Assessing the role of dispersed floral resources for managed bees in providing supporting ecosystem services for crop pollination

Study system

The Cape managed honeybee system in South Africa can be defined by the biogeographic distribution of the Cape honeybee (Apis mellifera capensis) (Fig. S1, Hepburn & Crewe, 1991; Hepburn & Guillarmod, 1991; Hepburn & Radloff, 1998). This endemic species has adapted to the local environmental conditions and co-evolved with the native flora. Its distribution predominantly overlaps with the floristically rich Fynbos biome but also includes some areas of the Nama-karoo, Succulent Karoo, Albany Thicket and Forest Biomes (Mucina & Rutherford, 2010). Within this system, we focused on the Western Cape deciduous fruit industry, an industry valued at US$ 688 million per year (Greef & Kotze, 2007; Hortgro, 2012; Melin et al., 2014) and that is largely dependent on managed honeybees for pollination services (Allsopp, De Lange & Veldtman, 2008; Melin et al., 2014). This service is supplied by an estimated 30,000 managed hives (Allsopp & Cherry, 2004). Within this system, beekeepers move their hives over hundreds of kilometres to access floral resources (for honey production, comb build-up, overwintering and swarm trapping) or to deliver pollination services. They utilise a range of floral resources, including natural vegetation (fynbos), stands of introduced trees (eucalypt), agricultural weeds (e.g., Echium plantagineum) and cultivated crops (e.g., canola, citrus, clover, and lucerne), and stock their apiary sites with between five and 180 hives, depending on the type and extent of floral resource being used. This study focused on three general floral resource-types (natural vegetation, eucalypt, and crops/canola, see Table 1) because they are considered most important by commercial beekeepers (Melin et al., 2014; Johannsmeier, 2001) and spatial data were available for these resource-types. These three floral resources become available over a temporal mosaic across the Western Cape. Eucalypt typically becomes available during summer months (November-February), when there is a shortage of alternative floral resources (see Fig. 1 in Melin et al., 2014). Fynbos, available all-year-round but fluctuates with rainfall-seasonality, is typically targeted in winter months (April–August). Canola typically flowers over the spring months (July–August). From year to year, the beekeeper may adjust the time hives spend at an apiary site depending on the quality of the floral resource and the requirements of the colony.

Beekeeper interviews and mapping apiary sites

Beekeepers were selected for one-to-one interviews using purposive sampling (De Vaus, 2002), in which respondents were selected because they were recognised as large commercial beekeepers (managing more than 600 hives) who provide pollination services to the Western Cape deciduous fruit industry. Respondents were identified using a snowballing technique (De Vaus, 2002) whereby beekeepers interviewed were asked to recommend the next beekeeper. The aim was to obtain a geographically representative sample of 6,000 managed hives representing approximately 20% of all hives used for pollination services in the Western Cape Province.

Six beekeepers were interviewed between June 2012 and October 2013. During the interviews, each of the beekeeper’s apiary sites were mapped using digital topographic maps (obtained from National Geo-Spatial Information, South Africa) with a scale of 1:50,000 in ArcGIS 10.1 (ESRI, Redlands, CA, USA).

Semi-structured interviews (Milner-Gulland & Rowcliffe, 2007) comprising open-ended questions, were used to obtain the following operational data: total number of hives; GPS coordinates of each apiary site; number of hives at each apiary site; the floral resources used and availability of floral resources (flowering time).

Assessment of floral resources that beekeepers indicated as being used

Information was obtained for 6,700 managed hives, representing approximately 22% of the managed hives in the Western Cape Province. Data for 708 apiary sites (ranging from 26 to 364 apiary sites per beekeeper) was captured across a wide geographic distribution of the Cape managed honeybee system (Fig. S1).

For each beekeeper, a random selection of 20 apiary sites (giving a total of 120 sites) was chosen for all further analyses. We tested whether the selected subset of apiary sites was representative of the full set of apiary sites by carrying out a Pearson’s chi-squared test. There was no significant difference in the proportion of resource-types between the full dataset and the subsample (χ² = 0.29, df = 2, p = 0.86).

