CUCURBITS AS A MODEL SYSTEM FOR CROP POLLINATION MANAGEMENT

Cucurbit crops have steadily increased in production over the last 50 years, particularly in Asia where pioneering technological advancements and genetic improvements have created new hybrid varieties. Generally, cucurbits are dependent on insect-pollination for fruit set and are popular species for pollination studies. This review systematically summarises pollination research conducted in the major food genera of cucurbits: Cucurbita, Cucumis, and Citrullus, to ask: 1) what are cucurbits’ requirement for pollination and their most effective pollinators? And 2) Does pollinator management increase pollinator visitation to, and yield of, cucurbit crops? These accounts of cucurbit pollination demonstrate that wild bee species such as Bombus terrestris, B. impatiens and Eucera spp. were frequently able to fulfil the pollination requirements of multiple cucurbit species. However, pollinator behaviour, pollen deposition on stigmas, and pollinators’ contribution to yield vary between cucurbit species and study site. Nonetheless, the provision of additional floral resources at both field and farm scales may help to encourage pollination of cucurbit species whilst supporting pollinators’ nutritional requirements beyond those already provided by the cucurbit crop. Synthesising studies on cucurbits’ requirement for pollination and how pollinators vary spatially and temporally in the landscape can extend beyond cucurbit systems to inform growers and pollination ecologists of other pollinatordependent crop species wishing to maximise pollination services, species conservation; or both.


INTRODUCTION
Cucurbits (Cucurbitaceae) are a large plant family which include major food crops within the Cucurbita (squash, pumpkin, courgette), Cucumis (cucumber, melon) and Citrullus (watermelon) genera (Kumar 2016). Over centuries, cucurbits have been domesticated for their fleshy fruits, roots, leaves, shoots, seeds, and flowers for food and commodity goods and thus are economically important crops (Bates et al. 1990;Bisognin 2002). Cultivated cucurbits are grown in a cultiar of agricultural environments from widespread monocultures to small-scale traditional garden systems and many species are able to persist in environmental conditions usually considered marginal for agriculture (Bates et al. 1990).
From a biological viewpoint, cucurbits' co-evolution with insects has inspired much scientific intrigue. For example, their ability to produce bitter cucurbitacins as a defence against insect herbivory has led to research into whether these compounds can be used for biological control, particularly against beetles (Metcalf et al. 1982;Adler & Hazzard 2009;Cavanagh et al. 2010). Likewise, cucurbits' dependency on pollination (Free 1993) means that cucurbit flowers offer large quantities of nectar and pollen as floral rewards to visiting insects, such as solitary bees, bumblebees and honeybees (Tepedino 1981;Vidal et al. 2006). In particular, the North American squash and gourd bees belonging to the genera Eucera ( Fig. 2A) and Xenoglossa are thought to rely exclusively on Cucurbita pollen to rear their offspring (Hurd, Linsley & Michelbacher 1974;Tepedino 1981).
From an agricultural viewpoint, various mechanisms have been explored to improve cucurbit yield such as improving the sex expression of flowers (Rodriguez-Granados et al. 2017) and producing F1 hybrid seed (Robinson 2000). Indeed, the yield (per hectare) of cucurbit crops has steadily increased over the last 50 years, particularly in Asia where pioneering technological advancements and genetic improvements, especially with seedless varieties not requiring pollination, have increased global production ( Fig. 1) (McCreight et al. 2013). Likewise, and most relevant to this review, cucurbit yield can also be increased by improving pollination (Hoehn et al. 2008;Kouonon et al. 2009).
