Introduction

Over the last 25 years there have been significant declines in the diversity of butterflies, bumblebees and other pollinating insects globally (Potts et al. 2010). Moreover, even once widely distributed species have suffered severe reductions in both geographic range and local abundance (Fox et al. 2001). Intensive agricultural management, loss of habitats and food plants, and increased pesticide use have all been cited as contributing factors to these declines (e.g. Steffan-Dewenter et al. 2005; Carvell et al. 2006; Brittain et al. 2010). The need to reverse these damaging impacts of modern agriculture on biodiversity has been recognised by reforms to the Common Agricultural Policy (Bignal 1998). The UK agri-environment schemes aim to achieve this by a combination of less intensive management practices within the crop and the complete removal of land from agricultural production to create wildlife habitat. Several agri-environment schemes promote the creation of pollen and nectar habitat for pollinating insects. The Countryside Stewardship Scheme (CSS) is a voluntary national scheme launched in 1991 with agreements running until 2014. CSS currently accounts for over 360,000 ha of farmland in England with an annual budget of £76.6 million (Natural England 2010a). In 2005 CSS was superseded by the Entry Level Stewardship Scheme (ELS) now covering 5 million hectares of farmland with an annual budget of £133.5 million (Natural England 2010b). Both schemes promote the sowing of simple, low-cost ‘nectar flower mixtures’, typically comprising agricultural varieties of legumes (Fabaceae), such as Trifolium pratense and Lotus corniculatus, in small habitat patches at the edges of fields to provide resources for pollinating insects within intensively managed landscapes. Recent research and monitoring has shown these to be an effective means of increasing the abundance and diversity of foraging bumblebees (Pywell et al. 2006; Carvell et al. 2007). However, there is currently little information regarding the potential value of this habitat for other insects, such as butterflies.

Bumblebee colonies require between 12 and 18 weeks of forage resources, depending on species, in order to complete their development and produce reproductives (queens and males) (Alford 1975). For many species, this period of annual activity begins in April with foraging activity reaching a peak in June and July. Most rare and declining bumblebee species tend to start their colonies later in the season with peak foraging activity between July and September (Edwards and Williams 2004). Indeed, this tendency for late commencement of annual activity has been shown to correlate with susceptibility to decline in bumblebee faunas from across the globe (Williams et al. 2009). Recent research suggests that the peak flowering of nectar flower mixtures sown under the UK agri-environment schemes is on average 7.4 (±0.5) weeks, between late-June and mid-August (Pywell 2008). There is also growing evidence that the agricultural legume varieties sown in these mixtures are short-lived (c.3 years) (Pywell et al. 2007). In the absence of alternative sources of mid- and late-season forage within intensively managed landscapes, it is therefore possible that such nectar flower margins may provide no benefit for bumblebee populations. Hence it is important to devise practical methods to prolong both the longevity of this habitat and its flowering season if benefits to bumblebees, butterflies and other pollinators are to be sustained in the longer term. Current management guidelines recommend cutting half this habitat in June to stimulate later flowering, with the remainder cut in the autumn (Natural England 2010b). The effectiveness of this guidance remains untested.

The aim of this experiment was to examine the effects of cutting regime on the provision of pollen and nectar resources for butterflies and bumblebees in intensively farmed landscapes. In order to achieve this aim we tested four main hypotheses:

H1

Cutting more frequently will increase the abundance of flowers, and therefore pollen and nectar resources for bumblebees and butterflies;

H2

Cutting in early summer will extend the flowering period of pollen and nectar species, and therefore enhance their forage value for bumblebees and butterflies, compared with the typical management of autumn cutting;

H3

Removal of cut material, especially under infrequent autumn cutting, will be beneficial in maintaining the abundance of pollen and nectar species on fertile ex-arable soils than leaving the cut material in situ;

H4

Autumn cutting and leaving the cut material in situ is essential for seed return and regeneration of pollen and nectar species.

