Prioritizing pollinators over pests: wild bees are more important than beetle damage for watermelon yield

Abstract

Insect pests and pollinators can interact directly and indirectly to affect crop production; however, impacts of these interactions on marketable yield are little known. Thus, the evaluation of interactions between pests and pollinators are needed to best prioritize management efforts. Over 2 years, we evaluated the impact of pollinator visitation and/or beetle (Acalymma vittatum) infestation on fruit set and yield in seedless watermelon production. In 2020, we tested the main effect of pollinator visitation: two or eight honeybee visits, two wild bee visits, hand pollinated and open pollinated. In 2021, we crossed wild and managed pollinator visitation (two or four honeybee visits, two or four wild bee visits, hand pollinated and open pollinated) with varying beetle infestation levels (0, 3, 6 and 9 beetles/plant). In both years, wild bees contributed significantly to high fruit yields, and exclusive visitation from wild bees increased yield by a factor of 1.5–3 compared to honeybees. In 2021, pollination was the only significant factor for fruit set and marketable yield even when compared to the varying beetle infestation levels. These data advocate for a reprioritization of management, to conserve and protect wild bee pollination, which could be more critical than avoiding pest damage for ensuring high yields.

1. Introduction

In agriculture, pests and pollinators can interact to impact crop production, yet research efforts and farm managers often consider these effects independently. When tested separately, the contribution of pests and pollinators to marketable yield are diametrically opposed—increasing pollinator visitation increases yield, whereas increasing pest densities decreases yield. Interactions between the two could occur through several mechanistic routes. For example, herbivore feeding tends to decrease pollination (e.g. damaging flowers), thereby limiting plant reproductive potential and indirectly exacerbating the direct negative effects of pests on yield (reviewed by [1] and [2]). In pollinator-dependent systems, this may indicate that growers tolerating higher pest damage could compromise crop pollination. However, the opposite also appears true since pest densities are often aggressively managed with insecticide applications that compromise pollinator health and can reduce pollination [3]. Relatively few studies have examined pest–pollinator interactions within an agricultural context, and most use simplified scenarios in which pollinators or herbivores are either present or absent (e.g. [4–11]).

In most pollinator-dependent systems, growers rely on annual supplementation of honeybees to secure high yields. However, honeybees are an average pollinator in most crops [12–16]. When compared to their wild counterparts, they often fall short in terms of both pollen deposition and fruit set [17]. While wild bees are known to contribute to marketable yields, growers are hesitant to rely on their free services [18–21]. The performance of wild bee pollination can be unpredictable, especially in the face of intensive crop management. Pesticides, namely insecticides, have been correlated with wild bee declines, including reductions in wild bee numbers and negative outcomes in bee health (e.g. [22–25]). These effects have been mostly associated with neonicotinoids [23,24,26–28]; however, similar results have been found with other compounds as well [29–31]. Thus, there is a critical need to ensure that pest management is properly balanced with pollinator health, such that needs are correctly prioritized within crop production. Agricultural systems fall within a continuum of pest and pollination management and require nuanced experimental designs to tease apart the relationship between pest damage and managed or wild pollinator visitation. These studies will be informative in the development of effective economic thresholds for pests that account for pollinators (e.g. [32]).

Seedless watermelon is an ideal crop to examine interactions between pest pressure and pollination. Seedless melons only result from the outcrossing between diploid (pollenizers) and tetraploid plants [33,34], making them highly dependent on pollination [35]. Walters [36] estimated that greater than 24 honeybee visits are needed to ensure complete watermelon fruit set. However, recent studies indicate that wild pollinators are frequent, if not primary, pollinators in commercial watermelon production [3,25,37,38]. Multiple studies have indicated that wild pollinators tend to deposit more pollen in cucurbits compared to honeybees [12,15,37,39,40]. For example, native bees—primarily Lasioglossum spp.—effectively pollinated watermelon but required 44% fewer visits compared to honeybees [40]. While wild bees have been identified as important pollinators in seedless watermelon production [15,39], their contribution to yield is poorly understood. Further, wild bees can be negatively impacted by agricultural practices in watermelon production [41,42].