Cucurbits are a popular plant family for pollination studies, particularly within the major food genera of Cucurbita (Hurd et al. 1971;Willis & Kevan 1995;, Cucumis (Adamson et al. 2012;Ali et al. 2015;Motzke et al. 2015), and Citrullus (Kremen et al. 2004;Winfree et al. 2008;Pisanty et al. 2015). This is likely because many cucurbit species are monoecious (Box 1) with many ovules, so manipulating pollen deposition to examine its effect on yield is relatively straight forward. Because cucurbits are such a large and genetically diverse plant family, synthesising studies on their pollination can extend beyond these systems to inform pollination ecologists and growers of other pollinator-dependent species on how to maximise pollination services, species conservation; or both. FIGURE 1. Regional production (primary y axis) and average global yield (secondary y axis) of Cucurbita species from 1961 to 2013. Data source: FAOSTAT (Aggregate, may include official, semi-official, estimated or calculated data). 18 1961 1965 1969 1973 1977 1981 1985 1989 1993 1997 2001 2005 2009 2013 Global Yield (T/Ha) This review encompasses multiple ways in which pollination is directly or indirectly quantified (Box 2). Broadly, there are two approaches for quantifying pollination: the first directly estimates pollinator performance in terms of pollinator behaviour and / or pollen deposition on stigmas, whilst the second indirectly estimates pollinators' contribution to yield, usually measured as seed set or fruit weight (Ne'eman et al. 2010). Therefore, in this review pollinator visitation to crop flowers, abundance at crop flowers, and pollen deposition (whilst stigmas are receptive) are all considered to be measures of pollinator performance. On the other hand, seed set, fruit weight, fruit weight per plant, fruit number, and percentage fruit set are all considered measures of yield.
Measures of yield should be interpreted carefully since they depend on multiple resources such as water, nutrients, and pollen availability. But broadly, fruit number and percentage fruit set reflect the number of flowers pollinated, whilst seed set, and average fruit weight reflect the quantity of pollen that a flower receives (affecting the number of seeds per fruit or the fruit size). Arguably, when other environmental factors that influence fruit production e.g. soil type, resource availability and cultivation practices cannot be standardised, single visit pollen deposition is the most direct measure of pollination success (Kremen et al., 2004). It should also be BOX 1. Scientific and common names of different cucurbit species included within this review. For each species the most common sex expressions, breeding systems, as well as average flower ratios and flower longevity are listed based on Bomfim et al. (2016) (Bomfim et al. 2016) and diurnal pollen transfer is likely to be important in cucurbits since stigma receptivity rapidly declines within a day (Bomfim et al. 2016). Thus, flowers on which pollen is deposited early in the day are more likely to set fruit. Also, the presence of already pollinated fruits has been shown to significantly decrease the number of pistillate flowers and increase the likelihood of new fruit aborting in Cucurbita pepo (Stephenson et al. 1988).

Pollinator dependency
Cucurbits have been described as having an 'essential need' for insect-mediated pollination (Free 1993;Klein et al. 2007) and research has shown that seed number (Roldán-Serrano & Guerra-Sanz 2005) and fruit set of Cucurbita pepo (Roldán-Serrano & Guerra-Sanz 2005;Vidal et al. 2010) and Cucumis sativus (Gingras, Gingras & DeOliveira 1999;Walters 2005) are positively correlated with the number of pollinator visits that each flower receives. Similarly, fruit has been shown to abort in the absence of pollination in Cucumis sativus (Motzke et al. 2015), Cucumis melo (Kouonon et al. 2009), Cucurbita moschata (Hoehn et al. 2008), and Cucurbita pepo (Martínez et al. 2014). Interestingly, the addition of honey bee colonies did not increase the yield of smaller-sized Cucurbita pepo varieties as much as larger ones; suggesting that alternative pollinator species may be more important in these varieties or that some smaller varieties may be able to set fruit with a smaller pollen load (Walters & Taylor 2006).
However, several varieties of Cucurbita pepo have been observed to set fruit in the absence of fertilisation, and therefore pollination, via parthenocarpy (Robinson & Reiners 1999;Kurtar 2003;Martínez et al. 2013;Knapp & Osborne 2017). This genetic trait is desirable for growers in crops that are usually pollinator-dependent because fruit is able to form is conditions that are adverse for pollinators, potentially extending geographic and climatic ranges of production (Knapp et al. 2016). Seedlessness in fruits, caused by the lack of fertilisation, can be an important measure of quality, for example in green-house grown Cucumis sativus, where seedlessness is generally appreciated by consumers (Knapp et al. 2016). However, evidence suggests that parthenocarpic varieties may still produce a greater quantity and quality of fruits, including a higher sugar content (Shin et al. 2007), when they are pollinated by insects (Martínez et al. 2013;Robinson & Reiners 1999;Nicodemo et al. 2013).