A secondary aim of the project was to make some more limited comparisons of the effects of seed mixture composition on the provision of pollen and nectar resources:

H5

Seed mixtures comprising a greater number of pollen- and nectar-rich plant species and larval host plant species attract a greater number and diversity of foraging bumblebees and butterflies than more simple mixtures.

Materials and methods

Experimental design

The study was conducted on an arable farm near Malton, Yorkshire, UK (54°5′2″ N 0°49′11″ W) on fertile clay soil. In April 2003 two seed mixtures (main treatments) were sown in contiguous strips each measuring 200 × 6 m along the margins of two separate cereal fields, providing two replicates of each seed mixture. Mixture 1 (‘complex mix’) conformed to that recommended under the Countryside Stewardship Scheme (prescription ‘WM2’), comprising six legumes and one non-leguminous dicotyledon (dicot) sown with six broad- and fine-leaved grass species at 20 kg ha−1 at a cost of £140 ha−1 (€160 ha−1) (Table 1). This seed mixture was designed to provide both nectar and pollen resources for insects, and larval host plants for a range of farmland butterfly species. Mixture 2 (‘simple mix’) conformed to that recommended under the Entry Level Stewardship Scheme (prescription ‘EF4’), comprising four legumes and three fine-leaved grasses at 20 kg ha−1 and a cost of £55 ha−1 (€63 ha−1). This simple mixture was designed to provide just pollen and nectar resources at very low cost. In the first year the vegetation was managed by cutting and removal of herbage on three occasions (including in October) to control undesirable weeds and encourage establishment.

Table 1 Seed mixture composition (kg ha−1)
Full size table

In April 2004 each main treatment was sub-divided into eight contiguous 25 × 6 m sub-treatment plots and the following cutting regimes were applied at random both with and without removal of cut herbage to test the main experimental hypotheses: A. Cut October (typical practice, the control treatment), B. Cut October + April, C. Cut October + June (scheme recommendation), and D. Cut April + June. An uncut treatment was not included in the experimental design because cutting is required under the agri-environment schemes, and some form of defoliation is essential in maintaining early successional habitats, such as flower-rich grassland. In 2004 cutting was carried out on 21 April, 30 June and 3 October. In 2005 cutting was carried out on 20 April, 1 June and 4 October. The June cutting date was deliberately advanced 30 days in 2005 in order to investigate the effects of earlier summer cutting on the provision of flower resources. This only allowed the effects of early- and late-June cutting to be investigated for a single year, respectively. However, it did provide valuable insights into the potential for manipulating the provision of flower resources through the simple advancement of the cutting date. Cutting was carried out using a 1.6 m wide Ryetec 1600C rear-mounted flail collector mower (www.ryetec.co.uk). The rear collector box was left open to deposit cut and macerated herbage evenly across the sub-treatment plots as required.

This split-plot factorial design provided modest replication (n = 2; df1,1) to detect the effects of seed mixture (secondary hypotheses), but good replication relative to many ecological studies (n = 16; df3,14) to detect the effects of cutting regime on the provision of pollen and nectar resources (primary hypotheses).

Monitoring

In August 2004 and 2005 the composition of the vegetation community was recorded from three 1 × 1 m quadrats placed at random within each sub-treatment plot. In each quadrat percentage cover of vascular plant species was estimated as a vertical projection. In addition, in November 2004 and April 2005 seedlings of sown dicots were counted in 20 quadrats measuring 0.2 × 0.2 m arranged in a permanent diagonal transect across each sub-treatment plot. Percentage cover of bare ground and chopped herbage were also recorded on each occasion.

The number of open flower units was recorded from four 0.5 × 0.5 m quadrats placed at random within each sub-treatment plot. Counts were coincident with pollinator transects and were made on seven occasions between 20 May and 13 September in 2004, and on eight occasions between 26 May and 5 September in 2005. On each visit the height of the vegetation in each sub-treatment was recorded from four drop disk measurements (diameter 300 mm, weight 200 g) (Stewart et al. 2001).

