The potential of electricity transmission corridors in forested areas as bumblebee habitat

Declines in pollinator abundance and diversity are not only a conservation issue, but also a threat to crop pollination. Maintained infrastructure corridors, such as those containing electricity transmission lines, are potentially important wild pollinator habitat. However, there is a lack of evidence comparing the abundance and diversity of wild pollinators in transmission corridors with other important pollinator habitats. We compared the diversity of a key pollinator group, bumblebees (Bombus spp.), between transmission corridors and the surrounding semi-natural and managed habitat types at 10 sites across Sweden's Uppland region. Our results show that transmission corridors have no impact on bumblebee diversity in the surrounding area. However, transmission corridors and other maintained habitats such as roadsides have a level of bumblebee abundance and diversity comparable to semi-natural grasslands and host species that are important for conservation and ecosystem service provision. Under the current management regime, transmission corridors already provide valuable bumblebee habitat, but given that host plant density is the main determinant of bumblebee abundance, these areas could potentially be enhanced by establishing and maintaining key host plants. We show that in northern temperate regions the maintenance of transmission corridors has the potential to contribute to bumblebee conservation and the ecosystem services they provide.

IB, 0000-0001-7893-4389 Declines in pollinator abundance and diversity are not only a conservation issue, but also a threat to crop pollination. Maintained infrastructure corridors, such as those containing electricity transmission lines, are potentially important wild pollinator habitat. However, there is a lack of evidence comparing the abundance and diversity of wild pollinators in transmission corridors with other important pollinator habitats. We compared the diversity of a key pollinator group, bumblebees (Bombus spp.), between transmission corridors and the surrounding semi-natural and managed habitat types at 10 sites across Sweden's Uppland region. Our results show that transmission corridors have no impact on bumblebee diversity in the surrounding area. However, transmission corridors and other maintained habitats such as roadsides have a level of bumblebee abundance and diversity comparable to semi-natural grasslands and host species that are important for conservation and ecosystem service provision. Under the current management regime, transmission corridors already provide valuable bumblebee habitat, but given that host plant density is the main determinant of bumblebee abundance, these areas could potentially be enhanced by establishing and maintaining key host plants. We show that in northern temperate regions the maintenance of transmission corridors has the potential to contribute to bumblebee conservation and the ecosystem services they provide.  network is owned, maintained and operated by Svenska kraftnät (SK), a state-owned public utility. SK's transmission corridors are subject to an easement that allows them the perpetual right to construct, keep and maintain the transmission corridor grid irrespective of the underlying land tenure. In the Uppland region, transmission corridors are maintained on an eight-year cycle. In year zero, transmission corridors are cleared of tall vegetation; in year three, trees threatening transmission lines are removed; in year four, transmission corridor access roads are cleared and in year seven, fast growing trees are felled. SK's maintenance is conducted by mechanical means (J Bjermkvist 2014, personal communication, SK).
To investigate the influence of transmission corridors on the surrounding area, we selected 10 sites of 4 km 2 (2 × 2 km) in Sweden's Uppland region (electronic supplementary material, figure S1). To minimize landscape composition confounding our results, we ensured that (i) all sites had at least 45% forest cover (range 45-70%); (ii) that the second most common land use was agriculture, and (iii) that all target habitats were represented (see table 1 for habitat description). Sites were between 3.2 and 6.4 km apart. There can be a wide variation in foraging distances between bumblebee species, with radiotracked Bombus terrestris (L, 1758) and Bombus ruderatus (Fabricius, 1775) workers foraging up to 2.5 km and 1.9 km, respectively, from their nests [43], while Bombus muscorum (L, 1758) has a much smaller foraging range of between 100 and 500 m [44]. Therefore, the distances between our sites minimized the chance that bumblebees recorded in one site were also recorded in another. Five sites were bisected by a transmission corridor section (widths ranging between 50 and 70 m), of which between 1.2 and 1.5 km was bordered by closed canopy forest. At the time of surveying, four sites were in year three of their maintenance schedule (all the tall vegetation was removed in 2011) and the remainder was in year six (all tall vegetation was removed in 2008). All corridors ran from north/northeast to south/southwest. The other five sites were at least 3 km from any other transmission corridors.
