Spatial and temporal variation in pollinator effectiveness: do unmanaged insects provide consistent pollination services to mass flowering crops?


  • Romina Rader,

    Corresponding author
    1. School of Marine and Tropical Biology, James Cook University, PO Box 6811, Cairns 4870, Australia
    2. The New Zealand Institute for Plant & Food Research Limited, Private Bag 4704, Christchurch, New Zealand
    • Correspondence author. Department of Entomology, Rutgers, The State University of New Jersey, 93 Lipman Drive, New Brunswick, NJ 08901, USA. E-mail:

    Search for more papers by this author
  • Bradley G. Howlett,

    1. The New Zealand Institute for Plant & Food Research Limited, Private Bag 4704, Christchurch, New Zealand
    Search for more papers by this author
  • Saul A. Cunningham,

    1. CSIRO Ecosystem Sciences, PO Box 1700, Canberra, ACT 2601, Australia
    Search for more papers by this author
  • David A. Westcott,

    1. School of Marine and Tropical Biology, James Cook University, PO Box 6811, Cairns 4870, Australia
    2. CSIRO Ecosystem Sciences, PO Box 780, Atherton, Qld 4883, Australia
    Search for more papers by this author
  • Will Edwards

    1. School of Marine and Tropical Biology, James Cook University, PO Box 6811, Cairns 4870, Australia
    Search for more papers by this author


1. Recent declines in honeybee populations have focused attention on the potential for unmanaged insects to replace them as pollinators of food crops. The capacity of unmanaged pollinators to replace services currently provided by honeybees depends on the spatial and temporal variability of these services, but few quantitative assessments currently exist.

2. We investigated spatial variation in pollinator importance by comparing pollinator efficiency and effectiveness in stigmatic pollen loads, stigmatic contact and visitation rate between honeybees and the seven most abundant unmanaged taxa in 2007. We assessed temporal variability in pollinator visitation using floral visits recorded three times a day over four consecutive years (2005–2008) in 43 ‘Pak Choi’Brassica rapa ssp. chinensis mass flowering fields in the Canterbury region of New Zealand. Further, we compared the aggregate effect of the unmanaged pollinator assemblage to the managed honeybee.

3. Pak Choi was visited by many insect species that vary in abundance and effectiveness as pollen transfer agents. There was spatial variation in the four measures of pollinator importance. Pollen deposited on stigmas and flower visits per minute were not significantly different comparing the unmanaged assemblage to honeybees, although stigmatic contact and visitor abundance per number of open flowers were greater in honeybees.

4. Unmanaged taxa were frequent visitors to the crop in all 4 years. The pooled services provided by the unmanaged assemblage did not differ within a day and were equal to or greater than those provided by honeybees in 2 of the 4 years. Pollinator importance changed little irrespective of the spatial and temporal variations among taxa.

5.Synthesis and applications. The results of this study suggest that some unmanaged insect taxa are capable of providing consistent pollination services over a 4-year period in a commercial mass flowering crop. As these taxa already contribute substantially to the pollination of food crops, they offer a safety net in the case of sudden collapse of managed honeybee hives. To optimize pollination services, we recommend pollinator-specific farm management practices that consider the needs of both managed and unmanaged pollinator taxa.


The current global pollination crisis highlights the advantages of the provision of pollination services by a suite of unmanaged pollinators (McCann 2000; Klein et al. 2007). This is because the intrinsic differences between taxa in their ecological tolerances should result in populations of individual species responding differently to environmental change, such that changes in abundance will not be temporally or spatially coordinated across species (Herrera 1990; Tylianakis, Klein & Tscharntke 2005; Hoehn et al. 2008).

Both abundance and behavioural-mediated mechanisms can enhance the stability of pollination services. Assemblages that contain a wide range of species with different ecological requirements could maintain pollination services as environmental conditions change over time because (i) declines in abundance of some taxa can be offset by increases in others (Yachi & Loreau 1999; McCann 2000; Elmqvist et al. 2003) and (ii) interspecific interactions can enhance net pollinator services (Greenleaf & Kremen 2006). This is the basis of the biological insurance hypothesis with respect to pollination as an ecosystem service (Walker 1992; Lawton & Brown 1993; Naeem & Li 1997; Naeem 1998).