On each visit the abundance and species richness of butterflies and foraging bumblebees was also recorded from a 25 × 6 m transect walked through the centre of each sub-treatment plot (Pywell et al. 2006). Walks were carried out between 10.00 am and 17.00 pm when weather conditions conformed to the Butterfly Monitoring Scheme rules (Pollard and Yates 1993). Foraging bumblebees were recorded to species level and caste (following Prŷs-Jones and Corbet 1991). Workers of Bombus terrestris and B. lucorum were collectively recorded as these cannot be reliably distinguished in the field.

Statistical analysis

The following summary variables were calculated for each sub-treatment: mean vegetation height, percentage cover and species richness of dicots m−2, density and richness of dicot flowers per visit m−2, and density of regenerating clover seedling m−2. In addition, counts of individual bumblebee species from each visit were summed for each sub-treatment plot for each year. Summary variables of species richness and abundance of bumblebees (of all castes, males only and queens only) and butterflies were calculated. Finally, the functional classification of ‘mobile’ or ‘immobile’ was applied to each butterfly species (Warren 1992; Table S1 Supplementary Material).

The effects of seed mixture and cutting regime on these variables were investigated using a split-plot analysis of variance (ANOVA) model with seed mix as the main treatment (tested against the block × seed mix mean square), sub-treatments of cutting date in factorial combination with leave or removal of cut material, and two-way interactions of each of the three factors (all tested against the error mean square). The 2 years were analysed separately. Vegetation cover values were arcsin transformed and counts of seedlings, flowers, butterflies and bumblebees were log transformed as necessary following an assessment of the normality of residuals. All analyses were carried out using GenStat® 11.0 statistical software. Pairwise comparisons of each cutting sub-treatment were made using the contrast function of GenStat. Finally, a split-split plot ANOVA model was used to examine the effects of seed mix and cutting regime on the distribution of flowers, bees and butterflies through each season by classifying each visit as either early season (May to mid-July) or late season (mid-July to September) following Carvell et al. (2007).

Results

Vegetation composition

Cover of sown dicots was generally greater in the simple seed mix than the complex mix in both years, however this effect was only significant in 2004 (Table 2). Cutting date had many more significant effects on vegetation composition, with cover of most sown species being greater under the April + June and October + June cutting regime than October + April and October cutting, especially in 2005. There was also evidence that the richness of sown dicots was greater under the April + June cutting regime. Herbage removal had no significant effects on composition in 2004, but in 2005 resulted in a significant increase in total cover and richness of sown dicots, with a corresponding significant decrease in cover of sown grasses. There were no significant treatment interactions. Cutting regime also had significant effects on seedling regeneration which is considered important in habitats dominated by short-lived perennial plants. Management regimes involving late June cutting were highly detrimental to clover seedling regeneration in the following autumn (Oct cut = 110 seedlings m−2, Oct + Apr = 228, Oct + Jun = 5.8, Apr + Jun = 1.4; ANOVA F3,14 = 23.60, P < 0.001). There were no effects of herbage disposal method on seedling regeneration.

Table 2 Effects of seed mixture and cutting regime on mean percentage cover and richness of sown dicots, and number and richness of dicot flowers per m2 per visit
Full size table

Vegetation height was higher in the complex mix than the simple mix in both 2004 (complex = 37 ± 2 cm vs simple = 33 ± 2 cm) and 2005 (complex = 41 ± 4 cm vs simple = 26 ± 1 cm.) However, this effect was only significant in 2005 (F1,1 = 379, P < 0.05). In 2004 cutting in April + June (25 ± 1 cm) and October + June (32 ± 1 cm) both resulted in a significantly lower sward height than October + April cutting (37 ± 2 cm). Cutting in October resulted in the tallest vegetation (45 ± 1 cm) (F3,14 = 33.77, P < 0.001). Similarly, in 2005 cutting in April + June (25 ± 1 cm) and October + June (25 ± 1 cm) resulted in a significantly lower sward height than October + April cutting (39 ± 5 cm). Cutting once in October resulted in significantly taller vegetation than any other treatment (45 ± 5 cm) (F3,14 = 49.98, P < 0.001). There was no significant effect of herbage removal on sward height in either year.