Pest management is also critical in seedless watermelon where a variety of pests can reduce yields. Striped cucumber beetles (SCB) are considered the most serious insect pest of watermelon and other cucurbits in the Midwestern United States. Transplanted watermelons are vulnerable to SCB leaf feeding, which can cause plant death. At the end of the season, during harvest, beetles feed on melon surfaces, thereby reducing marketability. Watermelon growers are recommended to use an economic threshold of five beetles/plant to determine when insecticide applications are needed [43]. Traditionally, SCB is managed using prophylactic insecticide applications, mostly in the neonicotinoid and pyrethroid classes. However, the repeated use of insecticides can reduce the visitation and performance of pollinators in watermelon [3]. SCBs have two generations in the Midwest where adults are active aboveground: the first in early June and then a second generation in mid-August. During these times, SCB and pollinators can interact either directly (SCB feeding damages plants during bloom) or indirectly (SCB feeding damages plants before pollinators arrive); however, little is known about how phenological overlap in pest and pollinator activity affects watermelon yield. Similarly, SCB larvae feed on watermelon roots belowground in mid-summer, which affected bee visitation and reduced yields in, cucumber, a related cucurbit system [44,45].

In this study, we evaluated (i) the impact of wild bees and honeybee visitation and (ii) the interactive effects of pest damage and pollinator visitation on fruit set and yield. We hypothesized that wild bees would be superior pollinators compared to honeybees and visitation would result in higher seedless watermelon yields. We also predicted that wild bee pollination would augment yield losses from high pest pressure (greater than or equal to six beetles/plant) through an increased fruit set, thereby indicating higher tolerance should be given to SCB in watermelon production.

2. Materials and methods

(a) . Maintenance and set-up of trials

Research was conducted at Purdue's Meigs Horticulture Research Farm in Lafayette, IN. Pollinator visitation treatments were evaluated in triploid watermelon production using cv. ‘Fascination’ and pollenizer ‘SP 7’ at 2 : 1 (triploid: pollenizer) ratio. Watermelons were planted within rows, each row approximately 3 m wide and either 60 m or 15 m long in 2020 or 2021, respectively (transplant dates: 19 May 2020 and 7 June 2021) (electronic supplementary material, figure S1). Pollenizers were interplanted with ‘Fascination’ for even distribution of the diploid pollen throughout the field and equal likelihood of seedless watermelon development. Fields were maintained using standard growing practices for the region [46]. Nitrogen, phosphorus and potassium fertilizers were applied to correct nutritional deficiencies per 2019 soil tests. Pesticides were occasionally applied to manage pest arthropods, pathogens and weeds, but no applications occurred prior to or during pollination treatments. Insecticide applications were limited to fenpyroximate and pymetrozine to control mites and aphids, respectively. Neither compound has known activity against SCB. To ensure honeybee visitation, the east side of the field was supplemented with two honeybee hives (approx. 20 000 bees per hive).

(b) . 2020 trial: evaluation of pollinator visitation on fruit set and quality

A trial was conducted during the summer of 2020 to determine the efficacy of different pollination treatments in the fruit set and quality of watermelon.

(i) . Pollination treatments

A total of five pollinator visitation treatments were tested to determine the effect of differing types and numbers of visitors on watermelon yield and quality: (i) two honeybee visits, (ii) two wild bee visits, (iii) eight honeybee visits, (iv) open pollinated and (v) hand pollinated. Non-honeybee visitors were deemed a wild bee. Visitation rates were determined from previous research conducted in seedless watermelon [36] and studies from IN watermelon production [3,25]. Prior to anthesis, pollination exclusion bags (Trimaco 11 511 Supertuff Paint Strainer, 1 gal) were placed over pistillate flowers using twist ties, and only removed for visit observations (electronic supplementary material, figure S1a). No bags were observed damaging the corolla or pistil. Observers monitored pistillate flowers until the necessary number of treatment visits was complete. Any non-treatment insect was removed from the area such that it did not contact the treatment flower (i.e. if a honeybee attempted to visit a ‘wild bee’ treatment it was obstructed before it contacted the flower). A visitation event occurred only when a bee contacted the reproductive organs of the flower. After treatment visits were observed, pistillate flowers were re-bagged to exclude additional pollinator visits. ‘Open pollinated’ flowers were left unbagged for a total of 24 h beginning at 07.00 and ending at 07.00 the following day. These times coincided with the first appearance of flower bloom in the field; pollinator activity began around 09.00. For ‘hand pollinated’ treatments, seven staminate flowers were harvested at 07.00. Pollen was removed and directly deposited onto the pistil of pistillate flowers and then bagged for 24 h. Pollinator visitation treatment observations occurred between 07.00 and 13.00 on days when the weather was sunny, mostly clear, and wind speeds below 2 m s^–1. All treatments were replicated at least seven times per row for a total of at least 63 replications. Pollinator visitation treatments were implemented within the first two weeks of active watermelon bloom (16–26 July). For a subset of pistillate flowers receiving the ‘two wild bee’ treatment (n = 44), the type of wild bee visitor was recorded to the lowest taxonomic level possible (species or genus). Fruit set assessments were completed 14 days after treatments were implemented in the field (i.e. number of fruit set from two honeybee visits/total number of treatment flowers with two honeybee visits). To ensure that pollination did not take place prior to implementing visitation treatments, a subset of bagged pistillate flowers remained covered and fruit did not set.

