1. The honeybee Apis mellifera is currently in decline worldwide because of the combined impacts of Colony Collapse Disorder and the Varroa destructor mite. In order to gain a balanced perspective of the importance of both wild and managed pollination services, it is essential to compare these services directly, a priori, within a cropping landscape. This process will determine the capacity of other flower visitors to act as honeybee replacements.
2. In a highly modified New Zealand agricultural landscape, we compared the pollination services provided by managed honeybees to unmanaged pollinator taxa (including flies) within a Brassica rapa var. chinensis mass flowering crop.
3. We evaluate overall pollinator effectiveness by separating the pollination service into two components: efficiency (i.e. per visit pollen deposition) and visit rate (i.e. pollinator abundance per available flower and the number of flower visits per minute).
4. We observed 31 species attending flowers of B. rapa. In addition to A. mellifera, seven insect species visited flowers frequently. These were three other bees (Lasioglossum sordidum, Bombus terrestris and Leioproctus sp.) and four flies (Dilophus nigrostigma, Melanostoma fasciatum, Melangyna novae-zelandiae and Eristalis tenax).
5. Two bee species, Bombus terrestris and Leioproctus sp. and one fly, Eristalis tenax were as efficient as the honeybee and as effective (in terms of rate of flower visitation). A higher honeybee abundance, however, resulted in it being the more effective pollinator overall.
6.Synthesis and applications. Alternative land management practices that increase the population sizes of unmanaged pollinator taxa to levels resulting in visitation frequencies as high as A. mellifera, have the potential to replace services provided by the honeybee. This will require a thorough investigation of each taxon’s intrinsic biology and a change in land management practices to ensure year round refuge, feeding, nesting and other resource requirements of pollinator taxa are met.
The value of managed vs. wild pollinator services has recently been the focus of much attention (e.g. Allsopp, de Lange & Veldtman 2008), particularly with reference to global food crops and their pollination requirements (Klein et al. 2007; Aizen et al. 2008; Winfree 2008). Honeybee Apis mellifera Linnaeus, 1758 colony viability is now a serious concern because many agricultural crops are reliant on this single pollinator species and consequently, the contribution of this species to food production is high (Morse & Calderone 2000; Klein et al. 2007). This has sparked renewed interest in the function of unmanaged or wild pollinator taxa, particularly because of their provision of ‘pollination insurance’ (Winfree et al. 2007a).
This premise has developed from two critical concerns: the first is that a majority of global food production does not in fact depend on animal pollination and hence the attention directed toward declining wild pollinators is currently not warranted (Ghazoul 2005a). The assumption that pollinator declines have yet to be translated into decreased food production is supported by a study by Aizen et al. (2008) which compared rates of yield increase between pollinator-dependent and pollinator-independent crops over the last 45 years. The second concern arises from the fact that studies that have assessed the value of wild pollinators, often fail to compare the managed and the unmanaged components to arrive at a balanced view of the net worth of their services (Allsopp et al. 2008). There is a need for empirical studies directly comparing managed and unmanaged pollinator services (Ghazoul 2005a,b; Steffan-Dewenter et al. 2005) in order to demonstrate unequivocally the relative significance of their services in pollinating global food crops (Allsopp et al. 2008).
Two components of pollination need to be assessed in order to directly compare the overall effectiveness of managed and unmanaged pollinator services, pollen transfer efficiency and visitation frequency. Pollen transfer efficiency describes the proficiency with which individual pollinators remove and transport pollen to conspecific stigmas (Primack & Silander 1975; Herrera 1987; Harder & Wilson 1998). Visitation frequency is a function of both the abundance of the pollinator and the number of flowers it visits in a given time interval (Herrera 1987, 1989; Vazquez, Morris & Jordano 2005; Madjidian, Morales & Smith 2008). The most effective insect pollinator would therefore be one that is present in high numbers and moves rapidly from flower to flower (i.e. has a high visitation rate). It would also frequently contact the stigma, transferring many pollen grains (i.e. has high pollen transfer efficiency). Conversely, the least effective insect pollinator would have low abundance and move relatively slowly from flower to flower (i.e. have a low visitation rate). It would rarely contact the stigma while visiting a flower and transfer few pollen grains when it did (i.e. have low pollen transfer efficiency).