Flower resources

Flowers of sown species comprised over 95% of the resource in both years. The abundance of sown dicot flowers was consistently higher in the simple mix than the complex mix (Table 2). However, this effect was only significant in 2005. Similarly, the simple mix consistently produced a greater abundance of flowers in early-season (May to mid-July) and late-season (mid-July to September) (Fig. 1a), but there was no significant season × seed mix interaction in either 2004 (F1,2 = 0.01, P > 0.05) or 2005 (F1,2 = 0.85, P > 0.05). Cutting date had a more marked effect on abundance and diversity of pollen and nectar resources. In 2004 flower abundance and richness was significantly higher following cutting in October and April + October than either of the June cuts (Table 2). Most individual species showed a similar response to cutting. There was also a highly significant season × cutting date interaction (F3,28 = 16.09, P < 0.001) showing a marked reduction in late-season flower resources as a result of late June cutting (Fig. 1b). Advancing the June cut date by 30 days in 2005 effectively reversed this pattern of response. Abundance of flowers was significantly higher following cutting in June than October and April + October cutting (Table 2). Once again there was a significant season × cutting date interaction (F3,28 = 9.25, P < 0.001), this time showing a marked increase in late-season flower resources as a result of early June cutting (Fig. 1b). In 2005 there was a large and consistent increase in flower abundance and richness following herbage removal. Similarly, there was a significant season × herbage disposal interaction (F1,28 = 7.02, P < 0.05) with removal also enhancing late-season flower abundance.

Fig. 1
figure 1

Effects of (a) seed mixture, (b) cut date and (c) herbage disposal on mean (±SE) number of sown dicot flowers per m2 on each visit. (□) Denotes early season (May to mid-Jul), (■) denotes late season (mid-Jul to Sep)

Full size image

Bumblebee and butterfly abundance

Between 2004 and 2005 we recorded 3379 bumblebees comprising six species and representing the assemblage typical of the study region (Table S2 Supplementary Material). Both the total abundance and species richness of bumblebees was consistently higher in the simple than complex mix. However, this effect was only significant for the richness of bees in 2004 (Table 3). Similarly, there were consistently more bees on the simple seed mix in both early- and late-season (Fig. 2a). This interaction effect was not significant in 2004 (F1,2 = 0.53, P > 0.05), but was significant in 2005 (F1,2 = 383, P < 0.01). The latter reflected numbers of bees on the simple mix in early season.

Table 3 Effects of seed mixture and cutting regime on species richness and total number of bumblebees per 150 m2
Full size table
Fig. 2
figure 2

Effects of (a) seed mixture, (b) cut date and (c) herbage disposal on mean (±SE) number of bumblebees per plot on each visit. (□) Denotes early season (May to mid-Jul), (■) denotes late season (mid-Jul to Sep)

Full size image

In 2004 there were highly significant effects of cutting date on all bee species and summary variables (Table 3). Late June cutting was highly detrimental to the abundance and richness of bees. This pattern of response to cutting was consistent for most individual bee species and the critically important reproductive components of bee colonies. Similarly, there were highly significant cut date × season interaction effects confirming the highly detrimental effects of late-June cutting on both bee species richness (F3,28 = 22.92, P < 0.001) and abundance (F3,28 = 34.37, P < 0.001) (Fig. 2b). Early June cutting in 2005 reversed these trends for total foraging workers (F3,28 = 2.97, P < 0.05) and the long-tongued bees B. pascuorum and B. hortorum (Table 3). In contrast, cutting in early June had a significant, detrimental effect on the abundance of male bumblebees, both overall and in the complex mix containing late-flowering forbs. The significant seed mix × cut date interaction reflected the greater number of bees attracted to the simple mix in response to early June cutting than the complex mix. Finally, there was a highly significant seasonal interaction effect confirming the beneficial effect of early June cutting on overall bee abundance in the late summer (F3,28 = 4.01, P < 0.05) (Fig. 2b). However, there was no significant effect on bee species richness (F3,28 = 0.63, P > 0.05).