Ascertaining the reliability of unmanaged, multi-species, pollinator assemblages requires an assessment of their overall contribution to pollination and the upper and lower limits of spatial and temporal variability associated with each species’ contribution (i.e. their reliability; Watanabe 1994; Allen-Wardell et al. 1998; Naeem 1998; Elmqvist et al. 2003; Memmott et al. 2007). While the potential for unmanaged insects to act as crop pollinators has been addressed (Klein et al. 2007; Winfree et al. 2007; Hoehn et al. 2008; Rader et al. 2009), few studies have assessed the consistency of pollination services in these systems (but see Kremen, Williams & Thorp 2002; Winfree & Kremen 2009). This is probably due to the inherent temporal variability in abundance estimates for unmanaged taxa (Cane & Payne 1993; Roubik 2001) that can obscure general patterns. This variability is especially prevalent in intensive agricultural systems because the ephemeral nature of floral resources (Corbet 2000; Williams & Kremen 2007; Ricketts et al. 2008) can alter pollinator foraging behaviours (Diekotter et al. 2010) and reduce the presence of habitat specialists (Tylianakis, Klein & Tscharntke 2005).

‘Pak Choi’Brassica rapa L. ssp. chinensis L. (Hanelt.) (Brassicaceae) in New Zealand is visited by many insect species that vary in abundance and effectiveness as pollen transfer agents (Rader et al. 2009). However, the capacity for a suite of unmanaged taxa to provide consistent pollination services to agricultural crops across a landscape, within a day and over a period of years, has to our knowledge not been demonstrated previously. We maintain that for this consistency to be applicable to industry and land managers, (i) unmanaged pollinators must be able to perform in commercial agricultural crops where productivity is directly linked to economic outcomes (Aizen et al. 2008; Allsopp, de Lange & Veldtman 2008; Gallai et al. 2009); (ii) unmanaged pollinator services need to be comparable to existing managed pollinator services if they are to offer a safety net in the case of managed pollinator collapse. Managed pollinators are, therefore, an ideal benchmark by which to assess any potential economic gain or loss provided by unmanaged taxa.

In this study, we compare the pollination services provided by the managed honeybee to unmanaged pollinator taxa in commercial mass flowering B. rapa seed crops to ask the following questions:

  • 1 Are unmanaged pollinator taxa capable of providing consistent pollination services to a mass flowering crop across several locations and over time?
  • 2 Are these services likely to provide pollination rates equal to that of the managed honeybee?

These questions are important to assess the likelihood that unmanaged pollinators can provide commercial level pollination services to a mass flowering crop.

Materials and methods

Study species

Brassica rapa is a mass flowering crop. This and other species of Brassicaceae are grown commercially in New Zealand for use as forage, seed, vegetable and oilseed production (Stewart 2002). Brassica rapa is an ideal crop to examine spatial and temporal changes in pollinator visitation as it attracts a diverse assemblage of insects (Feldman 2006; Howlett et al. 2009; Rader et al. 2009), displays increased seed set in the presence of insect pollinators (Free 1993) and is ubiquitous in most agricultural landscapes as a crop/environmental weed (Heenan, Fitzjohn & Dawson 2004; Feldman 2006; Sutherland, Justinova & Poppy 2006).

Frequently visiting pollinators

Eight pollinators were considered ‘frequent visitors’ in this study as they were responsible for 79·8% of all flower visits. These include four bees: Apis mellifera Linnaeus, 1758, Bombus terrestris (Linnaeus, 1758), Lasioglossum sordidum (Smith, 1853) and Leioproctus sp. and four flies, Dilophus nigrostigma (Walker, 1848), Melangyna novaezelandiae (Macquart, 1855), Eristalis tenax (Linnaeus, 1758) and Melanostoma fasciatum (Macquart, 1850).

Pollinator effectiveness

We used four measures to characterize overall pollinator effectiveness. Two of these were related to pollen transfer efficiency as follows: (i) amount of pollen deposited on stigmas per single visit and (ii) stigmatic contact, and two others related to the rate of visitation: (iii) flower visitation rate (number of flower visits by an individual pollinator per minute) and (iv) visitor abundance per number of available open flowers (measured by observed visits to a 10 × 10 m quadrat per 10 min and corrected for the no. open flowers in the quadrat; Rader et al. 2009). To simplify terminology relating to the rate of visitation, we refer to (iii) as ‘flower visits per minute’ and (iv) as ‘visitor abundance per number of open flowers’. As these measures have been used widely in the literature to quantify and rank individual pollinator contributions to overall pollination services (Primack & Silander 1975; Herrera 1987; Vazquez, Morris & Jordano 2005; Ne’eman et al. 2010), we use all four measures in an attempt to understand the spatial and temporal patterns relating to the provision of pollinator services by unmanaged taxa.