To capture the variability among the surveyed habitat, we conducted multiple transects per site in each habitat (mean of 2.25 transects per habitat and site). Some sites had no representation of particular habitat types. Overall, we surveyed 158 transects spread across seven habitat types (table 1; see photos in electronic supplementary material, figure S2). These habitat types were transmission corridors, semi-natural grasslands, maintained roadsides (hereafter roadsides), forest/semi-natural grassland boundaries, cereal crop edges, maintained ditches and forests. All these habitats, except forests, have been identified as valuable bumblebee habitat in the Uppland region [17]. Roadsides, ditches and crop edges were always embedded in grassy or shrubby areas and forest transects were often near clearings or trails.
To the best of our knowledge, none of the surveyed transects were in areas that had been ecologically enhanced. The surveyed roadsides (all quiet tertiary or quaternary roads) are mown once annually (M Lindqvist 2014, personal communication, Trafikverket) whilst ditches are maintained on an as-needed basis. The semi-natural grasslands surveyed met the EU's definition of permanent pasture and grassland [45].
Each transect included an area 50 m long and up to 3 m wide. All transects contained a representative density of flowering plants. Within each transect, we surveyed bumblebee abundance and diversity by slowly walking along the transect for 15 min (a method recommended in [46]). Transects were walked twice (back and forth) but always keeping the area surveyed and the survey time fixed.
Where possible, bumblebees were identified while foraging, but most individuals could not be readily identified on the wing and therefore, were caught by net, identified and released if possible. Caught specimens that were not identified in the field were killed then identified later. Owing to the difficulty distinguishing B. terrestris and Bombus lucorum (L. 1761) workers, all specimens were combined as B. terrestris [26]. Both species are common, extremely difficult to distinguish and are often grouped as they are ecologically similar. Hence, this grouping does not affect our distinction between ecosystem service providers and species of conservation concern. Collection handling time was not included in the 15 min survey time.
When possible, the host plant of each foraging bumblebee was identified to species level during the survey, otherwise plant specimens were identified later. To correspond with peak bumblebee activity in the Uppland region [17] each site was surveyed twice between 9 July 2014 and 25 August 2014, with at least two weeks between surveys. Each survey took 1 day and was undertaken between 9.00 and 17.30 and only during dry periods in temperatures above 15°C. Transects in transmission corridors were always in unshaded areas. Before beginning each survey within the respective transect, flower density was estimated as the total percentage of the transect area covered by flowers. The categories used were '<1%', '1-5%', '6-10%', '11-20%', '21-40%', '41-60%' and '>61%' coverage. Because all surveying was conducted by one person, this semi-quantitative measure enabled a quick yet consistent assessment of the flower density in all transects.

Statistical analysis
To compare species abundance and richness (alpha diversity) across sites and habitats, we built a generalized linear model (GLM) with species richness or abundance per transect as a function of site type (transmission corridors/no transmission corridor) and habitat type. Flower density was also included as a covariable. To account for the hierarchical structure of the data, transect nested within site was included as a random factor. Residuals were investigated to ensure they fulfilled the model assumptions and to meet the postulation of homoscedasticity we used a constant variance function. All models (see also below) were constructed using package nlme [47] in R [48]. The statistical power of the models to detect a 20% difference was calculated using package Simr [49].
Beta diversity was analysed on two scales. Firstly, we investigated if sites containing a transmission corridor had lower turnover rates among the different habitats. Secondly, we investigated beta diversity among different sites of the same habitat. To determine species turnover, we used additive partitioning of species richness [50][51][52][53]. Alpha diversity was defined as the mean number of species per transect (i.e. species richness). The beta diversity among sites with and without transmission corridors was calculated as the total number of species found within a transmission corridor site (gamma diversity) minus the mean number of species per transect on that transmission corridor site (alpha). Beta diversity among habitats was calculated as the rarefied number of species found across all transects of a given habitat type (gamma) minus the mean number of species per transect surveyed for that habitat type (alpha). Rarefication in gamma diversity was undertaken to 90 individuals to avoid difference in sampling intensity across habitats using the package vegan [54] (electronic supplementary material, figure S3).