Bee Landscape Unit (BLU) definition and floral resource classification

The landscape surrounding each site (n = 120) was classified into a BLU (Fig. 2). A BLU is defined here as the vegetation within a two km radius of an apiary site. Each BLU comprised 1256 ha and we classified the available floral resources within this area into three categories (natural vegetation, eucalypt, and canola fields; Table 1). A radius of two km is consistent with other similar studies (Steffan-Dewenter & Kuhn, 2003; Gallant, Euliss & Browning, 2014; Henry et al., 2012b) and takes into account a range of factors that determine honeybee foraging distance, including honeybee health, complexity of the landscape, floral resource availability and climatic conditions (Gallant, Euliss & Browning, 2014; Steffan-Dewenter & Kuhn, 2003; Beekman & Rathnieks, 2000; Nguyen et al., 2009; Visscher & Seeley, 1982; Waddington & Holden, 1979; Ghazoul, 2005; Steffan-Dewenter & Tscharntke, 2000; Steffan-Dewenter, Münzenberg & Bürger, 2002). However, most of these studies were based in the Northern Hemisphere and were mainly inferred from indirect methods such as translocation experiments, mark recapture experiments or pollen analysis (see Table 1 in Zurbuchen et al., 2010). The radius of two km, therefore, may not be ideal for natural vegetation (e.g., fynbos) alone but it is a reasonable and logistically practical estimate. The BLUs were mapped in ArcGIS 10.1 using multiple spatial data sources and approaches (Methods S1). It was not logistically possible to ground-truth the large number of apiary sites in this study (n = 120) and over such an extensive area (>150,000 ha). However, we used expert knowledge about the landscape surrounding the apiary site from the beekeeper to guide the digitisation of floral resources.

In finalising the BLU classification, areas unlikely to provide pollen and nectar resources for honeybees such as major water bodies (e.g., dams) and all other agricultural types (e.g., viticulture) were excluded. In addition, the BLU analysis is unlikely to be very effective for assessing floral resource availability of weeds or private gardens (e.g., in urban areas and rural farmhouses) which are patchily distributed and because of the lack of suitable land cover data to estimate landscape availability.

Data analysis

The analyses were based on three datasets: first, a table giving the number of times (counts) a floral-resource category was specified by each beekeeper (beekeeper dataset; Data S1); secondly, a table giving area in hectares of floral resources potentially available to bees, cross-classified by beekeeper and floral-resource category (landscape dataset; Data S2); and thirdly, a table was derived from the area of available resource-type adjusted according to the number of hives associated with the apiary sites (hive adjusted dataset; Data S3).

Floral resource use as indicated by beekeepers and BLU-analysis

The network figure (Fig. 3A) shows clearly that beekeepers in the Cape managed honeybee system say they are using proportionally more eucalypts (50%), than natural vegetation (37%), and canola fields (13%) (Fig. 3A).

Analysis of the resources available to beekeepers within BLUs provided a significantly different perspective to that provided by beekeepers (χ2 = 2237.60, df = 2, p < 0.001). Eucalypt was the least available resource (mean across all beekeepers = 168 ha; range 116–264 ha), as they typically occur in small stands across the region whereas natural vegetation (mean across all beekeepers = 11,221 ha; range 6,855–14,806 ha) was most common (Fig. 3B).

When the availability of different floral resources was adjusted according to the number of hives at an apiary site (hive adjusted dataset), eucalypt was found to be of primary importance (Fig. 3C). This finding was supported by the comparison of the BLUs with the regional average (T = 59.05, p < 0.001). This significant result was confirmed by Fig. 4A with the regional mean of eucalypt not falling within the confidence intervals of the BLU sample mean.

When analysing the beekeeper dataset against the hive adjusted dataset we find that these are significantly different (χ² = 1218.10, df = 2, p < 0.001), particularly with regards to canola (Fig. 3C). As with eucalypt, beekeepers used significantly more canola (Fig. 4B, T = 59.05, p < 0.001) than the regional average.

