The Field Experiment

We assessed the impact of manipulating conventional grassland sward diversity and management on pollinator communities using a replicated field-scale experiment from May to September 2011. This was carried out at Rothamsted Research, North Wyke, Devon, UK (50°46′N 3°54′W). A split-plot design was used with four replicate blocks to investigate the effect of two treatments: sward diversity and sward management (Figs S1 and S2 in Supporting information). Two plot sizes were used 0·1 ha (grazed plots) and 0·07 ha (cut plots).

Sward diversity was manipulated by sowing three seed mixes: grass only, grass–legume and grass–legume–forb (Tables S1 and S2 for species lists). Species were selected from a review of the biodiversity and agronomic value of grassland species by Mortimer et al. (2006) as potential target species for agri-environment scheme seed mixes. Each sward diversity treatment was split into two subplots which were subjected to one of two management regimes: (i) grazing: grazing by cattle from April/May to early June, no grazing from early June to August, and moderate grazing by cattle from August to October (two animals per 0·1-ha plot); (ii) cutting: cut early June, grazing by cattle from late August to October.

Farm Surveys

To increase the spatial scale of our study, and to measure pollinator population-level responses, we investigated the effect of pasture plant species richness across ten independent farms (separated by at least 6·5 km) across south-west England, scattered north and south of Bristol and Bath (map Fig. S3). All were mixed farms, with arable crops and pasture. The farms were selected as they were used in a previous study (Macfadyen et al. 2009), and therefore, data on their management and pasture plant diversity were available. Two pasture fields per farm (20 fields in total) were selected based on their plant species richness to cover a gradient of diversity. Richness was measured using two 30 × 2-m transects where 1–10 species = ‘low’ diversity, 11–20 species = ‘intermediate’ diversity and 21–30 species = ‘high’ diversity. Abiotic factors and surrounding landscape features, as well as differing management, are potential causes of each pastures' plant richness. The pastures had similar management; all were grazed by cattle, used herbicides to spot-spray undesirable species, including Rumex, Cirsium and Senecio sp., and used nitrogen-based fertilizers. All fields had hedgerows, and field sizes are included in Table S4. Whilst detailed information on the surrounding landscape was not recorded, we know it is dominated by agricultural land with some woodland cover. Data on local pesticide use were not available; however, none of the farms were organic so pesticides are likely to have been used. Plant–pollinator surveys were carried out on each pasture five times between May and August 2012, following the same survey protocol as the field experiment (i.e. zigzag walking within a central 500-m² area for 25 min followed by 30 × 2-m² transects to identify the plant species and count floral units).

Phytometer Experiment

Plant phytometers were used to bioassay the pollination service of the pasture on each of the ten farms. During their flowering period, five individuals of each phytometer species, strawberry variety ‘Symphony’, Fragaria × ananassa, (Duch.): Rosaceae; broad bean variety ‘Sutton’, Vicia faba (L.): Fabaceae and red campion Silene dioica ((L.) Clairv.): Caryophyllaceae, were positioned at the margin of one pasture field in each farm in early June 2013. As red campion is dioecious, five male plants were put out on the farms in addition to the five female plants. Prior to the experiment, the strawberry and bean plants were grown from seed in greenhouses until flowering and red campion plants, which were approximately 3 years old, were stored in polytunnels. Phytometers were relocated to the farms whilst in bud.

Strawberry is a crop commonly grown in the UK. The plants are self-compatible and whilst both wind and self-pollination occur, cross-fertilization is favoured (Free 1993). A wide diversity of insects visit strawberry flowers due to their open structure (Dimou et al. 2008; Klatt 2013). Strawberries have increased weight and fewer deformities if insect-pollinated (Chagnon, Gingras & Deoliveira 1989; Free 1993; Dimou et al. 2008; Klatt 2013). Broad bean is another crop commonly grown in the UK. It has partial cross-fertilization with significantly higher seed numbers and weight when insect-pollinated (Free 1966; Aouar-Sadli, Louadi & Doumandji 2008). Broad bean has closed papilionaceous flowers that only pollinators with long and strong mouthparts can access, predominantly bees (Free 1966, 1993; Aouar-Sadli, Louadi & Doumandji 2008). Red campion is a wildflower present in hedgerows and woodlands in the study region. It is dioecious and requires insect pollination and seed-set is related to the amount and identity of pollen deposited on the stigmas (Montgomery, Soper & Delph 2010). It is pollinated by insects with long mouthparts including bees, hoverflies and butterflies (Charlton 2013).