Whilst there have been extensive selective breeding programmes, use of growth hormones and even genetic modifications for parthenocarpy (Knapp et al. 2016), this review focuses on pollination of pollinator-dependent varieties. Nonetheless, evidence of parthenocarpy in cucurbit crops suggests that, to get a more complete picture of pollinator dependence in crops, varietal information is required -both in terms of pollinator dependence, but also in terms of choices that farmers are making (Klein et al. 2007;Melathopoulos et al. 2015). Realistically, the best way of obtaining this information is if the pollination requirements of each variety are tested by organisations or institutes conducting variety trials and that this information is made freely available alongside other trait details.

Relative importance of pollination to yield
Whilst pollination affects cucurbit yield, there are many other environmental factors which contribute and interact with each other to influence fruit set, such as nutrient and water availability, herbivore damage and weed competition (Tab. 1). If any of these factors are deficient then yield may decrease, widening the yield gap between actual and attainable yields (Bommarco et al. 2013). For example, Motzke et al. (2015) showed that weed control and fertilisation were able When tested independently, pollination, weed control, and fertilisation had a positive effect on fruit weight, however, herbivore control had no effect on fruit weight. When tested in combination, pollination and weed control had a positive effect on fruit weight. Cucumis sativus Monoecious variety, assumed to be dependent on insect pollination.
Field study, USA When tested independently, root herbivory had a negative effect on fruit weight, whilst leaf herbivory and pollination had no effect on fruit weight. When tested in combination, pollination and root herbivory, pollination and leaf herbivory, and pollination, root herbivory and leaf herbivory had no effect on fruit weight. When tested independently, root herbivory had a negative effect on seed set, leaf herbivory had positive effect on seed set and pollination had no effect on seed set. When tested in combination, pollination and root herbivory had a positive effect on seed set (pollination mitigated root herbivory), whilst pollination and leaf herbivory, and pollination, root herbivory and leaf herbivory had no effect on seed set. Monoecious variety, assumed to be dependent on insect pollination.
Field study, USA When tested independently, pollination and root herbivory had a positive effect on fruit weight, however, leaf herbivory had no effect on fruit weight. When tested in combination, pollination and root herbivory and pollination and leaf herbivory had no effect on fruit weight. Pollination, root herbivory and leaf herbivory, and their interactions had no effect on seed set. 'Pollination' compared hand pollination to open pollination. (Hladun & Adler 2009) to reduce the yield gap of Cucumis sativus by 45% and 18% respectively; however, these factors, even in combination, were unable to compensate for a total absence of pollination which increased the yield gap by 75% (the difference between open and non-pollinated flowers) (Tab. 1). In contrast, root herbivory had a negative effect on Cucumis sativus yield despite crop flowers receiving high levels of pollination, demonstrating that below-ground herbivores may have profound effects on plant performance (Barber et al. 2011) (Tab. 1). Soil nitrogen has also been shown to increase the number, weight and viability of pollen grains with flowers hand pollinated from pollen grown in higher nitrogen environments observed to produce Cucurbita pepo fruits with more seeds compared to hand pollinated flowers using pollen from lower nitrogen environments (Lau Tak-Cheung & Stephenson 1993).
Relatively few studies in cucurbits (Tab. 1) have examined how a plant's health interacts with the level of pollination it has received to influence fruit set. Drought stress, nutrient deficiencies and diseases such as cucumber mosaic virus and powdery mildew are relatively common in cucurbit production, often more so than direct pest damage (Agriculture and Horticulture Developement Board 2013). Since biological control may reduce aphid populations, which are common vectors of cucumber mosaic virus, and fungicides can reduce powdery mildew, fully-factorial experiments could be established to test the effect of disease control (biological and/or chemical) in relation to different levels of pollination (i.e. hand, open and no pollination), nutrient (e.g. fertiliser use), and water availability (e.g. irrigation and/or rain covers) on fruit set. Results from this type of experiment would help growers to realise the importance of pollination relative to other factors such as disease. A more complete understanding of the environmental factors affecting fruit set is vital to ensure that ecologists do not promote the conservation of one ecosystem service at the expense of another and that growers are able to prioritise key limiting services in their management for optimal crop yields.

