PEI is Canada’s smallest Atlantic province, located in the Gulf of St. Lawrence (Silva et al., 2005). The largest urban area of PEI is its capital, the city of Charlottetown (50 km² area), which has a population of roughly 44,000 individuals (Statistics Canada, 2018) ( Figure 1 ).

In order to compare the diversity and composition of bumble bee communities in urban and natural areas of PEI, a total of 20 sites were surveyed in PEINP and another 20 sites in Charlottetown ( Figure 1 ). Charlottetown sites (20/20) included walking trails, parks, public gardens, and open green spaces, in locations that could represent the interior and boundaries of the city. Sites typically consisted of native and non-native naturally occurring species (e.g., aster spp., autumn hawkbit, clover spp., dandelion spp., goldenrod spp., and thyme spp.) and/or human cultivated species (e.g., begonia spp., chrysanthemum spp., cinquefoil spp., phacelia spp., and sedum spp.). PEINP sites consisted mainly of open green spaces with native and non-native wildflowers (aster spp., clover spp., common eyebright, goldenrod spp., knapweed spp., wild rose, and tufted vetch). PEINP contains 27 km² of protected natural habitats but it is important to note that it also includes human infrastructures (e.g., campgrounds, visitor and interpretation centers) that are used predominantly during the summer. Given the substantial level of anthropogenic disturbance in Charlottetown, three major areas within PEINP were selected to encompass a broad spectrum, ranging from moderate anthropogenic disturbance to predominantly natural environments. The Cavendish/North Rustico sites (9/20) display moderate disturbance and the least natural characteristics. This area was selected for this study because of its similarity with Charlottetown, as it sustains a higher level of anthropogenic activities and human infrastructure while encompassing the least abundant amounts of green space compared to other PEINP locations. Disturbance is especially obvious during the summer when tourists visit PEINP and the town of Cavendish. In contrast, the Brackley/Dalvay sites (5/20) exhibit a moderate level of disturbance along with moderate instances of human infrastructure. Finally, the Greenwich/Fort Amherst sites (6/20), while geographically separated, demonstrate the least amount of infrastructure and rates of visitation, while displaying the highest influx of natural characteristics (forested areas, wetlands, and uncultivated fields), that could represent natural areas of PEI. In this study, we categorized the vegetation at all surveyed sites into three main groups: natural (including native and non-native plants), human-cultivated (such as ornamental plants), and mixed (a combination of natural and human-cultivated) based on the dominant plants present at each site. We used the provincial vegetation classification standards (e.g., Government of Prince Edward Island, 1977; Pollinator Partnership Canada, 2017) for this classification.

For this study, we opted for a non-lethal, non-invasive photographic survey method at the request of Parks Canada. This decision was made to avoid capturing live animals within PEINP and to minimize disturbance to bumble bee populations. We chose photographic surveys because we believe photographs are a reliable survey instrument, especially when complimented via citizen science platforms utilized or managed by field experts that verify identification. We conducted a literature review to identify the specific characteristics of species expected to occur in PEI and used the identification key published by Williams et al. (2014). In addition, we also consulted local and regional experts to ensure that our list of characteristics was accurate and exhaustive. Experts from the citizen science platforms were consulted for species that exhibited similar physical characteristics or if we were unsure of an individual species identification.

We surveyed each study site in August and September 2019, between 1000 and 1600, when the ambient temperature was at least 12°C, and no prolonged rain periods or strong winds during the survey time (sensu Pacific Northwest Bumble Bee Atlas, 2019). Even though bumble bees can fly in less suitable weather conditions relative to other bees, maintaining optimal weather conditions allows for high-quality photos of bumble bees, as strong winds can make selective focusing extremely challenging, and low light periods can greatly reduce photo quality. Two sites in Charlottetown and one in PEINP could not be re-surveyed in September for reasons out of our control. We opted not to use pre-determined transect lines to avoid the unintended exclusion of flora in bloom within the selected site (Nielsen et al. 2011). Therefore, transect lines varied slightly from August to September, as required, if the location of blooming floral resources had alternated. To reduce the possibility of double counting the same individual bumble bee, each survey was conducted for 10 minutes and involved continuously walking in a direction (e.g., meandering East to West) and only changing directions sharply if required (e.g., urban park site with flora positioned in a bordering L shape). Timers were paused if an individual bumble bee displayed uncharacteristic color patterns requiring significant photographs to further aid in accurate identification (e.g., individuals with lost hair, unusual pigmentation, and morphologically similar species).