Spatial differences in pollen transfer efficiency and visitation rate

To investigate spatial variation in pollen transfer efficiency and visitation rate, we observed flower visitors in four commercial B. rapa fields (two in Lincoln (Lincoln site 1: 43°37′53·92″S; 172°29′03·96″E; Lincoln site 2: 43°37′28·52″S; 172°28′12·46″E) and two in Gore (Gore site 1: 46°07′00·13″S; 168°52′58·74″E; Gore site 2: 46°06′28·89″S; 168°53′35·70″E) on the South Island of New Zealand between December 2006 and February 2007. Fields were selected according to their size (2 ha) and location adjacent to pasture (predominately Lolium perenne) without neighbouring flowering crops. Common flowering weeds were present within field margins at low densities. These included white clover Trifolium repens L., shepherds purse Capsella bursa-pastoris L. (Medik.), hedge mustard Sisymbrium officinale L. (Scop.) and Mallow Malva spp. Because of the large number of honeybee hives operating in close proximity to all four study sites (range 3–8), we assume all honeybee visits were from managed hives. All floral visitor observations were made on sunny or partly cloudy days when the temperature was >16 °C and wind speed <5 ms−1.

Pollen deposition was estimated by bagging (fine mesh 50 × 50 um) randomly selected virgin inflorescences in bud to exclude pollinators. When flowers opened, we removed the bag and observed until 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 placed on top. We counted pollen loads (after single visits) for 338 stigmas collected from four fields and observed 465 stigmatic contact events for 13–25 individuals of each of the frequent visitor species. To evaluate the absolute contribution of insects to pollination, additional ‘control’ stigmas (n = 338) were processed using the same methods. See Rader et al. (2009) and Walker et al. (2009) for further details.

We determined the proportion of all flower visits that resulted in stigmatic contact per insect by following individual insects for each of the frequently visiting species. A hand-held video camera was used to record the behaviour of each insect for the period required for them to visit 10 flowers. This recording enabled the determination of the proportion of occasions in which an insect successfully touched the stigmatic surface. We collected data for 5–23 individuals for each of the eight taxa. We did not examine changes in pollen deposition or stigmatic contact within a day or between years as this was beyond the scope of our study; hence, we assume per visit efficiency measures are equal across years.

Temporal differences in pollinator visitation

To investigate temporal variation in pollinator visitation rate, we observed pollinators in 8–12 commercial B. rapa fields in the same locations each year for 4 years (2005–2008; 43 fields in total). To calculate 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 voice recorder. We observed flower visits for 21–132 individuals of each of eight species for a total of 436 visits.

To calculate visitor abundance per number of open flowers, five observation quadrats (10 × 10 m) were established per field, one near each of the four field corners in four directions (North West, North East, South West, South East) and one in the field centre. Observations of floral visitors were made by walking along each of the four boundaries of the quadrat (i.e. 10 m) and recording all insect species and abundances within 1 m of the boundary observed during a 10-min period. Where species identity was not determined at the time of observation, specimens were collected and taken back to the laboratory for identification. It took approximately 1 h to complete observations in all five locations. All five quadrats in each field were observed for three observation periods (3 h) throughout the day, 10·00–11·00, 12·00–13·00 and 14·00–15·00 h for 3 days.

In each of these observation quadrats, three randomly located 1 m2 quadrats were used to estimate flower density by counting the number of individual plants within each quadrat, the number of inflorescences per plant and the number of flowers per inflorescence on 10 randomly selected plants. The mean 1 m2 quadrat values were then extrapolated to estimate the number of flowers in each of the observation quadrats. Insect visitor frequencies were then divided by the estimates of open flowers (from the mean 1 m2) to remove the confounding effect of differences in floral density between fields on visit frequency (see Ivey, Martinez & Wyatt 2003).

