Elsevier

Ecological Modelling

Volume 267, 10 October 2013, Pages 1-10
Ecological Modelling

Modelling estimates of honey bee (Apis spp.) colony density from drones

https://doi.org/10.1016/j.ecolmodel.2013.07.008Get rights and content

Highlights

  • Honey bee feral colony densities can be measured via new genetic techniques.

  • We construct an agent-based model to simulate the capture of congregating drones.

  • Colony numbers in captured drone samples have a log-linear relationship to density.

  • Colony densities from past surveys using drone traps are likely overestimated.

Abstract

Given reports of declines in populations of pollinators globally, it is increasingly important to develop efficient procedures to assess the density and distribution of honey bee colonies in both agricultural and natural landscapes. One such procedure utilises the fact that drone honey bees from different colonies congregate in mating leks where they can be conveniently sampled. Genetic analysis of the captured drones can determine the number of colonies contributing to the sampled population. Here, through the use of sampling distributions derived from an agent-based model, we provide an improved procedure for estimating the density of colonies from the number of unique colonies identified from the sampled drones. We present simulations for different spatial environments and densities, and show that the number of unique colonies observed in a sample of drones collected at a drone trap covaries with the density of colonies in range of the sampled drone congregation area in a log-linear manner. As a consequence of this relationship, we find that colony densities from past surveys are likely to be lower than previously reported.

Introduction

Honey bees (Apis spp.), are the pre-eminent pollinators of crops and natural ecosystems worldwide (Klein et al., 2007). One species, Apis mellifera, originally from Africa and Europe has been introduced to temperate and tropical regions throughout the world for both pollination and honey and wax production. In many areas, A. mellifera has established large feral populations (Crane, 1999, Whitfield et al., 2006), but in other areas, notably the United States and Europe, domestic, wild and feral honey bee populations are in decline (Lebuhn et al., 2013, Potts et al., 2009).

It is increasingly important to understand the density and distribution of wild and feral honey bee colonies in both agricultural and natural landscapes. First, wild and feral colonies provide incidental pollination to crops as a classic ‘ecosystem service’ (Cunningham et al., 2002). The value of insect (and particularly honey bee) pollination has been estimated as €153 billion (Gallai et al., 2009) worldwide. The decline of honey bee populations in some areas is a worrying trend that will likely increase the cost of food production (Gallai et al., 2009, Southwick and Southwick, 1992). Some Asian species are threatened by local extinction wrought by a combination of deforestation, over-hunting, pesticide use and competition with introduced species (Oldroyd and Nanork, 2009, Oldroyd and Wongsiri, 2006). Second, despite the immense value of honey bee pollination to agriculture and many natural ecosystems, honey bees can be invasive. In some circumstances they compete with native wildlife for food resources and nesting sites, and disrupt the pollination of native flora (Dupont et al., 2004, Goulson, 2003, Goulson et al., 2008, Kato et al., 1999). Understanding the density and demography of feral honey bee populations is essential to risk assessment and management in the conservation context. Third, wild and feral colonies can serve as a vector and reservoir for the spread of honey bee pathogens that affect the commercial industry (Morse et al., 1990, Taylor et al., 2007). Finally, wild and feral populations may play an important role as reservoirs of genetic diversity for the commercial industry (Chapman et al., 2008, Oldroyd, 2007).

To understand the ecological significance of feral honey bees it is necessary to determine the density of feral colonies in different environments over time. A number of different approaches have been developed for measuring wild and feral colony densities. The simplest approach is a systematic search for nests (Baum et al., 2005, Baum et al., 2008, Galton, 1971, Goodman and Hepworth, 2004, Manning et al., 2007, McNally and Schneider, 1996, Morse et al., 1990, Oldroyd et al., 1994, Oldroyd et al., 1997, Paton, 1996, Ratnieks et al., 1991, Schneider and Blyther, 1988, Wenner, 1989). This method is labour intensive, severely limiting the size and number of areas that can be surveyed, and the survey method itself is prone to underestimating the number of actual colonies due to the difficulties in locating all nest sites in the survey area (Baum et al., 2005).

More recently, alternative techniques have emerged that are based on the use of microsatellite loci to identify the number of colonies that are contributing gametes to the sampled population. Two different approaches are available. In the first, unmated queens are placed at the survey location and allowed to mate with drones from the colonies that are in flight range (Jaffé et al., 2010, Moritz et al., 2007, Moritz et al., 2008). The worker offspring are analysed genetically to identify the number of unique colonies with which the queens mated, and from this the density of colonies in the area is inferred. In the second approach, drones from wild colonies are captured in a Williams (1987) drone trap baited with sex pheromone and genetically analysed to determine the number of unique source colonies represented (Baudry et al., 1998, Hinson et al., 2013, Jaffé et al., 2009, Jaffé et al., 2010, Moritz et al., 2007, Moritz et al., 2008, Moritz et al., 2013). The focus of this paper is how best to infer the density of colonies in an area from the counts of unique source colonies obtained in a sample of drones captured using a Williams trap.