Pollinator performance
Cucurbits are pollinated by multiple species of honeybees, bumblebees, and solitary bees (Tab. 2). Consequently, several studies have compared pollinator performance in cucurbit crops by exploring aspects of pollinator behaviour (visitation rate and diurnal activity patterns, single visit pollen deposition) and/ or their contribution to yield (fruit weight, fruit number, seed number) (Tab. 2). Cucumis melo Andromonoecious variety, however, perfect flowers assumed to be dependent on pollination.
No difference between A. mellifera and B. impatiens at single-visit pollen deposition, however, A. mellifera had a lower pollen removal to pollen deposition ratio, suggesting it is a more effective pollinator.
Field study, USA (Goodell & Thomson 2007) Cucurbita maxima Monoecious variety, assumed to be dependent on insect pollination.
Bombus spp. were more effective than A. mellifera at pollen deposition, however, A. mellifera had a higher visitation frequency than Bombus spp.
Field study, Germany (Pfister et al. 2018) Cucurbita moschata Monoecious variety, no fruit set in no pollination controls.
High species richness of pollinators increased seed set.
Field study, Indonesia (Hoehn et al. 2008) Cucurbita moschata   observed Bombus impatiens C., depositing more than three times the number of pollen grains per stigma and nearly always contacting the stigma compared to A. mellifera and Eucera pruinosa S, in Cucurbita pepo. On the other hand, Eucera limitaris was more abundant and more effective than A. mellifera at single-visit pollen deposition (no Bombus spp. present) in Cucurbita moschata (Canto-aguilar & Veterinaria 2000). Indeed, P. pruinosa has been shown to forage for Cucurbita pollen (Willis & Kevan 1995) and Lassioglossum spp. forage for Citrullus pollen (Njoroge et al. 2010) early in the morning and thus have a higher pollination potential. However, these foraging visits will only contribute to pollination if pistillate flowers are also visited. Outside of intensive production systems in North America and Europe, more diverse assemblages of pollinators have been observed to fulfil the pollination requirements of Cucumis melo, Citrullus lanatus (Ali et al. 2015), Cucurbita pepo (Ali et al. 2014) and Cucurbita moschata (Hoehn et al. 2008) (Tab. 2). Likewise, Pisanty et al. (2015) observed spatial and temporal variation in pollinator species' visitation to Citrullus lanatus, suggesting niche complementarity.
Since the majority of cucurbit species are monoecious (Box 1), consideration must also be given to whether some pollinator species show a preference for either staminate or pistillate flowers which could significantly influence pollen transfer. Indeed, bee species have been shown to preferentially choose and forage for longer in staminate flowers as their nectar has a higher sugar content (Knapp et al. 2018) and the nectaries are harder to access than pistillate flowers (Tepedino 1981;Phillips & Gardiner 2015). In the United Kingdom, Bombus terrestris L. has been shown to have a more equal preference for staminate (Fig. 2B) and pistillate Cucurbita pepo flowers (Fig. 2C) and carried more loose pollen grains than A. mellifera, thought to be desirable for optimum pollen transfer and therefore, fruit set (Knapp et al. 2018). This study and others from outside North America (Tab.2 and Tab. 4) have also shown that maximal Cucurbita yields can be achieved without Eucera and Xenoglossa species, North American bee species which are frequently cited as the most effective pollinators of Cucurbita crops (Hurd et al. 1974;Tepedino 1981).

Supplementing fields with managed pollinators
To mitigate potential fluctuations in wild bee abundance some cucurbit producers choose to supplement fields with managed bee species (Mader et al. 2010). Despite managed bee species being present in fields, visitation rates to crop flowers can be variable and benefits to crop yields sometimes not seen (Tab. 3). This could be because of a high abundance of wild pollinators and therefore little difference in pollination between control sites and those where managed pollinators have been added Petersen et al. 2013. This suggests that in many cases pollination services by wild bees may be sufficient for optimal yields (Pfister et al. 2018). Therefore, before supplementing fields with managed pollinators, it would be advantageous to determine if a site is experiencing a pollination deficit. This can be done by comparing yields from open and hand pollinated flowers  or by combining pollinator visitation (Julier & Roulston 2009) or pollen deposition data at sites (Xie & An 2014;Pfister et al. 2018) with published data on the pollination requirements of the crop (Tab. 3). If farmers could use similar techniques (Fig. 3) then they could potentially reduce the amount that they spend on hiring in temporary pollination services.

