Bombus terrestris in a mass‐flowering pollinator‐dependent crop: A mutualistic relationship?

Abstract Bumblebees (Bombus spp.) rely on an abundant and diverse selection of floral resources to meet their nutritional requirements. In farmed landscapes, mass‐flowering crops can provide an important forage resource for bumblebees, with increased visitation from bumblebees into mass‐flowering crops having an additional benefit to growers who require pollination services. This study explores the mutualistic relationship between Bombus terrestris L. (buff‐tailed bumblebee), a common species in European farmland, and the mass‐flowering crop courgette (Cucurbita pepo L.) to see how effective B. terrestris is at pollinating courgette and in return how courgette may affect B. terrestris colony dynamics. By combining empirical data on nectar and pollen availability with model simulations using the novel bumblebee model Bumble‐BEEHAVE, we were able to quantify and simulate for the first time, the importance of courgette as a mass‐flowering forage resource for bumblebees. Courgette provides vast quantities of nectar to ensure a high visitation rate, which combined with abundant pollen grains, enables B. terrestris to have a high pollination potential. While B. terrestris showed a strong fidelity to courgette flowers for nectar, courgette pollen was not found in any pollen loads from returning foragers. Nonetheless, model simulations showed that early season courgette (nectar) increased the number of hibernating queens, colonies, and adult workers in the modeled landscapes. Synthesis and applications. Courgette has the potential to improve bumblebee population dynamics; however, the lack of evidence of the bees collecting courgette pollen in this study suggests that bees can only benefit from this transient nectar source if alternative floral resources, particularly pollen, are also available to fulfill bees’ nutritional requirements in space and time. Therefore, providing additional forage resources could simultaneously improve pollination services and bumblebee populations.


| INTRODUC TI ON
Loss of floral resources due to changes in land management is generally thought to be the primary driver of reported declines in pollinator populations Brown and Paxton (2009). This is because generalist flower visitors such as bumblebees (Bombus spp.) rely on an abundant and diverse selection of floral resources for nectar and pollen to meet their energy requirements: nectar is rich in sugars, a source of energy, and pollen is rich in protein which is essential for growth and development (Rotheray, Osborne, & Goulson, 2017).
In farmland, mass-flowering crops are often the intended forage resource because insect visitation can result in pollination, and therefore, increased yield (Pufal, Steffan-dewenter, & Klein, 2017). This is the case for courgette (Cucurbita pepo L.) where pollination, particularly by bumblebee species has been shown to increase yield by 39% (Knapp & Osborne, 2017). Indeed Bombus impatiens C. (a North American species) has been observed to be a highly effective pollinator in Cucurbita crops, depositing more than three times the number of pollen grains per stigma compared to Apis mellifera L. and Peponapis pruinosa S. (Artz & Nault, 2011). Quantifying the effectiveness of individual pollinator species can help growers target their pollination management to species most likely to increase yields (Ne'eman, Jürgens, Newstrom-Lloyd, Potts, & Dafni, 2010).
While mass-flowering crops may enhance pollinator densities (Westphal, Steffan-Dewenter, & Tscharntke, 2003), it is largely unknown if this is due to a transient movement of bees between patches of forage or due to an actual increase in colony growth (Holzschuh et al., 2016). This is because mass-flowering crops only provide temporary pulses of nectar and pollen unlike natural areas, with higher floral species richness, which are able to provide resources that are more stable over time (Montero-Castaño, Ortiz-Sánchez, & Vilà, 2016). Nonetheless, intense flowering periods and large areas of mass-flowering crops in the landscape may still benefit pollinators spatially and temporally, potentially improving pollination and boosting bee populations.
Since accurately studying bumblebee colony development in a field setting can be difficult (Westphal, Steffan-Dewenter, & Tscharntke, 2009;Wood, Holland, Hughes, & Goulson, 2015), this study uses an in-silico approach to simulate the population dynamics of Bombus terrestris L. in landscapes with and without courgette fields using the agent-based model Bumble-BEEHAVE (Becher et al., 2018). Although other bumblebee models exist (Crone & Williams, 2016;Häussler, Sahlin, Baey, Smith, & Clough, 2017;Olsson, Bolin, Smith, & Lonsdorf, 2015), Bumble-BEEHAVE is uniquely able to simulate the effects of multifactorial stressors on bumblebee survival at individual, colony and population levels on a daily basis, based on nectar and pollen sources which are approximated from real landscape maps of study sites.
This study explores the mutualistic relationship between B. terrestris, a common bumblebee visitor to courgette fields in the United Kingdom (Knapp & Osborne, 2017), and the mass-flowering crop courgette to ask: (a) How much pollen and nectar do courgette crops provide? (b) Is B. terrestris an effective pollinator (in terms of visitation rate and pollen transfer) of courgette? and (c) How does courgette affect B. terrestris colony development at a landscape scale (using Bumble-BEEHAVE)?
To answer these questions, we quantified the potential pollination efficiency of B. terrestris in courgette as well as the extent to which courgette fulfills bees' requirements for pollen and nectar ( Figure 1) explored at different spatial (flower/crop) and temporal (day/year) scales ( Figure 1).