(c) . 2021 Trial: assessment of interaction between pollinator visitation and beetle infestation

In 2021, pollinator visitation treatments were crossed with varying SCB density levels to determine impact on fruit set, quality and overall yield.

(i) . Pest treatments

In total, four beetle infestation levels were chosen to reflect a wide spectrum of pest densities that are below/above the economic threshold (five per plant), including: (I) zero beetles per plant, (II) three beetles per plant, (III) six beetles per plant and (IV) nine beetles per plant. Based on observations in other watermelon field trials at research sites and commercial farms in our area, 0–6 SCB per plant is relatively common, whereas nine per plant (treatment IV) is extremely high and outside the range typically observed, even in the absence of control measures. Beetle infestation treatments were applied to an entire row of watermelons and organized in a randomized complete block design for a total of five row replications per beetle infestation level (=20 rows) (see electronic supplementary material, figure S2 for additional detail). SCB were collected from untreated cucurbit fields 3–5 days prior to infestation and kept in screened cages. Beetles were uniformly distributed throughout treatment rows and then confined to the row using cover (GG-17, 0.5 oz) (electronic supplementary material, figure S1b). Prior to row covering, the ‘zero beetles per plant’ treatment received a water drench of Wrangler® insecticide (imidacloprid; Loveland Products, Inc., Greeley, CO, USA) at transplant at a rate of 10 fl. oz acre^–1 (0.35 kg ha^–1) to remove ambient beetles. Rows remained covered for a total of three weeks (15 June–6 July). Upon uncovering, watermelon plants were assessed for SCB pressure and damage. Six plants per row were scouted for beetles and assigned a score from 0–100 based on leaf damage to each plant (n = 30 plants trt⁻¹). After harvest of melons (15 September), three plants per row were carefully uprooted, washed with water to remove soil and weighed (n = 15 plants trt⁻¹).

(ii) . Pollination treatments

A total of six pollination treatments were examined: (i) two honeybee visits, (ii) two wild bee visits, (iii) four honeybee visits, (iv) four wild bee visits, (v) open pollinated and (vi) hand pollinated. Pollination treatments were nested within beetle infestation rows (electronic supplementary material, figure S2). Each pollinator visitation treatment was replicated at least three times per row for a minimum total of 15 replications per beetle infestation × pollinator visitation treatment cross. Due to the low fruit set of the two-honeybee treatment in 2020, ‘2HB’, we increased replication of this treatment in 2021 to increase the likelihood for a yield assessment. Similar to 2020, we also recorded the species of wild bee visitors for a subset of flowers (n = 105), however in addition, we also recorded whether these fruit set as a means to determining if pollinator species impacted fruit yield. Pollination treatments were observed over 14 days in 2021 (6–20 July) in accordance with the first two weeks of watermelon bloom. In 2021, pollinator observations were completed to determine if infestation impacted pollinator visitation. All plots were observed for 5 min for a total of three observations per plot (=60 min trt⁻¹). No significant differences were detected between beetle infestation treatments and pollinator visitation numbers (electronic supplementary material, figure S3). Execution of visitation treatments and trial maintenance was the same as described above in 2020 ‘Pollination treatments’.