In contrast, natural vegetation was used significantly less than the regional average despite it making up a large proportion (Fig. 3B) of the BLUs (T = 46.16, p < 0.001). When the landscape availability of natural vegetation was adjusted relative to hive number, its importance was reduced (Fig. 3C).

Based on the composition of BLUs, the population of managed honeybees in the Western Cape uses the following areas of floral resources: 700,818 ha of natural vegetation (13% of study system), 10,485 ha of eucalypt (0.2% of study system), and 73,910 ha of canola fields (1.4% of study system) (Table 2).

Floral resources required to support managed pollination services

One rationale for studies of ecosystem services is to identify and evaluate those natural assets that are required to provide services to people (Millennium Ecosystem Assessment, 2003; Daily, 1997) and various models and studies of pollination services have highlighted the value of natural vegetation adjacent to crop fields (Ricketts et al., 2008; Kennedy et al., 2013; Garibaldi et al., 2011; Kremen et al., 2007). Assessing the services required to support managed honeybees can illustrate the value of landscapes that are not contiguous with crops. This study showed that the largest estimated floral resource area was natural vegetation, which also formed a major part of the floral resource mix in BLUs. Nevertheless, results of surveys showed that beekeepers placed less value on natural vegetation and field data showed there were fewer hives at apiary sites where natural vegetation predominated. Beekeepers value eucalypts for honey production and canola as it allows colonies to build-up strength prior to pollination services (Melin et al., 2014). Beekeeper operational responses reinforced the view that natural vegetation can support fewer hives than alternative floral resources. Baseline data on the carrying capacity of natural vegetation to support managed honeybees for pollination services in unknown. Floral abundance within natural vegetation, such as fynbos, fluctuates seasonally (Johnson, 1993; Hepburn & Guillarmod, 1991) and there may be relatively little flowering at certain times despite the large area of natural vegetation within BLUs. The general perception in the beekeeping literature is that natural vegetation, such as fynbos, is only marginally productive for commercial beekeeping in South Africa (Johannsmeier, 2001) but may be targeted for honey production (Johannsmeier, 2005; Melin et al., 2014). In the current mix of floral resources, natural vegetation plays a supplementary role when beekeepers have access to eucalypts but what is not clear from this study is whether natural vegetation provides critical resources at times of year when other resources are not available. This requires more detailed studies of bee behaviours at apiary sites (Hawkins et al., 2015; De Vere et al., 2017; Steffan-Dewenter et al., 2017). It should also be noted that in other systems where managed honeybees occur, natural vegetation may be more productive than fynbos (for beekeepers) so that supporting services provided by vegetation that is separated from crop fields should still be evaluated as an ecosystem service.

Human-modified landscapes played a significant role in sustaining a large number of hives. The relative importance of these different floral resources therefore needs to be considered when estimating the landscape components required to provide the supporting ecosystem services for managed honeybees. Our study could only give an approximate indication of the floral resource requirements to support managed pollination services and further refinement through fine-scale field-based research is needed to determine what honeybees are using in the landscape (e.g., see Russo et al., 2013 for detailed plant–pollinator phenological assessments at the landscape-level). The estimate for each floral resource-type could be improved by incorporating floral cover (e.g., relationship between plants size and the number of flowers) and the amount of nectar/pollen provided per flower per plant (Williams et al., 2011; Saifuddin & Jha, 2014). The flowers that honeybees choose to forage on may not correspond with the beekeeper’s choice. An analysis of pollen loads collected by honeybees would provide essential information on how honeybees are actually using the landscape and if beekeepers are making choices that allow honeybees to save energy by placing hives close to preferred floral resources.

Of particular interest is that the area of eucalypt needed to support the current number of hives for managed pollination services is estimated at a little over 10,000 ha. This is probably an under-estimate due to the exclusion of riverine species of eucalypt in BLUs, which could not be digitised with any confidence. The results are especially relevant to regulations regarding invasive Eucalyptus species and invasive species management programmes which coordinate ongoing clearing of eucalypts in the Western Cape Province (Van Wilgen et al., 2012).