| Study species
Courgette is monoecious with predominate staminate flowers until pistillate flowers gradually dominate over a season. Within a single day, both types of flower start opening around 05:30 hours before closing around 12:00 hours on the same day, and they do not open again. Flower anthesis is not thought to be directly affected by climatic events such as rainfall (Nepi & Pacini, 1993).
In the United Kingdom, courgette is usually grown over two cropping periods with flowering and harvesting lasting around 5 weeks at two separate sites, often several kilometers apart, to ensure a constant supply of courgette from the beginning of June until the end of August. Hereafter, the first cropping period is referred to as "early courgette" and the second cropping period is referred to as "late courgette.".
Although all bee species visiting courgette were recorded during pollinator surveys, B. terrestris was the focus of this study because of their natural abundance at study sites and availability as commercial colonies (Biobest Biological Systems, Belgium) which were required to quantify the proportion of courgette pollen in B. terrestris' diet ( Figure 1). Colonies were placed in each field, with sugar water but no additional pollen at a density of three colonies per field.

| Study sites
The empirical data for this study ( Each field had an average field size of 3.6 ± 0.3 ha SE and was situated at least 2 km from any other courgette field so that pollinator communities were unlikely to be shared between fields (Vaissière, Freitas, & Gemmill-Herren, 2011

| Quantifying nectar and pollen resources in courgette flowers (2017)
The standing crop of nectar and pollen and the 24-hr secretion rate of nectar (Corbet, 2003) were quantified to show the availability of pollen and nectar over time as well as to parameterize Bumble-BEEHAVE ( Figure 1). The weight of sugar (mg; nectar) and pollen (mg) were calculated per flower. Detailed information about pollen and nectar measurements are in Supporting Information Appendix S1.

| Bee visitation to courgette and wild flowers (2016 and 2017)
To quantify B. terrestris abundance at courgette flowers, and therefore, their potential pollination efficiency (Figure 1 In 2017, additional transects in the crop and the field margins were simultaneously surveyed by two observers from 08:15 to 15:30 hours at ten sites, resulting in an additional 640 transects.
This was to capture pollinator activity in the 4 hr either side of courgette senescence, which occurs around 12:00 hours.
All bee species and the plant species they were feeding on, for nectar or pollen, were recorded to species level. However, B. terrestris and bees belonging to the Bombus lucorum L. complex were all recorded as "B. terrestris" due to difficulties in reliably distinguishing workers in the field (Murray, Fitzpatrick, Brown, & Paxton, 2008).
Since colonies of B. terrestris were added to all fields in 2017, foragers from these colonies are highly likely to have been recorded on pollinator transects.

