Hairiness: the missing link between pollinators and pollination

Submitted 8 September 2016 Accepted 2 November 2016 Published 21 December 2016 Corresponding author Jamie R. Stavert, [email protected] Academic editor Chris Cutler Additional Information and Declarations can be found on page 13 DOI 10.7717/peerj.2779 Copyright 2016 Stavert et al. Distributed under Creative Commons CC-BY 4.0 OPEN ACCESS Hairiness: the missing link between pollinators and pollination Jamie R. Stavert1, Gustavo Liñán-Cembrano2, Jacqueline R. Beggs1, Brad G. Howlett3, David E. Pattemore4and Ignasi Bartomeus5 1Centre for Biodiversity and Biosecurity, School of Biological Sciences, The University of Auckland, Auckland, New Zealand 2Instituto de Microelectrónica de Sevilla CSIC/Universidad de Sevilla, Sevilla, Spain 3The New Zealand Institute for Plant & Food Research Limited, Christchurch, New Zealand 4The New Zealand Institute for Plant & Food Research Limited, Hamilton, New Zealand 5Integrative Ecology Department, Estación Biológica de Doñana (EBD-CSIC), Sevilla, Spain ABSTRACT Background. Functional traits are the primary biotic component driving organism influence on ecosystem functions; in consequence, traits are widely used in ecological research. However, most animal trait-based studies use easy-to-measure characteristics of species that are at best only weakly associated with functions. Animal-mediated pollination is a key ecosystem function and is likely to be influenced by pollinator traits, but to date no one has identified functional traits that are simple to measure and have good predictive power. Methods. Here, we show that a simple, easy to measure trait (hairiness) can predict pollinator effectiveness with high accuracy. We used a novel image analysis method to calculate entropy values for insect body surfaces as a measure of hairiness. We evaluated the power of our method for predicting pollinator effectiveness by regressing pollinator hairiness (entropy) against single visit pollen deposition (SVD) and pollen loads on insects. We used linear models and AICCmodel selection to determine which body regions were the best predictors of SVD and pollen load. Results. We found that hairiness can be used as a robust proxy of SVD. The best models for predicting SVD for the flower species Brassica rapa and Actinidia deliciosa were hairiness on the face and thorax as predictors (R2=0.98 and 0.91 respectively). The best model for predicting pollen load for B.rapa was hairiness on the face (R2=0.81). Discussion. We suggest that the match between pollinator body region hairiness and plant reproductive structure morphology is a powerful predictor of pollinator effectiveness. We show that pollinator hairiness is strongly linked to pollination—an important ecosystem function, and provide a rigorous and time-efficient method for measuring hairiness. Identifying and accurately measuring key traits that drive ecosystem processes is critical as global change increasingly alters ecological communities, and subsequently, ecosystem functions worldwide. Subjects Biodiversity, Ecology, Ecosystem Science, Entomology, Zoology Keywords Pollination, Pilosity, Entropy, Functional trait, Pollen deposition, Ecosystem function, Image analysis, Pollen load, SVD How to cite this article Stavert et al. (2016), Hairiness: the missing link between pollinators and pollination. PeerJ 4:e2779; DOI 10.7717/peerj.2779 INTRODUCTION Trait-based approaches are now widely used in functional ecology, from the level of individual organisms to ecosystems (Cadotte, Carscadden & Mirotchnick, 2011). Functional traits are defined as the characteristics of an organism’s phenotype that determine its effect on ecosystem level processes (Naeem & Wright, 2003;Petchey & Gaston, 2006). Accordingly, functional traits are recognised as the primary biotic component by which organisms influence ecosystem functions (Gagic et al., 2015;Hillebrand & Matthiessen, 2009). Trait-based research is dominated by studies on plants and primary productivity, and little is known about key traits for animal-mediated and multi-trophic functions, particularly for terrestrial invertebrates (Didham, Leather & Basset, 2016;Gagic et al., 2015; Lavorel et al., 2013). Most animal trait-based studies simply quantify easy-to-measure morphological characteristics, without a mechanistic underpinning to demonstrate these ‘‘traits’’ have any influence on the ecosystem function of interest (Didham, Leather & Basset, 2016). This results in low predictive power, particularly where trait selection lacks strong justification through explicit ecological questions (Gagic et al., 2015;Petchey & Gaston, 2006). If the ultimate goal of trait-based ecology is to identify the mechanisms that drive biodiversity impacts on ecosystem function, then traits must be quantifiable at the level of the individual organism, and be inherently linked to an ecosystem function (Bolnick et al., 2011;Pasari et al., 2013;Violle et al., 2007). Methodology that allows collection of trait data in a rigorous yet time-efficient manner and with direct functional interpretation will greatly enhance the power of trait-based studies. Instead of subjectively selecting a large number of traits with unspecified links to ecosystem functions, it would be better to identify fewer, uncorrelated traits, that have a strong bearing on the function of interest (Carmona et al., 2016). Selecting traits that are measurable on a continuous scale, would also improve predictive power of studies (McGill et al., 2006;Violle et al., 2012). However, far greater time and effort is required to measure such traits, exacerbating the already demanding nature of trait-based community ecology (Petchey & Gaston, 2006). Animal-mediated pollination is a multi-trophic function, driven by the interaction between animal pollinators and plants (Kremen et al., 2007). A majority of the world’s wild plant species are pollinated by animals (Ollerton, Winfree & Tarrant, 2011), and over a third of global crops are dependent on animal pollination (Klein et al., 2007). Understanding which pollinator traits determine the effectiveness of different pollinators is critical to understanding the mechanisms of pollination processes. However, current traits used in pollination studies often have weak associations with pollination function and/or have low predictive power. For example Larsen, Williams & Kremen (2005) used body mass to explain pollen deposition by solitary bees even when the relationship was weak and non-significant. Many trait-based pollination studies have subsequently used body mass or similar size measures, despite their low predictive power. Similarly, Hoehn et al. (2008) used spatial and temporal visitation preferences of bees to explain differences in plants reproductive output. They found significant relationships (i.e., low Pvalues) Stavert et al. (2016), PeerJ, DOI 10.7717/peerj.2779 2/18 between spatial and temporal visitation preferences and seed set, but with small R2values, suggesting these traits have weak predictive power. To advance trait-based pollination research we require traits that are good predictors of pollination success. Observational studies suggest that insect body hairs are important for collecting pollen that is used by insects for food and larval provisioning (Holloway, 1976;Thorp, 2000). Hairs facilitate active pollen collection, e.g., many bees have specialised hair structures called scopae that are used to transport pollen to the nest for larval provisioning (Thorp, 2000). Additionally, both bees and flies have hairs distributed across their body surfaces which act to passively collect pollen for adult feeding (Holloway, 1976). Differences in the density and distribution of hairs on pollen feeding insects likely reflects their feeding behaviour, the types of flowers they visit, and whether they use pollen for adult feeding and/or larval provisioning (Thorp, 2000). However, despite anecdotal evidence that insect body hairs are important for pollen collection and pollination, there is no proven method for measuring hairiness, nor is there evidence that hairier insects are more effective pollinators. Here, we present a novel method based on image entropy analysis for quantifying pollinator hairiness. We define pollination effectiveness as single visit pollen deposition (SVD): the number of conspecific pollen grains deposited on a virgin stigma in a single visit (King, Ballantyne & Willmer, 2013;Ne’eman et al., 2010). SVD is a measure of an insects’ ability to acquire free pollen grains on the body surface and accurately deposit them on a conspecific stigma. We predict that hairiness, specifically on the body parts that contact the stigma, will have a strong association with SVD. We show that the best model for predicting pollinator SVD for pak choi Brassica rapa is highly predictive and includes hairiness of