From the recorded set of bumblebee species, we determined which habitats were used by bumblebees listed as threatened in Europe by the IUCN [14] (B. muscorum) and species listed as declining by Scheper et al. [16]. These included Bombus humilis (llliger, 1806), Bombus sylvarum (L, 1761) and Bombus soroeensis (Fabricius, 1777) and are hereafter termed 'threatened species'. We also recorded which habitats were used by the species that are the main providers of crop pollination in Europe: B. terrestris, Bombus lapidarius, Bombus pascuorum (Scopoli, 1763), Bombus hypnorum (L, 1758), Bombus pratorum (L, 1761) and Bombus hortorum (L, 1758) [6], and are hereafter termed 'provider species'. We constructed a GLM with abundance of both threatened species and provider species per transect as a function of habitat and flower density. Transect nested within site was also included as random factor. To meet the model assumptions of homoscedasticity, we used a constant variance function.
Finally, to assess the importance of each host plant species for every recorded bumblebee species in the surveyed habitats, we calculated the plant species' strengths [55] for the pool of transects of transmission corridor habitats, semi-natural grassland habitats and all habitats combined. For each plant, strength is defined as the sum of all pollinators' dependencies on that given plant. Pollinator dependence is the fraction of all pollinator recorded visits performed on that given plant species. Therefore, a plant species could have high strength values if it attracted many pollinator species that had low dependency on it, or if it attracted few pollinators which were highly reliant on it. Note that this metric measures plant species use, not preference; a plant species could be visited by a given pollinator simply because it was the most abundant, not because it was preferred.

Cost of managing and/or enhancing roadsides, semi-natural grasslands and transmission corridors
These managing costs were gathered from EU member material [56][57][58]    the initial currency in which these costs were originally stated. Therefore, no attempt was made to adjust these costs to inflation or currency fluctuations. Consequently, to enable an approximate comparison of these costs, all are expressed in euros per hectare per annum, with the conversion of the original currency to euros being carried out in June 2015.

Results
In total, we recorded 1016 bumblebee specimens, comprising 20 species (electronic supplementary material, table S1). These were recorded foraging on 24 plant species. Transmission corridor bisecting a site did not change bumblebee abundance (table 2 and figure 1a) or species richness (table 2 and  figure 1b). Similarly, we found no differences among habitats in terms of total bumblebee abundance or species richness (table 2 and figure 2a,b). As we predicted, flower density was the strongest predictor of bumblebee abundance and richness (table 2). While the power to detect a 20% difference among sites that were bisected and not bisected by a transmission corridor is low (power ranges from 19% for abundance model to 31% for richness model), our power to detect a 20% difference between semi-natural grasslands and transmission corridors is higher (67% for the abundance model; 89% for the richness model).
Patterns of species beta diversity reveal that sites bisected by a transmission corridor did not have more homogeneous species composition compared with sites not bisected by a transmission corridor (test for differences in beta diversity: n = 10, F 1,8 = 0.03, p = 0.85; figure 1b). We also found that species turnover among transects of the same habitat was similar, with all habitats having between 11 and 15 rarefied species (i.e. gamma diversity; figure 2b).
We found that provider species were present in most habitats. B. pascuorum and B. terrestris were present in all habitats and were also the most abundant, while B. lapidarius was found in all habitats except forest. Overall, the abundance of provider species was not different across habitats (figure 3a and  table 3). Interestingly, threatened species were not limited only to semi-natural grasslands (B. sylvarum and soroeensis), but were also found in roadsides (B. humilis, soroeensis and sylvarum) and transmission corridors (B. muscorum and humilis). However, threatened species were rarely found in the other habitat types (figure 3b and table 3). Flower density did not explain threatened species abundance (table 3).
Throughout all the sites Carduus crispus (L., 1753), Trifolium pratense (L., 1753) and Centaurea jacea (L., 1753) were the most important host plants for sustaining both threatened and provider species (table 4 and figure 4). However, the importance of plant species measured as its strength varied between transmission corridors and semi-natural grasslands. For example, due to their abundance, species in the genus Trifolium were more important in semi-natural grasslands than in transmission corridors. Overall, important plant species sustained both bumblebee species that were not overly reliant on them and threatened species (e.g. B. sylvarum, B. humilis: figure 4).
