A novel farmland wildflower seed mix attracts a greater abundance and richness of pollinating insects than standard mixes

Wildflower strips are a popular agri‐environment scheme (AES) implemented on farmland to provide forage for insect pollinators. The standard seed mixtures were often formulated without a clear evidence base, and subsequent field trials to assess their attractiveness to insects are commonly carried out at low taxonomic resolution (e.g., pooling all ‘solitary’ bees). We created two novel wildflower mixes: a wild bee mix based on primary research (WB) and one on literature‐based evidence (LT). We trialled our novel mixes against two standard AES wildflower mixes: a Fabaceae‐heavy mix (FAB); a diverse wildflower mix (WF); plus a fallow plot (control). Our aim was to determine which mix attracted the highest overall insect pollinator abundance and highest species richness for wild bees. Our WB mix attracted both the highest number of total insect visitors, and the highest wild bee abundance and richness. WB attracted significantly more bumblebees (abundance and richness) than the typical low diversity, Fabaceae‐heavy mix (FAB); and significantly greater solitary bee abundance, than all other treatments. Only 11 ‘key’ wildflower species were required to cater to all wild bee species recorded during the study, eight of which were sown species. Taraxacum officinale agg., Cirsium vulgare, Daucus carota and Geranium pyrenaicum received the highest numbers of wild bee species visits. In conclusion, we suggest a novel wildflower seed mix based on primary research has the potential to provide more attractive forage for both wild bees and other insect pollinators compared to current AES mixes.


INTRODUCTION
Land-use change brought about by agricultural intensification is generally considered to be the main driver of farmland biodiversity decline, as semi-natural habitats and weeds have become increasingly scarce (Foley et al., 2005;Potts et al., 2010). To counteract the negative effects of modern farming practices, agri-environment schemes (AES) have been implemented throughout the Western world in an attempt to restore biodiversity in agricultural landscapes (Harmon-Threatt & Hendrix, 2015;Howlett et al., 2021;Lindenmayer et al., 2012;Scheper et al., 2015).
Wildflower strips are a popular AES employed on farmland throughout Europe to provide forage for wild pollinators and insect predators of crop pests (Ouvrard et al., 2018;Tschumi et al., 2016).
These strips have the potential to provide a diverse resource for the insect community, as both food for adults foraging on pollen and nectar (e.g., bees), or as host plants for insect larvae (e.g., Lepidoptera; Curtis et al., 2015). Different flower species produce pollen and nectar of varying nutritional components, qualities, and quantities. The main nutritional component of nectar is carbohydrate in varying quantities (Hicks et al., 2016), while pollen is a source of protein and lipids (Jeannerod et al., 2022). Insects such as bees must not only consume sufficient quantities of food to meet their own energy expenditure requirements, but must also obtain an adequate and diverse range of nutrients to feed their larvae. Where nutritional components are missing, evidence suggests reproductive success is decreased (Brunner et al., 2014;Génissel et al., 2002).
Wildflower strips are typically created adjacent to crops in field margins, comprised of non-competitive grasses and wildflower species. These wildflower seed mixes differ in the number of wildflower species and in their target taxa. For example, mixes can contain as few as 4-6 species (DEFRA, 2013), or in excess of 50 species (Warzecha et al., 2018) and may not always include grasses depending on the country and region. Some mixes contain a high proportion of Fabaceae species that have the potential to attract specialist pollinators Kleijn et al., 2018), while others contain a higher proportion of Apiaceae and Asteraceae species that attract shorttongued generalists (Kleijn et al., 2018). Mixes can also be specialised to a geographical region, for example, there is a specific list of species allowed to be sown on the Swiss plateau through AES (Tschumi et al., 2016).
Studies measuring the impact of wildflower strips on pollinating insect abundance or diversity typically assess a single seed mix (Carvell et al., 2011;Haaland & Bersier, 2011;Ouvrard et al., 2018), or compare multiple seed mixes that are available locally through AES agreements (Grass et al., 2016;Warzecha et al., 2018). Examples include early research showing that a (Fabaceae-heavy) pollinatortargeting mix was more effective at providing bumblebee forage than a grass-only seed mix or natural regeneration ; while recent research comparing multiple wildflower seed mixes found that it was the presence of 'key plant species' rather than floral diversity that attracted the highest insect richness (Warzecha et al., 2018). Few studies have considered novel or bespoke seed mixes (though see Griffiths-Lee et al., 2022;Uyttenbroeck et al., 2017), nor do they include the evidence base for selecting certain species for a flower mix. Uyttenbroeck et al. (2017) created four bespoke seed mixes of varying functional diversity (FD) levels and found that FD had no effect on pollinator abundance or richness, but that pollinator visitations could occasionally be explained by floral abundance of specific plant species. Moreover, studies have historically focused on single insect groups, such as just Bombus species or butterflies Haaland & Bersier, 2011). More recently, solitary bees, hoverflies, and non-bee species (Howlett et al., 2021;Warzecha et al., 2018;Wood et al., 2017), along with the wider insect community (Grass et al., 2016;Ouvrard et al., 2018) have also been taken into consideration. For example, Scheper et al. (2021) trialled two seed mixes, one targeting long-tongued pollinators such as bumblebees, and another targeting more generalist species and predators such as hoverflies. Their study found that bumblebee abundance was positively associated with the amount of Fabaceae cover, while hoverflies were positively associated with Apiaceae cover. Therefore, it is timely to further consider the wider insect community and its interaction with novel seed mixes.
In this study we assess the whole flower-visiting insect community as well as focusing on wild bees. We compared two standard AES wildflower seed mixes with our own two novel wildflower seed mixes and with unsown fallow plots. Our novel seed mixes were designed to attract maximum wild bee visitations, but had the potential to also attract a broad range of insect pollinators (Nichols et al., 2022a). The aims of our study, therefore, were to: 1. Identify which wildflower mix attracted (a) the highest number of insect pollinators; and (b) the highest wild bee abundance and richness; 2. Determine which wildflowers were key resources for wild bees.