Management for wild pollinators
Cucurbit yield can also increase if a crop is surrounded by more diverse habitats where increased species richness and abundance of wild pollinators can improve pollination services (Hoehn et al. 2008;Garibaldi et al. 2011) and provide insurance against any pollinator loss (Shuler et al. 2005). Improving the quantity and quality of pollen and nectar resources available for pollinators, and allowing areas to remain undisturbed for the insects to nest, mate, and hibernate could benefit pollinator populations and therefore reduce pollination deficits (Bommarco et al. 2013). This can be done at a field scale by providing additional areas for food and nesting and following principles of integrated pest management, as well as at a landscape scale by promoting larger areas of natural habitat (Tab. 4). However, the effectiveness of field scale pollinator-supporting practices are often variable (Tab. 4), with more simplistic landscapes generally showing greater benefits after interventions than complex landscapes (Batáry, Báldi, Kleijn, & Tscharntke, 2011;Scheper et al., 2013, Tab. 4).
Wild flowers co-flowering with crops have been shown to increase solitary bee abundance in Cucumis melo and Citrullus lanatus (Winfree et al. 2008) as well as Bombus abundance in Cucurbita pepo (Knapp et al. 2018) (Tab. 4). Naturallyoccurring wild flowers such as agricultural weeds and hedgerow flowers are frequently overlooked floral resources for pollinators (Bretagnolle & Gaba 2015) despite being free and sustainable, and thus easily available to growers. These wild flowers are unlikely to be competing with cucurbit flowers for pollination services as both B. terrestris and A. mellifera have been observed to visit crop flowers more often than wild flowers in the morning when Cucurbita pepo flowers were open, before 'switching' to wild flowers after C. pepo senescence (Knapp et al. 2018). These findings provide the first evidence of bee fidelity (from non-specialists) to a Cucurbita crop.
Tillage has been shown to have variable results on bee abundance and visitation (Tab 4.), despite P. pruinosa preferentially laying their eggs in crop areas at depths around 12 to 30 cm (Julier & Roulston 2009;Hurd et al. 1974). For example, no-tillage farms have been shown to have an almost three-fold increase in P. pruinosa density (Shuler et al. 2005), whilst other studies have observed no effect of tillage on P. pruinosa abundance (Julier & Roulston 2009, Minter & Bessin 2014. This may be a result of survey timings, as Julier & Roulston (2009) surveyed in mid-August, compared to Shuler et al. (2005) who surveyed in July when P. pruinosa is thought to emerge (Mathewson 1968). Given P. pruinosa requirement for cucurbit pollen, allowing areas to remain undisturbed for nesting and hibernation near to the crop may be particularly important (Mathewson 1968). A. mellifera had no effect on A. mellifera visitation to crop flowers and fruit weight.
B. impatiens had no effect B. impatiens visitation to crop flowers and fruit weight. (Petersen et al. 2013) Cucurbita pepo Monoecious variety assumed to be dependent on pollination.
Field study, USA Pollination not limiting to yield-no difference in yield between hand-and openpollinated flowers.
B. impatiens had no effect on B. impatiens visitation, fruit weight, and seed set.  Interestingly, in the studies we found, pesticide /insecticide use had no effect on bee visitation (Shuler et al. 2005;Motzke et al. 2015), pollen deposition (Kremen et al. 2004), or yield (Motzke et al. 2015 Cavanagh et al. (2010) showed that trap crops could reduce insecticide use by 97% compared to control fields. Importantly, both studies found no effect of trap crops on pollinator visitation to Cucurbita moschata, suggesting no competition between trap crops and focal crops for pollination services (Adler & Hazzard 2009;Cavanagh et al. 2010, Tab. 4).