The lack of studies comparing managed and unmanaged services in situ, in intensive agricultural systems, is surprising as there is a clear link between diverse pollinator guilds and improved pollen loads, high fruit and seed set and increased offspring vigour (Schemske & Pautler 1984; Herrera 1987; Klein et al. 2003; Gomez et al. 2007). Intensive agricultural systems that support diverse unmanaged pollinator assemblages co-existing with managed honeybee hives prior to honeybee decline, are ideal systems to identify potential alternative taxa that might be used if honeybees decline.
We use a highly modified landscape in the Canterbury region of New Zealand, to ask the following questions: (1) Does pollen transfer efficiency (as measured by stigmatic pollen loads and the proportion of visits in which the stigma is contacted) differ between the honeybee and other flower-visiting taxa? (2) Does the rate of flower visitation (measured as both visitor abundance per number of available open flowers, and the number of flower visits per minute) differ between the honeybee and other taxa? (3) How do these differences translate into overall pollinator effectiveness? (4) Are any of the alternative pollinator taxa directly, or as a group, capable of replacing honeybee services in a mass flowering crop?
Materials and methods
Brassica rapa var. chinensis (Brassicaceae) or ‘Pak Choi’ is a mass flowering vegetable and forage crop grown commercially in New Zealand as a seed crop for export. It is an ideal study species for comparing alternative pollinator assemblages with the honeybee because of its mass flowering habit and its attractiveness to a generalist pollinator assemblage. At the time of this study, Varroa destructor was not present on the South Island of New Zealand and an average of two managed honeybee hives (range 1–4) were located within 0·75–3 km from the 11 fields sampled in this study. This study assumes that honeybee visits are from honeybees in managed hives (as opposed to feral honeybees) because of the large number of hives operating in close proximity to the study plots.
We measured two components of pollination that are important in determining which alternative species may have the potential to replace the honeybee in providing pollination services: pollen transfer efficiency and pollinator rate of visitation. A number of methods were used to assess these two components. To determine pollen transfer efficiency, we measured the quantity of pollen transferred by each flower-visiting taxon and the likelihood that pollen was transferred to stigmatic surfaces. To determine pollinator rate of visitation, we observed how often flowers were visited by different taxa within a specified time.
Observations of flower visitors
From 15 December 2006 to 20 February 2007, we observed flower visiting insects in 11 commercial B. rapa fields (range: 0·75–2·0 ha) in the Canterbury region of New Zealand. We selected fields for observations at the time of peak flowering, defined as the period during which mature receptive female flower density exceeded 1000 m−2. Five observation quadrats (10 × 10 m) were established per field; one near each of the four corners in four directions and one in the field centre. In each of these observation quadrats, flower density was estimated within three smaller randomly located 1 m2 quadrats by counting the number of individual plants within each quadrat, the number of inflorescences per plant on 10 randomly selected plants, and the number of flowers per inflorescence (on the same 10 randomly selected plants). Using these quadrat level estimates (mean number of flowers per 1 m2), we extrapolated values to estimate the number of flowers observed in visitor survey transects (i.e. multiplied mean flowers per 1 m2 by 10 to calculate density per 10 m2). On average, flower density across all fields was estimated to be (mean ± SE) 1596·6 ± 149 flowers m−2. Flower density differed significantly between fields (F1,10 = 359·61; P <0·001).
Pollen transfer efficiency
We measured pollen transfer efficiency by recording pollen deposition on stigmatic surfaces and the proportion of flower visits that resulted in stigma contacts per insect visit.