| Swabbing B. terrestris for pollen grains
To quantify the number of courgette pollen grains carried on B. terrestris, and therefore, their potential pollination efficiency (

| Pollen grains on stigmas
To quantify courgette pollination, pollen accumulation per stigma was quantified ( Figure 1). A total of 20 stigmas were removed from pistillate flowers and placed into centrifuge tubes every 90 min from 05:30 to 12:00 hours over 2 days at two different sites (10 stigmas per time point per day). In the laboratory, 1/6 of the stigma (one half of a lobe) was dissected and gently squashed onto a microscope slide; fuchsin jelly was then melted over the stigma, under a coverslip (Kremen et al., 2002). The number of courgette pollen grains were then counted with a 20× magnification and multiplied by six to achieve an estimate of pollen deposition for the whole stigma.

| Yield
To further quantify courgette pollination (Figure 1), yield measurements were also taken. To do this, commercial colonies of B. terrestris were closed at one field site to quantify courgette yield without managed B. terrestris present. This was done over five nonconsecutive days for a total of 100 pistillate flowers (20 flowers per day), following the methodology for "open pollination" in Knapp and Osborne (2017).

| Pollen loads from B. terrestris
To quantify the proportion of courgette pollen in B. terrestris' diet ( Figure 1), "forager trap modules" (Martin et al., 2006) were placed onto all commercial colonies within a field for around 45 min, between 07:00 and 09:00 hours. Once trapped on returning from a foraging trip, workers were narcotised in situ using CO 2 for 30 s and the number of bees carrying (and not carrying) pollen loads were recorded. One pollen pellet from one of the corbiculae on each bee, that is, half of their total pollen load, was placed into a centrifuge tube and taken back to the laboratory. Here, all pollen loads (n = 394) were sorted to color and all yellow pollen loads checked to see if they were courgette, which has large (180-200 µm in diameter) and distinctive pollen grains (Nepi & Pacini, 1993). A subset (n = 56) of all pollen loads were identified to species where possible using Sawyer (1981) and a microscope. All foragers were returned to their colony within an hour of being caught. Pollen loads were taken from 42 colonies across the 14 sites and each site was surveyed on a separate day.

| Habitat maps
Habitat maps for each study site were required to estimate the amount of forage and nesting sites, that is, seminatural habitat and mass-flowering crops, available to bumblebees in the landscape ( Figure 1) Figure S1).

| Bumble-BEEHAVE simulations using BEE-STEWARD
Simulations were run in BEE-STEWARD (www.beesteward.co.uk), a software tool that combines in a user-friendly way the bumblebee model Bumble-BEEHAVE and the landscape defining features of BEESCOUT (Becher et al., 2016). BEESCOUT was developed as the landscape module for the honeybee model BEEHAVE (Becher et al., 2014) and for Bumble-BEEHAVE (Becher et al., 2018), and creates input files from images of landscape maps. These input files define the number and specification of food sources such as, nectar and pollen, flowering phenology, and therefore, represent landscapes in the BEEHAVE and Bumble-BEEHAVE models. BEE-STEWARDS' interface also enables users to simulate the effects that different management options, such as changing crop types will have on bumblebee population dynamics.  Table S1). Courgette fields were specified as either "early courgette," flowering from the beginning of June until the middle of July, or "late courgette" flowering from the middle of July until the end of August, to reflect the cropping practices of courgette production in the UK.
A map of each study site was separately input into the model and manually edited (if needed) using the functions available within the program (Becher et al., 2016).
In order to reduce computational time and to ensure that simulations were based solely on populations in equilibrium (Hui, 2006),  Table S1). All simulations were run 10 times per landscape and cropping scenario, totaling 420 simulations.
The average number of overwintering queens, colonies, and adult workers were calculated daily for each landscape over 11 years.