the face and thorax dorsal regions as predictors, and the face region alone explains more than 90% of the variation. Similarly, the best model for predicting SVD for kiwifruit Actinidia deliciosa includes the face and thorax ventral regions and has good predictive power. Our novel method for measuring hairiness is rigorous, time efficient and inherently linked to pollination function. Accordingly, this method could be applied in diverse trait-based pollination studies to progress understanding of the mechanisms that drive pollination processes. MATERIALS AND METHODS Imaging for hairiness analysis We photographed pinned insect specimens using the Visionary Digital Passport portable imaging system (Fig. 1). Images were taken with a Canon EOS 5D Mark II digital camera (5,616 ×3,744 pix). The camera colour profile was sRGB IEC61966-2.1, focal length was 65 mm and F-number was 4.5. We used ventral, dorsal and frontal shots with clear illumination to minimise reflection from shinny insect body surfaces. All photographs were taken on a plain white background. Raw images were exported to Helicon Focus 6 where they were stacked and stored in .jpg file format. Image processing and analysis We produced code to quantify insect pollinator hairiness using MATLAB (MathWorks, Natick, MA, USA), and functions from the MATLAB Image Processing ToolBox. We Stavert et al. (2016), PeerJ, DOI 10.7717/peerj.2779 3/18 50 100 150 200 250 ab Figure 1 Entropy image of the face of a native New Zealand solitary bee Leioproctus paahaumaa (A) and the corresponding entropy image (B). Warmer colours on the entropy image represent higher entropy values (shown by the scale bar on the right). Black dots on the entropy image are near-round and small objects that have been removed from the analysis by the pre-processing function. quantified relative hairiness by creating an entropy image for each insect photograph, and computed the average entropy within user-defined regions (Gonzales, Woods & Eddins, 2004). To calculate entropy values for each image we designed three main functions. The first function allows the user to define up to four regions of interest (RoIs) within each image. The user can define regions by drawing contours as closed polygonal lines of any arbitrary number of vertexes. All information about regions (location, area and input image file name) is stored as a structure in a .mat file. The second function executes image pre-processing. We found that some insects had pollen grains or other artefacts attached to their bodies, which would alter the entropy results. Our pre-processing function eliminates these objects from the image by running two filtering processes. First, the function eliminates small objects with an area less than the user definable threshold (8 pixels by default). For the first task, each marked region is segmented using an optimized threshold obtained by applying a spatially dependant thresholding technique. Once each region has been segmented, a labelling process is executed for all resulting objects and those with an area smaller than the minimum value defined by the user are removed. Secondly, as pollen grains are often round in shape, the function eliminates near-circular objects. The perimeter of each object is calculated and its similarity to a circle (S) id defined as: S=4π·Area Perimeter2. Objects with a similarity coefficient not within the bounds defined by the user (5% by default) are also removed from the image. Perimeter calculation is carried out by finding the object’s boundary, and computing the accumulated distance from pixel centre to pixel centre across the border, rather than simply counting the number of pixels in the border. The entropy filter will not process objects that have been marked as ‘‘deleted’’ by the Stavert et al. (2016), PeerJ, DOI 10.7717/peerj.2779 4/18 pre-processing function. This initial pre-processing provides flexibility by allowing users to define the minimum area threshold and the degree of similarity of objects to a circle. Users can also disable the image pre-processing by toggling a flag when running the entropy filter. Once pre-processing is complete, each image is passed to the third function, which is the entropy filter calculation stage. The entropy filter produces an overall measure of randomness within each of the user defined regions on the image. In information theory, entropy (also expressed as Shannon Entropy) is an indicator of the