(d) . Watermelon harvest and quality assessment for 2020 and 2021 trials

Watermelons were continuously harvested over a three-week period (10–24 August 2020; 16 August–6 September 2021). Treatment watermelons were individually weighed and graded. Hollow heart is a physiological watermelon disease, which has been correlated with poor pollination [47]. To understand if these treatments in either year compromised the internal quality of watermelon, we also assessed the amount of carpel separation within each melon. The severity of hollow heart was determined by cutting open melons and assessing severity of carpel separation. Hollow heart severity scores were directly related to the degree of separation within watermelon heart tissue; 0 = no carpel separation, 1 = slight carpel separation (less than 1.2 cm), 2 = mild carpel separation (1.3–2.5 cm), 3 = moderate carpel separation (2.5–5 cm) and 4 = severe carpel separation (greater than 6.4 cm). Any separation greater than 2.5 cm was considered unmarketable (electronic supplementary material, figure S4a). Similarly, melons with greater than 45 cm² rind feeding (caused by SCB) were considered unmarketable per USDA standards [48] (electronic supplementary material, figure S4b).

(e) . Statistical analysis

Data were analysed using generalized linear mixed models (GLMM) (R v.3.5.2) (R package; ‘lme4’, [49]). Separate GLMMs were generated to analyse fruit set, weight and hollow heart severity in either year. All data were analysed assuming visitation (in 2020) or visitation and beetle infestation treatment (in 2021) as the fixed effect and row or block as a random effect. Fruit set was analysed assuming a binomial distribution (n fruit set from observed flowers/n total observed flowers). Total fruit yield (combined weight of melons) was summed within rows for each pollination or pollination × beetle treatment in the 2020 or 2021 field season, respectively. Fruit yield was analysed with a normal distribution, and due to the low occurrence of hollow heart, severity scores were analysed assuming a Poisson distribution. In 2021, additional analysis was completed to determine the impact of beetle densities on associated plant metrics (plant defoliation and root biomass) as well as number of beetles scouted. Similar to previous analyses, block was considered a random effect and beetle infestation treatment the fixed effects. Plant defoliation was assessed using a binomial distribution, root biomass a normal distribution and beetle densities as a negative binomial distribution. In a subset of the pollinator analyses in 2021, the number of large-bodied bees was also regressed against the proportion of fruit set. Large-bodied bees were counted as a bee belonging to either the Bombus or Melissodes genera, whereas small-bodied bees were in either Lasioglossum genus or other smaller halictid genera (e.g. Agapostemon). All models were tested for overdispersion using package blmeco and function dispersion_glmer [50].

3. Results

(a) . Trial 1 (2020): impact of pollinator visits on fruit set and quality

(i) . Fruit set

Pollinator visitation treatments significantly impacted the fruit set of watermelon in 2020 (X² = 98.9, d.f. = 4, p < 0.0005). Flowers that were restricted to two honeybee visits had the lowest fruit set (19.1 ± 1.8%) compared to all other treatments. Highest proportion of fruit set (82 ± 3%) was recorded in the hand-pollinated treatment; however, this was similar to flowers that were either visited by two wild bees or open-pollinated (figure 1a). Melissodes species were the most frequent wild bee visitor to treatment flowers (45% of visits) followed by Lasioglossum spp. and Bombus spp. (figure 2a).

Figure 1.

Effect of pollinator visitation treatments on fruit set (a) and total fruit weight (b) in 2020. Number along x-axis indicates the number of treatment visitors. Visitors were either honeybees (HB) or wild bees (W). ‘Open’ indicates open pollinated and ‘hand’ indicates hand pollinated. Mean values were determined by averaging across treatments in each replication (n = 9). (Online version in colour.)

Figure 2.

Wild pollinators contributing to ‘2 W’ and ‘4 W’ treatments that received either two wild bee visits or four wild bee visits respectively in 2020 (n = 44 subsample) and 2021 (n = 105). (Online version in colour.)

(ii) . Fruit weight and quality

Due to excessive rainy weather, a large number of melons were lost prior to harvest due to plant disease. Nonetheless, we found that flowers visited by two honeybees had lowest melon weights (5.53 ± 1.79 kg). All other treatments had significantly greater total weights except those melons resulting from eight honeybee visits (X² = 20.8, d.f. = 4, p = 0.0003) (figure 1b).

