Results

To assess if floral community composition affects bumble bee community composition I ran direct, indirect, and constrained gradient analyses. More specifically, I ran direct and indirect redundancy analyses and canonical correlation analyses on both sample rounds. I ran all three gradient analyses because each has their own advantages and disadvantages. Direct redundancy analyses may lose important bumble bee community structure, indirect redundancy analyses may lose floral community structure, and canonical correlation analyses can lose structure for both communities. Together these three analyses are informative. 

Regarding the redundancy analyses, only Yellow-banded bumble bee in the direct redundancy analysis on the third sample round was significantly explained by Tansy (p = 0.046; Figure 17C; Table 5). Otherwise, floral community structure did not significantly drive bumble bee community structure according to the redundancy analyses (Table 5).

Regarding the canonical correlation analysis, we have eight significant canonical functions for the first sample round and nine for the third sample round. Due to the sheer number of canonical functions, I will be explaining correlations between bumble bees and flowers with absolute scores ≥0.4. For the full scores report, see Table 6. 

Each of the eight significant canonical functions for the first sample round have at least one bumble bee and one flower score ≥|0.4| (Table 6). In the first canonical function, Yellow-banded bumble bee and Nevada bumble bee were correlated with Alfalfa, and Northern amber bumble bee was associated with Trefoil clover. Frigid bumble bee was correlated with Trefoil clover while Orange-belted bumble bee and Sanderson bumble bee were correlated with Harebell, Common hemp-nettle, Sage, Dandelion, Field pennycress, and Tufted vetch in the second canonical function. For the third canonical function Red-belted bumble bee, Half-black bumble bee, and Sanderson bumble bee were associated with Harebell, Yellow sweet clover, Sage, and Tufted vetch. Sanderson bumble bee was also correlated with Harebell, Sage, and Tufted vetch in the fourth canonical function. Yellow-fronted bumble bee, Red-belted bumble bee, and Nevada bumble bee were correlated with Prairie thermopsis in the fifth canonical function. The sixth canonical function associated Central bumble bee with Siberian yarrow, Common hemp-nettle, Night-flowering catchfly, and Field pennycress, as well as Yellow-fronted bumble bee with Dandelion. In the seventh canonical function Orange-belted bumble bee, Half-black bumble bee, Frigid bumble bee, and Central bumble bee to be correlated with Yellow avens, Common hemp-nettle, Night-flowering catchfly, and Field pennycress. Finally, the eight canonical function correlated Frigid bumble bee with Canada goldenrod (Table 6).

The third sample round had nine significant canonical functions but only eight had at least one bumble bee and one flower score ≥|0.5| (CAN1, CAN2, CAN3, CAN4, CAN5, CAN6, CAN7, CAN8; Table 6). Central bumble bee was correlated with Trifoil clover and both Yellow and White sweet clover, while Half-black bumble bee was associated with Western showy aster for the first canonical function. Cryptic bumble bee was associated with Alfalfa, and Sanderson bumble bee was correlated with Tansy in the second and third canonical function respectively. In the fourth canonical function Nevada bumble bee was associated with Canada thistle, Sow thistle, Canada hawkweed, Canola, Sheppard's purse, Night-flowering flycatch, Dandelion, and Field penny cress. The fifth canonical function correlated Red-belted bumble bee with Oxeye daisy and Tansy. American bumble bee was correlated with Common toadflax, and Half-black bumble bee was correlated with Canada goldenrod in the sixth canonical function. For the seventh canonical function Northern amber bumble bee and Nevada bumble bee were correlated with Canola, Sheppard's purse, Night-flowering catchfly, Dandelion, Canada hawkweed and Field pennycress, while Half-black bumble bee was associated with Field sowthistle and Canada thistle. Lastly, Orange-belted bumble bee was correlated with Siberian yarrow and Oxeye daisy (Table 6).

In the introduction page I predicted that treed crop borders, high roadside floral abundance, and high roadside floral richness would increase bee abundance and richness, and that the effects of treed crop borders and high floral abundance would be greater than additive. Additionally, I had also predicted that the roadside floral community would drive the bee community composition. Since only the bumble bees had been processed in the laboratory, these objectives were assessed only using bumble bee data so the results are not definitive. Regardless, these predictions were largely not supported by the results. 

I have two leading hypotheses as to why floral abundance and richness did not influence bumble bee abundance and richness. The first is vehicle exhaust. Since the sites were next to moderately busy highways we can assume that exhaust levels were high, especially on non-windy days when bees are more likely to be foraging. This is a problem because vehicle exhaust has been shown to either degrade or mask floral scents preventing pollinators from being able to locate flowers in the first place [24]. This would mean that the bees were essentially "blind" at the study sites, therefore they would not be influenced by the floral abundance and richness. The second hypothesis is that blue vane traps are not a good method for estimating bee abundance. Mathis et al. 2024 found that blue vane trap sampling does not correlate to bee abundance [16]. They recommended that researchers couple blue vane traps with hand netting which I did do in this study. The hand netting samples have not been processed yet, however, so results may change once those samples are included in the analysis. 

As for why the treed crop borders did not result in higher bumble bee abundance or richness, bumble bees in Alberta all nest below ground [1]. This means that they likely do not benefit one way or the other if crops are bordered by trees. If trees do increase bee abundance or richness, it will mainly affect above ground nesting bees which will be evident once the complete dataset has been analyzed.

While crop border type, floral abundance, and floral richness had no significant effect on bumble bee abundance or richness (Figure 16), the results for the bumble bee community drivers is a little more complicated. The direct and indirect redundancy analyses yielded non-significant results except for a single bumble bee species, Yellow-banded bumble bee, which was correlated with Tansy but only in the third sample round (Figure 17). The canonical correlation analysis, however, yielded seventeen significant canonical functions (Table 6). This combination suggests that floral community did drive the bumble bee community, but explained very little variation between the communities. This may improve once all three sample rounds have been processed and every bee species has been identified and analyzed statistically. Bumble bees are usually generalists so it is unsurprising that there was not much variation in their community composition [1, 11]. 

Given the current results from the limited dataset, I do not have any recommendations for landowners or roadside managers. Bumble bees in Alberta nest below ground and are flower generalists, therefore they would not benefit more so from any one crop border type or from specific flower species [1, 11]. Other ways to improve their conservation in agricultural landscapes have been more broadly addressed using other methods in the literature [5, 25]. As for roadside habitats, following current management practice recommendations is the best course of action for now [26]. Once the complete dataset has been analyzed I may have different recommendations.