Overall pollinator effectiveness

We define overall pollinator effectiveness as a product of pollen transfer efficiency for each taxon (median pollen deposited on stigmas × proportion of successful stigma contacts) and the frequency of visits per h (visitor abundance per number of open flowers × flower visits per min × 10 min−1 × 6).

Data analyses

We used three different models to investigate spatial and temporal differences in pollen transfer efficiency and visitation rate. All statistical analyses were calculated with R Statistical Software (R Development Core Team 2011). We used Akaike’s information criterion (AIC) as a guide to select the best models in all analyses. First, we used a mixed effects model (R nlme package; Pinheiro et al. 2009) to examine differences in pollen deposited on stigmas in a single visit and visitation rate (i.e. flower visits per minute and visitor abundance per number of open flowers). The fixed factors considered were ‘field’ and ‘pooled taxa’ where ‘field’ represented four levels (Gore site 1, Gore site 2, Lincoln site 1 and Lincoln site 2) and ‘pooled taxa’ represented two levels (the honeybee and the pooled visits of all unmanaged taxa present at each field). In this model, time of day was a random effect. We then ran a similar model where the fixed effect ‘taxon’ comprised a variable with eight levels comparing the seven individual unmanaged taxa to the honeybee. Violation of homoscedasticity required the square root transformation of the dependent variables (i.e. pollen deposited on stigmas, flower visits per minute and visitor abundance per number of open flowers).

Secondly, we investigated spatial differences in stigmatic contact between managed and unmanaged taxa among fields using a generalized linear model (McCullagh & Nelder 1989) with number of stigmatic contacts as the response variable and ‘pooled taxa’ and ‘field’ as fixed factors. We employed a model based on a binomial distribution and controlled for overdispersion by correcting the standard errors using a quasi-GLM model (Zuur et al. 2009).

Thirdly, to examine temporal changes in visitor abundance per number of open flowers across fields, within a day and among years, we used a generalized least squares model in the nlme package (Pinheiro et al. 2009) with the factors ‘taxon’ (with separate models for individual frequent visitors and all pooled unmanaged taxa), time of day (10·00, 12·00 and 14·00 h) and year (2005–2008). To allow for temporal correlation within a day and between years, we compared the model fit under different residual autocorrelation structures (Crawley 2007; Zuur et al. 2009) and selected the final model based on the lowest AIC. To investigate the stability of visitation services provided by all unmanaged taxa alone (i.e. all 43 taxa), we ran this same model after exclusion of honeybee records. Where possible post hoc pair wise comparisons were conducted using Tukeys HSD test. Heteroscedastic variation within treatments was controlled using the method outlined in Herberich, Sikorski & Hothorn (2010) or where this could not be applied we used default treatment contrasts in R.


We observed 42 032 visits to B. rapa flowers over a 4-year period. These visits were made by a total of 43 insect species in addition to honeybees. Of these, seven unmanaged insect species were frequent visitors (i.e. observed in >10 time periods per year) in all 4 years. The seven frequent visitors in addition to A. mellifera were one introduced bee, B. terrestris, two native bees, L. sordidum and Leioproctus sp. and four flies, D. nigrostigma, M. novaezelandiae, E. tenax and M. fasciatum. Honeybees were responsible for 40·6% of all visits, and the seven unmanaged taxa were responsible for 39·2%.

Visitation rates of the remaining 35 species (see Table S1, Supporting Information) were highly variable. Fourteen species visited with high frequency (>10 time periods per year) in two or three of the 4 years, 12 were frequent in 1 year only, while the remaining nine species were always observed in low numbers (<5 time periods) and occurred in one or more years (Table S1, Supporting Information).

Comparisons between the honeybee and unmanaged taxa

The amount of pollen deposited on stigmas by the honeybee was marginally higher than that of the unmanaged pollinator assemblage, but the difference was not significant (F1,130 = 3·12; = 0·079; Fig. 1a). The proportion of stigmatic contacts was significantly greater in honeybees than for unmanaged taxa (t = 6·55; < 0·0001; Fig. 1b).

Figure 1.

 Boxplots of spatial differences in four pollinator measures used to determine overall effectiveness. Box indicates quartiles with median marked as a horizontal line; points are outliers, and whiskers (error bars) above and below the box indicate the 90th and 10th percentiles. (a) Stigmatic pollen; (b) stigmatic contact; (c) flower visits per min; (d) visitor abundance per number of open flowers.