Numerically, hollow heart severity followed a similar pattern as observed with fruit set and weight, but this finding was only marginally significant (table 1). Melons with greatest severity of hollow heart were pollinated by two honeybees, and lowest severity ratings were assigned to melons in either the ‘hand’- or ‘open’-pollinated treatments. Despite variable beetle pressure in the trial (3–12 beetles per plant), no melons were marred enough to be deemed unmarketable.

Table 1.

Mean hollow heart severity across pollination treatments in seedless watermelon production in 2020 and 2021.

year	treatment (number of pollinator visits)	hollow heart severity ± s.e. (severity rating from 0 to 4)
2020	two honeybees	1.7 ± 0.8
	eight honeybees	0.5 ± 0.3
	two wild bees	0.4 ± 0.2
p = 0.055	open pollinated	0.2 ± 0.2
	hand pollinated	0.1 ± 0.1
2021	two honeybees	0.23 ± 0.07
	four honeybees	0.05 ± 0.04
	two wild bees	0.04 ± 0.03
	four wild bees	0.04 ± 0.03
p = 0.048	open pollinated	0.06 ± 0.03
	hand pollinated	0.08 ± 0.03

(b) . Trial 2 (2021): impact of the interaction between pollinator visitation and beetle infestation on fruit set and quality

(i) . Pest infestations and crop damage

Beetle infestation treatments applied under row covers translated to their intended density levels upon final assessment. Significantly more beetles were scouted on plants infested with higher densities of beetles (figure 3a) (X² = 112.4, d.f. = 3, p < 0.0005), although uniformly across all treatments final counts were slightly lower than the initial stocked density. Similarly, leaf damage increased in direct proportion with increasing beetle densities, from virtually none on the beetle-free control to ca 10% leaf area removal with the highest density of nine beetles per plant (figure 3b) (X² = 25.2, d.f. = 3, p < 0.0005). Last, we observed dramatic reductions in final root biomass in a stepwise pattern with increasing beetle densities, indicating that the stocked adults likely oviposited on plants, leading to much higher levels of root herbivory (figure 3c) (X² = 22.4, d.f. = 3, p < 0.0005). The highest density beetle treatment had ca 50% smaller roots than the beetle-free control.

Figure 3.

Effect of SCB densities on (a) number of beetles scouted on plants 21 days after infestation, (b) per cent leaf defoliation, (c) weight of roots and (d) total harvestable weight of watermelon fruits in 2021. Mean values were determined by averaging across treatments in each replication (n = 5). (Online version in colour.)

(ii) . Fruit set

Watermelon fruit set was significantly impacted by visitation treatment (X² = 65.5, d.f. = 5, p < 0.0005) but not beetle infestation or the interaction of beetle infestation and pollination (p = 0.50 and p = 0.90, respectively; see electronic supplementary material, table S1 for fully crossed pest × pollinator data). Highest fruit set was in the hand-pollinated treatments (87 ± 2%) and lowest in the two honeybee treatment (19 ± 3%) (figure 4a). Similar to 2020, Melissodes bimaculatus (Lepeletier) was the most frequent wild bee visitor (51% of visits) followed by Lasioglossum spp., Bombus spp. and Eucera (Peponapis) pruinosa (Say) (figure 2b). Visits from large-bodied bees rather than small-bodied bees were significantly associated with fruit set (X² = 13.05, d.f. = 1, p = 0.0003) (figure 5).

Figure 4.

Effect of pollinator visitation treatments on fruit set (a) and total fruit weight (b) in 2021. Number along x-axis indicates the number of treatment visitors. Visitors were either honeybees (HB) or wild bees (W). ‘Open’ indicates open pollinated and ‘hand’ indicates hand pollinated. In this trial, pollination treatments were crossed with different beetle infestation intensities; however, beetle infestation did not significantly impact fruit set or fruit yield (p > 0.05), which is why only main effects are shown. Mean values were determined by averaging across treatments in each replication (n = 5). (Online version in colour.)

Figure 5.