There were no significant effects of herbage disposal technique. However, both abundance and species richness of bees were significantly higher following herbage removal in 2005. In contrast, abundance of new emerged queens was higher where the cut material was left in situ. There were no significant season × herbage disposal interaction effects in 2004 on bee abundance (F1,28 = 0.00, P > 0.05) or richness (F1,28 = 1.46, P > 0.05) (Fig. 2c). Similarly, in 2005 there were no seasonal interactions on abundance (F1,28 = 0.44, P > 0.05) or richness (F1,28 = 1.94, P > 0.05.

We recorded 490 butterflies representing 14 species typical of the study region (Table S1 Supplementary Material). The effects of seed mixture were less pronounced on butterflies compared with bumblebees (Table 4). Abundance of mobile species was consistently higher in the simple mix and this effect was significant in 2005. Richness of all butterflies was significantly higher in the simple mix in 2005. Similarly, there were no consistent seasonal interactions with seed mix (Fig. 3a) in either 2004 (F1,2 = 0.24, P > 0.05) or 2005 (F1,2 = 0.22, P > 0.05).

Table 4 Effects of seed mixture and cutting regime on species richness and total number of butterflies per 150 m2 plot
Full size table
Fig. 3
figure 3

Effects of (a) seed mixture, (b) cut date and (c) herbage disposal on mean (±SE) cumulative number of butterflies per plot on each visit. (□) Denotes early season (May to mid-Jul), (■) denotes late season (mid-Jul to Sep)

Full size image

In 2004 there were strong and consistent detrimental effects of late June cutting on the abundance and diversity of butterflies (Table 4). Similarly, there were highly significant cut date × season interaction effects confirming the highly detrimental effects of late June cutting on both butterfly species richness (F3,28 = 8.95, P < 0.001) and abundance (F3,28 = 14.77, P < 0.001) (Fig. 3b). The pattern of response to cutting management was less clear cut in 2005 due to fewer butterflies being recorded. Total butterflies were significantly higher following April + June and October cutting than April + October cutting. Total mobile butterflies were significantly higher following April + June cutting than April + October cutting. There were no significant season × cutting date interactions for either butterfly abundance (F3,28 = 0.83, P > 0.05) (Fig. 3b) or butterfly richness (F3,28 = 1.60, P > 0.05).

Finally, there were no significant effects of herbage disposal technique on butterfly abundance or richness in either year (Table 4). However, removal of cut material in April + June appeared to have a significant beneficial effect on abundance and richness of butterflies (particularly mobile species), whereas this effect was reversed for October cutting. There were no significant herbage disposal × season interaction effects on abundance (2004 F1,28 = 0.16, P > 0.05; 2005 F1,28 = 0.18, P > 0.05) or richness (2004 F1,28 = 0.56, P > 0.05; 2005 F1,28 = 0.02, P > 0.05) (Fig. 3c).

Discussion

Pollen and nectar seed mixtures

Given the limited replication of the seed treatments, the significant effects are suggestive of large and consistent differences in the composition and floral resources of the two mixtures. Cover of sown dicots, especially Trifolium spp. and Lotus corniculatus, was higher in the less costly seed mixture based on finer grasses than the complex, tall grass mix. Consequently the simple mix produced a consistently higher density of dicot flowers (35–45%), particularly those of Trifolium sp., in both years. These effects can be explained in terms of differences in seed mixture composition and sowing density, and the performance of individual species (Pywell et al. 2003). The seed mixtures differed in several important respects, namely the complex mix contained two highly competitive, broad-leaved grasses (Phleum pratense and Festuca pratensis) which were both sown at relatively high rates and achieved higher cover. This resulted in a significantly taller, and presumably more competitive sward. In contrast, only low-growing, fine-leaved and therefore less competitive grasses were sown in the simple mix. Also, seed sowing density of several dicot species was significantly higher in the simple seed mix, for example individual seed rates of T. hybridum and Lotus corniculatus were 33 and 50% higher, respectively. Finally, two of the seven dicots sown in the complex mix either failed to establish (Vicia sativa ssp. sativa), or established poorly and did not persist (Medicago lupulina). For these reasons, the complex and more costly seed mixture containing a greater number of pollen- and nectar-rich plant species did not attract a greater number and diversity of foraging bumblebees and butterflies than the low cost, simple mixture. However, the long-term sustainability of pollination services within intensively managed landscapes requires the provision of resources for both the foraging workers and the reproductive components of bumblebee colonies. The complex mix included Centaurea nigra, a dicot which flowers in late summer and which, along with other members of the Asteraceae, is known to attract male bumblebees (Carvell et al. 2007), and be an important nectar species for farmland butterflies, such as Melanargia galathea. This species attracted a significantly greater number of male bumblebees to the complex mix than the simple mix, provided the margin had not been cut in June.