The Williams trap approach relies on the observation that mating between queens and drones occurs at or near drone congregation areas (DCAs). DCAs have a limited spatial extent and tend to persist in the same location from year to year (Loper et al., 1992). Drones from each colony fly to DCAs, preferring nearer DCAs over those further away (Baudry et al., 1998, Jaffé et al., 2009, Koeniger et al., 2005). At the DCAs, the drones fly in circles waiting for the arrival of virgin queens (Loper et al., 1987, Loper et al., 1992). Queens advertise their presence at the DCA by a sex pheromone secreted from their mandibular gland, of which (2E)-9-oxodecenoic acid (9-ODA) is the major component. 9-ODA is effective at attracting drones within an area roughly 100 m in diameter (Brockmann et al., 2006, Winston, 1987). To construct a Williams trap, blackened cigarette filters impregnated with 9-ODA are placed within a net with an opening at the bottom, and this is then attached to a weather balloon and raised to an altitude of between 8 and 40 m (Baudry et al., 1998, Koeniger et al., 1989). Flying drones are difficult to trap outside DCAs (Taylor, 1984). Thus to successfully sample drones in large numbers it is necessary to locate the trap within a DCA. This can be achieved through either previous observation of the DCA locations, or a search of the environment using the trap as a detector (Brockmann et al., 2006, Jaffé et al., 2009). Under ideal conditions, several hundred drones can be caught in a few minutes.

There are usually several thousand drones present at a DCA. Typically, researchers genotype a sample of 96 drones from each site (96 being the number that conveniently fits in a standard plate used for PCRs). The drones are then genotyped at a set of tightly linked microsatellite loci (Moritz et al., 2008). As each diploid queen produces haploid males, and recombination events between the tightly linked loci are rare, each male offspring will be one of two possible haploid genotypes (haplotypes). Thus the number of unique colonies represented in the sample is the count of unique haplotypes divided by two. An alternative approach involves using a maximum likelihood algorithm to assign drone genotypes to the minimum number of colonies that could explain the observed drone genotypes (Wang, 2004). Both these approaches may underestimate the number of colonies contributing drones to the sampled DCA due to non-sampling error. Thus, regardless of the method used to determine the number of colonies represented in the sample there is uncertainty in the estimate of colony density. In many surveys, once the estimate of the count of unique colonies has been made, the density of the colonies has been estimated based on assumptions of the area from which the drones occupying a sampled DCA are drawn (Table 1).

In previous studies (Hinson et al., 2013, Jaffé et al., 2010, Moritz et al., 2007, Moritz et al., 2008), the density has been determined by dividing the count of unique colonies by 2.5, derived from converting the average distance of a drone (900 m; Taylor and Rowell, 1988) to an area of 2.5 km2. The assumptions underpinning this calculation are that (i) all colonies represented at the DCA are detected in the genotyped drones, (ii) the drone flight range measured in a single experiment (Taylor and Rowell, 1988) is representative of drone flight ranges in different environments and insensitive to variations in underlying densities of colonies or DCAs.

In this study, we examine the validity of these assumptions and make suggestions that will enhance the value of the drone trap survey method. Our approach extends the work of Arundel et al. (2012) who used a simulation of the spatial environment of honey bee colonies to aid the interpretation of survey results. We showed that for surveys using unmated queens the relationship between the number of colonies observed in the sample and the underlying colony density is log-linear. Here we apply the “virtual ecologist” approach (Zurell et al., 2010) and simulate the drone trap capture process in environments characterised by different spatial distributions and densities of colonies and DCAs. We generate distributions which can be used to determine the relationship between the count of unique drone haplotypes found in a sample and the underlying parameters used to generate the simulation environment. In particular, we explore the hypotheses that the count of unique haplotypes in a sample of drones captured using a Williams trap is primarily related to the underlying density of colonies, and that this relationship is non-linear. To assess the implications of our findings, we use this new method to estimate colony densities from field experiments. Finally, we make recommendations for the design of field experiments, and provide sampling distributions that should be useful for the interpretation of data.

Section snippets

Materials and methods

The model description below follows the ODD (Overview, Design concepts, Details) protocol for describing individual- and agent-based models (Grimm et al., 2006, Grimm et al., 2010). The submodel descriptions are included in the process overview described in Section 2.3.