Location of the phytometer plants was prioritized to ensure that surrounding features were similar between farms, for example hedgerows and tree cover. Plants were put in areas where disturbance by cattle/tractors was minimal. Wild specimens of the phytometer species were not found in close proximity. Chicken wire fences protected the plants from grazing animals, and the phytometers were left in the field for 2 weeks to allow pollination. Four plant–pollinator surveys were carried out in the centre of the associated pasture field during this period following the same protocol as the plot experiment surveys where a zigzag walk was carried out within a 500-m² area over 25 min and associated 30 × 2-m² plant transect surveys.

In late June, the phytometer plants were collected from the farms and kept in enclosed polytunnels to allow fruits to mature. Any new flower buds were removed. Strawberry fruits were picked when ripe and weighed. Mean fruit weight was calculated per farm. Each fruit was classed based on commercial deformity grades (European Commission (2007)) where fruits without or with slight aberrations were sorted into Class 1, whereas severe aberrations lead to Class 2 classification. Broad bean seed pods were collected at maturity, and the seeds were counted and weighed; mean seed count (per pod) and seed weight were calculated per farm. Bean plants were still young when harvested. Seed capsules of red campion were collected, the number of seeds per capsule was counted, and a mean was calculated per farm.

Analysis

Objective 1: The Impact of Sward Diversity and Management on Pollinator Functional Diversity and the Resulting Temporal Stability of Flower Visitation

In the field experiment, 4169 flower visitors were collected comprising 166 insect species: 12 bee species, 34 hoverfly species, 90 non-hoverfly Diptera species, 18 Coleoptera species and 2 Lepidoptera species (Table S5). In the 2012 farm surveys, a gradient of 9–36 plant species per 60 m² was recorded and 1530 flower visitors were collected, comprising 146 insect species: 15 bee species, 15 hoverfly species, 76 non-hoverfly Diptera species and 23 Coleoptera species (Table S6).

Pollinator functional diversity significantly increased as sward diversity increased in the field experiment (χ² = 125·57, d.f. = 1, P = 0·0052) (Fig. 1a). The difference was between grass only and grass–legume–forb plots (z = 3·61, d.f. = 7 P < 0·001) (Fig. 1a). There was no significant difference in pollinator functional diversity between cut and grazed management (χ² = 123·22, d.f. = 1, P = 0·13). No significant difference in pollinator species richness was found between sward types (χ² = 155·97, d.f. = 1, P = 0·069) or management type (χ² = 153·25, d.f. = 1, P = 0·099). Sward type did have a significant effect on pollinator abundance (χ² = 264·24, d.f. = 1, P = 0·026) being higher in grass–legume–forb plots than grass only plots (z = 2·98, d.f. = 7, P = 0·0081). Management type did not have any significant effect on pollinator abundance (χ² = 261·01, d.f. = 1, P = 0·072).

In the farm surveys, pollinator functional diversity was significantly positively associated with plant species richness (χ² = 14·542, d.f. = 1, P < 0·001) (Fig. 1b) as was pollinator species richness (χ² = 10·831, d.f. = 1, P < 0·001) and pollinator abundance (χ² = 9·178, d.f. = 1, P = 0·002).

In both the field experiment and farm surveys, the response variables, pollinator functional diversity, species richness and abundance, were collinear (r ≥ 0·77; Table S7). However, they responded differently to some of the treatments (e.g. pollinator functional diversity c.f. pollinator species richness) and so are considered separately.

It could be argued that high pasture plant species richness indicates management that is generally ‘sympathetic’ to biodiversity across the farm and so pollinator communities could have been responding to farm-scale rather than field-scale management. We used a Wilcoxon matched-pairs signed-ranks test where fields were paired per farm to test whether the fields differed in pollinator functional diversity, thereby removing the farm effect. There was a significant difference between fields of the same farm (V = 50, n = 10, P = 0·02); hence, variation in pollinator communities was not due to farm-scale management but to individual fields.