average amount of information contained in a message (Shannon, 1948). Therefore, Shannon Entropy, H, of a discrete random variable Xthat can take npossible values {x1,x2,...,xn}, with a probability mass function P(X) is given by: H(X)= − n X i=1 P(xi)·log2(P(xi)). When this definition is used in image processing, local entropy defines the degree of complexity (variability) within a given neighbourhood around a pixel. In our case, this neighbourhood (often referred to as the structuring element) is a disk with radius r (we call the radius of influence) that can be defined by the user (7 pixels by default). Thus for a given pixel in position (i,j) in the input image, the entropy filter computes the histogram Gij (using 256 bins) of all pixels within its radius of influence, and returns its entropy value Hij as: Hij = −Gij ·log2(Gij), where Gij is a vector containing the histogram results for pixel (i,j) and (·) is the dot product operator. Using default parameters, our entropy filter employs a 7 pixel (13 ×13 neighbourhood) radius of influence, and a disk-shaped structuring element, which we determined based on the size of hairs. Therefore, in the entropy image, each pixel takes a value of entropy when considering 160 pixels around it (by default). We determined the optimal radius of influence for the entropy filter by running our entropy function with the radius of influence set as a variable parameter. We then visually compared the contrast in areas of low vs. high hairiness in the resulting entropy images (i.e., Fig. 1). We found that a 7 pixel radius of influence gave the best contrast between low and high hairiness areas for our species set. Hair thickness values across species typically ranged between 3.5–4.5 pixels and therefore, the 7 pixel radius of influence is approximately two times the width of a hair. The definition of the optimum radius of influence depends on the size of the morphological responsible for the complexity in the RoI. This is defined not only by the physical size of these features but also by the pixel-to-millimetre scaling factor (i.e., number of pixels in the sensor plane per mm in the scene plane). Thus, although 7 pixels is the optimum in our case to detect hairs, the entropy filter function takes this radius as an external parameter which can be adjusted by the user to meet their needs. Stavert et al. (2016), PeerJ, DOI 10.7717/peerj.2779 5/18 The entropy filter function is a process that runs over three different entropy layers (ER,EG,EB), one for each of the camera’s colour channels (Red, Green, and Blue), for each input image. These three images are combined into a final combined entropy image ES, where each pixel in position (i,j) takes the value ES(i,j): ES(i,j)=ER(i,j)·EG(i,j)·EG(i,j) Once entropy calculations are complete, our function computes averages and standard deviations of ESwithin each of the regions previously defined by the user, and writes the results into a .csv file (one row per image). Entropy values produced by this function are consistent for different photos of the same region on the same specimen (Supplemental Information 5). The scripts for the image pre-processing, region marking and entropy analysis functions are provided, along with a MATLAB tutorial (Supplemental Information 1–4). Hairiness as a predictor of SVD and pollen load Model flower floral biology and pollinator collection We used pak choi Brassica rapa var. chinensis (Brassicaceae) and kiwifruit Actinidia deliciosa (Actinidiaceae) as model flowers to determine if our measurement of insect hairiness is a good predictor of pollinator effectiveness. Both B.rapa and A.deliciosa are important mass flowering global food crops (Klein et al., 2007;Rader et al., 2009). B.rapa has an actinomorphic open pollinated yellow flower with four sepals, four petals, and six stamens (four long and two short) (Walker, Kinzig & Langridge, 1999). The nectaries are located in the centre of the flower, between the stamens and the petals, forcing pollinators to introduce their head between the petals. B.rapa shows increased seed set in the presence of insect pollinators and the flowers are visited by a diverse assemblage of insects that differ in their ability to transfer pollen (Rader et al., 2013). A.deliciosa is dioecious with individual plants producing either male or female flowers. Flowers are large (4–6 cm in diameter) and typically have 5–9 white/cream coloured petals (Devi, Thakur & Garg, 2015). Flowers have multiple stamens and staminodes with yellow anthers. Female flowers have a large stigma with multiple branches that form a brush-like structure. Both male and female flowers do not produce nectar but both produce pollen, which acts as a reward to visitors. Like B.rapa,A.deliciosa flowers are visited by a diverse range of insects that differ in their ability to transfer pollen, and seed set is increased in the presence of insect pollinators (Craig et al., 1988). We collected pollinating insects for image analysis during the summer of December 2014–January 2015. Insects were chilled immediately and then killed by freezing within 1 day and stored at −18 ◦C in individual vials. All insects were identified to species level with assistance from expert taxonomists. Image processing We measured the hairiness of 10 insect pollinator species (n=8–10 individuals per species), across five families and two orders. This included social, semi-social and solitary bees and pollinating flies. Regions marked included: (1) face; (2) head dorsal; (3) head ventral; Stavert et al. (2016), PeerJ, DOI 10.7717/peerj.2779 6/18 (4) front leg; (5) thorax dorsal; (6) thorax ventral; (7) abdomen dorsal and (8) abdomen ventral. All entropy analysis was carried out using our image processing method outlined above. For estimates of body size, we took multiple linear measurements (body length, body width, head length, head width, foreleg length and hind leg length) of each specimen using digital callipers and a dissecting microscope. Single visit pollen deposition (SVD) and pollen load For B.rapa we used SVD data for insect pollinators presented in Rader et al. (2009) and Howlett et al. (2011); a brief description of their methods follows. Pollen deposition on stigmatic surfaces (SVD) was estimated using manipulation experiments. Virgin B.rapa inflorescences were bagged to exclude all pollinators. Once flowers had opened, the bag was removed, and flowers were observed until an insect visited and contacted the stigma in a single visit. The stigma was then removed and stored in gelatine-fuchsin and the insect was captured for later identification. SVD was quantified by counting all B.rapa pollen grains on the stigma. Mean values of SVD for each species are used in our regression models. To quantify the number of pollen grains carried (pollen load), sensu Howlett et al. (2011), collected insects while foraging on B.rapa flowers. Insects were captured using plastic vials containing a rapid killing agent (ethyl acetate). Once dead, a cube of gelatine-fuchsin was used to remove all pollen from the insect’s body surface. Pollen collecting structures (e.g., corbiculae, scopae) were not included in analyses because pollen from these structures is not available for pollination. Slides were prepared in the field by melting the gelatine-fuchsin cubes containing pollen samples onto microscope slides. B. rapa pollen grains from each sample were then quantified by counting pollen grains in an equal-area subset from the sample and multiplying this by the number of equivalent sized subset areas within the total sample. We measured SVD for A.deliciosa (n=8–12 per pollinator species). SVD measurements were taken for insect movements from staminate to pistillate flowers, using a method that differed from B. rapa. Individual pistillate buds were enclosed within paper bags 2–3 days prior to opening, and were later used as test flowers to evaluate pollen deposition by flowering visiting species. Each bag was secured using a wire tie (coated in plastic) that was gently twisted to exclude pollinators from visiting the opening flowers. Following flower opening, the bag was removed and the flower pedicel abscised where it joined the vine. The test flower was then carefully positioned using forceps to hold the pedicel 1–2 cm from a staminate flower containing a foraging insect, avoiding any contacting between flowers. If the test flower was visited by an insect, we allowed it to forage with minimal disturbance until it moved from the flower on its own accord. The first stigma touched by the foraging insect was then lightly marked near its base using a fine black felt pen. We then placed the marked stigma onto a slide and applied a drop of Alexander stain (Dafni, 2007). Alexander stain was used due to its effectiveness to stain staminate and pistillate pollen differently (pistillate pollen—green-blue, staminate pollen—dark