Honey bee success predicted by landscape composition in Ohio, USA

Introduction

Honey bees (Apis mellifera, L.) exist in large, eusocial colonies that require massive and sustained inputs of floral nectar and pollen. They meet this demand by foraging at an extremely large spatial scale and with rapid responsiveness to changes in the surrounding floral community (Visscher & Seeley, 1982; Seeley, 1995). Depending on local floral availability, colonies may routinely forage over an area of more than 100 km² (Seeley, 1995), and much larger ranges have been reported under extreme conditions (Eckert, 1931; Beekman & Ratnieks, 2001).

Because honey bee foraging is a decidedly landscape-scale process, one should expect landscape composition to interact meaningfully with colony nutrition and overall colony success. While the plausibility of such a relationship is widely acknowledged (Steffan-Dewenter & Kuhn, 2003; Naug, 2009; Van Engelsdorp & Meixner, 2010; Härtel & Steffan-Dewenter, 2014), and the importance of apiary location is axiomatic among practicing beekeepers, there are very few published studies that quantitatively measure colony success in response to local landscape variables. As rapid landscape conversion continues as a global phenomenon, and beekeepers in many regions continue to suffer unsustainable losses, the task of refining and expanding our knowledge of honey bee landscape ecology takes on obvious urgency.

Several studies have indirectly explored the relationship between landscape and colony success by analyzing the spatial information encoded in the honey bee dance language (von Frisch , 1967). Waddington et al. (1994) found that colonies located in two suburban landscapes tended to forage over a smaller area and with a less clumped distribution than a previously studied colony located in a temperate deciduous forest (Visscher & Seeley, 1982), suggesting that suburban landscapes might provide richer and more evenly distributed resource patches. Similarly, Garbuzov, Schürch & Ratnieks (2014) found that colonies in the city of Brighton, UK, concentrated most of their foraging within city limits rather than venturing into surrounding countryside that was well within their foraging range. Conversely, Beekman & Ratnieks (2001) observed remarkably long-distance foraging under conditions of apparently scarce local resources in a suburban landscape and highly rewarding resources in outlying seminatural heather moors. In agricultural landscapes, honey bee foraging patterns suggest that pollen sources can be scarcer and floral patches less spatially and temporally variable in highly simplified cropping systems compared to more structurally complex habitats (Steffan-Dewenter & Kuhn, 2003), while conservation management within farmlands can increase the availability of bee-attractive flora (Couvillon, Schürch & Ratnieks, 2014).

Landscape composition can also influence the type and quality of pollen foraged by honey bees. Donkersley et al. (2014) found that the protein content of “beebread” (processed pollen stored by honey bees) was negatively correlated with agricultural land cover and positively correlated with broad-leaf forest, improved grassland, and urban land cover.

Two recent studies have directly related colony success to local landscape variables (Sande et al., 2009; Odoux et al., 2014). In the dry coastal forest habitat of southeastern Kenya, Sande et al. (2009) found that a colony’s honey production was positively correlated with its proximity to forest patches. Odoux et al. (2014) similarly found that colony size was positively correlated with forest land cover in the intensively agricultural landscape of central-western France.

Among non-peer-reviewed sources, there is a widely circulated opinion that honey bee success is favored by urban/suburban landscapes, especially in comparison to cropland (Graham, 1992; Anonymous, 2008; Wilson-Rich, 2012). These claims remain unsubstantiated but plausible given the ostensibly positive effects of suburban land use suggested by Waddington et al. (1994) and the more direct evidence supporting the favorability of suburban land use for bumble bees (Hymenoptera: Bombus, Latreille) living in predominantly agricultural areas (Goulson et al., 2002; Goulson et al., 2010).

Here, we present a quantitative study of honey bee colony success in relation to landscape composition in the State of Ohio, USA, a region characterized by a mixture of intensive cropland, deciduous forest, grassland, and urban development. While there are many ways to measure colony success, we focus on four metrics that are highly relevant to beekeepers and easily assessed through simple hive inspection: honey and pollen accumulation, wax production, adult population, and brood population. Using a citizen-science survey, we investigate the relationship between colony success and the landscape as a whole, accounting for all major land cover types and also for the potential influence of hive management variables that vary between beekeepers. Then, we specifically evaluate the putative favorability of urban land use using a subset of sites dominated by urban development and/or cropland.