| Statistical analysis
All analyses were carried out using R (R Core Team, 2017). For empirical data, independent sample t tests were used to compare the differences in mean sugar production (g) between staminate and pistillate flowers (over 24 hr and every 90 min), pollen depletion (mg/ flower) between 05:30 and 10:00 hours, pollen accumulation on stigmas (grains/stigma) between 05:30 and 11:30 hours, and B. terrestris abundance in the margin and cropped area per hour.
For simulated data, the effect of cropping scenario (fixed effect) was explored in relation to the peak number of hibernating queens (day 365), adult workers (day 149), and colonies (day 149) in year 11 using linear mixed-effects models with site specified as a random effect. Post hoc Tukey tests were calculated using the multcomp package (Hothorn, Bretz, & Westfall, 2008). All means are presented with their associated standard error unless otherwise stated.

| Nectar and pollen measurements from courgette
The secretion rate of nectar, that is, the weight of sugar produced

| Visitation to courgette and wildflowers
Apis mellifera and B. terrestris were the most abundant pollinator species observed visiting courgette flowers across the 2 years of this study, although commercial colonies of B. terrestris were added to fields in 2017 (Figure 3). Bombus terrestris showed a more equal preference to staminate and pistillate flowers then A. mellifera (Figure 3). In

| Pollination of courgette flowers
Bombus terrestris carried an average of 1,866 ± 476 (n = 13) pollen grains on their bodies, more than A. mellifera which carried an average of 122 ± 39 (n = 4) pollen grains on their bodies.
The percentage of open-pollinated pistillate flowers setting fruit was very high across the 5 days of surveying at 97% ± 2% (n = 96).

| Pollen loads
None of the 394 pollen loads collected from the 42 colonies of B. terrestris contained courgette pollen (Supporting Information Table S2).
Brassica spp. (15), bramble (11), and common poppy (seven) were the most common pollen species identified out of a subsample (n = 56) of pollen loads (Supporting Information Table S2). Consequently, all courgette flowers were specified as having a pollen resource value of zero in BEE-STEWARD (Supporting Information Table S1).  August; Figure 6).