red) (Goodwin & Perry, 1992). Stavert et al. (2016), PeerJ, DOI 10.7717/peerj.2779 7/18 Statistical analyses We used linear regression models and AICC(small sample corrected Akaike information criteria) model selection to determine if our measure of pollinator hairiness is a good predictor of SVD and pollen load. We constructed global models with SVD or pollen load as the response variable, body region as predictors and body length as an interaction i.e., SVD or pollen load ∼body length * entropy face +entropy head dorsal +entropy head ventral +front leg +entropy thorax dorsal +entropy thorax ventral +entropy abdomen dorsal +entropy abdomen ventral. We included body length in our global model as a proxy for body size as it had high correlation coefficients (Pearson’s r> 0.7) with all other body size measurements. Global linear models were constructed using the lm(stats) function. AICCmodel selection was carried out on the global models using the function glmulti() with fitfunction =‘‘lm’’ in the package glmulti. We examined heteroscedasticity and normality of errors of models by visually inspecting diagnostic plots using the glmulti package (Crawley, 2002). Variance inflation factors (VIF) of predictor variables were checked for the best models using the vif() function in the car package. All analyses were done in R version 3.2.4 (R Core Team, 2014). RESULTS Body hairiness as a predictor of SVD For SVD on B.rapa, the face and thorax dorsal regions were retained in the best model selected by AICC, which had an adjusted R2value of 0.98. The subsequent top models within 10 AICCpoints all retained the face and thorax dorsal regions and additionally included the abdomen ventral (adjusted R2=0.98), head dorsal (adjusted R2=0.98), and thorax ventral (adjusted R2=0.97) and front leg (adjusted R2=0.97) regions respectively (Table 1;Fig. 2). The model with the face region included as a single predictor had an adjusted R2value of 0.88, indicating that this region alone explained a majority of the variation in the top SVD models. The best model for predicting SVD on A.deliciosa included the face and thorax ventral regions as predictors (adjusted R2=0.91) (Table 1;Fig. 3). However, the subsequent top four models were within two AICCpoints of the best model and therefore cannot be discounted as the potential top model. The face, thorax ventral, head ventral and abdomen ventral regions were retained in four of the five top models, which indicates that hairiness of the face and ventral regions is important for pollen deposition on A. deliciosa. For both B.rapa and A.deliciosa, body length and the body length interaction were not included in the top models. Body hairiness as a predictor of pollen load The best model for pollen load retained the face region only and had an adjusted R2value of 0.81 (Fig. 4;Table 1). The subsequent best models retained the abdomen dorsal (adjusted R2value of 0.73), the face and head dorsal (adjusted R2=0.83), the face and abdomen dorsal (adjusted R2=0.82) and the abdomen dorsal and front leg (adjusted R2=0.8) regions respectively. For pollen load, body length and the body length interaction were not included in the top models. Stavert et al. (2016), PeerJ, DOI 10.7717/peerj.2779 8/18 Table 1 Regression models examining the effect of entropy on SVD and pollen load. Top regression models examining the effect of insect body region entropy on single visit pollen deposition (SVD) for Brassica rapa and Actinidia deliciosa and pollen load for B. rapa. Models are presented in ascending order based on AICCvalues. Top models for each response variable are highlighted in bold. Response variable Model Adj R2AICc1i wiacc wi Face +Thorax dorsal 0.98 88.29 0.00 0.82 0.82 Face +Thorax dorsal +Abdomen ventral 0.98 93.09 4.80 0.07 0.89 Face +Head dorsal +Thorax dorsal 0.98 93.81 5.52 0.05 0.94 Face +Thorax ventral +Thorax dorsal 0.97 96.59 8.29 0.01 0.96 SVD (B.rapa) Face +Thorax dorsal +Front leg 0.97 97.02 8.72 0.01 0.97 Face 0.81 168.47 0.00 0.64 0.64 Abdomen dorsal 0.73 171.59 3.12 0.13 0.78 Face +Head dorsal 0.83 173.59 5.12 0.05 0.83 Face +Abdomen dorsal 0.82 173.76 5.29 0.05 0.87 Pollen load (B.rapa) Abdomen dorsal +Front leg 0.80 174.86 6.39 0.03 0.90 Face +Thorax ventral 0.91 74.18 0.00 0.15 0.15 Abdomen dorsal 0.81 74.21 0.03 0.15 0.30 Face 0.80 74.35 0.17 0.14 0.45 Head ventral 0.79 74.84 0.66 0.11 0.56 SVD (A.deliciosa) Abdomen ventral 0.78 75.08 0.90 0.10 0.65 Notes. 