Effect of large-bodied wild bee visits on seedless watermelon mean proportion fruit set. Mean proportions were determined by averaging across all wild bee treatments in each replication, with a possibility of five total replicates. Analysis was conducted on a subset of wild bee treatments (n = 105). (Online version in colour.)

(iii) . Fruit weight and quality

Despite high levels of beetle feeding on melon leaves and roots (figure 3b,c), melon yield was not significantly impacted by SCBs at the infested densities. Similar to fruit set data, fruit weight and quality were only impacted by pollination (F = 83.8, d.f. = 5, p < 0.0005) and not beetle infestation or the interaction between beetle infestation and pollination (p = 0.24 and p = 0.76, respectively; see electronic supplementary material, table S2 for fully crossed pest × pollinator data). Melons that only received ‘two honeybee visits’ had lowest melon weights whereas ‘open-pollinated’ or ‘hand-pollinated’ flowers had highest melon weights (figure 4b). Although hollow heart was low in 2021, trends followed similar patterns to those in 2020. Hollow heart was significantly higher in ‘two honeybee’ treatments compared to all other visitation treatments (table 1). No melons were deemed unmarketable due to high levels of rind feeding by beetles.

4. Discussion

Pest management is predicated on a negative relationship between the crop and pest. However, pest management does not act in a vacuum and practices may compromise pollination. Wild bees, in particular, are negatively impacted by intense pesticide programmes [23,24,27,51]. In seedless watermelon production, SCBs and wild pollinators may interact to impact yield. However, the degree to which pollination may offset yield losses from SCB or how SCB may compromise pollination efforts is unknown. In this study, we examined the impact of pollinator visitation and interactive effects of SCB densities and pollinator visitation on seedless watermelon fruit set and yield. In line with our hypotheses, we found that wild bees had higher levels of fruit set when compared with honeybees in both years of our trial. Contrary to our hypothesis, we found no interactive effects of pollination and beetle pressure on watermelon yield. Instead, pollinator visitation was the only significant effect on fruit set and yield.

Surprisingly, SCB had little to no effect on watermelon production. While our leaf defoliation ratings and root mass weights were significantly associated with beetle infestations, indicating that stocked beetles indeed fed on the crop, total melon yield was unaffected. SCB is a well-known pest of watermelon but can have a variable impact on melon yield [52,53]. Some studies have found negative impacts of SCB on yield [52,54], others have found little to no effect [52,53]. While we applied densities of beetles that were well over the economic threshold of five beetles/plant, it may be possible that these densities were not high enough to affect melon yield. However, it should be noted that grower tolerance for SCB is much lower than five beetles/plant in most commercial watermelon farms [55]. In many cases, growers make multiple insecticide applications to keep populations low (less than two beetles per plant; [55]), but our study indicates that this may not be necessary since watermelons appear highly tolerant of beetle feeding damage (i.e. no changes in fruit production, despite 50% root loss). Watermelon is naturally resistant to the deadly bacterial wilt that SCB transmits in other cucurbits like squash and cucumber (reviewed by [56]), which likely is a key factor driving its tolerance of high beetle densities. Similar to Sawe et al. [57], these results suggest that pollination on average should have a much more significant consideration in dictating management interventions and that insecticide sprays targeting SCB may, in fact, do more harm than good.

Pollinator visitation treatments significantly impacted yield, and unsurprisingly more pollinator visitation equated to higher fruit set. On par with previous reports [15,40], wild pollinators performed better than honeybees especially at lower visitation frequencies. Low honeybee visitation was also significantly correlated with a reduction in fruit quality. Treatments visited only by two honeybees had 4–5 times more severe hollow heart when compared to all other treatments. While this physiological disorder has been attributed to a number of variables, including weather and cultivar [58], pollination is thought to be a leading factor in hollow heart severity [59], although data mechanistically linking the two are tenuous.

