Results & Discussion
Albertan genetic stocks are under-performing
Figure 6. Distribution of honey production by genetic stock
Figure 7. Percent of colonies that successful overwintered by genetic stock
Figure 8. Violin plot demonstrating the distribution of longevity by genetic stock
Our results show that, contrary to our prediction, Albertan queens were not the best performing group for either honey production (Fig. 6 & Table 2) or overwintering success (Fig. 7 & Table 2). Albertan overwintering success is the lowest of all 3 groups, with only about 35% of queens surviving to April 2023. This is compared to ~36.5% in Hawaiian queens and ~45% in Québécois queens. This is also shown in Figure 8, using longevity (note the bimodal distribution resulting from overwinter death). Albertan colonies tended to have a lower longevity, as shown by the wider base, than the other genetic stocks. In terms of honey production, Albertan queens produced an average of about 8 kg less honey than Québécois queens, though both stocks produced more than the Hawaiian genetic stock (Table 2).
A Chi-square test of independence of independence was performed on overwinter success, finding no significant differences between the three stocks (Χ²: 1.406, p-value: 0.495). An ANOVA test was run for honey production, as well as all other variables in Table 2, with the effects of location accounted for as a blocking variable. Preliminary tests were done to evaluate any interactive effects there may be between location and stock, finding no significant interactive effects. Following this, we chose to use location as a blocking variable to reduce statistical noise from location differences.
P-values were adjusted for multiple inferences using the Holm-Boneferonni method, resulting in a p-value of 0.715 for Honey production. This indicates no significant difference between the three groups in honey production. With this in consideration, while there appear to be trends in our success metrics, they cannot be ascribed to a difference between the three genetic stocks.
Table 2. Summary statistics for each genetic stock of queens. Includes estimated mariginal means(emmeans) and standard errors(SE), and Bonferroni-Holms adjusted p-values. 95% confidence interval Emmeans and SE of field variables (Honey, Longevity, Temperature, Varroa Levels) were calculated using a model that integrated the effect of both stock and location, while queen specific behavioural and physical traits were evaluated only according to stock.
Of all our tested variables, only head width, body weight, and sperm count were found to be significantly different between the three genetic stocks (Table 2). Québécois queens had the highest value for all three variables, with Albertan queens having the second highest body weight and sperm count and Hawaiian queens having the second highest head width. These results demonstrate that there are some phenotypic difference between the three stocks, even if they did not result in significantly greater colony success.
Queen physical and behavioral traits are poor predictors of overwintering success
Table 3. Summary statistics by overwintering success. Includes estimated marginal means and standard errors for selected physical, behavioural and environmental traits.
While we can see that there are trends in the genetic background of queens and overwintering success, most of the predictor variables we tested had no distinct difference between the groups that survived the winter and those that did not. Distribution of head width, body width, geotaxis running score, and number of turns (Table 3) were all virtually identical between the two groups. P-values all linear models in Table 3 were already insignificant prior to adjustment (Holms-Boneferonni) for multiple inference. The lack of any notable difference for the behavioral and physical traits suggest that those traits may not strongly impact overwintering success.
Kovačić et al. (2020) found that, despite foreign queens scoring higher on behavioral tests, they had a lower overwintering index. This both contradicts our findings that foreign queens are more likely to successfully overwinter, and supports our findings that behavioral traits are poor predictors of overwintering success.
There did appear to be a notable difference in temperature variance between the different genetic stocks, with a corresponding change in the rate of winter survival (Table 2). This supports the popular understanding of thermoregulation's role in honeybee overwintering (Seeley, 1985; Schäfer, 2011). Considering the lack of differences when looking at all colonies, this may also indicate that there is are other underlying variables affecting winter survival. Additionally, the sensor error prevalent in our temperature data is likely impacting the accuracy of our analysis. Mitigating these errors may provide further insight on the impacts of thermoregulation.
Varroa mite presence impacts overwintering success
Figure 9. A box plot depicting the distribution of Varroa mite count for colonies grouped by overwintering success
Varroa mite count, along with temperature variance, appears to also have a notably different distribution between groups that did and did not successfully overwinter. Colonies that perished over the winter had, on average, 0.089 more mites per 100 bees than those that survived (Table 3).
One location, Beaverlodge Research Farm, appears to have had a much greater issue with Varroa mites, having an emmean of 0.633 mites per 100 bees, compared to ~0.02 for both other locations. This corresponds with a very low rate of overwintering success, with only ~14% of Beaverlodge's colonies surviving the winter compared to ~74-80% at the other locations. This dramatic difference is also likely impacting the emmeans shown in table 2.
Few clear trends are present in predicting honey production
Table 4. A table summarizing the results of regression analysis for select physical, behavioral, and environmental traits.
Linear regression analysis was run on physical, behavioural, and environmental traits in relation to honey production. Overall correlation with honey output was weak, with body weight and running score having the strongest correlation (Table 4). Though none of the predictor variables had a significant (p > 0.05) relationship with honey production, some trends can still be analyzed. Most variables were positively correlated (however weakly) with honey production, with one exception (Table 4). Queen body weight, number of turns during behavioral testing and geotaxis running score display a positive correlation, while queen head width was found to have a negative correlation. However, the points do not well fit the linear regression model and all regression analysis found insignificant (p>0.05) relationships. Our findings do not support the use of our measured physical or behavioral queen tests as predictors of colony success. As seen in Figures 10-11 and Table 2, colonies with queens (and therefore workers) originating from the Quebec breeder (Stock 3) tend to have the highest honey outputs, indicating a potential genetic basis for differences in honey production.
Figure 10. A scatter plot depicting the relationship between honey production and geotaxis running score for all three genetic stocks.The overall trend line is depicted in grey.
Figure 11. A scatter plot depicting the relationship between honey production and queen body weight for all three genetic stocks. The overall trend line is depicted grey.
Conclusion
Our findings do not support our hypothesis that the locally bred Albertan queens would perform better in our success metrics (honey production and overwintering success), as they would be better adapted to their local environment. Queens originating from Quebec had the greatest mean honey production, with almost 8 kg more honey produced on average than Albertan queens, though both genetic stocks outperformed Hawaiian queens. Quebec queens also had the highest rate of overwintering success, with Albertan queens having the lowest. While these trends are present, they were not found to be statistically significant, so no conclusive claims can be made about genetic stock's impact on colony success.
The behavioural and physical characteristic of the queen that were measured were poor predictors of both overwintering success and honey production. We did, however, observe an effect of Varroa mites and temperature variance on honeybee overwintering, suggesting that outside environmental factor may have a greater impact. These findings were not statistically significant, but they do warrant further investigation. The strong presence of varroa mites in Beaverlodge likely played a significant role in our statistical analysis, as location was used as a block in our models.
Further research is needed in quantifying the thermoregulation capability of a colony, as the sensor error we observed is likely impacting the quality of this data. Given the performance of Quebec queens, further research looking at other genetic, physiological, and behavioral differences may be warranted.