Creating wildflower mixes
We created two novel wildflower mixes (see Nichols et al., 2022a, for details on the selection process used, which is summarised here). First, we used the existing literature to select wildflowers that were shown to be key resources for a diversity of solitary bees within agricultural settings (Howlett et al., 2021). Second, we created a novel mix through primary research on a wildflower farm (Nichols et al., 2019), selecting the wildflowers that attracted the highest richness of wild bees. Plant species were placed into categories according to their flowering phenology: late-spring, early-summer, mid-summer, and late-summer to ensure we had at least 1-2 species flowering within each period (Williams et al., 2015), and any species that were unavailable were removed. We also included four annual cornfield species to both mixes to act as a 'nursery' in the first year: poppy (Papaver rhoeas), corn marigold (Glebionis segetum), cornflower (Centaurea cyanus) and corncockle (Agrostemma githago). Our final literature-based mix (LT) contained 17 wildflower species, and our primary research-based wild bee mix (WB) contained 16 wildflower species. We also used two standard AES mixes available in the United Kingdom (UK): a Fabaceae-focused mix of six species (FAB), and a typical wildflower mix of 12 species (WF). All mixes were created with 20:80 ratio of wildflowers to non-competitive grasses, except for FAB which was made with 100% wildflower seeds following normal farming practice (see Supplementary Information S1 for seed mix details).

Study site
The study was conducted in a single field on two farms: Church Farm, Oxfordshire (51 38 0 14.4348 00 N, 1 11 0 5.4528 00 W) and Lee Farm, West Sussex (50 53 0 0.2112 00 N, 0 28 0 24.1464 00 W). The farms were of different soil type and the fields had received different management prior to our experiment commencing. Church Farm has freely-draining, base-rich loamy soil of high fertility, and the field had recently been in production. It also had a high abundance of aggressive weeds (e.g., Cirsium vulgare, Alopecurus myosuroides). Lee Farm had lime-rich loamy soils over chalk, the field had been out of production for 2-3 years, and any natural regeneration had been cut and ploughed yearly, breaking up any perennial thistles and encouraging annuals.
Seed mixes were sown in 2018 on 4th and 5th September respectively. Both sites were cultivated and sprayed-off using glyphosate herbicide as ground preparation to aid establishment. The ground was then rolled, seeds were broadcast sown by hand, and then rolled again to ensure sufficient seed-to-soil contact.
Seed mixes were sown in 20 Â 5 m contiguous plots. Our treatment consisted of four different seed mixes: FAB, WF, LT, WB; and a fallow (control). Each treatment was replicated 5 times on each farm, therefore each farm had 25 20 Â 5 m plots. Seed mixes were allocated to plots using a Latin-square-type design to ensure no mix was next to itself. Plots were then marked by a GPS device to store and re-find the plot locations for surveys.
All plots on a farm were managed the same way, but cuts were performed according to need on each farm. Church Farm was cut in May 2019 when thistles began to take over the plots, and then in June during both 2020 and 2021 when plants started to collapse under their own weight. Lee Farm was cut in July in both 2019 and 2020 when growth began to collapse. Both farms were also then cut during the autumn months in 2019 and 2020 once the majority of plants had seeded and growth could be removed.