1i is the difference in the AICCvalue of each model compared with the AICCvalue for the top model. wiis the Akaike weight for each model and acc wiis the cumulative Akaike weight. DISCUSSION Here we present a rigorous and time-efficient method for quantifying hairiness, and demonstrate that this measure is an important pollinator functional trait. We show that insect pollinator hairiness is a strong predictor of SVD for the open-pollinated flower B.rapa. Linear models that included multiple body regions as predictors had the highest predictive power; the top model for SVD retained the face and thorax dorsal regions. However, the face region was retained in all of the top models, and when included as a single predictor, had a very strong positive association with SVD. In addition, we show that hairiness, particularly on the face and ventral regions, is a good predictor of SVD for A.deliciosa, which has a different floral morphology, suggesting our method could be suitable for a range of flower types. Hairiness was also a good predictor for pollen load, and the face region was again retained in the top model for B.rapa. The abdomen dorsal, head dorsal and front leg regions were also good predictors of pollen load and were retained in the subsequent top models. Our results validate the importance of insect body hairs for transporting and depositing pollen. Surprisingly, we did not find strong associations between SVD and body size, and top models did not contain the body length interaction. Similarly, body length was not retained in the top models for pollen load. This indicates that our measure of hairiness has far greater predictive power than body size for both SVD and pollen load. When deciding on which body regions to measure hairiness, researchers may first need to assess additional pollinator traits, such as flower visiting behaviour. This is because Stavert et al. (2016), PeerJ, DOI 10.7717/peerj.2779 9/18 Goodwin RM, Perry JH. 1992. Use of pollen traps to investigate the foraging behaviour of honey bee colonies in kiwifruit orchards. New Zealand Journal of Crop and Horticultural Science 20:23–26 DOI 10.1080/01140671.1992.10422322. Herrera CM. 1987. Components of pollinator ‘‘quality’’: comparative analysis of a diverse insect assemblage. Oikos 50:79–90. Hillebrand H, Matthiessen B. 2009. Biodiversity in a complex world: consolidation and progress in functional biodiversity research. Ecology Letters 12:1405–1419 DOI 10.1111/j.1461-0248.2009.01388.x. Hoehn P, Tscharntke T, Tylianakis JM, Steffan-Dewenter I. 2008. Functional group diversity of bee pollinators increases crop yield. Proceedings of the Royal Society B: Biological Sciences 275:2283–2291 DOI 10.1098/rspb.2008.0405. Holloway BA. 1976. Pollen-feeding in hover-flies (Diptera: Syrphidae). New Zealand Journal of Zoology 3:339–350 DOI 10.1080/03014223.1976.9517924. Howlett BG, Walker MK, Rader R, Butler RC, Newstrom-Lloyd LE, Teulon DAJ. 2011. Can insect body pollen counts be used to estimate pollen deposition on pak choi stigmas? New Zealand Plant Protection 64:25–31. Javorek S, Mackenzie K, Vander Kloet S. 2002. Comparative pollination effectiveness among bees (Hymenoptera: Apoidea) on lowbush blueberry (Ericaceae: Vaccinium angustifolium). Annals of the Entomological Society of America 95:345–351 DOI 10.1603/0013-8746(2002)095[0345:CPEABH]2.0.CO;2. King C, Ballantyne G, Willmer PG. 2013. Why flower visitation is a poor proxy for pollination: measuring single-visit pollen deposition, with implications for pollination networks and conservation. Methods in Ecology and Evolution 4:811–818 DOI 10.1111/2041-210X.12074. Klein A-M, Vaissiere BE, Cane JH, Steffan-Dewenter I, Cunningham SA, Kremen C, Tscharntke T. 2007. Importance of pollinators in changing landscapes for world crops. Proceedings of the Royal Society of London B: Biological Sciences 274:303–313 DOI 10.1098/rspb.2006.3721. Kremen C, Williams NM, Aizen MA, Gemmill-Herren B, LeBuhn G, Minckley R, Packer L, Potts SG, Roulston T, Steffan-Dewenter I, Vázquez DP, Winfree R, Adams L, Crone EE, Greenleaf SS, Keitt TH, Klein AM, Regetz J, Ricketts TH. 2007. Pollination and other ecosystem services produced by mobile organisms: a conceptual framework for the effects of land-use change. Ecology Letters 10:299–314 DOI 10.1111/j.1461-0248.2007.01018.x. Kremen C, Williams NM, Thorp RW. 2002. Crop pollination from native bees at risk from agricultural intensification. Proceedings of the National