Cutting regime

The effects of April and October cutting on the flower resource were relatively consistent over both years (2004 first flower 8 June, peak 29 July; 2005 first flower 9 June, peak 11 July). This suggests that the effect of advancing the June cut by 30 days was unlikely to be just an artefact of differences in growing conditions and phenology between years. The key finding of this study suggests that timing rather than the frequency of cutting had the most marked effects on the performance and regeneration of the sown species, and the provision of pollen and nectar resources through the season. Cutting has direct effects on the vegetation and associated invertebrates through the removal of the vegetation structure and flowers (Morris 2000). It will also have indirect effects on the plant community by encouraging vegetative growth, enhancing flower production, creating gaps for germination and by altering the competitive balance between species (Bullock 1996). We found that adding a summer cut to the typical autumn cut appeared to enhance the cover and flower abundance of T. pratense and T. hybridum at the expense of competitive grasses. Further research is required to determine the precise mechanism of this observed effect, and whether it is an effective means of maintaining sown pollen and nectar species in the longer term. However, tall, late-flowering dicots, such as C. nigra, were reduced by June cutting. Timing of summer cutting also appeared to be the critical factor controlling numbers of bumblebees and butterflies attracted to the sown mixtures. Cutting in late June delayed the re-flowering of T. hybridum by c.50 days and that of T. pratense by c.70 days. This effectively removed the dicot flower resource available to bumblebee colonies during the July and August period which is considered critical for reproductive success (Goulson 2003). Cutting at this time would have also removed both nectar and potential oviposition sites for farmland butterfly species, such as Polyommatus icarus and Maniola jurtina (Feber et al. 1996). Furthermore, the very low densities of Trifolium seedlings emerging in the autumn and following spring suggested that this de-synchronisation of flower resources and pollinators may have resulted in potentially catastrophic consequences for seed production and regeneration of these species (Hawkins 1958).

While early June cutting in the following year delayed re-flowering of Trifolium species for a similar period of time, in this case peak re-flowering was coincident with peak abundance of foraging bumblebees in late July and early August. This therefore suggests a highly effective means of extending the flowering period of pollen and nectar species and enhancing their abundance by two- to seven-fold. This significantly increased the abundance of foraging workers, and long-tongued bumblebee species (Bombus pascuorum and B. hortorum) which specialise on T. pratense. Seed counts from 5 random T. pratense flower heads collected from each sub-treatment plot in September 2005 also found that early June cutting and delayed flowering had significant, beneficial effects on seed production (April + June cut = 77.3 ± 6.7 viable seeds head−1; April + October cut = 50.4 ± 6.7 seeds head−1; ANOVA F1,14 13.26; P = 0.003**, R.Pywell, unpublished data).