Floral and insect surveys
Surveys were conducted from April to August in 2019, 2020 and 2021. They were conducted every 2-3 weeks, providing a total of eight survey rounds each year. Farms were surveyed on two separate days close to one another for each of the eight survey periods. Surveys were not conducted on a farm if they would occur immediately after a cut, since nothing would be in flower. During each survey, each plot was walked centrally lengthways and insects visiting flowers seen 2 m either side were identified to group-level on the wing. Bees and hoverflies were then further identified to species-level. Those that could not be identified to species-level on the wing were captured for identification in the lab by RNN. Specimens that were not able to be identified by RNN were sent to Steven Falk and Ellen Rotheray for identification. Flower species seen 2 m either side were also noted and the estimated abundance of open flowers in the plot recorded (Campbell et al., 2017;Wood et al., 2017). 'Flower' was defined as either a single flower, flowers on an umbel or spike, or a capitulum (Heard et al., 2007). A maximum of 10 minutes was allowed for each plot. Surveys were conducted between 08:30 and 17:00 when the temperature was above 13 C with at least 60% clear sky, or above 17 C in any sky conditions, and not raining (Pollard & Yates, 1993).

Data analysis
All data analysis was handled in R version 4.0.3 (R Core Team, 2020).
Zero-inflated Generalised Linear Mixed Models were built using the glmmTMB package (Brooks et al., 2017). Shannon's Diversity index was calculated for the 'plant diversity' of each plot for every survey conducted and included as an explanatory variable (Griffiths-Lee et al., 2022) in each model to improve model fit. 'Survey round' was also included as an explanatory variable in each model. 'Replicate' nested within 'farm' was included as a random variable in each model. Models testing insect 'richness' were built with a 'Poisson' log link, and models testing insect 'abundance' were built using a 'negative binomial' family after conducting residual plot diagnostic checks. An ANOVA was then performed on each model and its null, reported as χ 2 values, and post hoc Tukey tests were conducted to see where the significance lay when appropriate. All figures were created using ggplot2 (Wickham, 2016).
First, the plant and insect abundance heatmap was created by calculating the total abundance of the flowers and all insects visiting each flower, per plot, averaged across replicates, survey periods, mixes, farms, and years.
Next, we assessed the treatment effect on insect visitation. All insect visits were summed for each plot, per survey round, per farm, per year (hereon referred to as total insect abundance). A model was built to test the effect of 'treatment' and 'survey year', and their interaction, as predictor variables.
Following this, we built models to determine the effect of 'treatment' and 'survey year', and their interaction, on wild bee abundance and richness (also considering solitary bees and bumblebees separately). The number of wild bees (abundance), and the number of wild bee species (richness) were summed for each plot, per survey round, per farm, per year.
To identify the plant species most significantly visited by wild bees, we calculated their species strength. First, insect visits to each wildflower species were pooled across treatments and years. 'Species strength' is defined as the sum of dependencies (proportion of visits) of flower visitors relying on a specific plant species, and was calculated on the pooled data using the 'strength' function in the bipartite package (Dormann, 2011). To identify the minimum plant species composition needed to cover all wild bee species (hereafter 'key wildflower species'), we first took the plant species attracting the highest number of wild bee species and subsequently added plant species attracting most of the remaining wild bee species until all wild bee species were covered. Seed mixture potential was defined as the percentage of key wildflower species within each plot (Warzecha et al., 2018). A Linear Model was built to test the effect of 'treatment' on the 'percentage' of key plant species in each plot, and the result is reported as an F-statistic.
The wild bee-plant visitation network was calculated after identifying abundance of each wild bee species visiting each plant species over the whole 3-year experiment, removing any species that were seen less than three times, and then analysed and visualised with the 'compute-Modules' and 'plotModuleWeb' functions, in the bipartite package.

