provincial results
There were several patterns that emerged in the wild bee communites at the provincial level. Firstly, in general the Grasslands and Parklands natural regions tend to be more species rich and have a higher abundance of bees than the Boreal region in the north. In addition, some of the most abundant groups of bees, including Lasioglossum and Agapostemon, tend to be more common at sites in the Grasslands natural region than at other sites. In contrast, larger-bodied bees of the Apidae family, including Bombus and Apis, tend to be more common further north in Parkland and Boreal sites. While there is overlap in many of the bee genera occurring across regions, the communities are different across these three natural regions (perMANOVA p-value=0.001), with the regions separating out based largely on latitudinal position within the province. In contrast, when sites are considered based on the land-use type, there isn't a visible difference between rangeland and canola sites at the provincial level (Figure 13). However, the result of the perMANOVA indicates potential differences that are not detected by the ordination (perMANOVA p-value=0.06). The potential differences between rangeland and canola may be driven by select groups of bees that are strongly associated with rangeland sites. Species indicator analysis identified 3 genera of bees--Osmia, Hoplitis, and Megachile--as indicator species for rangeland sites at the provincial level (p-values: 0.001, 0.034, 0.005). In contrast, only one genera, Agapostemon, was identified as being strongly associated with a particular region, with the majority of Agapostemon occurrences in the Grasslands natural region (p-value: 0.012).
grasslands region
Within the Grasslands natural region, differences in bee communities occurred between both natural sub-regions and land-use type (perMANOVA p-values:=0.001, 0.01). Four sub-regions were sampled within the Grasslands, with overall species diversity levels highest at sites in the Mixedgrass (Figure 14). Species indicator analysis identifed Halictus bees (p-value=0.01), and their cleptoparasites Sphecodes (p-value=0.03) as strongly associated with the Northern Fescue and Mixedgrass regions of the province. Anthorphora was also identified as an indicator species for the Northern Fescue (p-value=0.038), and Bombus was identified as an indicator for the Foothills Fescue, Northern Fescue, and Mixedgrass (p-value=0.018). Surprisingly, despite having a high occupancy level in the Dry Mixedgrass (80% of observations), Agapostemon was not identified as an indicator for the Dry Mixedgrass. Agapostemon was the second most common genera of bee, so these bees were fairly abundant at most sites in the Grasslands in comparison with other groups of bees. When comparing between rangeland and canola sites, most genera of bees co-occured at both types of land-use (Figure 15). However, the same 3 genera that were identified as rangeland indicators at the provincial level--Osmia, Hoplitis, and Megachile--were also identified as rangeland indicators within the Grasslands region (p-values: 0.006, 0.01, 0.014). These 3 genera are all members of the Megachilidae family, indicating that this association potentially occurs at a fairly high taxonomic level. No groups of bees were identified as indicators for canola crops.
parklands region
The majority of sampling in the Parklands took place in the Central Parkland, with the furthest north sampling sites occurring in the Peace River Parkland. These two sub-regions did have differences in the composition of the wild bee community (perMANOVA p-value=0.05). There was overlap in the types of bees in these two different regions, but overall diversity levels were higher in the Central Parkland (Figure 16). Despite no overall difference between rangeland and canola sites, Nomada was identified as an indicator species for canola fields in the Parklands region (p-value=0.037), but no specific indicator species were identifed for the different sub-regions. Andrena was identified as an indicator species for three of the sampling clusters within the Parklands which occurred primarily in the southeastern part of the Central Parklands.
regional vs local effects
Figure 17. Multivariate regression tree for the 13 most common genera with the following factors as potential explanatory categories: Site type (rangeland vs. canola), natural region, natural subregion, and sampling cluster. Split 1 explains 15% of the variation in the bee community, split 2 explains 11%, and split 3 explains 5%.
Differences between wild bee communities were observed at multiple scales, including regional, sub-regional and in some cases between rangeland and canola sites. A multivariate regression tree was used to determine which factors overall explained the most variation in the bee community data (Figure 17). Data was split initially into two different groups based on clusters, with one side representing sites generally from the northwest part of the sampling area, and the other side representing sites generally in the southeast (Figure 18). This initial split expained 15% of the variation in bee communities. Interestingly, this split does not correspond to the ecological groupings of the natural regions or subregions. However, this split does represent sites that are higher in latitude or elevation gradient and are therefore in closer proximity to large areas of forested habitat which experience different climate patterns. These sites tended to be higher in larger-bodied bees like Bombus and Megachile. The subsequent splits separated sites out by natural sub-region and explained a further 16% of the variation in bee communities.
conclusion
There is a great deal of variation in Alberta's wild bee communities that occurs at multiple scales and with multiple factors affecting that variation. Differences occur across broad regions of the province, including natural regions and sub-regions, and following agro-climatic gradients. Overall, little of the variation is explained by day-to-day weather patterns, likely as a result of restricting sampling to fair-weather days. There are a few fairly widespread and common genera that occur across the province, but many genera are fairly rare or restricted in their current detection levels. In particular, certain species in the southern part of the province tend to associate with particular sub-regions or habitat types. Overall, differences between rangeland and canola sites were low or relatively undetectable, with the exception of three genera in the grasslands region of the province that favoured rangeland sites. These results represent the first description of broad-scale patterns and differences in Alberta's wild bee community. Rangeland sites in this system represent the closest available representation of reference habitat for wild bees. These results suggest that at a very coarse scale, there are minimal differences between bee communities in relatively undisturbed rangeland sites and in active canola fields.