Pollen deposition on stigmatic surfaces was estimated via manipulation experiments. We bagged virgin inflorescences in bud (fine mesh 50 × 50 μm) to exclude pollinators. At flower opening we removed the bag and observed flowers for the period required before an insect visited the flower and contacted the stigma in a single visit. After identifying each insect, we removed the stigma by carefully severing it from the style using finely pointed forceps. The stigma was placed on a cube of gelatine-fuchsin (c. 3 × 3 × 3 mm) and a coverslip was placed on top of the gelatine cube. Gentle pressure was applied, after which the gelatine was melted onto a microscope slide by applying heat (Dafni 1992; Kearns & Inouye 1993). Pollen loads were estimated by counting all B. rapa pollen grains surrounding the stigma under 20× magnification. In total, we estimated pollen loads for each of the 456 stigmas collected from 11 fields. We did this for 18–31 individual insect contacts for each of the eight frequent visitor species (in addition to other species).
For each stigma sampled in this way, we also collected a second control stigma at a similar developmental stage, from flowers to which insect access had been excluded by bagging. This allowed us to assess the possible influence of pollen drift between neighbouring flowers and hence on total pollen counted. The control stigma was processed using the same method.
To assess self-pollen movement due to insect foraging behaviour (i.e. pollen transfer within a flower) we removed anthers from filaments in a second treatment of control flowers (n = 75) and compared the pollen loads of these stigmas with those flowers having intact anthers.
The second component of pollinator efficiency is the reliability with which individuals transfer pollen during floral visitation. We determined the proportion of all flower visits that resulted in stigma contact per insect individual. We followed individual insects for each of eight species of visitor (the most frequent visitors in the entire assemblage) over a period required for them to visit 10 flowers and we used a hand-held video camera to record their behaviour while attending each flower visited. The recording allowed us to identify the number of occasions (/10) in which an insect landed on a flower and successfully touched the stigmatic surface.
It was not possible to manipulate the experimental protocol in order to ensure that equal representation of taxa across each field for either measure of pollen deposition or stigmatic contact. We could not control which insects visited flowers and therefore sampling all taxa in all fields with equal frequency was not possible. Stigmatic pollen deposition and stigma contact data thus represent the product of natural variability across the fields. In total, we collected data for 6–30 individuals for each of eight species of visitor (Table 2).
Table 2. Efficiency of the eight most frequent flower visitors to 11 Brassica rapa fields in the Canterbury region, New Zealand
Stigmatic pollen grains transferred (mean ± SE)
Post-hoc comparisons with A. mellifera Tukey’s HSD (P)
Proportion of successful stigma contacts (mean ± SE)
Post-hoc comparisons with A. mellifera LSD test (P)
Numbers in parentheses represent n. Post-hoc pair-wise comparisons for pollen grains transferred are the P-values from Tukey’s HSDs and for successful stigma contacts are the P-values from LSD tests.
123·03 ± 16·42 (32)
0·88 ± 0·010 (230)
236·63 ± 43·69 (16)
0·97 ± 0·006 (110)
153·20 ± 40·41 (10)
0·95 ± 0·01 (80)
30·14 ± 12·07 (7)
0·28 ± 0·023 (50)
106·64 ± 19·83 (22)
0·93 ± 0·011 (80)
16·13 ± 7·78 (8)
0·24 ± 0·014 (80)
6·36 ± 2·52 (11)
0·43 ± 0·015 (70)
68·29 ± 15·43 (24)
0·45 ± 0·029 (80)
We estimated visitation rate using two measures: visitor abundance per number of available open flowers (measured by visits to quadrat/no. open flowers observed per 10 min) and flower visitation rate (number of flower visits by an individual pollinator per minute). To determine visitor abundance in the plots, visitors were surveyed for 2 days at each field for three observation periods; 10.00–11.00, 12.00–13.00 and 14.00–15.00 h. The five observation quadrats (10 × 10 m) already established to determine flower density were used to conduct these visitor observations. Observations of floral visitors were made by walking along each of the four boundaries of each quadrat (i.e. 10 × 1 m) and recording all insect species and abundances within the boundary during a 10-min time period. The time taken to complete five quadrats was therefore 50 min per time interval. All floral visitor observations were made on sunny or partly cloudy days when the temperature was >16 °C and wind speed <5 ms−1. We observed flower visitors for a total of 55 h. The order that fields, and quadrats-within-fields were examined was randomized throughout the study period. Frequencies were then divided by the number of open flowers (estimated from the open flower estimates in each quadrat) to remove the confounding effect of differences in floral density between fields on visit frequency (see Ivey, Martinez & Wyatt 2003). We then divided the unmanaged taxa into two groups; all flies and all bees (other than the honeybee) and compared visitation frequencies among these groups.