Overall visitation network
A total of 4002 insects were recorded making flower visits to the plots over the 3 years. Visits to flowers were dominated by Diptera (flies; excluding hoverflies; 28.1% of all visits), followed by Coleoptera (beetles; 20.6%) and Bombus spp. (bumblebees; 20.0%). Non-corbiculate wild bees (solitary bees) accounted for 14.8% of visits, and Syphridae (hoverflies) made up 8.6% of visits. The remaining 7.9% of visits were made up of Apis mellifera (honeybees), Lepidoptera (butterflies), and solitary wasps.
We recorded 79 flower species across 21 families in the plots, including 28 sown and 51 spontaneous species (see Nichols et al., 2022a, for details on success of sown plant species). Insect visits were recorded to 55 flower species across 15 families, with 75.5% of visits to 26 sown species and 24.5% of visits to 29 spontaneous species (Figure 1). Insect visits were predominantly made to Asteraceae species (58.0%), followed by Apiaceae (14.2%) and Fabaceae (7.5%) species (see Supplementary Information S2 for full list of plant-insect interactions).
There were 1390 visits made to flowers by wild bees. Of the bees recorded, 1124 (80.9%) were identified to species level: 798 bumblebees across six species, and 326 solitary bees across 34 species. Bumblebee F I G U R E 1 Flower and insect counts for each plant species visited. Counts summed for each plant species within each plot for the survey period, and then averaged between replicates, farms, and years. Mean number of flowers observed (square-root-transformed) and mean insect abundance (log-transformed) per flower species (each calculated per 80 m 2 ). visits were recorded mainly to Fabaceae species (29.7%), Asteraceae species (27.9%), and Papaveraceae (26.0%); whereas solitary bee visits were heavily focused on Asteraceae species (58.9%), followed by Apiaceae (12.4%) and Geraniaceae (9.5%).
Hoverflies visited the second highest number of flower species (37)

Treatment effect on visitations
Treatment had a significant effect on total insect abundance (GLMM: Figure 2), with the WB mix attracting significantly more insect visitors than all other treatments, when controlling for plant diversity. Additionally, although WF and LT attracted significantly lower insect abundances than WB, they attracted significantly greater numbers than the fallow control. There was no significant difference between the fallow plots or FAB mix in terms of insect abundance (see Supplementary Information S4 for insect abundance of each insect group).
There was also a significant interaction effect of treatment Â year on total insect abundance (GLMM: χ 2 = 15.9, p = 0.044), though no effect of year alone on total insect abundance (GLMM: χ 2 = 3.36, Treatment also had a significant effect on total wild bee abundance and richness, with the WB mix attracting significantly greater abundance (GLMM: χ 2 = 42.0, p < 0.001), and richness (GLMM: were significantly more bumblebee counts on the WB mix than on the Fallow, FAB and LT treatments (GLMM: χ 2 = 27.8, p < 0.001). There was no significant difference in bumblebee abundance between the fallow plots and the FAB or LT mixes.
F I G U R E 2 Total insect visits to each treatment. Mean total insect abundance per treatment, summing values for each plot, averaging across replicates, survey periods, farms and years. Significance (post hoc Tukey) of treatments denoted by lettering (±SE).
Likewise, there were significantly more solitary bee species on the WB treatment than on the Fallow or WF plots (GLMM: χ 2 = 13.7, p = 0.008), once more with mixes FAB, WF and LT attracting no more solitary bee species than the fallow plots. There were also significantly more bumblebee species on the WB mix than on the Fallow, FAB and LT treatments (GLMM: χ 2 = 26.7, p < 0.001).

Key wildflower species for wild bees
Wildflower species visited by wild bees were ranked according to their species strength (

