We calculated each effect size in R using the escalc function in the ‘metafor’ package . We fit meta-analytic and meta-regression multilevel linear mixed-effects models, using the rma.mv function in the ‘metafor’ package . We used three random effects to control for non-independence of effect sizes collected from the same study or plant species: study ID, plant species, and an observation-level ID for individual SVE measures. We used phylogenetic comparative methods to account for non-independence that may arise due to shared evolutionary history of focal plants by including a phylogenetic covariance matrix. The phylogeny and branch lengths used to compute a phylogenetic covariance matrix came from a recently published, dated megaphylogeny contained in the package ‘V.Phylomaker’ , which combines the seed plant phylogeny from Smith and Brown with the pteridophyte phylogeny from Zanne et al. . Despite slightly higher AIC values and larger P values , we present results from models including phylogenetic controls to fully account for non‐ independence due to shared ancestry . Both analyses produced qualitatively similar results and neither the magnitude nor the sign of SMD estimates changed when phylogenetic controls were included . However, uncertainty around SMD estimates was consistently smaller in models without phylogenetic controls, such that marginally significant effects became significant when phylogenetic controls were removed. Thus, inclusion of phylogenetic controls renders our analysis more conservative. With this mixed-effects structure, we specified four models, growing strawberries vertically which include an intercept only model , and three meta-regression models for different fixed effects/moderators: pollinator taxonomic group, whether the plant was a crop plant , and for native plants, whether it was in the honeybee’s native range .
We follow Hung et al. and define the West Palearctic as the honeybee’s native range . For the analysis comparing honeybee comparative effectiveness inside and outside of the honeybee’s native range, we excluded non-native plants from the analysis.To test whether there was a relationship between a pollinator taxon’s single visit effectiveness and visit frequency, we calculated Pearson’s correlation coefficients for the relationship between visit frequency and pollinator effectiveness for each unique study, plant, site, and year combination in which there were at least five pollinator taxa represented. We filtered data because sample variances cannot be confidently estimated when fewer than five observations are used to calculate correlation coefficients. In total, 26 studies of 50 plant species had visit frequency and effectiveness data for at least five taxa, and 62% of studies were fully excluded. After calculating correlation coefficients, we used the escalc function in the metafor package to calculate Fisher’s r-to-Z transformed correlation coefficients and corresponding sampling variances. Using the same multilevel linear mixed-effects model structure and phylogenetic controls as described above we generated three models. The first model was an intercept-only model to test for the overall relationship between a pollinator’s visit frequency and single visit effectiveness. The second model compared three categories against one another: studies where honeybees were present, studies where honeybees were absent, and studies where we artificially removed all points corresponding to honeybees . We generated this third category to determine whether the patterns we observed were solely driven by honeybees themselves or whether there might also be indirect effects of honeybee presence on the relationship between visit frequency and single visit effectiveness. The third model tested whether there was an interaction between crop status and honeybee presence.
Our meta-analysis supports the hypothesis that honeybees are frequently not the most effective pollinator of plants globally. Across six continents and hundreds of plant species, honeybees showed significantly lower single visit effectiveness than the most effective pollinator . This general pattern is likely driven by comparison of honeybees against birds and other bees. The most effective bird and bee pollinators were significantly more effective than honeybees, as were the average bird and bee pollinators. The finding that birds are more effective than honeybees is based on only six studies that were likely focused on flowers frequently pollinated by birds. Nevertheless, it supports the idea that plants adapted to bird pollination have traits that enhance pollination by birds at the expense of pollination by bees . Although data for non-bee taxa were relatively sparse, honeybees were as effective as the average and most effective ant, beetle, butterfly, fly, moth, and wasp pollinators, confirming that non-bee insects can be important pollinators . Our results bolster initial work summarizing honeybee pollination effectiveness and demonstrate that honeybees are less effective than many other visitors and at best average. Analysis of crop plants also revealed important differences between honeybees and nonApis pollinators. Despite their abundance in commercial cropping systems, honeybees are less effective crop pollinators than the most effective bee pollinators and the average non-honeybee bees . This finding supports the idea that the importance of honeybees as crop pollinators derives largely from their numerical dominance as crop visitors .Our analysis adds robust evidence to a growing consensus that wild bees have the potential to contribute greatly to agricultural pollination. Indeed, wild bee species richness, functional diversity, and visit rates increase crop yield , and the use of managed honeybee hives might not compensate for losses in wild bee species richness and abundance .