| D ISCUSS I ON
This study clearly demonstrates a mutualistic relationship between courgette flowers and B. terrestris that is beneficial to both, improving pollination success and colony dynamics (Bailes, Ollerton, Pattrick, & Glover, 2015;Holzschuh et al., 2016). Courgette, offers an abundant source of nectar to attract pollinators to its flowers for pollination (Vidal, Jong, Wien, & Morse, 2006). Indeed per m 2 , courgette offers more nectar (0.35 ml) than oilseed rape (0.30 ml), and field bean (0.092 ml) (Becher et al., 2016), and is therefore a high value mass-flowering crop in terms of nectar production.
Results showed that over 24 hr pistillate flowers produce significantly more sugar than staminate flowers. This overall higher sugar content combined with nectaries which are harder to access than staminate flowers is thought to be why bee species show a preference for, and spend longer at pistillate flowers (Artz & Nault, 2011;Nepi & Pacini, 1993;Phillips & Gardiner, 2015;Tepedino, 1981). At a field scale, B. terrestris also showed a strong fidelity to courgette, visiting crop flowers more often than wildflowers in the hedgerows, in the morning when courgette flowers were open, providing the first empirical evidence of B. terrestris fidelity to a Cucurbita crop (Petersen, Reiners, & Nault, 2013).
However, personal observations showed B. terrestris removing excess courgette pollen grains from their bodies early in the morning, supporting the findings of Nepi and Pacini (1993). Nonetheless, B. terrestris was still observed to carry more loose pollen grains on their body, and therefore, have a higher pollination potential than A. mellifera. Indeed pollen was still transferred to stigmas well after anthesis and by the end of the morning, stigmas had received an adequate number of pollen grains (4,749 ± 441) for optimum fruit set as ~1,200 are thought to be required for maximal fruit set in pumpkin (Vidal, Jong, Wien, Morse, & a., 2010 (Petersen et al., 2013), its large sticky grains may make it difficult for B. terrestris to collect (Vaissière & Vinson 1994). Bombus terrestris may also avoid collecting Cucurbita pollen, since as a generalist species it can visit alternative, more easily obtainable pollen, unlike Peponapis and Xenoglossa spp.
which as Cucurbita specialists are thought to rear their offspring exclusively on Cucurbita pollen (Tepedino, 1981). This may be why no pollen loads from returning B. terrestris foragers contained courgette pollen.
After courgette flower senescence (within a day) B. terrestris appeared to "switch" from courgette to hedgerow flowers, evidenced by the diverse range of pollen loads collected from returning B. terrestris foragers. While some of these plant species may occur in hedgerows immediately surrounding courgette fields, others may be from species located further away. This highlights the importance of maintaining wildflowers at different spatial scales to fulfill bees' requirements for nectar and pollen beyond that of the focal crop.
Indeed flower-rich areas have been shown to increase colony density (Wood et al., 2015) and food supplementation shown to increase colony development, particularly of queen and male bumblebees (Pelletier & McNeil, 2003). However, the extent to which pollinators are attracted into mass-flowering crops will vary depending on the relative quality and quantity of floral resources in the mass-flowering crop and nearby seminatural habitat. In this study, it appears that wildflowers near to mass-flowering courgette facilitate pollination services to courgette, supporting bumblebee nutrition without distracting bees from courgette flowers. Indeed, wildflower species richness in courgette fields has been shown to be the most important factor for determining bumblebee abundance at courgette flowers (Knapp, Shaw, & Osborne, 2018). Therefore, wildflowers around courgette fields could attract bumblebees to courgette flowers also provide additional forage.
Given courgette's bountiful, yet transient supply of nectar, bumblebee population dynamics were shown (using Bumble-BEEHAVE) to improve in landscapes with early flowering courgette compared to a no courgette baseline. As bumblebee foragers are generally most active mid-summer, early courgette was the best cropping scenario for concurrently achieving more forager visits (pollination potential) and more food (nectar only) to be brought back to the colony.
However, bees can only benefit from the additional energy provided by courgette nectar, which will help to reduce foraging efforts, if protein-providing pollen is also available to raise their brood. Empirical data showed that, within a day, bees were able to utilize courgette and wildflowers for nectar ( Figure 2b) as well as wildflowers for pollen (Supporting Information Table S2). This supports model results which showed at a coarser temporal scale that with more nectar, colonies were able to grow and subsequently forage on more, additional resources for pollen. Subsequently, early courgette supports more adult workers (foragers), colonies, and hibernating queens for subsequent years compared to late, and no courgette landscapes.
Thus, planting early courgette and late courgette in fields adjacent to each other could improve forager numbers in late courgette and further improve bumblebee populations for subsequent years (Riedinger, Renner, Rundlöf, Steffan-Dewenter, & Holzschuh, 2014 (Westphal et al., 2009). This is because while oilseed rape can improve colony establishment and growth of bumblebees, the lack of resources later in the season mean there is no increase in the number of males or queens produced (Herrmann et al., 2007;Westphal et al., 2009). This lack of phenological matching is also true of late courgette which despite offering resources later in the season (unlike oilseed rape) still misses the key period of bumblebee foraging. However, Rundlöf, Persson, Smith, and Bommarco (2014) observed more queen and male bumblebees on transects around fields of late-flowering red clover, suggesting results could be specific to flower and pollinator species.
Interestingly, the average number of colonies per landscape decline around day 119, which may be a result of willow species, common to hedgerows and scrub in Bumble-BEEHAVE's input files, no longer flowering.

ACK N OWLED G M ENTS
We would like to thank all the farmers that allowed us to use their land during this study. We would also like to thank Ellie Brown,

AUTH O R S' CO NTR I B UTI O N S
JK and JO conceived the ideas and designed methodology; JK and CR collected the data; JK and MB analyzed the data; MB and GTD created the BEE-STEWARD software and provided support on simulations. JK led the writing of the manuscript. All authors contributed critically to the drafts and gave final approval for publication.

DATA ACCE SS I B I LIT Y
The research data supporting this publication are openly available from the University of Exeter's institutional repository at: https:// doi.org/10.24378/exe.823.