Academy of Sciences of the United States of America 99:16812–16816 DOI 10.1073/pnas.262413599. Larsen TH, Williams NM, Kremen C. 2005. Extinction order and altered community structure rapidly disrupt ecosystem functioning. Ecology Letters 8:538–547 DOI 10.1111/j.1461-0248.2005.00749.x. Lavorel S, Storkey J, Bardgett RD, De Bello F, Berg MP. 2013. A novel framework for linking functional diversity of plants with other trophic levels for the Stavert et al. (2016), PeerJ, DOI 10.7717/peerj.2779 16/18 quantification of ecosystem services. Journal of Vegetation Science 24:942–948 DOI 10.1111/jvs.12083. Mayfield MM, Waser NM, Price MV. 2001. Exploring the ‘most effective pollinator principle’ with complex flowers: bumblebees and Ipomopsis aggregata.Annals of Botany 88:591–596 DOI 10.1006/anbo.2001.1500. McGill BJ, Dornelas M, Gotelli NJ, Magurran AE. 2015. Fifteen forms of biodiversity trend in the Anthropocene. Trends in Ecology & Evolution 30:104–113 DOI 10.1016/j.tree.2014.11.006. McGill BJ, Enquist BJ, Weiher E, Westoby M. 2006. Rebuilding community ecology from functional traits. Trends in Ecology & Evolution 21:178–185 DOI 10.1016/j.tree.2006.02.002. Naeem S, Wright JP. 2003. Disentangling biodiversity effects on ecosystem functioning: deriving solutions to a seemingly insurmountable problem. Ecology Letters 6:567–579 DOI 10.1046/j.1461-0248.2003.00471.x. Ne’eman G, Jürgens A, Newstrom-Lloyd L, Potts SG, Dafni A. 2010. A framework for comparing pollinator performance: effectiveness and efficiency. Biological Reviews 85:435–451 DOI 10.1111/j.1469-185X.2009.00108.x. Ollerton J, Winfree R, Tarrant S. 2011. How many flowering plants are pollinated by animals? Oikos 120:321–326 DOI 10.1111/j.1600-0706.2010.18644.x. Pasari JR, Levi T, Zavaleta ES, Tilman D. 2013. Several scales of biodiversity affect ecosystem multifunctionality. Proceedings of the National Academy of Sciences of the United States of America 110:10219–10222 DOI 10.1073/pnas.1220333110. Petchey OL, Gaston KJ. 2006. Functional diversity: back to basics and looking forward. Ecology Letters 9:741–758 DOI 10.1111/j.1461-0248.2006.00924.x. Potts SG, Dafni A, Ne’eman G. 2001. Pollination of a core flowering shrub species in Mediterranean phrygana: variation in pollinator diversity, abundance and effectiveness in response to fire. Oikos 92:71–80 DOI 10.1034/j.1600-0706.2001.920109.x. R Core Team. 2014. R: a language and environment for statistical computing. Vienna: R Foundation for Statistical Computing. Available at http://www.R-project.org/. Rader R, Edwards W, Westcott DA, Cunningham SA, Howlett BG. 2013. Diurnal effectiveness of pollination by bees and flies in agricultural Brassica rapa: implications for ecosystem resilience. Basic and Applied Ecology 14:20–27 DOI 10.1016/j.baae.2012.10.011. Rader R, Howlett BG, Cunningham SA, Westcott DA, Newstrom-Lloyd LE, Walker MK, Teulon DAJ, Edwards W. 2009. Alternative pollinator taxa are equally efficient but not as effective as the honeybee in a mass flowering crop. Journal of Applied Ecology 46:1080–1087 DOI 10.1111/j.1365-2664.2009.01700.x. Rathcke B. 1983. Competition and facilitation among plants for pollination. In: Pollination biology. New York: Academic Press, 305–329. Roubik DW. 2000. Deceptive orchids with Meliponini as pollinators. Plant Systematics and Evolution 222:271–279 DOI 10.1007/BF00984106. Shannon C. 1948. A mathematical theory of communication. Bell System Technical Journal 3:379–423 DOI 10.1002/j.1538-7305.1948.tb01338.x. Stavert et al. (2016), PeerJ, DOI 10.7717/peerj.2779 17/18 Sorensen AE. 1986. Seed dispersal by adhesion. Annual Review of Ecology and Systematics 17:443–463. Thorp RW. 2000. The collection of pollen by bees. Plant Systematics and Evolution 222(1):211–223 DOI 10.1007/BF00984103. Violle C, Enquist BJ, McGill BJ, Jiang L, Albert CH, Hulshof C, Jung V, Messier J. 2012. The return of the variance: intraspecific variability in community ecology. Trends in Ecology & Evolution 27:244–252 DOI 10.1016/j.tree.2011.11.014. Violle C, Navas M-L, Vile D, Kazakou E, Fortunel C, Hummel I, Garnier E. 2007. Let the concept of trait be functional! Oikos 116:882–892 DOI 10.1111/j.0030-1299.2007.15559.x. Walker B, Kinzig A, Langridge J. 1999. Plant attribute diversity, resilience, and ecosystem function: the nature and significance of dominant and minor species. Ecosystems 2:95–113 DOI 10.1007/s100219900062. Stavert et al. (2016), PeerJ, DOI 10.7717/peerj.2779 18/18