To determine the number of flower visits per minute we followed individual insects from flower to flower and recorded all the flower visits made by this individual within a 1-min period using a digital voice recorder. We recorded observations of 20–50 individuals of each taxon (n =479) to calculate the mean number of visits per flower per min for each field. All flower visitors were described to species level in the field where possible. Where species identity was not determined at the time of observation, specimens were collected between observation periods and taken back to the laboratory for identification.
To calculate overall pollinator effectiveness per day we multiplied pollen transfer efficiency for each taxon (median stigma load × proportion of successful stigma contacts) by the frequency of visits/h (visitor abundance per number of open flowers × number of flowers visited per minute × 10 min−1 × 6) (see Madjidian et al. 2008).
Stigmatic pollen loads were log transformed to improve normality and means were compared between taxa using univariate general linear models. Where significant differences were revealed, means were compared using Tukey’s Honestly Significantly Different (HSD) tests, which control the experiment-wise error rate to α = 0·05. The variation between fields was included as a random factor in all analyses.
The proportion of successful flower visits were compared between taxa using generalized mixed models with a binomial error distribution. We compared visitation frequency between taxa (abundance per flower per 10 min), between groups of taxa, and per flower visitation rate between taxa (number of flowers visited per minute per focal animal) using mixed models restricted estimates maximum likelihood variance analysis (REML) with taxa as a fixed effect. We also compared flower density estimates between fields using mixed models REML analysis. Post-hoc pair-wise comparisons were performed using Least Significant Difference (LSD) tests. All analyses were conducted using version 17, SPSS statistical package (SPSS, 2008).
In total, we observed 31 species attending flowers of B. rapa (Table 1). In addition to A. mellifera, seven insect species visited flowers often enough to be included in analyses. These were three other bees: Lasioglossum sordidum (Smith, 1853), Bombus terrestris (Linnaeus, 1758) and Leioproctus sp.; and four flies: Dilophus nigrostigma (Walker, 1848), Melanostoma fasciatum (Macquart, 1850), Melangyna novae-zelandiae (Macquart, 1855) and Eristalis tenax Linnaeus, 1758 (Table 1).
Table 1. Taxa recorded visiting flowers in Brassica rapa fields in the Canterbury region, New Zealand
In 63% of control (i.e. unvisited) flowers, there were no pollen grains on stigmas. In the remaining control stigmas, there were <2 pollen grains (mean = 1·96 ± 0·01). We thus have no reason to expect that pollen movement occurred without insect pollinators. There was no significant difference between emasculated flower stigmatic loads and intact flower pollen loads (t = −1·14, P =0·27). This suggests pollen transfer estimated in stigmatic pollen load calculations was not likely to be self-pollen. In the absence of detailed data regarding pollinator behaviour, however, (i.e., extent of grooming, amount of self pollen carried on insect body (Harder 1990; Harder & Wilson 1998; Aizen & Harder 2007) we cannot verify with certainty that self-pollen was excluded from pollen transfer estimates in stigmatic pollen load calculations.
There were significant differences in the mean pollen load (log transformed) deposited onto stigmatic surfaces between species (F9,132 = 7·646, P <0·0001, Fig. 1, Table 2). In comparisons between taxa representing the unmanaged component of pollinating fauna, A. mellifera transferred significantly greater amounts of pollen per stigmatic contact than four of the native species; Dilophus nigrostigma, Melanostoma fasciatum, Melangyna novae-zelandiae and Lasioglossum sordidum (Table 2, Fig. 1). Three species from the unmanaged assemblage were not different in this respect to A. mellifera; B. terrestris and Leioproctus sp. and the fly E. tenax, Fig. 1, Table 2).