Treatment effect on visitations
Wildflower seed mixes created through regional primary research could be considered for future AES. Seed mixes are typically produced to either follow specific AES guidelines (Schmidt et al., 2020;Warzecha et al., 2018), the local abiotic conditions (e.g., soil type; Haaland & Bersier, 2011;Nowakowski & Pywell, 2016), or to target a specific taxa or insect group (Carvell et al., 2011;Kleijn et al., 2018). Here we showed that a mix created through primary research to attract a target insect group (wild bees) was successful, despite being sown on two farms with different soil types, management histories, and different seed banks (Nichols et al., 2022a). Studies have previously shown that targeted AES are better at producing optimal abundance and diversity of a target taxa than general AES (Carvell et al., 2011;Wood, Holland, & Goulson, 2015a;Wood, Holland, Hughes, & Goulson, 2015b). Our WB mix was not only successful at attracting the highest abundance and richness of wild bees compared to the other treatments, but also the highest total insect abundance, suggesting it was attractive to other taxa as well (Ouvrard et al., 2018). Although conducting region-specific primary research to identify key species is far costlier than adhering to a generic national seed mix, it should be considered if we are to better support the wider pollinator community. We suggest that Fabaceae-heavy seed mixes such as the one trialled in this study and that have been widely used by farmers should be updated to include a broader range of species. When composed of only 4-6 wildflower species (as recommended; DEFRA, 2013), this mix not only lost floral abundance rapidly as grasses became predominant Nichols et al., 2022a), but it performed poorly in terms of overall abundance of insect visitors, and wild bee visitors. Early research showed that the Fabaceae-heavy mixes could attract high bumblebee abundance and support specialist species Pywell et al., 2006). By contrast, our results showed the Fabaceae mix was statistically no better than a fallow plot for attracting bumblebees (though see Cole et al., 2022 for contrasting evidence of a low diversity Fabaceae mix). Only one species typically included in the mix, Tr. hybridum, was found to be an attractive plant for wild bees (in the 'top 11' species). Therefore, we suggest that the mix is updated, and instead, species that attracted very few wild bees (e.g., O. vicifollium) are removed, and replaced with a Fabaceae that did attract bumblebees (e.g., Anthyllis vulneraria). Additional key species that attracted high diversity of other wild bees across the whole spring-summer season (e.g., D. carota, G. pyrenaicum, T. officinal agg.) could replace the less attractive M. moschata and C. nigra. This would create a low diversity, low-cost AES seed mix that delivers greater benefit for a wider range of pollinators.
The nutritional components of pollen and nectar varies widely between plant species and families (Hanley et al., 2008;Jeannerod et al., 2022), which can in turn limit the growth and survival of bee broods. Brunner et al. (2014) found that a diet consisting of just Taraxacum spp. resulted in failure to lay eggs in B. terrestris micro-colonies; while Austin and Gilbert (2021) found Osmia bicornis larvae survival was positively correlated with carbohydrate quantity. Therefore, bee species benefit from a diverse diet in which pollen and nectar are sourced from a variety of plant species in order to obtain all essential amino acids and sufficient quantities of sugar nectar (Hanley et al., 2008;Jeannerod et al., 2022;Vaudo et al., 2015). Although our novel WB mix was found to attract the greatest abundance of insect pollinators, we did not record if visits were made for pollen or nectar foraging/collection. Therefore, it is difficult to attribute a nutritional benefit to the seed mix or specific plant species within it. Additionally, all foragers recorded were adult forages, therefore it is unclear if they were foraging for their own consumption, or collecting to provide for offspring. As the protein and carbohydrate contents of food can have a big effect on the reproductive success, this should be considered when forming a wildflower seed mix (Jeannerod et al., 2022).  Klecka et al., 2018;Nichols et al., 2019;Ouvrard et al., 2018;Warzecha et al., 2018;Wood et al., 2017). Asteraceae species have been shown to hold some of the highest nectar and pollen quantities due to their capitulum of tiny florets (Hicks et al., 2016).
Similarly, D. carota has relatively low quantities of nectar sugar per flower, but as the flowers are grouped en masse in umbellifers, and flower in high abundance, it provides large overall quantities of nectar (Hicks et al., 2016). This in turn reduces flight time and energy expenditure for adult foragers, allowing them to provide large quantities of nectar to the brood in a shorter time frame, making these species valuable resources to bees as well as other insects.
Bumblebees were particularly reliant on sown species, with few spontaneous species receiving high numbers of visits. The only plant species visited by bumblebees in the trial plots during the early season surveys were sown species (Tr. hybridum, A. vulneraria, T. officinale), indicating that spontaneous species found on farmland during this period could be relatively unattractive to bumblebees (Falk & Lewington, 2015;Wood, Holland, & Goulson, 2015a;Wood, Holland, Hughes, & Goulson, 2015b).
Solitary bees relied upon a diverse range of sown species within Asteraceae, Apiaceae, Geraniaceae and Ranunculaceae. Although solitary bees are known to often rely upon spontaneous plants in farmland (McHugh et al., 2022;Wood et al., 2017), our study shows that they will also readily use sown plants. Geranium pyrenaicum was particularly significant to species within the Halictidae family, while T. officinale agg.
provided forage for a wide range of Andrena spp. Taraxacum officinale agg., although sown in our mixes, is historically considered a horticultural (Tilman et al., 1999) and organic-farming weed (Carr, 2017), and not a weed of concern in modern farming, and is not regularly included in European wildflower seed mixes (Nichols et al., 2022a). Here we showed the importance of including it in a seed mix, as it not only had the highest species strength for wild bees, but it was the only sown species in relatively high abundance during the early-season surveys (Nichols et al., 2022a), therefore providing substantial nectar and pollen to early emerging species (Hicks et al., 2016).
We suggest that cornfield annuals are added to future wildflower seed mixes. Papaver rhoeas and G. segetum were shown to be attractive to wild bees and other insect pollinators, likely due to their high yielding pollen rewards (Hicks et al., 2016). As they are not long-lasting, they should not be relied upon to provide forage for insects yearon-year. By contrast, they can act as a 'nursery', suppressing the growth of weeds and allowing the perennial mix to establish during the first year (Emorsgate Seeds, 2021;Nichols et al., 2022a). They may be a better nurse crop than grasses as they are unlikely to persist, unlike grasses which can become too predominant, especially on fertile soils. Further research is needed as to whether grasses should be included in seed mixes, as many recent European studies trialling seed mixes appear to not include grasses (Scheper et al., 2021;Schmidt et al., 2020;Schoch et al., 2022) or are recommended at a lower proportion (pers. comm., J. Bijkirk).