A significant linear negative relationship existed between the functional diversity of the pollinator community and the temporal variability of insect–flower visitation in both the field experiment (χ² = 21·70, d.f. = 1, P < 0·001) (Fig. 1c) and farm surveys (χ² = 11·86, d.f. = 1, P < 0·001) (Fig. 1d). Thus, as pollinator functional diversity increases, the temporal stability of flower visitation increases.

Objective 2: Which Grassland Plant Species have Disproportionately Positive Effects on Pollinator Abundance and Diversity?

In the field experiment, Taraxacum sp. (F. H. Wigg.) was the most important species for supporting pollinators (per floral unit), attracting 35% of all pollinator visits and 33% of all pollinator species, followed by Ranunculus acris (L.) and Cardamine pratensis (L.). Surprisingly, grass species including Alopecurus pratensis (L.) and Dactylis glomerata (L.) were commonly visited by pollinators (Fig. S4a, Table S3), predominantly by Diptera within the Syrphidae and Muscoidea families. To confirm that the insects were feeding on the pollen and verify the grasses as a protein source, a stratified random sample of 60 individuals of 23 of the Diptera species caught on the grasses in the field experiment were dissected. In 72% of cases, Poaceae pollen was present in the abdomen, suggesting the dietary importance of grasses to Diptera.

In the farm surveys, Cirsium arvense ((L.) Scop.) was found to be the most important species, attracting 17% of all pollinator visits and 18% of all pollinator species followed by Cirsium palustre ((L.) Scop.) and Crepis capillaris ((L.) Wallr.) (Fig. S4b). Taraxacum sp. floral units accounted for 0·02% of all floral units within the plots and C. arvense floral units accounted for 0·08% of floral units in the farm surveys. Therefore, the results are not necessarily a consequence of these species' abundance.

Objective 3: The Impact of Pasture Plant Species Richness on the Pollination of Crop and Wildflower Species

In the 2013, pasture surveys (associated with the phytometer experiment), 349 insects were collected comprising 72 insect species: 9 bee species, 17 hoverfly species, 36 non-syrphid Diptera species, 5 Coleoptera species and 3 Lepidoptera species. A gradient of 9–28 plant species per field was recorded per transect (30 × 2-m). With regard to seed/fruit production of the phytometer plants, 161 strawberries were harvested, 136 broad bean seeds were collected from 44 pods, and 39 280 red campion seeds were collected from 274 seed capsules.

Pasture plant species richness was significantly and positively associated with strawberry fruit weight (t = 2·86, d.f. = 9, P = 0·021, Fig. 2) and proportion of Class 1 fruits (t = 4·62, d.f. = 9, P = 0·002, Fig. 3). Plant species richness was also significantly positively associated with seed count per capsule of red campion (z = 2·79, d.f. = 9, P = 0·005, Fig. 4). For broad bean, there was no significant relationship detected between pasture plant species richness and seed count per pod (t = −1·28, d.f. = 8, P = 0·24) or seed weight (t = −1·43, d.f. = 1,7, P = 0·20). There was a significant positive relationship between pasture plant species richness and pollinator functional diversity (t = 4·031, d.f. = 1,8, P = 0·004) as in Objective 1. Pollinator functional diversity, richness and abundance were all responsible for enhanced pollination of the phytometer plants to varying extents (Table 1, Figs 2 and 3).

Can Higher Sward Diversity Enhance Ecosystem Functioning and Services?

In the field experiment, both legumes and forbs were needed to create a suitable ecological infrastructure to enhance pollinator functional diversity. Sward richness in the farm surveys was also positively associated with pollinator functional diversity. Complementarity in resource use of the more functionally diverse pollinator communities is a potential mechanism behind the lower temporal variability in flower visitation found at both scales. This has potential implications for a more temporally stable ecosystem service. The fact that relationships found in the field experiment held true in the farm-scale studies, where population responses were measured, supports the use of small-scale experiments with pollinators.