The proportion of times that individuals contacted stigmatic surfaces when visiting flowers, differed between species (d.f. = 7, Wald = 434·405, P <0·0001, Fig. 2). The honeybee and three unmanaged taxa (B. terrestris, Leioproctus sp. and E. tenax) contacted stigmatic surfaces on significantly more occasions than non-contact occasions. Stigma contact was low in the remaining taxa (Table 2, Fig. 2).
Taxon level visitation frequency (visitor abundance per number of available flowers) varied significantly (F7,134 = 15·587, P <0·0001, Fig. 3a). Honeybees visited flowers at a significantly higher rate than all other taxa (LSD tests: P <0·0001). When taxa were grouped, honeybee visitation frequencies were still significantly higher than both fly and bee groups (F2,69 = 29·835, P <0·0001).
Visitation rate (no flower visits per min) also varied between taxa (F1,416 = 2·013, P =0·052, Fig. 3b) but significance was marginal at P =0·052. Post hoc analysis suggests that this effect is due to B. terrestris visiting significantly more flowers per minute than the honeybee (LSD test: P =0·047) while other taxa did not differ from the honeybee in this respect (LSD test: P >0·05; Table 3).
Table 3. Effectiveness of the eight most frequent flower visitors to 11 B. rapa fields in the Canterbury region, New Zealand
Post-hoc comparisons with A. mellifera LSD test (P)
Visit rate: floral visits per minute (mean ± SE)
Post-hoc comparisons with A. mellifera LSD test (P)
Numbers in parentheses represent n. Post-hoc pair-wise comparisons for visit frequency and visit rate are the P-values from LSD tests.
2·35 × 10−2 ±0·0001
33·83 ± 3·07
1·28 × 10−3 ±0·00006
69·12 ± 20·03
4·83 × 10−4 ±0·00007
64·31 ± 20·16
3·74 × 10−4 ±0·00003
10·03 ± 2·71
3·39 × 10−3 ±0·0003
19·42 ± 1·71
2·92 × 10−3 ±0·0003
7·99 ± 1·07
1·87 × 10−3 ±0·0006
6·38 ± 1·39
4·08 × 10−3 ±0·0003
6·10 ± 1·67
All flies combined
8·20 × 10−3 ±0·004
All bees combined (except Apis mellifera)
9·25 × 10−4 ±0·004
When both efficiency and visitation frequency were combined to produce an estimate of effectiveness (median stigma pollen load per visit × proportion of successful stigma contact × hourly rate of visitation), honeybees were the most effective single pollinator species. We estimated that honeybees accounted for the deposition of 7879 pollen grains per hour which is more than three times greater than the next highest pollinator taxa, (B. terrestris: 2247 pollen grains transferred per hour). The overall effectiveness of the remaining taxa was as follows in pollen grains transferred per hour: D. nigrostigma, 22; E. tenax, 968; L. sordidum, 2; Leioproctus sp., 300; M. fasciatum, 1; M. novae-zelandiae, 13.
This study revealed a diverse unmanaged component of the pollinator assemblage in B. rapa crops. We found that in terms of pollen transfer efficiency, the unmanaged component of the pollinator assemblage includes taxa that are capable of providing pollination services equal to those currently performed by honeybees. First, we found that mean pollen loads deposited on stigmas of virgin flowers by two bee species, B. terrestris and Leioproctus sp. and one fly species, E. tenax, were not significantly different to that deposited by the honeybee (Table 2, Fig. 1). Secondly, these three species were as likely to touch stigmatic surfaces when attending flowers as A. mellifera (Table 2, Fig. 2).
Rate of visitation is an important component affecting pollination success and determining the overall contribution of individual taxa to total pollination services (Vazquez et al. 2005). In this study, we consider both visitation frequency (abundance per number of available flowers) and visitation rate (number of flower visits per min) separately for the purpose of demonstrating the potential of alternative pollinator taxa. Although honeybees visited flowers at significantly higher frequencies than any of the other visitors, they did not differ significantly in the number of flowers visited per minute when compared with all other taxa. We suggest that in the above measures of pollen transfer efficiency and floral visits per min, several alternative taxa are equal to the honeybee but are not common enough to make them more effective overall.