Materials and Methods

Results

Landscape analysis

The landscapes surrounding the colonies in our survey represented a broad range of landscape composition in terms of the major land cover classes: Crop, Forest, Grassland, and Urban (Fig. 1). Principal components analysis of these four variables yielded two readily interpretable axes that explained greater than 96% of total variance (Fig. 2). PC1 was essentially an urban-rural axis, with sites dominated by Urban scoring low and sites dominated by combinations of Crop, Forest, and/or Grassland scoring high. PC2 partitioned non-urban landscapes into those characterized by Crop and those characterized by Forest and, to a lesser extent, Grassland.

Modeling colony success metrics by landscape principal components

Food and Wax were best modeled with PC2 as the only explanatory variable. Almost all competing models (ΔAIC_c <2) included PC2 alongside other explanatory variables, further supporting the conclusion that PC2 was the single most important predictor (Table 1). For Food, the optimal spatial scale was a 1 km radius, while Wax was best predicted at a 2 km radius. In both cases, the relationship was negative and the linear regression models were statistically significant (Food: F = 4.796, df = 48, p = 0.033; Wax: F = 6.184, df = 48, p = 0.016) (Fig. 3). AdultPop was best modeled with the coupled management variables experience and syrup as the only explanatory variables. The relationship was positive and the linear regression model was significant (F = 6.128, df = 47, p = 0.004), with significant contributions from both experience (t = 2.98, df = 47, p = 0.005) and syrup (t = 2.474, df = 47, p = 0.017) (Fig. 4). BroodPop was best predicted by the intercept-only model, indicating that none of our measured explanatory variables were good predictors of this success metric.

Modeling colony success metrics by urban landcover

In the subset of sites for which Urban + Crop was greater than 50% of total land cover, we found a significant (p < 0.05) negative relationship between Food and Urban (Fig. 5) at all spatial scales except for the two extremes of 0.5 km and 5 km; the relationship was strongest at the 2 km scale (F = 6.041, df = 29, p = 0.02). Wax was not significantly related to Urban (p > 0.05).

Discussion

The negative responses of Food and Wax to PC2 indicate that food accumulation and wax production increase with surrounding cropland and decrease with forest/grassland. This finding seems to contradict the conventional wisdom that agricultural land conversion threatens honey bee nutrition through the depauperation of floral resources relative to semi-natural environments (De La Rúa et al., 2009), but is consistent with studies that have found honey bees to be notably resilient to natural habitat loss compared to other bee taxa (Ricketts et al., 2008; Winfree et al., 2009). The productivity of honey bees does not depend so much on the presence of undisturbed natural floral communities as it does on the availability of rich resources that can be exploited efficiently by cooperative foraging (Visscher & Seeley, 1982), and agricultural environments can offer honey bees surprisingly rich floral resources in the form of “weeds” (Odoux et al., 2012; Requier et al., in press). In Ohio, the largest honey yield is believed to come from non-native clovers (Trifolium spp. L.) (Pellett, 1920; Bailey, 1955; Goltz, 1975); these plants grow abundantly along roadsides, in field margins, and in grassy yards, but they are scarce in habitats shaded by forest canopy or dominated by the dense herbaceous vegetation of unmowed grassland. In addition to the clovers, Erickson (1984) observed that, under some conditions, honey bees will forage very productively on soybean (Glycine max (L.) Merr.), and corn/soybean rotations comprise the vast majority of Ohio cropland. Dandelion (Taraxacum officinale FH Wigg.), one of the most important spring flora for honey bees in the Midwest (Jaycox, 1976) during the period of peak wax production, is distributed in much the same pattern as the clovers, thus favoring wax production in cropland over seminatural forest and grassland.