At a landscape scale, pollination services to cucurbit crops by wild bees have been shown to relate to the amount of natural habitat surrounding a site (Tab. 4). For example, Kremen et al. (2004) found that pollination by native bees in Citrullus lanatus were strongly associated with the proportion of natural habitat within a 1 to 2.5 km radius of farm sites, meaning that based on the area of natural habitat, pollination services to a given site could be estimated (Kremen et al. 2004). The effect of natural habitat was also more important than organic versus conventional farming for predicting pollen deposition and pollinator abundance in Citrullus lanatus (Kremen et al. 2004, Tab. 4). As many agricultural systems are isolated from natural habitats, crop producers may need to provide floral resources and nesting sites suitable for pollinators. In the UK, farm stewardship schemes provide guidance on hedgerow and field margin management, particularly favoured by bumblebee species (Osborne et al. 2008;Carvell et al. 2015;Dicks et al. 2015;Wood et al. 2015). Costs can also be directly offset by increased profit from improved quality and quantity of yields. For example, considering an example of a non-cucurbit crop: the economic benefit of improved Vaccinium corymbosum L. yields following wild flower establishment has been shown to exceed the original cost of implementation (Blaauw & Isaacs 2014).
Large areas of mass-flowering crops, including the cucurbit crop itself, may dilute pollinator densities or, if the area is small, concentrate pollinator densities (Holzschuh et al. 2016). This will be especially pronounced if additional food and nesting sites are not provided, meaning that pollinators move transiently between available forage rather than increasing their population size (Holzschuh et al. 2016).
In a UK study of Cucurbita pepo, whilst B. terrestris showed a strong fidelity to the crop flowers' bountiful nectar, no foragers were found returning to their colonies with C. pepo pollen loads (Knapp et al. 2018). Despite C. pepo pollen being relatively high in protein (Petersen et al. 2013), its large sticky grains may make it difficult for the bees to collect (Vaissière & Vinson 1994). Indeed, B. terrestris has been observed to remove excess pollen grains from their bodies early in the morning (Personal observations). Whilst B. terrestris has been observed to collect Cucurbita pollen in flight cages (Vaissière & Vinson 1994), no studies have observed B. terrestris collecting cucurbit pollen in open fields. Therefore, B. terrestris may avoid collecting Cucurbita pollen, since as a generalist species it can visit alternative, more easily obtainable (in open field settings) or more desirable pollen.
Combining this empirical data on Cucurbita pepo nectar and pollen with model simulations using the novel bumblebee model Bumble-BEEHAVE (Becher and Twiston-Davies et al. 2018), showed that populations of B. terrestris can only benefit from the transient nectar source of C. pepo if alternative floral resources (particularly pollen) are also available to fulfil bees' nutritional requirements in space and time (Knapp et al. 2018).
The complexity of pollinator-supporting practices can become further complicated when species-level responses are taken into consideration. For example, Apis spp. are less likely to increase their colony size in response to pollinatorsupporting practices because their populations are artificially maintained by beekeepers (Tab. 4). Since previous research has shown that Cucurbita crops are primarily serviced by long range, generalist pollinators: B. impatiens and A. mellifera in the USA ) and B. terrestris and A. mellifera in the UK (Knapp and Osborne 2017) increasing forage and nesting may be less important to cucurbit crops, particularly in already heterogeneous environments (Tab. 4). This highlights the need to match pollinator-supportive management practices with the surrounding landscape and crops' individual requirements for pollination, since an increase in pollinator abundance and/or species richness may not necessarily be required for yield to be improved (Kleijn et al. 2015;Winfree et al. 2015). However, relying on a few generalist species for pollination services may be risky as any declines in their populations are also likely to affect specialist and/ or less common bee species which may also be effective pollinators.

Value of pollination to growers
Quantifying the economic value of pollination can be useful for informing decision making at farm and policy levels (Hanley et al. 2014). Differentiating between the economic value of a crop's dependence on insect pollination and its pollination deficit will show the economic benefit of increasing pollinator populations as well as the economic cost that a decline in pollinator populations may have. Consequently, quantifying the economics of pollination is a fundamental way for growers to understand the implications that changes in pollinator populations may have on their yield and economic return. Economic valuations in the Cucurbita pepo variety 'Tosca' showed that whilst 56% of fruit set could be achieved via parthenocarpy, the total economic value of insect pollination to production was estimated to be worth approximately £3,398/ha (Knapp & Osborne 2017). However, this economic valuation was based on the pollinator dependency and pollination deficit of just one Cucurbita pepo variety, thus inter-variety differences in pollinator dependence, or site-specific levels of pollination deficit may increase or decrease this economic value. An analogous example of this comes from apple crops: Malus domestica in the UK, where the variety 'Cox' is estimated to have a pollination deficit of £146/ha, compared to the variety 'Gala' which had a much higher pollination deficit of £6,459/ha (Garratt et al. 2013). This is because 'Gala' is more pollinator-dependent and has a larger pollination deficit due to high yields from handpollinated flowers, compared to 'Cox'. Whilst economic valuations are based on relatively simple estimates of pollinator dependence, pollination levels, and growing practices, which may not be representative of a larger spatial scale, they do clearly demonstrate the importance of pollinators to crop production (Gallai et al. 2009).