Increased pollinator functional diversity, species richness and abundance were associated with increased pollination of strawberry. As these pollinator community variables were correlated, it is difficult to determine the causative factor behind enhanced pollination. However, AIC values suggested pollinator functional diversity and richness to be equally effective in increasing strawberry weight, whilst richness appeared to have the biggest positive effect on strawberry quality (class) followed by abundance and functional diversity. Spatial complementarity of pollen deposition has been highlighted as a mechanism behind increased strawberry pollination and resulting quality; large and average-sized bees pollinate the apical stigmata, and small-sized bees pollinate the basal stigmata (Chagnon, Gingras & Deoliveira 1993). Maximizing fruit weight and quality will achieve the highest prices for growers providing an incentive to encourage these natural ecosystem processes. However, given we use phytometers, the results are not directly related to estimates of farm-scale crop production.

The seed-set of red campion was positively associated with sward richness; this however could not be explained by pollinator functional or species diversity or abundance. The tubular flower structure of red campion is likely to lead to a more specialized pollination syndrome than strawberry. Therefore, the diversification of pollinator feeding niches may be redundant. Although many studies focus on the value of natural systems in providing benefits to managed systems, few have considered the value of managed systems in maintaining wildflower pollination (Blitzer et al. 2012).

Pasture plant species richness was not associated with improved pollination of broad bean. The flower of broad bean is even more specialized than red campion and is predominantly pollinated by large bees (Free 1966; Aouar-Sadli, Louadi & Doumandji 2008) and so the issue of redundancy is raised again. It is likely that large bees forage at a scale greater than that of individual pastures and consequently the local effect of increased botanical richness may not translate into enhanced pollination.

An unexpected outcome of the field experiment was that the pollinator community parameters did not significantly differ between the cut and grazed treatments. One possible reason could be that surrounding landscape features provided a refuge for the pollinators during cutting. The realized plant species composition of the plots (Table S2) show there is not a great difference in the species richness between the cut and grazed plots.

It is important to note that a limitation of our work was that we utilized a pre-existing gradient of pasture species richness in the farm surveys and therefore a correlative approach; manipulative experiments at the farm scale are needed to really prove the relationship between plant richness and pollination.

Which Species should be Introduced into Seed Mixes?

Taraxacum sp. and C. arvense were the most valuable floral resources to pollinators in the visitation networks of the field experiment and farm data, respectively. It could be that these species have a high visual appearance to pollinators due to their large flowers. Unfortunately, these species have little agronomic value and may even be detrimental (Mortimer et al. 2006). This trade-off between agronomic and biodiversity benefits must be considered in agri-environment schemes and species that provide benefits to the farmer as well as the environment must be identified. We highlight Cichorium intybus a species sown into the experimental plots, as a possible target species. It had high visitation providing resources for pollinators and also agronomic value; it has antihelminthic properties which result in increased weight gain in lambs (Marley et al. 2006) and a deep tap root that captures fertilizers (Moore, Sanford & Wiley 2006).

Conclusion

Without widespread changes in the management of improved grasslands, the decline of many pollinator species is likely to continue (Tscharntke et al. 2005; Carvell et al. 2006). Manning et al. (2015) show that increasing the diversity of grassland plants is likely to be associated with increases in the diversity of a wide range of taxa, with possible conservation and ecosystem service benefits. Our work adds a new aspect to this evidence by demonstrating positive effects on pollination services. Techniques to improve pasture plant species richness are achievable with the expertise and resources available to most grassland farmers. These can include sowing desirable seed mixes, spreading green hay cut from species-rich sites, sward disturbance (e.g. turf removal, harrowing or use of herbicides), sowing hemiparasitic species, for example Rhinanthus minor and reducing phosphorous and potassium levels (Pywell et al. 2012). A desirable balance between agronomic performance of the grassland and its diversity must be considered when choosing such management options. The cascading bottom-up effects of plants, seen at two spatial scales here, demonstrate that modest increases in grassland floral richness is an option for land managers wanting to improve the value of their land for pollinators and ultimately enhance pollination in agricultural habitats.