A higher abundance of honeybees ultimately resulted in overall greater effectiveness of honeybees as a single taxon. This result is similar to a study by Madjidian et al. (2008) in Argentina comparing a native and exotic bumblebee species. The higher visitation frequency of the exotic bumblebee Bombus ruderatus resulted in it being a more effective pollinator than the native bumblebee B. dahlbomii, even though the native bumblebee was more efficient.
The higher abundance of honeybees per number of available flowers relative to unmanaged taxa probably reflects managed and unmanaged status. By definition, honeybee populations are managed to maintain high population sizes, whereas unmanaged taxa are not. This does not necessarily preclude currently unmanaged taxa from performing the same services. It is possible that even though the effectiveness of unmanaged taxa was lower, it may still result in maximum seed set. In the absence of seed set data, we cannot test this assumption.
If lower effectiveness results in lower seed set, unmanaged pollinators would need to be managed in order to increase population sizes in accordance with those of the honeybee, beyond that which exists naturally at this time in this system.
Managing a range of naturally existing pollinators in-situ is likely to be challenging for several reasons; First, we have become accustomed to ‘mobile’ as opposed to ‘in-situ’ pollination services. Honeybees are efficient, versatile and easily managed within transportable hives (Morse & Calderone 2000; Klein et al. 2007; Winfree 2008). In contrast, unmanaged pollinators are not as transportable and hence not as versatile (at present). Nonetheless, the effectiveness of flies (Syrphidae in particular) as crop pollinators is becoming increasingly evident (e.g. Feldman 2006; Pontin et al. 2006). For example, pollination by E. tenax was shown to improve the shape and weight of sweet peppers in Canada (Jarlan, De Oliveira & Gingras 1997) while Episyrphus balteatus (also Syrphidae) significantly increased seed set and yield of an oilseed rape crop in cage experiments when compared with control cages (Jauker & Wolters 2008).
Secondly, managing alternative pollinator taxa in situ to achieve high densities is challenging as it requires a thorough investigation of each taxon’s intrinsic biology (Cane et al. 2006). For instance, in this study system (and across most of New Zealand) the indigenous pollinating fauna lacks large social bees and is dominated by solitary bees and flies (Lloyd 1985; Donovan 2007). Solitary bees in particular have a short, fixed flying season, which is synchronized with the flowering time of certain host plants (Minckley et al. 1994; Westerkamp & Gottsberger 2000). This means that timing may not always be compatible with the flowering crop in need of pollination. In contrast, eusocial bees are capable of recruiting foragers quickly to high quality resources (Brosi et al. 2007; Winfree 2008).
Fundamental research into the intrinsic biology and life history traits of both solitary bees and flies is currently lacking in this, and most other systems (Klein et al. 2007). In order to understand which resources are needed for these taxa to maintain stable populations in agricultural landscapes we need to first understand their role and function in their current system.
In conclusion, the results of this study demonstrate that three species that currently exist as part of the unmanaged pollinator assemblage of B. rapa in the South Island of New Zealand are equally as efficient as the honeybee in providing pollination services. Effectiveness was higher in honeybees but this probably reflects the higher population sizes of a managed species giving rise to higher rates of visitation in honeybees. Our results suggest that there is potential for other species to fulfil the pollination role of honeybees under management strategies that increase local population sizes and thus visitation rates.
The authors wish to thank Steve Griffiths, Stewart Armstrong, Smiths Seeds Ltd., Jan Grant, Laura Mesa, Corina Till, Anna-Marie Barnes, Randal Toonen and Emma Parr for their assistance with this project. This manuscript was greatly improved by the comments of the editor and two anonymous reviewers. This project was supported by the New Zealand Foundation for Research Science and Technology and James Cook University Cairns.