if they were drawn from a single underlying assemblage. The same null model test was applied to compare rarefaction curves based on iNaturalist observations with those generated from blue vane trap samples, to determine whether these two sampling methods reliably converged on estimates of species richness for the contemporary bumble bee community. In these tests, Z scores represent the cumulative area between the observed sample rarefaction curve and the composite rarefaction curve; P-values represent the probability of Z given the distribution of simulated Z values; low P-values imply that observed differences among samples in species composition, richness and/or relative abundance are improbable if the samples were drawn from the same assemblage (Cayuela and Gotelli 2014). For historical data, we used individual-based rarefaction to implement null model tests comparing rarefaction of historical data with rarefaction of data from iNaturalist observations and blue vane samples. While historical voucher specimens may be analysed as sample-based data, we followed Osazuwa-Peters et al. (2018) in treating these data as individual-based to control for differences in random sampling effort and to facilitate comparison of rarefaction curves using null model tests. Taking this approach, bumble bees collected by blue vane traps were necessarily treated as individuals (n = 47,896) for comparison with individual-based rarefaction curves in null model tests, though a sample-based curve was also considered for these data, which converged on the same estimate of species richness.

Figure 1.

Choropleth map of bumble bee species richness based on historical records (1970 to 2010) and recent iNaturalist observations (2016 to 2021) from Galiano Island, BC, Canada. Field sites from 2018 surveys are also indicated, as well as historical collection sites for three bumble bee species (Bombus insularis. B. occidentalis, and B. suckleyi) reported as extirpated in this study, and two species represented by singletons in the historical record (B. fervidus, B. flavidus). The six species (B. flavifrons, B. melanopygus, B. mixtus, B. sitkensis, B. vancouverensis, B. vosnesenskii) presently known to Galiano Island were widespread in 2018, occurring in samples from all 24 field sites. Grid scale = 0.01°; at this latitude, each cell represents about 82 ha.

Extirpation: The Case of Galiano Island's Missing Bumble Bees

We documented significant shifts in the composition of an insular bumble bee community, with the apparent loss of half of the historically occurring fauna and the recent arrival of Bombus vosnesenskii, making Galiano Island an important case study in ecological change. These findings coincide with regional trends documenting the historical decline of B. occidentalis (Colla and Ratti 2010, COSEWIC 2014) and the subsequent expansion of B. vosnesenskii in British Columbia (Fraser et al. 2012).

The case for the extirpation of B. fervidus, B. flavidus, B. insularis, B. occidentalis, and B. suckleyi remains tentative, however, with several caveats to be considered. According to the International Union for Conservation of Nature (IUCN 2012), “a taxon is presumed Extinct (in our case, equivalent to Extirpated – locally extinct) when exhaustive surveys in known and/or expected habitat, at appropriate times (diurnal, seasonal, annual), throughout its historic range have failed to record an individual. Surveys should be over a time frame appropriate to the taxon's life cycle and life form.” The evidence we present, including one year of intensive blue vane sampling targeting Bombus, as well as multiple years of iNaturalist data, establishes the strong likelihood that at least five bumble bee species historically known to Galiano Island no longer occur. Our sampling focused on a diverse range of habitats representative of the island landscape (e.g., woodlands, wetlands, clearcuts, rural areas, and gardens, all within a forested matrix), including areas in the vicinity of historical collection sites (Figure 1). Studies have consistently demonstrated blue vane traps to be highly effective at sampling large insects, particularly bumble bees, though additional sampling methods may be required to capture a complete picture of bee communities (Stephen and Rao 2005, 2007; Rao and Stephen 2007; Wilson et al. 2008; Geroff et al. 2014; Buchanan et al. 2017; Gibbs et al. 2017; McCravy and Ruholl 2017; Rhoades et al. 2017). Studies have also shown blue vane traps to be effective at passively sampling the species concerned in this study, including B. fervidus, B. flavidus, B. insularis, B. occidentalis, and B. suckleyi (Kimoto et al. 2012, Pampell et al. 2015, Rhoades et al. 2016, Gibbs et al. 2017, Rivers et al. 2018). Nevertheless, additional targeted sampling efforts may be necessary to conclusively establish this case of bumble bee extirpation.

TABLE 2.

Specimen counts and proportional abundances of bumble bee species represented by: 1) historical collections; 2) iNaturalist observations; and 3) individuals captured in blue vane traps. Note: the proportional representation of B. vosnesenskii in iNaturalist observations contrasts strongly with what was found in blue vane samples, indicating a bias in community science observations toward disturbed environments (see Table 3).

TABLE 3.

Bumble bee community composition, expressed as the relative abundance of species per habitat type, based on intensive blue vane sampling from April through August 2018. Bumble bee species: B. flav. = B. flavifrons, B. mel. = Bombus melanopygus, B. mix. = Bombus mixtus, B. sit. = Bombus sitkensis, B. van. = Bombus vancouverensis, B. vos. = Bombus vosnesenskii.

