Data

When plotting bumble bee abundance over floral abundance there was no clear trend and each point was clustered close to zero as expected based on the histograms shown in figures 10 and 13 (Figure 14A). The points were more spread out after both bumble bee and floral abundances were log-transformed and still no clear trend was visible (Figure 14B). Models that fit the data (controlled for trap hours) can be found on the Results page of this website (Figure 16). 

Similarly, no trend was clearly visible when plotting bumble bee abundance over floral richness (Figure 14C, 14D), bumble bee richness over floral abundance (Figure 14E, 14F), or bumble bee richness over floral richness (Figure 14G). Likewise, models were fitted to the data in the Results page (Figure 16). 

To visualize community structures I ran principal component analyses on the flower and bumble bee species matrices for both sample rounds (Figure 15). Before I ran the principal component analyses I removed species with no observations in a given sample round and Hellinger transformed the data to reduce the importance of common species and to meet the principal component analysis assumption of no outliers. I also attempted Wisconsin double standardization which is another common transformation that is useful for making rare species more important. The Wisconsin double standardized principal component proportion of variance explained was far lower than Hellinger transformation principal component proportion of variance explained which is another reason I opted for Hellinger transformed data. 

The principal component proportions explained for the flower communities and the bumble bee communities were quite low (<40%) indicating that there was low species correlations within their respective communities. Regardless, we can still interpret what we have from the principal component analyses, we just have to take the correlations in Figure 15 with a grain of salt. Note that Figure 15 only shows species with loading values above 0.1, Table 5 (Results page) includes the principal component loading values for each species.

In the first floral survey Alfalfa and Yellow sweet clover were positively correlated with one another. Trefoil clovers were not positively correlated with any other flowering species. However, they, along with Alfalfa and Yellow sweet clover, were slightly negatively correlated with Canada goldenrod, Field pennycress, Dandelion, and Tufted vetch, which they themselves were all positively correlated with one another. There was no clear grouping for the treed and herbaceous sites which indicates that trees did not affect the floral community in this setting (Figure 15A). 

For the third floral survey, Canada thistle and Field sowthistle were positively correlated and both were negatively correlated with Tansy and somewhat negatively correlated with Alfalfa. Alfalfa was also negatively correlated with Western showy aster and Canada goldenrod, but slightly positively correlated with Tansy and Trefoil clovers. Trefoil clovers were strongly negatively correlated with Western showy aster and Canada goldenrod (Figure 15C). 

Like in the first floral survey, the third floral survey also observed no flower species being correlated with either crop border type (Figure 15C). This is where the similarities end, however. The floral communities observed changed over time. The first floral survey had 8 unique species observed, meanwhile the third floral survey had 5 unique species (Table 4). Notably, no Tufted vetch was observed in the third floral survey which drove it's community in the first floral survey (Figure 15A). Another difference between the two is that in the first floral survey Trefoil clovers and Yellow sweet clover were not correlated positively or negatively, but in the third floral survey, the two flowering species were slightly positively correlated as mentioned in the previous paragraphs (Figure 15A, 15C). 

As for the first sampling round of bumble bees, Yellow-banded bumble bee and Yellow-fronted bumble bee are highly positively correlated with one another and are slightly positively correlated with Central bumble bee and Half-black bumble bee but slightly negatively correlated with Red belted bumble bee. Central bumble bee and Half-black bumble bee are highly positively correlated to one another but slightly positively correlated with Red belted bumble bee. Central bumble bee, Half-black bumble bee, and Red belted bumble bee are slightly negatively correlated with both Frigid bumble bee and Northern amber bumble bee which they themselves are strongly positively correlated with one another (Figure 15B).

The principal component analysis for the second bumble bee sampling round was far simpler than the first. Here Sanderson bumble bee, Yellow-fronted bumble bee, and Central bumble bee were strongly positively correlated and were all slightly negatively correlated with Yellow-banded bumble bee, Red-belted bumble bee, and Northern amber bumble bee. Red-belted bumble bee and Northern amber bumble bee were positively correlated with each other, with Northern amber bumble bee being slightly negatively correlated with Half-black bumble bee and Yellow-banded bumble bee which they themselves were positively correlated with one another. No bumble bee species appeared to be associated with either crop border type (Figure 15D). 

Like the flower communities, the bumble bee communities changed over time. Both the first and third sample rounds had 2 and 3 unique bumble bee species, respectively (Table 4). Interestingly, the species correlations changed between the two sample rounds as well. For example, in the first sampling round Northern amber bumble bee and Red belted bumble bee are negatively correlated (Figure 15B) but are quite positively correlated in the third sample round (Figure 15D).