Photographs were taken with a smartphone (Samsung S8 generation; 2268 x 4032 pixels) to provide an optimal view of the abdomen, face, and thorax of bumble bees ( Figure 2 ). An initial attempt was made to distinguish each bumble bee at the species level during the photographing process in the field. Confirmation of the identification was made by examining the photographs on a laptop and noting distinct morphological features (e.g., abdomen/tail color, facial hair, and the number/pattern of bands) (sensu Laverty and Harder, 1988). We could confidently recognize most species in our study sites by assessing morphological traits. Field observations and photographs did not permit us to distinguish confidently between B. vagans and B. sandersoni (two species expected to occur in PEI), therefore we grouped all potential observations of individuals from these two species into one group for data analyses (B. sandersoni/B. vagans). The photographs depicting bumble bees with un-characteristic color patterns were posted on iNaturalist (iNaturalist, 2020) and/or Bumble Bee Watch (The Xerces Society et al., 2020) as well as individuals from each species with typical patterns to confirm surveyor validation.

To assess community structure at each site in Charlottetown and PEINP, we computed species richness and Simpson’s index of diversity. We used evenness as a measure of relative abundance. Evenness values were calculated as the inverse of the Simpson’s index for each study site in both August and September (Krebs, 1989). Evenness reflects the relative abundance of species, indicating whether a community is dominated by a few species or if species are more evenly distributed. For example, if a community has high evenness, species are in similar abundance. Conversely, low evenness indicates that a few species are much more abundant than others. All community measurements, including species richness, species diversity, and evenness, were computed using Stata. Log transformation was applied to variables as needed to meet the assumptions of parametric statistical analyses. Multiple regression analyses were employed to investigate the influence of vegetation types and surveyed months on species diversity and evenness, while Poisson regression analysis was utilized to explore their effects on species richness.

In our study sites, we observed nine bumble bee species, six of which were also recorded by Laverty and Harder in 1988. Two expected species, B. vagans and B. sandersoni, are difficult to differentiate without close inspection. Therefore, we only confirmed the subgenus Pyrobombus as representing one species, but both species may be present, which would account for seven of the nine species expected. In addition, we found two species, B. perplexus, and B. bimaculatus, only in Charlottetown sites, which had not previously been reported in PEI. This is an important finding suggesting that B. bimaculatus may be expanding its range into Atlantic Canada, confirming observations made by Dominey (2021) in the neighboring province of Nova Scotia. Furthermore, based on “Research Grade” iNaturalist records, B. bimaculatus is most abundant between July and early September. This may explain why all 12 individuals we observed from this species were photographed only in August.

We only observed one individual of B. perplexus, documented in urban site (Site 10) on the southern border of the survey area and on an ornamental plant (chrysanthemum sp.). While listed as secure for the research location (NatureServe, 2024), limited documentation is not completely surprising given with the restricted survey period. Yet, we observed relatively higher abundances of B. terricola; while B. terricola is considered a common species (NatureServe, 2024), they are listed as Species of Special Concern by the Committee on the Status of Endangered Wildlife in Canada (COSEWIC, 2015). Although the seasonal activity of B. perplexus typically reaches peak abundance slightly earlier than B. terricola, our observations showed overlap. The higher number of B. terricola (n=10) compared to B. perplexus (n=1) suggests that B. perplexus may be experiencing declines. However, additional studies are necessary to confirm this trend throughout the species’ active life cycle in PEI.

No specimens of B. citrinus or B. insularis were observed at any of our urban or natural sites. The most likely explanation for the absence of these two species is that they are parasitic species without worker castes, making them difficult to observe. According to both iNaturalist and Bumble Bee Watch, there are sufficient “Research Grade” sightings of individuals from both B. insularis and B. citrinus in early and mid-summer to confirm that these two species occur on PEI (iNaturalist, 2020; The Xerces Society et al., 2020). As our photographic surveys were conducted in August and September, it is evident that potential opportunities to capture images of these two species at our study sites may have been missed. To mitigate this issue, it is recommended that future studies incorporate surveys that encompass the entire seasonal activity of Bombus, preferably spanning from May to October. The acquisition of a more comprehensive dataset would enable scientists and conservationists to effectively gauge the focal areas for their endeavors and identify species warranting a conservation management program.

One species not previously recorded in PEI but expected to be present was B. impatiens, given its presence in nearby Canadian provinces and American states (Matteson and Langellotto, 2009). Our study revealed the species’ presence in urban (n=558) and natural sites (n=412). B. impatiens dominated most survey areas irrespective of month and site, indicating its adaptability to various conditions. These findings are consistent with those of Dominey (2021), indicating that while B. impatiens was not historically recorded in the study area, it is now firmly established within the region. Contrary to other species, B. impatiens populations typically reach peak abundance in September and continue their colony cycle into the fall (Colla and Dumesh, 2010), explaining its prevalence in our study sites during that period. The fact that Laverty and Harder (1988) did not document the presence of this species in PEI may be attributed to their study’s reliance on museum specimens rather than field surveys. Alternatively, the species may have expanded its distribution since their research, approximately 36 years ago. Future studies should explore other areas of PEI to assess the potential widespread distribution of this species throughout the province.