Flower visits per minute did not vary significantly between the honeybee and the unmanaged pollinator assemblage (F1,6 = 0·86, = 0·39; Fig. 1c). Honeybees visited more frequently than the unmanaged pollinator assemblage in the second measure of visitation: visitor abundance per number of open flowers (F1,94 = 57·15, < 0·001; Fig. 1d).

Spatial variation in pollen transfer efficiency

There were no significant differences between fields in the amount of pollen deposited on stigmas (F3,130 = 3·42; = 0·20; Fig. 1a). Stigmatic contact varied significantly between fields with Gore site 1 having a higher proportion of flower visits resulting in stigmatic contacts than Gore site 2 (t = −2·75; = 0·007; Fig. 1b) and Lincoln site 2 (t = −4·07; < 0·001; Fig. 1b), and Lincoln site 1 had significantly greater proportion of flower visits resulting in stigmatic contacts than Lincoln site 2 (t = 3·5, < 0·001; Fig. 1b).

Spatial variation in pollinator visitation rate

Flower visits per minute varied between field locations (F3,408 = 5·63, < 0·001). Flower visits per minute at Lincoln site 1 were significantly higher than Lincoln site 2 (t = 3·57, = 0·0004; Fig. 1c) and at Gore site 1 were significantly higher than Lincoln site 2(t = 3·15, = 0·001; Fig. 1c).

Visitor abundance per number of open flowers was significantly higher at Lincoln site 1 than the other three sites (Lincoln site 2: t = 11·49; Gore site 1: t = 16·20; Gore site 2: t = 18·6; < 0·01; Fig. 1d).

Overall pollinator importance

When efficiency estimates were compared between frequent visitors, the honeybee was the most effective pollinator overall in comparison with each individual unmanaged taxon (Table 1). Variations in different measures of pollination among taxa and across sites, however, did little to alter the relative order of importance of unmanaged pollinators, particularly the four most effective taxa (Table 1). When unmanaged taxa were pooled as a group, they were more effective overall than the honeybee at two of the four field locations (Table 2).

Table 1.  Spatial differences in the four pollinator efficiency measures recorded in a single year of observation for each of the frequent flower visitors. Median values presented per field
SpeciesStigmatic contact (proportion of visits)Stigmatic pollen (per grains visit)Flower visits (min−1)Visitor abundance (per no. of open flowers)Overall effectiveness (pollen grains per hour)Ranking*
  1. *Ranking of taxa within sites is based on overall effectiveness.

Lincoln site 1
 Apis mellifera 0·999·527·460·0151313391
 Bombus terrestris 150233·100·000442652
 Lasioglossum sordidum 0·31412·870·000140·287
 Leioproctus sp.110422·300·00021174
 Dilophus nigrostigma0·522·55·290·0009526
 Eristalis tenax 15023·620·00197843
 Melanostoma fasciatum 0·30·52·260·000330·0047
 Melangyna novaezelandiae 0·6146·010·0014635
Lincoln site 2
 A. mellifera 0·912721·700·00042371
 Bo. terrestris 149·517·500·0001543
 L. sordidum 0·424·55·030·000010·018
 Leioproctus sp.124716·130·0000124
 D. nigrostigma 0·4662·110·000050·16
 E. tenax 0·71599·590·00025102
 Me. fasciatum 0·05145·070·000230·037
 M. novaezelandiae 0·335·55·610·000120·255
Gore site 1
 A. mellifera 16927·400·00261771
 Bo. terrestris 124655·550·000083412
 L. sordidum 0·3348·080·0000100·047
 Leioproctus sp.14716·130·00006924
 D. nigrostigma 0·4182·160·00010·066
 E. tenax 0·9571·55·450·0025333
 Me. fasciatum 0·371·410·0000410·0048
 M. novaezelandiae 0·427·270·000470·15
Gore site 2
 A. mellifera 0·9107·550·850·00274861
 Bo. terrestris 133643·400·000087452
 L. sordidum 0·324·51·390·0000140·0058
 Leioproctus sp.11041·310·00009464
 D. nigrostigma 0·2971·110·00010·086
 E. tenax 0·951266·320·001403
 Me. fasciatum 0·373·800·0000510·017
 M. novaezelandiae 0·31411·810·00125
Table 2.  Comparison of spatial and temporal differences in pollinator efficiency measures between Apis mellifera and all pooled unmanaged taxa. Median values presented per field. Stigmatic pollen loads, stigmatic contact and flower visits per minute recorded in a single year of observation. Visitor abundance per number of open flowers recorded across 4 years of observation
TaxaStigmatic contact (proportion of visits)Stigmatic pollen (per grains visit)Flower visits (min−1)Visitor abundance (per no. of open flowers)Overall effectiveness (pollen grains per hour)
Lincoln site 1
 A. mellifera 0·999·527·170·015413 488
 Unmanaged taxa pooled0·82618·960·00150212
Lincoln site 2
 A. mellifera 0·912716·330·000589395
 Unmanaged taxa pooled0·4586·910·004230
Gore site 1
 A. mellifera 16950·8470·002421
 Unmanaged taxa pooled15511·80·0163738
Gore site 2
 A. mellifera 0·9107·527·390·0027842655
 Unmanaged taxa pooled0·412629·170·0136880