Because historical collection efforts are spatially limited in extent, individual-based rarefaction of historical biological specimen data could be spatially biased. However, recent work has shown individual-based rarefaction to be preferable to spatially explicit rarefaction to control for differences in random sampling effort (Osazuwa-Peters et al. 2018). Moreover, recent sampling across a broad range of habitats revealed that bumble bee species comprising the contemporary fauna are relatively widespread on Galiano Island. Species occurred throughout all habitats, and community composition overall was relatively even across the landscape, though two species varied significantly in abundance between certain habitats. Bumble bees are known to have broad foraging ranges, from 1.5 km (Osborne et al. 2008) to as far as 11.6 km (Rao and Strange 2012). Hence, in the past, bumble bees were likely pervasive across the extent of this relatively small island (27.5 km in length and 1.6 km at its narrowest point) as they are in the present day. Based on the overall evenness of the communities sampled in 2018, spatial autocorrelation is not likely to have been a significant source of bias in historical collection efforts, further justifying individual-based rarefaction as an approach to comparing historical and contemporary species occurrence data in this study. With that said, historical collections are biased toward the north end of the island, and the occurrence of singletons suggests possible unevenness in the distribution of the historical fauna. Differences in habitat diversity between the north and south ends of Galiano Island (with more modified rural environs, open woodlands, and rock outcrops toward the south end) could have resulted in a biased picture of the historical community, especially with respect to species having narrow habitat requirements, such as B. fervidus.

In this study, the historical bumble bee sample size is small (n = 278). Our estimation of species decline should thus be considered a minimum estimate, as other species may well have occurred on the island in the past. Additionally, while three of the undetected species occurred in relatively large numbers historically, two species (B. fervidus and B. flavidus) are represented by singletons in the historical record. These singletons may represent small historical populations that have since disappeared, or vagrants that never successfully established on the island. The cuckoo bumble bee B. flavidus is a widespread holarctic species (Lhomme et al. 2021), demonstrating a high potential for dispersal. Bombus fervidus, on the other hand, has a range limited to North America, where it has shown a poorly understood but consistent pattern of decline in relative abundance since 1990 (Hatfield et al. 2015b). In this case, given its habitat preferences (open grassland, old fields, and tallgrass habitats), it may have struggled to get established on this forested island. Yet it is also possible that certain habitats (e.g., rural areas) were undersampled in the past, where B. fervidus may have been well established.

The disappearance of B. occidentalis along with the cuckoo bumble bees B. insularis and B. suckleyi is interesting to note in light of known parasitic relationships between these species (Thorp et al. 1983, Williams et al. 2014). Yet while B. insularis is a versatile parasite associated with multiple species in the local fauna, several of which persist today, B. suckleyi is not known to associate with any species reported for the island other than B. occidentalis. Bombus suckleyi thus appears to have vanished as an obligate parasite along with its host, consistent with trends elsewhere (Hatfield et al. 2015a, COSEWIC 2019). Bombus suckleyi has not been detected in British Columbia since 2013, despite extensive surveys, though it may persist in northern parts of the province where surveys have been less intensive (COSEWIC 2019).

Parasitic associations among the Galiano Island bumble bee fauna could have also resulted in higher rates of infection by Vairimorpha bombi, compounding the stressors afflicting these species. Four out of the five species reported extirpated from Galiano Island (B. fervidus, B. flavidus, B. occidentalis, B. suckleyi) are frequently infected by V. bombi (Gillespie 2010, Cameron et al. 2011, Lozier et al. 2011, Cordes et al. 2012, Pampell et al. 2015, McArt et al. 2017). Parasitic interactions among cuckoo bumble bees and their hosts could result in pathogen spillover involving V. bombi, which here may have contributed to their mutual demise. More work is necessary to evaluate the threat of pathogens such as V. bombi, as some bumble bee populations have demonstrated high pathogen prevalence yet no indication of decline (Koch and Strange 2012).

Other potential sources of environmental stress on Galiano Island historically include apiculture, logging, and reforestation, the last of which may have resulted in the loss of habitat for some bumble bees. Indeed, logging and subsequent reforestation represents the most significant landscape change that has occurred on Galiano Island over the last half century (MES 2008, 2017). Disturbance events such as forest fire and clearcuts can potentially create habitat for pollinators (Hanula et al. 2015, Korpela et al. 2015, Ponisio et al. 2016, Roberts et al. 2017, Mola and Williams 2018); subsequent forest succession may then result in declining bee biodiversity (Rivers and Betts 2021). That said, the importance of forests has largely been overlooked to date in terms of the resources they provide for bumble bees (Mola et al. 2021). The implications of forest succession for bumble bee population dynamics thus remain poorly understood.