Wild pollinators appear to differentially contribute to watermelon quality. While Lasioglossum spp. are a ubiquitous pollinator in watermelon production, exclusive visitation from this species resulted in low fruit set in our trial. We also tended to observe more irregularly shaped fruit when pollinated by Lasioglossum spp. Similar results were found in Pisanty et al. [37] in which small-bodied bees contributed less fruit set than honeybees. Campbell et al. [15] also found that Lasioglossum pollen deposition was less than or similar to that of honeybees. Bee body hair has also been implicated in the transfer of pollen [60], and many of the large-bodied bees we observed (e.g. Melissodes spp. and Bombus spp.) are hairier than our small-bodied bees. Nevertheless, Lasioglossum spp. visitation in watermelon fields is likely to exceed the restrictions we placed on visitor number in this trial. Thus, our trial may prove a poor proxy for the performance of numerous small-bodied bees in seedless watermelon production. The dominant wild pollinator in both years of our trial was M. bimaculatus, which has a reputation for being an efficient pollinator in watermelon [15,39] and elsewhere [61,62]. In our observations, we noticed that this species moved quickly among watermelon flowers, manoeuvred around stigmas, and was frequently coated in pollen. While this genus is commonly encountered in cucurbit production [3,15,25], relatively little is known about its life history, making conservation and protection difficult beyond the traditional recommendation to support solitary bees. Other dominant wild pollinators in our trial included Bombus spp., which also deposit large amounts of pollen and significantly contribute to high fruit sets [12,15,39].

While we observed a fairly diverse suite of wild pollinators, our observations revealed that relatively few bees visited flowers over the course of the day. Specific treatments (eight honeybees in 2020, four wild bees and four honeybees in 2021) required hours of observation. In some cases, one female flower would be observed for the entire day (08.00–14.00) and still be short of necessary treatment observations. This seems to contrast with findings from other studies (e.g. [36]) in that we observed higher fruit sets with fewer bee visits. However, Walters [36] only tested honeybee visitation rates and wild bee visitation should, on average, be more efficient that honeybees. We also posit that these differences are a result of trial design differences. Our trial featured higher pollenizer ratios (2 : 1 versus 3 : 1) compared to Walters [36]. Additionally, significant headway has been made in the development of diploid pollen donors (i.e. ‘pollenizer plants’) for seedless watermelon production [63–66]. Pollenizer watermelons have been bred to feature large numbers of staminate flowers, which can increase likelihood of fruit set, and have been shown to significantly increase yield [63]. Therefore, in addition to the wild bee community, there are other characteristics intrinsic to production systems that can alter yield probabilities.

Our study lies inherently along a continuum between pest and pollinator management, and the consequences of imbalance therein. Our study suggests that a balance may not always be warranted but rather a reprioritization of management efforts. Like many other studies, we show that wild bees can be critical in pollinator-dependent crops. However, unlike other studies, we found that when crossed with pest infestations, they were still the most important factor in securing high yields. Further, grower investment into wild bee conservation may improve adoption of integrated pest and pollination management practices (IPPM). For example, one key IPPM recommendation urges growers to only apply insecticides at specific pest densities or economic thresholds to limit negative effects on pollinators [67,68]. The positive impact of these threshold-based insecticide programs was recently shown by Pecenka et al. [3], in which insecticide applications were reduced by more than greater than 90% while effectively controlling target pests, including SCB. Threshold-based treatments also boasted higher yields, a finding largely due to a more rich and robust wild bee community. This collective work, including the data reported here, strongly suggest that in watermelon, and potentially other pollinator-dependent crops, insecticide applications, counterintuitively, have a higher likelihood of reducing than increasing yields due to interference with bee foraging and the lack of threat posed by the pest community. Given this imbalance (i.e. pollinators outweigh pests), insecticide use decisions by growers should consider wild bee presence and activity to optimize yield, which is currently not common practice.

Data accessibility

All data and code are publicly available on GitHub under the primary author's username, aleach379 (https://github.com/aleach379/dataforpestpol).

The data are provided in the electronic supplementary material [69].

Authors' contributions

A.L.: conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, visualization, writing—original draft and writing—review and editing; I.K.: conceptualization, data curation, funding acquisition, investigation, methodology, resources, supervision, writing—original draft and writing—review and editing.

Both authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Conflict of interest declaration

We declare we have no competing interests.

Funding

This research was supported with funding awarded to I.K. from the USDA/National Institute of Food & Agriculture grant no. 2016-51181-25410. A.L. was supported by a USDA/National Institute of Food & Agriculture Fellowship grant no. 2020-67034-31783.