Temporal variation in visitation rate

The generalized least squares model revealed no significant three- (taxa : time : year: F6,272 = 2·00; = 0·16) or two-way interactions (taxa : time: F6,272 = 0·01; = 0·84; time : year: F1,272 = 2·08; = 0·16) with the exception of individual unmanaged taxa and year (F6,272 = 1·80; = 0·001) and pooled taxa and year (F1,986 = 35·00; < 0·001). This interaction arose because unmanaged visitor abundance per number of open flowers was the same or higher, than the honeybee in 2005 (t = −3·56; < 0·0001) and 2006 (t = −0·78; = 0·44), while honeybees visited at a higher rate in 2007 (t = 4·53; < 0·001) and 2008 (t = 4·11; < 0·001; Fig. 2b). Visitor abundance per number of open flowers did not differ significantly between taxa within a day (F1,986 = 0·56; = 0·41).

Figure 2.

 Boxplots of temporal differences between managed and unmanaged taxa in visitor abundance per number of open flowers. Box indicates quartiles with median marked as a horizontal line; points are outliers, and whiskers (error bars) above and below the box indicate the 90th and 10th percentiles. (a) Within a day; (b) over a 4-year period.


In this study, we compared unmanaged pollination services to that provided by managed honeybees in space and over time. We found no differences between honeybees and unmanaged pollinators in two of four measures of pollinator effectiveness. The unmanaged pollinator assemblage visited flowers with equal frequency to the honeybee and transferred equal quantities of pollen to stigmas per visit. Unmanaged pollinators were not, however, as effective as the honeybee when efficiency was expressed in terms of the number of stigmatic contacts or visitor abundance per number of open flowers. Thus, while unmanaged taxa were capable of efficient pollen transfer in single interactions between flower and insect relative to the honeybee, the frequency in which contacts occurred was significantly lower for the unmanaged assemblage.

Pollination services differed spatially among fields. The significant differences observed among fields are likely to be due to differences in pollinator community composition. For example, the significantly greater proportion of flower visits resulting in stigmatic contacts at Gore site 1 (compared to the three other fields) was likely to be due to the presence of three highly efficient bee taxa (i.e. A. mellifera, Bo. terrestris and Leioproctus sp.) and fewer inefficient fly taxa (M. novaezelandiae and Me. fasciatum). Further, as A. mellifera is managed to maintain high population sizes, the high number of honeybee visits per number of open flowers at Lincoln site 1 was most probably because a larger number of hives were located within 1 km of this field site compared to the other sites (i.e. eight hives as opposed to a maximum of four).

In our study area, floral resources such as weeds were present in field margins potentially providing the variation in local resources to impact upon the observed spatial variations among the four pollinator measures (Backman & Tiainen 2002; Marshall, West & Kleijn 2006). It is clear from other studies that resource heterogeneity can impact upon plant–insect interactions in highly modified systems (Isaia, Bona & Badino 2006; Tylianakis et al. 2008); however, we lack the data to explore the influence of local resources on variations in pollinator effectiveness as the principal aim of our design was to test for landscape level field differences rather than micro-spatial heterogeneity within fields.