There was a large range in the costs of maintaining and/or ecologically enhancing transmission corridors, roadsides and semi-natural grasslands. The current maintenance of transmission corridors in Uppland costs approximately e60 ha -1 yr -1 (J Bjermkvist 2014, personal communication). Mowing Uppland roadsides similar to those surveyed costs between e500 and 1000 ha -1 yr -1 (M Lindqvist 2015, personal communication). In comparison, the EU funding of Swedish AES for semi-natural grassland maintenance and enhancement, depending on inputs ranges between e121 and 506 ha -1 yr -1 [59]. Where funding is awarded, implementation of the AES is only required for 5 years [59].

Discussion
We found that SK's current maintenance regime resulted in transmission corridors having bumblebee abundance and diversity equivalent to that in semi-natural grasslands. This supports the increasing recognition of transmission corridors as valuable wild pollinator habitat [17,60].  Table 3. Abundance differences across habitats for ecosystem service providers and threatened species. While provider species mirror the general abundance pattern, for threatened species we found habitat differences, but flower cover is no longer significant. transmission corridors providing bumblebee habitat equivalent to that supplied by semi-natural grasslands.
The fact that both transmission corridors and roadsides can sustain high numbers of bumblebees is remarkable because it indicates that they are suitable habitat, specially for threatened species. Eighteen of the 41 bumblebee species in Sweden are in decline and seven more are threatened with extinction [14] and the area of semi-natural grasslands in Sweden is estimated to be less than 10% of what it was one century ago [61]. Hence, areas of transmission corridors in forested areas could provide some mitigation to the loss of semi-natural grasslands for these species.
Roadsides also provided valuable habitat for threatened and provider species, with numbers of individuals per transect in both groups ranking higher than semi-natural grassland and forest/grassland boundaries. Roadsides tended to have high flower cover (30% density on average) which is similar to that of semi-natural grasslands. Maintained ditches and cereal crop edges also had flower coverage similar to transmission corridors (13-20%), but sustained fewer bumblebee individuals, particularly those of threatened species. Dense grass swards were observed in many of the maintained ditches. These swards possibly limited the habitat available for the favoured host species such as T. pratense, which are light demanding and low growing [62]. Overall, cereal crop edges were the narrowest habitat, with some being less than or equal to 1 m wide, and hence provided the least suitable area for host plants. As forested areas of tall evergreen trees (predominantly Pinus sylvestris (L., 1753) and Picea abies (L. 1753)) had little flower cover (average of 5% density), it is not surprising that this habitat type hosted few bumblebees.
In comparison, transmission corridors and roads bisecting those forest patches were flower rich and may have an aggregation effect, concentrating pollinators into these resource-rich areas [26]. However, it is important to note that flower density did not explain threatened species abundance, which suggests other factors, such as nesting sites, may be more limiting for these species [26]. It is not known what the effects of electrical and magnetic field radiation from high-voltage powerlines have on bees [36] and quiet roads potentially represent a minor threat to bumblebees [33]. It is possible that these risks are countered by providing suitable habitat for rodents, thereby potentially increasing nesting availability for bumblebees using abandoned rodent cavities as nesting sites [63]. Similarly, roadsides often contain areas of withered grass and tussocks that are crucial for nesting sites [17].
Overall, our results do not indicate that transmission corridors enhance bumblebee abundance or species richness by increasing connectivity of non-forested habitats or by having a spill-over effect into surrounding habitats. However, with only 10 sites the power to detect such landscape effects in our dataset is limited. The intrinsic variability in bumblebee populations between years [64] suggests that long-term data in different boreal countries are needed to confirm our results.
Within transmission corridors the main host plants for bumblebees are mostly limited to small areas that are not dominated by shading shrubby vegetation (B Hill 2015, personal observation). Floral density is an important predictor of bumblebee diversity and abundance. The large areas of herbaceous vegetation and shrubs within transmission corridors could provide considerable potential to enhance bumblebee habitat. Such actions could also assist in providing the approximately 2% of flower-rich habitat within farmland that is required to maintain provider bumblebee species colonies [65].