It was surprising that beekeepers ranked crop floral resources (different to the target crop requiring pollination) as the least important resource. Beekeepers are thought to be shifting towards a greater use of crops such as canola to “boost” managed honeybee populations before hives are moved to farms for pollination (Melin et al., 2014) and the estimated use of canola fields (74,000 ha) was almost the same as the currently estimated area of production of the crop in the Western Cape (78,050 ha) (Crop Estimates Committee, 2015). However, beekeepers expressed concern about the use of pesticides on canola. This was also noted in other beekeeper surveys in the region (Allsopp & Cherry, 2004) and is supported by mounting scientific evidence that certain pesticides are harmful to bees (Pettis et al., 2013; Staveley et al., 2014; Nguyen et al., 2009; Van der Sluijs et al., 2015; Mao, 2013; Henry et al., 2012b).

An important question to consider in terms of overall sustainability is what would happen if particular “supporting” resources, such as eucalypt was removed or lost. Would this constrain crop production in the Western Cape or could hive numbers be sustained on other resources such as natural vegetation despite its seemingly lower carrying capacity for managed hives. It is not clear how and if beekeepers would adapt to changes in floral resource availability, i.e., would beekeepers switch to alternative resources or would such losses result in decreased profitability and ultimately closure of their business? Conducting detailed behavioural studies to investigate beekeeper’s responses to changing resource availability could provide insights into the future direction of the industry and how it relates to the provision of pollination services. The analyses in this study do not fully answer this question and it is clear that we are only beginning to understand the importance of different floral resources for managed honeybees in South Africa and other regions that are heavily reliant on managed pollinators.

The results of this study are not unique to the South African context. Many agricultural systems rely on managed honeybees (Klein et al., 2007; Potts et al., 2010) and beekeepers need to move their hives to access floral resources (e.g., Beekman & Rathnieks, 2000; Oldroyd, 2007), which may be separated from the crop field where they provide pollination (Gallant, Euliss & Browning, 2014). The ideas tested here therefore apply generally to systems where managed honeybees are important for crop pollination. As such, landscape models which seek to understand and provide a general framework for the conservation of pollination services, would need to consider the incorporation of floral resources away from the target crop.

Inclusion of supporting services brings in additional complexity to the management of ecosystem services for pollination because the managers of land where supporting services for pollinators are provided may not derive any direct benefit. In addition, the managed bees being supported at remote locations may provide pollination services to multiple crops. As a result, the management of floral resources for managed bees may benefit from studies of other ecosystem services such as water provision where there is a need to incentivise appropriate land management in water catchments to provide water to downstream users (Stosch et al., 2017).

An additional complexity that would need to be considered in landscape models incorporating managed honeybees is the importance of both natural and human-modified landscapes. One of the main conclusions of our study is that human-modified components of the landscape (eucalypt stands) can play an integral role in supporting managed honeybees (Melin et al., 2014; De Lange, Veldtman & Allsopp, 2013; Johannsmeier, 2001). Again, this result is not unique within the South African context but adds to a growing body of literature that shows how human-modified landscapes benefit both managed pollinators (Rollin et al., 2013; Sponsler & Johnson, 2015) and wild pollinators in the absence of natural vegetation (Samnegárd, Persson & Smith, 2011). How one balances the value of these “novel ecosystems” with conservation programmes (e.g., invasive species management programmes) is much debated but options for management of these systems are available (Murcia et al., 2014; Hobbs et al., 2014; Hobbs, Higgs & Harris, 2009; Hobbs, Higgs & Harris, 2014). In most studies to date, the main focus has been on contribution from natural habitats (Samnegaard et al., 2011). Some models using managed honeybees have also considered the importance of “indirect” ecosystem services provided by natural vegetation away from where the pollination service is delivered (Mouton, 2011). A critical result of our study, however, is the recognition that several different floral resources provided by natural and/or human-modified landscapes may be spatially dispersed across a region and the way in which beekeepers manage hives determines the landscape use of managed pollinators (Fig. 1).