Herbage disposal

The method of herbage disposal also had important ecological effects on vegetation composition, seedling regeneration and the provision of pollen and nectar resources. The removal of cut material once or twice a year resulted in a significant increase in the cover (45%) and richness (25%) of sown dicot species at the expense of grasses. Cutting and removal of herbage resulted in an instant reduction in competition for space and light than the more gradual reduction resulting from leaving the cut material in situ. It is also likely that removal resulted in nutrient off-take and reduction in soil fertility (Walker et al. 2004). Similarly, removal also resulted in significant increases in the density of sown dicot flowers, particularly Trifolium spp. (Wells and Cox 1993), and therefore increased the abundance and richness of bumblebees. Cut and macerated herbage left in situ will act as a physical barrier to light reaching the underlying plant species. Plant species will vary greatly in their ability to tolerate the stress induced by this type of shading. This will have indirect effects on plant community composition by altering the competitive balance between species. Removal of the cut material did not appear to affect the ability of the sown dicots to regenerate from seed, suggesting that sufficient seed is returned to the soil surface either before or during the cutting operation (Sakanoue 2005). However, it is likely that the thick mat of cut and macerated vegetation rich in viable seed resulting from cutting and leaving in situ is a poor medium for germination and survival. Further research is required into the factors constraining the regeneration of Trifolium species in productive habitats managed for the provision of pollen and nectar resources.

Conservation management recommendations

Field margins sown with the ‘nectar flower mixture’ under the current UK agri-environment scheme guidance (Natural England 2010b) are only effective for 3–4 years despite intensive cutting management (Pywell et al. 2007). The most practical measure to guarantee a continuity of pollen and nectar resources is to re-establish this habitat either in situ or preferably elsewhere on the farm after 3 years. It is recommended that the composition of future nectar flower mixtures is revised and simplified to include a combination of good performing dicot species which flower in mid-season (e.g. T. pratense, T. hybridum, L. corniculatus) and late-season (e.g. C. nigra, Malva moschata). The latter is critically important to provide foraging resources for the reproductive stages of bumblebee colonies, and may also help to buffer against the potential negative effects of global warming on the flowering phenology of sown mixtures (Memmott et al. 2010). Further research is required to determine if the exclusion of grass species will reduce competition, increase the regeneration of sown dicots by seed and therefore improve the longevity of these habitats. However, the potential value of the sown grass species as larval food plants for some butterfly species, such as Melanargia galathea, should also be considered. It may be more practical and less damaging to butterfly populations to create separate breeding habitat comprising larval host grasses and dicots, and manage these to provide a varied sward structure, with suitable edges and recesses for egg laying. Neither seed mixture provided early season forage for nest-searching queen bumblebees or early emerging butterfly species. It may also be more efficient to provide these resources in separate habitats scattered throughout the landscape, for example by either planting or conserving early flowering hedgerow species (Salix cinerea, Malus sp., Lamium album, Glechoma hederacea) and patches of novel crop species (e.g. Lunaria annua). In this small plot experiment mobile pollinating insects were easily able to track the availability of flower resources in the cut and uncut plots. However, the amount and spatial distribution of resources required at the landscape scale for central place foragers such as bumblebees are still not fully understood (Heard et al. 2007), but are key factors in the design of management plans for the conservation and sustainability of pollination services (Lonsdorf et al. 2009).

The current management guidelines of cutting pollen and nectar margins in summer are effective in increasing the cover and extending the flowering period of sown dicots. Cutting only half of the sown margin also reduces the potential damage to breeding populations of butterflies, and will retain some tall vegetation habitat for nesting bumblebees and other invertebrates. However, we recommend that the date of cutting is advanced to May or early June at the latest in order to ensure the provision of pollen and nectar resources are synchronised with peak foraging for farmland bumblebees and butterflies. The precise date of cutting will require adjustment for different latitudes and seasonal variation in growing conditions. Ideally cut material from the pollen and nectar margins should be removed. However, this increases the time taken, cost and complexity of the cutting operation for land managers. To overcome this problem the practical and ecological effects of traditional hay-making and baling require further investigation.

The UK agri-environment schemes have been developed from a relatively strong evidence base, but there remains a strong bias in option uptake by farmers, towards less targeted options that generally require less management intervention (Boatman et al. 2007). This may be in part related to the limited guidance available to farmers on how to successfully establish and manage nectar flower mixtures to provide high quality habitat. Our study therefore provides an important step in experimentally testing a range of practical management regimes and demonstrating how the benefits of such conservation measures could be maximised for pollinating insects in intensively farmed landscapes.