Results

We simulated five densities (0.1/km2, 0.5/km2, 1.0/km2, 5.0/km2 and 10.0/km2) of colonies and DCAs with the four spatial distributions (random, aggregated, hyper-aggregated and overdispersed). For each combination of parameters, we executed 100 simulations and we present the results of these simulations as distributions.

Discussion

The count of unique haplotypes present in a sample of drones captured in a trap is strongly related to the underlying density of colonies in the environment. We find a log-linear relationship between the count of unique colonies present in a trap sample and the true density of colonies in the environment. Crucially, the most likely density that would result in the count of unique colonies observed from the sample must be estimated from a distribution specific to the sample size. Thus, the use

Conclusion

The number of unique colonies observed in a sample of drones collected at a drone trap varies in a log-linear relationship with the density of colonies in range of the sampled DCA. This finding validates the usefulness of the drone trap technique as an important survey tool, but highlights the challenges of interpreting field survey results. Typically, unless the sample size is large and the underlying density small, the proportion of colonies captured by the sample will be low. Our sampling

Acknowledgements

We acknowledge the financial support of the Rural Industries Research and Development Corporation, and a seed grant from the Sustainable Futures Program of the University of Sydney.

References (62)

  • N.C. Chapman et al.

    Population genetics of commercial and feral honey bees in Western Australia

    Journal of Economic Entomology

    (2008)
  • E. Crane

    The World History of Beekeeping and Honey Hunting

    (1999)
  • S.A. Cunningham et al.

    The future of pollinators for Australian agriculture

    Australian Journal of Agricultural Research

    (2002)
  • G. Deodikar et al.

    Nesting behaviour of Indian honeybees, III. Nesting behaviour of Apis dorsata Fab

    Indian Bee Journal

    (1977)
  • D. Galton

    Survey of a Thousand Years of Beekeeping in Russia

    (1971)
  • R. Goodman et al.

    Densities of feral honey bee Apis mellifera colonies in Victoria

    Victorian Naturalist

    (2004)
  • D. Goulson

    Effects of introduced bees on native ecosystems

    Annual Review of Ecology, Evolution, and Systematics

    (2003)
  • D. Goulson et al.

    Decline and conservation of bumble bees

    Annual Review of Entomology

    (2008)
  • Hinson, E.M., Duncan, M., Lim, J., Oldroyd, B.P., 2013. Density of feral honey bee (Apis mellifera) colonies in South...
  • R. Jaffé et al.

    Temporal variation in the genetic structure of a drone congregation area: an insight into the population dynamics of wild African honeybees (Apis mellifera scutellata)

    Molecular Ecology

    (2009)
  • R. Jaffé et al.

    Estimating the density of honeybee colonies across their natural range to fill the gap in pollinator decline censuses

    Conservation Biology

    (2010)
  • M. Kato et al.

    Impact of introduced honeybees, Apis mellifera, upon native bee communities in the Bonin (Ogasawara) Islands

    Researches on Population Ecology

    (1999)
  • A.M. Klein et al.

    Importance of pollinators in changing landscapes for world crops

    Proceedings of the Royal Society B: Biological Sciences

    (2007)
  • G. Koeniger et al.

    Assortative mating in a mixed population of European honeybees, Apis mellifera ligustica and Apis mellifera carnica

    Insectes Sociaux

    (1989)
  • N. Koeniger et al.

    The nearer the better? Drones (Apis mellifera) prefer nearer drone congregation areas

    Insectes Sociaux

    (2005)
  • G. Lebuhn et al.

    Detecting insect pollinator declines on regional and global scales

    Conservation Biology

    (2013)
  • G.M. Loper et al.

    Detection and monitoring of honeybee drone congregation areas by radar

    Apidologie

    (1987)
  • G.M. Loper et al.

    Honey-bee drone flyways and congregation areas: radar observations

    Journal of the Kansas Entomological Society

    (1992)
  • R. Manning et al.

    Survey of feral honey bee (Apis mellifera) colonies for Nosema apis in Western Australia

    Australian Journal of Experimental Agriculture

    (2007)
  • L.C. McNally et al.

    Spatial distribution and nesting biology of colonies of the African honey bee Apis mellifera scutellata (Hymenoptera: Apidae) in Botswana, Africa

    Environmental Entomology

    (1996)
  • R.F.A. Moritz et al.

    The size of wild honeybee populations (Apis mellifera) and its implications for the conservation of honeybees

    Journal of Insect Conservation

    (2007)
  • View full text