It would be useful to know how growers of different species of pollinator-dependent crops perceive pollination in relation to their crops' level of pollinator dependence, as well as how the different factors affecting fruit set are prioritised in farm management, relative to the empirical evidence. A social survey of growers would tell us how their attitudes FIGURE 3. Experimental process for a grower to discover the extent to which their cucurbit variety depends on pollination, and to investigate the adequacy of the existing pollinator community (adapted from Corbet et al. 1991). towards pollinators correspond to their management actions and if growers identify any key barriers or opportunities to integrating pollination in their management for optimal crop yields. To date, farmer surveys have mostly focused on ecosystem services in isolation, although see Andersson et al. (2015). For example, surveys identified that achieving consistent and reliable pollination is a priority for V. corymbosum growers in Michigan and Florida in the USA (Integrated Crop Pollination Project, 2016). However, there is no way of knowing how much of a priority pollination is to these growers, relative to all the other factors which may affect yield such as pest control or soil quality. This information is critical to understand the likelihood of growers adapting existing, or adopting new, sustainable pollination management.
Since growers are unlikely to know pollination rates for their crops and therefore their potential pollination deficit, it would be useful to develop a predictive model to determine if managed pollinators are required and/ or if longer-term pollinator habitat creation is warranted. Although other pollination service models exist (e.g. Olsson et al. 2015), BEE-STEWARD (www.beehave-model.net) has an interface which already enables users to simulate the effects that different management options, such as wild flower strips will have on bumblebee population dynamics and pollination services.

Conclusion
These studies highlight the importance of pollination for improving cucurbit yield, despite pollinator behaviour, pollen deposition on stigmas, and pollinators' contribution to yield varying across cucurbit species and study. From a biological view, this may be due to the relative attractiveness of flowers, and/or the pollinator dependency of the cucurbit species, whilst from a management view, the spatial and temporal context of study sites could affect the abundance and richness of wild pollinator species. Nonetheless, these accounts of cucurbit pollination demonstrate that wild bee species such as B. terrestris, B. impatiens and Eucera spp. are frequently able to fulfil the pollination requirements of multiple cucurbit species. Sufficient abundance of these wild bee species may be why in several cases, the addition of managed pollinator species had little or no effect on increasing yield. The provision of additional floral resources at both field and farm scales may help to encourage pollination of cucurbit species whilst supporting pollinators' nutritional requirements beyond those already provided by the cucurbit crop.
These findings extend beyond cucurbit systems (Fig. 3) to demonstrate how understanding a crop's requirement for pollination and how pollinators vary spatially and temporally in the landscape can aid growers in their decision making about what varieties and sites should be used. In doing so, growers may be able to increase their agricultural resilience and further their economic advantage. Nonetheless, further work is needed to understand how other environmental factors interact with pollination to influence fruit set so that growers can prioritise key regulating services in their management for optimal crop yields. Insecticide use had no effect on pollen deposition. Amount of natural habitat surrounding a site had a positive effect on pollen deposition. Field size and organic farming had no effect on pollen deposition. Flowering weeds had a positive effect on B. terrestris abundance and no effect on A. mellifera abundance.
Amount of natural habitat surrounding a site had a negative effect on A. mellifera abundance and no effect on B. terrestris abundance. Field size had no effect on A. mellifera or B. terrestris abundance. Pesticide use had no effect on Bombus spp. or P. pruinosa visitation to crop flowers.
Tillage had a negative effect on P. pruinosa visitation to crop flowers and no effect on B. impatiens visitation to crop flowers. (Shuler et al. 2005) Canto-Aguilar A, Veterinaria M (2000) Importance of conserving alternative pollinators: assessing the pollination efficiency of the squash bee, Peponapis limitaris in Cucurbita moschata