Visitor abundance per number of open flowers varied among years for both the honeybee and unmanaged taxa. Irrespective of these variations among taxa and years, the seven most frequent unmanaged visitors nonetheless visited B. rapa flowers each year over a four period and the pooled unmanaged assemblage provided visitation services with equal frequency or higher than the managed honeybee in 2 of the 4 years (i.e. 2005 and 2006). Further, when unmanaged visitation services over this 4-year period were combined with median values of pollen transfer efficiency across different fields, unmanaged pollinators as a group performed better than the honeybee in two of the four fields (Table 2). Other studies have demonstrated that spatial and temporal differences in the rate of visitation among taxa can impact upon the ranking of pollinator importance (Fishbein & Venable 1996; Hoehn et al. 2008; Madjidian, Morales & Smith 2008; Olesen et al. 2008). In this study, however, the significant differences between field locations (Fig. 1) revealed in the four pollinator measures did little to alter the ranking of individual pollinator importance (Table 1). Even though these findings were detected across only four fields, the consistency in the rankings of pollinator importance suggests that the relative differences in pollinator measures among taxa are consistent, probably reflecting taxon-specific foraging behaviours (Thomson & Goodell 2001; Goodell & Thomson 2007; Jha & Vandermeer 2009) such as response to floral resource availability (Jha & Vandermeer 2009; Diekotter et al. 2010) and/or the extent to which pollen is groomed from pollinator bodies (Rademaker, deJong & Klinkhamer 1997).

Our findings imply that a diverse unmanaged pollinator assemblage can provide a consistent pollination service similar to the services provided by the managed honeybee. First, in this system, declines in honeybee abundance because of pests or diseases could be offset by alternative species if they are unaffected by these conditions. This study system is in contrast to others conducted on unmanaged pollinators in agricultural systems in that unmanaged pollinators are abundant and diverse irrespective of intensive land use and little remnant vegetation. This is probably due to the abundance of solitary bee and fly pollinator fauna in this system and across much of New Zealand (Lloyd 1985; Donovan 2007) that are capable of nesting and foraging in agricultural land use types.

Secondly, the diversity of the assemblage may indicate a measure of resilience in pollination services, because individual population sizes in diverse species assemblages are unlikely to covary as a function of temporal environmental change (Yachi & Loreau 1999; McCann 2000; Elmqvist et al. 2003), in support of the biological insurance hypothesis (Walker 1992; Lawton & Brown 1993; Naeem & Li 1997; Naeem 1998). This ensures some species are present when others are not (Kremen, Williams & Thorp 2002; Ricketts 2004). For example, Kremen, Williams & Thorp (2002) found that diversity was essential for sustaining the pollination services to watermelon (Citrullus lanatus) crops and that year-to-year variation in community composition had the capacity to buffer crops against population fluctuations of any one given pollinator species. In our study, although the majority of the 43 unmanaged taxa also fluctuated in their visitation rates within a day, as a group they provided consistent services between fields and on a daily and annual basis, equivalent with that of the honeybee in 2 of 4 years.

In conclusion, our study demonstrates the potential for an unmanaged pollinator assemblage to provide efficient and consistent pollination services similar to those provided by the managed honeybee in 2 of the 4 years. Pollen transfer efficiency varied between field locations, yet this did little to alter the rankings of importance among unmanaged pollinators. As unmanaged pollinators performed equally to honeybees in two of the four pollinator measures determining overall effectiveness, this study demonstrates the capacity for unmanaged insects to be consistent pollinators across several locations and over a period of years. Future research should be directed toward (i) determining the specific resources required by unmanaged pollinators to increase populations of efficient taxa; (ii) investigations concerning spatio–temporal interactions in pollinator efficiency and (iii) examining unmanaged pollinator consistency in other agricultural crops. This information is necessary to understand how nutritional/nesting demands change during the season and the associated impacts upon pollinator consistency, identify broad patterns that may emerge across a range of mass flowering crops and to elucidate the best possible land management strategies to maintain or increase the populations of these species. To ensure the consistent delivery of pollination services to mass flowering seed crops across fields and over time, we recommend farm management practices that include pollinator-specific planning. This will ensure strategies of management (i.e. application and timing of pesticides, field margin composition and structure, method of tillage, etc.) account for both managed and unmanaged taxa so that optimal pollination services are provided.


The authors wish to thank M. Walker, S. Armstrong, S. Griffiths and C. Till for technical assistance. D. Teulon provided advice and support. This project was supported by the New Zealand Foundation for Research Science and Technology (project code CO2X0221) and James Cook University, Cairns, Australia.