The study used photographic surveys to examine bumble bee populations in urban and natural habitats and sought expert opinion via citizen science platforms to aid accurate identification. The effectiveness and cost-efficiency of this approach were successfully demonstrated, indicating its potential for application in areas where different species can be distinguished based on physical characteristics. However, our findings also revealed that photographic surveys of bumble bees had limitations when two or more species showed morphological resemblances, such as in the example of B. vagans and B. sandersoni. Individuals from these two species were indistinguishable based on photographs because of their similar morphological traits. Other photographic surveys have also found that with two or more morphologically similar species, only 94-98% of sightings could be identified to the species level (Armistead, 2023; Flaminio et al. 2021). To enhance taxonomic identification, we suggest researchers consider integrating netting to temporarily capture individuals, especially morphologically similar species. This would more effectively enable the differentiation between species that share similar morphological characteristics. Yet, in some cases, identification may be impossible without lethal capture (microscopic inspection and/or genetic analysis), for example, differentiating B. vagans and the subspecies B. vagans bolsteri. While more labor intensive than passive capture, this combination of surveys ensures minimal disturbance to populations. With the advancements posed by artificial intelligence, it is perceivable that future citizen science platforms may not require humans to dedicate time to confirm species identification (Montero-Castaño et al., 2022; Suzuki-Ohno et al., 2017).

Like any other citizen science approach, species identification based on photographs of bumble bees requires a certain level of taxonomic knowledge or expertise from the part of the surveyor. To verify surveyor expertise in accurately identifying bumble bee species, one should refer to published keys to test abilities before conducting field surveys (MacPhail et al., 2020; Montero-Castaño et al., 2022). However, the effectiveness of this approach also largely depends on what Bombus species are present. If the temporal region includes multiple species that are not easily distinguishable from field surveys, another approach may be deemed more appropriate. Limitations will still be present when conducting photographic surveys. Aside from morphologically similar species, some temporal regions support higher volumes of expected species, which can further complicate accurate identification in the field. When considering other pollinator species, such as solitary bee species, it can be impossible to differentiate without lethal capture and genetic analyses. Nonetheless, it is noteworthy that photographic surveys of bumble bees continue to offer high-quality data for preliminary assessments or to acquire informative diversity and richness indexes without necessitating permits or intrusive techniques. We suggest considering all possible limitations given the expected species for the region, and comparing the photographic survey approach to other capture methods (Armistead, 2023) to properly facilitate research objectives.

Our study found that PEINP sites exhibited the highest levels of species richness, relative abundance, and diversity compared to Charlottetown urban sites. The highest species richness observed in a single survey was within a Greenwich site (n=7). This site documented the presence of seven of the nine total species observed. Cavendish/North Rustico, while experiencing increased levels of anthropogenic activity, is predominantly surrounded by agriculture. This survey area would benefit from a comparative analysis to assess how exactly bumble bees use and adapt to each landscape use. A monthly analysis revealed a decrease in both species richness and abundance across all study sites in September. These findings are consistent with previous studies that have documented seasonal variations in the abundance of bumble bees, including some of the species found in PEI (e.g., Novotny et al., 2021). A plausible explanation for this decline is that most bumble bee species may have concluded their colony cycles by late August.

Interestingly, sites characterized by a mix of natural and human-cultivated (ornamental) vegetation—largely, sites within urban Charlottetown—showed greater bumble bee species diversity and abundance when compared to sites with solely natural vegetation or human-cultivated. Supporting this, other studies (Baldock et al. 2015; Kaluza et al. 2016; Marín et al. 2020; Nakamura and Kudo 2019; Sirohi et al., 2015) propose that urban green spaces with varied land use and moderate human activity promote species diversity by enhancing habitat and foraging diversity. For bumble bees, the mix of natural and human-cultivated vegetation offers a continuous and diverse supply of floral resources throughout the growing season, able to support a wider range of species with different foraging preferences and needs (Nakamura and Kudo 2019; Sikora et al., 2020). Furthermore, moderate disturbances like occasional mowing, planting, or construction in urban sites may prevent any single species from dominating the ecosystem. It is also possible that while ornamental plants may be visited, naturally occurring species (e.g., aster, clover, goldenrod) are the drivers for visitation within the area. To further this finding, studies should be conducted to encapsulate the entire active life cycle of bumble bees, as well as documenting floral preferences and visitation rates among each species.