Given the limited extent of agriculture on the island, pesticide use is not likely an important factor contributing to this case of bumble bee extirpation. Other potential stressors include the effects of bumble bee sex determination on genetic diversity at low population sizes (Zayed and Packer 2005, Lozier et al. 2011). Due to its proximity to other land masses, however, Galiano Island is not a strictly insular system, remaining subject to periodic migration or colonization by outside populations. Indeed, local Indigenous knowledge indicates that long-distance dispersal of bumble bees does occur in this archipelago. According to Rosemary Georgeson, an Indigenous resident of Galiano Island, a large bumble bee was once observed flying across the Salish Sea, landing on her boar near the mouth of the Fraser River (Rosemary Georgeson, personal communication; see supplemental material). The details of this account, including the size of the bumble bee, the time of year, and apparent distance traveled, are consistent with the long-range dispersal of a new queen.

Colonization: The Arrival of bombus vosnesenskii

Bombus vosnesenskii was first observed on Galiano Island in 2017 and subsequently collected using blue vane traps, marking another historical change in the local fauna. Shifts in bumble bee community composition may be expected to occur in the wake of local extinction (extirpation) events, as in the case of B. vosnesenskii's recent range expansion following the decline of B. occidentalis (Fraser et al. 2012). Bombus vosnesenskii has become dominant in many urban environments (McFrederick and LeBuhn 2006, Cole et al. 2019), though we found it occurred relatively infrequently on Galiano Island (Table 3). Researchers have previously suggested a threshold of urbanization that must be crossed before B. vosnesenskii assumes prominence in a community (McFrederick and LeBuhn 2006). Thus, this species may be struggling to establish on this largely forested island. On the other hand, B. vosnesenskii's low abundance may simply be the result of its recent colonization of Galiano Island.

Potential interactions between B. vosnesenskii and B. sitkensis were also noted in this study. We found B. sitkensis to be least abundant in dry modified habitats where B. vosnesenskii was most prevalent; conversely, B. sitkensis was most abundant in wetland habitats where B. vosnesenskii was least common. These findings are consistent with previous research citing the negative influence of B. vosnesenskii on bumble bee community richness in urban environments, where populations of B. sitkensis have been found to be particularly negatively affected (McFrederick and LeBuhn 2006, Cole et al. 2019). Researchers have postulated that this effect may be due to competitive exclusion; both B. sitkensis and B. vosnesenskii are subterranean nesters, making them potential competitors for nesting habitat (McFrederick and LeBuhn 2006). However, the significant habitat differences that we detected for these species could simply indicate a preference for wetlands on the part of B. sitkensis and for disturbed environments on the part of B. vosnesenskii. From this perspective, our results indicate that the effect of competitive exclusion previously reported for these species could be confounded by, coincide with, or be mitigated by niche segregation. Further research is required to understand interactions between these species given the recent arrival of B. vosnesenskii on Galiano Island.

Conclusion

Baseline data are rare for many insect groups, including bumble bees, making population analysis difficult (MacPhail et al. 2019). As a result, important pollinator species may undergo dramatic declines unnoticed (Buchmann and Nabhan 1996). Our study demonstrates the efficacy of two forms of search effort in detecting ecological change in a bumble bee community with reference to a historical baseline dataset comprising 278 museum specimens. Rarefaction curves generated from intensive blue vane sampling and iNaturalist observations converged on the same estimate of bumble bee species richness. Comparison of both sources of contemporary species occurrence data against historical data enabled the detection of changes in community composition, though as a caveat it should be noted that the local species pool (maximum of 10 species) and study area (57 km²) were small.

Blue vane traps are optimal for sampling large insects such as bumble bees (Stephen and Rao 2005), yet as the results of this study demonstrate, they can also result in high mortality. This mortality is of particular concern with respect to social insects such as bumble bees, for which passive sampling of queens during the early spring period could negatively affect populations (Gezon et al. 2015, Gibbs et al. 2017). Our findings demonstrate that bumble bee surveys need not necessarily be so intensive, however, depending on research goals, as well as the size of the species pool and the area under study. Further research deploying alternative sampling methods alongside blue vane traps and iNaturalist is necessary to compare their efficacy in estimating species diversity, richness, and evenness in bumble bee communities. Developing a non-lethal methodology for reliably surveying bumble bee communities is critical to promote more sustainable and compassionate wildlife research practices in the future (Tepedino and Portman 2020, Zemanova 2020).

In this study, iNaturalist observations crowd-sourced over the timespan of 5 years produced a reliable estimate of species richness, comparable to estimates obtained through intensive sampling using blue vane traps. These results indicate that iNaturalist might be harnessed to detect changes in the composition of ecological communities in limited cases where: a) adequate historical baseline data are available; b) the study area and local species pool of the target taxonomic group is relatively small; and c) there is low spatial autocorrelation of species occurrences. Collection of physical specimens may still be required to validate difficult taxa such as Bombus vosnesenskii, which requires careful examination to discriminate from B. caliginosus (Williams et al. 2014). Further inventory work combining methods to sample bumble bees on different spatial scales is warranted to determine the reliability of community science data for monitoring ecological change.