Spontaneous species
Spontaneous species played a particularly important role for solitary bees, however, the inclusion of 'weeds' in a wildflower seed mix should be carefully considered. Both Cr. capillaris and Cs. vulgare grew spontaneously in the plots and were highly visited by wild bees, as seen in other studies (Balfour & Ratnieks, 2022;Carvell et al., 2007;McHugh et al., 2022;Nichols et al., 2019). This is most likely due to high nectar sugar quantities (Hicks et al., 2016), however, some spontaneous species such as Cs. vulgare, are injurious weeds, competing with crops (DEFRA, 2003;Maskell et al., 2020). Nevertheless, many species of annual arable plants arable are relatively uncompetitive, even at high densities, (Marshall et al., 2003). Herbicides also differ in their efficacy on different species offering an opportunity for selective weed control, while crops also differ in their ability to compete with weeds. The challenge is therefore to manage weed populations to benefit biodiversity (Storkey & Westbury, 2007) and ensure that they do not cause a greater ecosystem disservice than benefiting biodiversity (Tschumi et al., 2018).

Recommendations for further research
Insect-plant interactions are shaped by complex ecological mechanisms related to the functioning of the entire ecosystem. These mechanisms influence the quantity and quality of food resources offered to pollinators by plants that grow in different habitats. We conducted this study on only two farms, with particular geology, climate, and flora. Therefore, it is unknown if similar results would be obtained in other agricultural environments. We suggest primary research into flower species favoured by insect visitors is carried out in other areas, both nationally and globally, before 'pollinator targeting' seed mixes are sown. This ensures that the species sown are more likely to attract their target taxa, not only encouraging farmers to continue conservation efforts, but also saving financial resources by only planting the most useful species. Additionally, the nutritional benefits of these plant species should also be considered when designing seed mixes in the future.
We suggest our mix is trialled across a wider range of soil types, assessed over a longer time period, using larger sized plots that are more akin in real field margins (Ouvrard et al., 2018). Additionally, short-term trials such as these are primarily assessing the attractiveness of the mix to local insect pollinator communities.
More extensive and longer-term trials would be needed to determine if they can increase the community diversity and abundance.
There is always the possibility that wildflower strips are simply redistributing local insects in an area, or if they are boosting population numbers. Therefore, this mix should be trialled as an AES to test its cost, viability, and its long-term impact on insect pollinator populations.
Finally, although we have recorded non-bee visitors during our study, giving us a better picture of plant species required to support the greater insect pollinator community on farmland, these were adult foragers. Non-bee pollinators also require suitable resources for larvae, such as hostplants (Curtis et al., 2015). Over 65% of insect visits recorded to flowers were from non-bee species, and although most non-bee species provide a lower pollen deposition rate per visit than bees, due to their greater numbers they are likely to be important for the continued pollination service provided to both wildflowers and crops (Rader et al., 2016).
Therefore, to achieve a continued pollination service, it is vital all lifestages are taken into consideration when forming a seed mix, as the functioning of the population depends on the nutritional requirements of juveniles being met. It may be that larval resources limit their populations on farmland, so that providing more floral resources may not increase the population. We identified plant species that attracted high abundances of adult non-bee insect pollinators (e.g., D. carota, T. inodorum, Lc. vulgare, Ln. hispidus, G. segetum, and P. rhoeas), and it would be interesting to investigate further the resources required to also support additional life-stages of non-bee insect pollinators.
sown at a rate of 20% wildflower to 80% grasses, and FAB was sown at 100% wildflower. For ease, only the wildflower species are included in the table.