For example, increases in honeybee visitation only occasionally increase crop pollination whereas wild insect visitation universally increases fruit set . As such, managed honeybees alone may be insufficient to meet the increased pollination demands of global agricultural production and our results validate the importance of actions to promote resilient native bee communities within agricultural lands . Honeybees were equally effective as pollinators of plants inside and outside of their native range and were less effective compared to the most effective other bees in both regions . This result is not entirely surprising based on what we know about the co-evolution of plants and pollinators. The non-honeybee bee community may contain specialists sympatric with their host plants. Meanwhile, if honeybees are broad generalists, selective pressure might be less consistent, even within the native range of honeybees. Furthermore, if the morphological features relevant to pollination are relatively consistent across plants within the same genus or family, insects may be capable of pollinating novel plant species. For example, Prunus spp. occur in Europe and North America and Osmia spp. are highly effective pollinators of Prunus tree crops in both regions , despite the fact that North American Osmia spp. do not have shared evolutionary history with the Prunus species introduced as tree crops.We found an overall positive relationship between visit frequency and single visit pollinator effectiveness, but this relationship was largely driven by data from systems in which honeybees were absent . The overall positive correlation suggests that more frequent visitors are also more effective, but this result should not be interpreted to indicate that visitation frequency is an adequate proxy for overall pollination importance . This positive correlation may suggest that pollinators which visit frequently do so to the exclusion of other plant species, such that they display high floral constancy. High floral constancy may indicate that visitors gather and transport more conspecific pollen . Although the pollen loads of visitors do not always adequately predict effective pollination , high conspecific pollen transport likely predisposes visitors to higher pollination effectiveness on average. Another possible explanation is that, drainage planter pot for pollen-collecting visitors, more frequent visitors could be more efficient at extracting large quantities of pollen and might therefore transfer more pollen depending on how well pollen is groomed. Addressing whether more frequent visitors transport more conspecific pollen or deliver fewer heterospecific pollen grains are ripe questions for further study. The finding that honeybees erode this otherwise positive correlation suggests that this hyper-generalist species is often a numerically dominant visitor with modest effectiveness and may modify the pollination context for plant communities. Interestingly, when comparing systems with and without honeybees, visit frequency and pollination effectiveness do not positively correlate even when we artificially remove the data on honeybees and re-calculate correlation coefficients. This result suggests that honeybee presence may indirectly influence the relationship between visitation frequency and pollination effectiveness by altering the visitation patterns and effectiveness of other plant visitors. High honeybee visitation frequencies may indicate that honeybees efficiently extract nectar and pollen without also efficiently depositing the pollen they extract . If honeybees deplete floral nectar, this could make plants less attractive to other common visitors and alter their visit behavior and effectiveness . If they extract large amounts of pollen , this could reduce the amount available for collection and deposition by other pollinators . Indeed, honeybees can outcompete and reduce visits from other pollinators, reducing wild pollinator abundance and the diversity of plant species visited by non-Apis species .
Honeybee competition can also decrease interaction diversity by causing pollinators to become more specialized . Such changes in plant-pollinator interaction patterns can ultimately reduce the reproductive success of plants species frequently visited by honeybees and change the pollination context for other species. There are several potential limitations of our study and possibilities for future work. First, we only included measures of female reproductive success in assessing pollination effectiveness . The proportion of extracted pollen that is successfully transferred to stigmas may be a better assessment of the overall reproductive contribution of different taxa , because pollen that is removed but not successfully transferred represents a loss to male fitness . Unfortunately, data on such transfer dynamics are much rarer in the literature. Second, there are likely other factors about plant and pollinator taxa that moderate the effects we observe but which we do not test in this study, for example, functional traits such as plant and pollinator specialism. We hope our study will motivate other researchers to pair our data with trait databases and information on single visit pollen removal to further investigate the factors that influence effective pollination.Managing urban runoff and its associated pollutants is one of the most challenging environmental issues facing urban landscape management. The conversion of naturally pervious land surfaces to buildings, roads, parking lots, and other impervious surfaces results in a rapid surface runoff response for both time of concentration and peak flow. Impervious land surfaces adversely impact the quantity and quality of surface runoff because of their effects on surface water retention, infiltration, and contaminant fate and transport. Large volumes of storm runoff from urbanized areas cause flooding, sewer system overflows, water pollution, groundwater recharge deficits, habitat destruction, beach closures, toxicity to aquatic organisms, and groundwater contamination. Traditional urban runoff management focuses on removing the surface runoff from urban areas as soon as possible to protect public safety. However, as excess surface water is quickly drained from urban areas, it is no longer available for recharging groundwater, irrigating urban landscapes, sustaining wildlife habitat and other uses. Green infrastructure uses natural or engineered systems that mimic natural processes to control stormwater runoff. For example, traditional detention ponds have been widely used to treat storm runoff and permeable paving promotes infiltration of rain where it falls. Importantly, decentralized green infrastructure strategies control runoff and contaminants at their source. Vegetation is a green infrastructure strategy that can play an important role in surface runoff management. Large-scale tree planting programs have been established in many cities to mitigate the urban heatisland effect, improve urban air quality, and reduce and treat urban runoff . There are municipal stormwater credit programs in a growing number of cities that promote retaining existing tree canopy, as well as planting new trees. Although these programs encourage planning and management of urban forests to reduce runoff impacts, fertilizer is required to promote plant growth, and these added nutrients may contribute to contamination of surface runoff . Thus, reducing nutrients in storm runoff is a challenging task for landscape and water managers. Bioswales are shallow drainage courses that are filled with vegetation, compost, and/or riprap. As a part of the surface runoff flow path, they are designed to maximize the time water spends in the swale, which aids in the trapping and breakdown of certain pollutants. Bioswales have been widely recognized as an effective decentralized stormwater BMP to control urban runoff. Their effects are threefold; vegetation intercepts rainfall reducing net precipitation; plant uptake of water via transpiration reduces soil moisture, thereby increasing subsurface water storage capacity, and root channels improve infiltration. New bioswales are being developed for harvesting surface runoff and supporting urban tree growth. Bioswales that integrate engineered soil mixes and vegetation are being used to enhance treatment and storage of surface runoff.