Maintaining and enhancing the abundance of early flowering Salix species such as Salix caprea (L. 1753) is a way of potentially improving the quality of bumble habitat in transmission corridors. Early flowering Salix species provide critical forage for early emerging bumblebee queens and subsequently, successful colony establishment. It has been shown that more than 1000 m 3     influenced bumblebee abundance [17]. Flower abundance later in the season is also critical for late emerging species, because many of these are threatened [16]. In Sweden, bumblebees are mostly active up to early September, after which the new queens hibernate underground [17]. As we surveyed almost to this period, we assume that we captured the peak phenology of most bumblebee species, including the threatened species. For most bumblebee species, legumes and other nectar-rich flowers are a significant resource [62] and our results support this observation. Although we did not separate nectar and pollen foraging trips, it is likely that different plant species are important for different reasons. For example, while T. pratense is a rich source of nectar and pollen, most thistle species may be used only for nectar [62]. However, in comparison to semi-natural grasslands, the transmission corridors we surveyed had a lower abundance of key plants such as T. pratense. The sowing of nectar-rich flower seeds is a proven way of enhancing bumblebee abundance and diversity [28]. This is a possible means of enhancing bumblebee habitat in transmission corridors and would cost approximately e42 ha -1 yr -1 [58]. Suitable open areas include access roads as these are not dominated by shading shrubby vegetation, and the additional areas of bare earth exposed during their maintenance.

Carduus arvense
Increasing the amount of open habitat within transmission corridors is another potential way of increasing host plant habitat and consequently, bumblebee diversity and abundance [29,37,65]. Removal of existing shrubs on transmission corridors would cost approximately e14 ha -1 yr -1 [66]. Host plants might then naturally colonize these areas or seeds of suitable species could be sown.
Funding the enhancement of bumblebee habitat within transmission corridors could be an effective way to both benefit bumblebee conservation and increase the pollination services they provide. It might also augment the ecological value of these areas. Depending on the location, enhancing the ecological value of transmission corridors could be conducted in tandem with the protection of ecological focus areas as prescribed by the EU [45]. The opportunity cost of producing an ecological focus area via converting productive agricultural land to unproductive biodiversity-rich areas can be considerable. For example, winter wheat which is a major crop in Uppland region, can provide gross returns of between e565 and e1505 ha -1 [67,68]. The establishment and maintenance of biodiversity-rich areas within transmission corridors, like those studied here, would avoid any such opportunity cost. The permanence of transmission corridors in the landscape also means that any enhancement within these is likely to provide long-term benefits. Such actions might well aid in meeting the EU's Biodiversity Strategy to 2020 Target 2, as well as the 2020 headline target [20]. 1 However, areas of transmission corridors do not meet the EU's CAP, enabling definitions of either 'eligible hectare' or 'ecological focus area'. Therefore, funding via EU AES for the ecological enhancement of such areas is not currently possible [45].
Pollinator habitat within transmission corridors is spatially limited to certain areas. Moreover, we only tested for the effect of transmission corridors in forested landscapes. The ability of transmission corridors to sustain pollinators in non-forested landscapes is still unexplored. Consequently, transmission corridors cannot substitute AES, but can complement it. In other situations it has been shown that tailoring inputs for specific results is possible. Application of AES to simple resource-poor landscapes, e.g. croplands, had the greatest benefit to provider species, while applying AES in more complex landscapes provided more benefit to threatened species [69]. The widespread geographical extent of transmission corridors through many Northern Hemisphere landscapes provides valuable but yet to be fully exploited opportunities for bumblebee conservation. However, the benefit of transmission corridors for biodiversity other than bumblebees has not yet been explored.

Conclusion
Bumblebee abundance and diversity is threatened by many factors. Given both the intrinsic value of bumblebees and the ecosystem service they provide, actions are being taken to counter these threats. Studies, including ours have shown that the maintenance of transmission and other infrastructure corridors may unintentionally create valuable habitat for pollinators. Our study also shows that SK's current transmission corridor maintenance regime is a cost-effective way of producing such habitat when compared to other maintenance regimes. The permanence and extent of transmission corridors means that any wild pollinator habitat created due to their maintenance is likely to be present in the long term. There are simple, proven management practices to enhance bumblebee richness and abundance but further research is needed to evaluate and optimize conservation approaches. Funding is needed for such work. Any future reviews of the Europe 2020 Strategy, CAP, or similar policy may provide opportunities to promote incentives to enhance the valuable pollinator habitat provided by maintaining infrastructure corridors.