Interestingly, our finding that colony productivity is favored by cropland relative to forest/grassland is strikingly consistent with an anecdotal description of regional honey production in Ohio published nearly forty years ago (Goltz, 1975). In Goltz’ account, the areas of “primary” and “secondary” importance for honey production are in the heavily cultivated glacial plains that comprise most of the state, while the forest-dominated Appalachian Plateaus in the southeast are described as only “marginally” productive.

The positive response of AdultPop to the management variables experience and syrup is difficult to interpret. In early spring, when new colonies are very small and limited in their foraging ability, it is standard practice to supplement colony nutrition with sugar syrup. All workers produced during the period of spring build-up, though, died long before colonies were inspected in the fall, so any positive effect of springtime management on adult population at time of inspection would have to be mediated by factors that allow colonies to increase reproduction later in the year. An alternative interpretation is plausible if we allow that significant feeding may have occurred later in the year. While supplemental feeding is normally concentrated in early spring, some Ohio beekeepers also feed their colonies in mid-late summer, a period of perceived dearth in natural forage. Feeding during the summer dearth period might trigger a population increase that would persist until fall inspection. Our survey did not distinguish between feeding at different times during the season. The effect of beekeeper experience on adult population is difficult to parse, as all aspects of hive management would be expected to improve with increasing experience. Somewhat ironically, a positive relationship between colony success and beekeeper experience might be explained by the tendency of more experienced beekeepers to perform less colony management; the enthusiasm of new beekeepers can lead to unnecessary interventions that do more to disturb natural colony function than to ameliorate ills (J Tew, pers. comm., 2014).

By late September and early October, when beekeepers were inspecting their colonies for the fall survey, the bees had likely already begun to reduce brood rearing in preparation for winter (Graham, 1992). This would explain the failure of both landscape and management variables in predicting BroodPop.

The negative relationship observed between Food and Urban in the subset of our sites strongly characterized by urban and/or agricultural land use does not support the popular opinion that urban landscapes favor honey bee success relative to agricultural landscapes. At least in Ohio, the relationship appears to be the opposite, and the fact that Food was the only success metric to respond to Urban ratio suggests a likely mechanism. The last major nectar and pollen flow in Ohio is usually from goldenrods (Solidago spp. L.) (Morse, 1972; DB Sponsler, 2014, unpublished data), which bloom prolifically from late summer into fall, roughly the same period during which beekeepers in our study were conducting fall hive inspections and filling out the fall survey. At this time of year, honey bees rarely produce additional wax (Lee & Winston, 1985), and brood rearing has begun to slow down in preparation for winter (Graham, 1992), so incoming food is stored rather than being invested in brood or wax production. Goldenrods occur abundantly in uncultivated fields and conservation strips throughout agricultural landscapes, but they are relatively scarce in developed areas where vegetation is more often subject to mowing and weed control. This is consistent with the anecdotal observation of Burgett, Caron & Ambrose (1978) that urban hives tend to have poor late-season honey production, which the authors attribute to scarcity of late-blooming “weeds,” including goldenrods.

We conclude that both landscape composition and colony management contribute to the success of nascent honey bee colonies in our study region. Due to complexities not explored in this study, the prediction of colony success was partitioned such that landscape predicted food accumulation and wax production, while colony management predicted only adult worker population. We find no support for the opinion that honey bees in urban landscapes are more successful than those in cropland. To the contrary, we find that colony food accumulation responds negatively to urban land cover in landscapes dominated by urban or agricultural land use, a pattern that we attribute to the influence of late-season floral availability, particularly goldenrods.

It is important to note that while model selection identified landscape composition as the best predictor of colony food and wax accumulation, the amount of unexplained variation in our models was high, indicating that factors other than the ones we measured are also at play in the determination of colony success. Such factors may include (1) fine-scale landscape variables that were not measurable using the NLCD dataset, (2) hive management variables not accounted for in our beekeeper survey, and (3) “in-hive” determinants of colony success like queen fertility, disease prevalence, and parasite load. We also suggest caution in generalizing our results beyond our study region. While the landscape of Ohio is broadly similar to much of the American Midwest, it would be premature to extend our findings to other ecoregions that may differ strongly both in natural flora and in agricultural practices.

Supplemental Information