ArticlePDF Available

Conservation ecology of bees: Populations, species and communities


Abstract and Figures

Recent concerns regarding the decline of plant and pollinator species, and the impact on ecosystem functioning, has focused attention on the local and global threats to bee diversity. As evidence for bee declines is now accumulating from over broad taxonomic and geographic scales, we review the role of ecology in bee conservation at the levels of species, populations and communities. Bee populations and communities are typified by considerable spatiotemporal variation; whereby autecological traits, population size and growth rate, and plant-pollinator network architecture all play a role in their vulnerability to extinction. As contemporary insect conservation management is broadly based on species- and habitat-targeted approaches, ecological data will be central to integrating management strategies into a broader, landscape scale of dynamic, interconnected habitats capable of delivering bee conservation in the context of global environmental change.
Content may be subject to copyright.
Apidologie 40 (2009) 211–236 Available online at:
INRA/DIB-AGIB/EDP Sciences, 2009
DOI: 10.1051/apido/2009015 Review article
Conservation ecology of bees: populations, species and
Tomás E. Murray1, Michael Kuhlmann2,SimonG.Potts3
1Crops Research Centre, Teagasc, Oak Park, Carlow, Co. Carlow, Ireland
2Department of Entomology, Natural History Museum, Cromwell Road, London SW7 5BD, UK
3Centre for Agri-Environmental Research, School of Agriculture, Policy and Development, University of
Reading, Earley Gate, Reading RG6 6AR, UK
Received 30 October 2008 Accepted 13 February 2009
Abstract Recent concerns regarding the decline of plant and pollinator species, and the impact on ecosys-
tem functioning, has focused attention on the local and global threats to bee diversity. As evidence for bee
declines is now accumulating from over broad taxonomic and geographic scales, we review the role of
ecology in bee conservation at the levels of species, populations and communities. Bee populations and
communities are typified by considerable spatiotemporal variation; whereby autecological traits, popula-
tion size and growth rate, and plant-pollinator network architecture all play a role in their vulnerability
to extinction. As contemporary insect conservation management is broadly based on species- and habitat-
targeted approaches, ecological data will be central to integrating management strategies into a broader,
landscape scale of dynamic, interconnected habitats capable of delivering bee conservation in the context
of global environmental change.
conservation /biodiversity /population /community /plant-pollinator
Concerns about a potential pending global
‘pollinator crisis’ have been raised for over a
decade, based on the evidence available at the
time (Buchmann and Nabhan, 1996; Kearns
et al., 1998; Kremen and Ricketts, 2000;Diaz
et al., 2005), though such a crisis was ques-
tioned by Ghazoul (2005); but see Stean-
Dewenter et al., (2005). Prior to 2006, doc-
umentation of bee declines were limited to
a few case studies and were generally taxo-
nomically restricted and considered only spe-
cific locations (e.g. bumblebees, Peters, 1972,
Williams P.H., 1986; honeybees, Williams I.H.
et al., 1991; solitary bees, Westrich, 1989), and
were not often published in mainstream scien-
tific literature (e.g. Rasmont, 1988;Falk,1991;
Corresponding author: T.E. Murray,
* Manuscript editor: Robert Paxton
Banaszak, 1995).Theneedformoreacom-
prehensive understanding of the global sta-
tus and trends of pollinators triggered a num-
ber of projects and studies aiming to provide
large-scale and taxonomically diverse assess-
ments. In Europe, the ALARM project1pro-
vided the first national-scale assessments of
shifts in whole bee communities in the UK and
Holland (Biesmeijer et al., 2006) and demon-
strated severe declines in bee diversity, abun-
dance and ranges and also concurrent declines
in bee-pollinated flowering plants. In North
America, a synthesis of existing information
was undertaken (National Research Council,
2006) and concluded that there were long-term
declines in honeybees and some wild bees
(most notably bumblebees). Recognition of
widespread loss of pollinators and pollinator
1Assessing Large-scale Risks to biodiversity with
tested Methods,
Article published by EDP Sciences
212 T.E. Murray et al.
services by the Convention on Biological Di-
versity (through the Agricultural Biodiver-
sity programme and International Pollinator
Initiative2) resulted in FAO coordinating the
‘Rapid Assessment of Pollinators’ status’ re-
port (2008) which aims to compile global ev-
idence of the extent of pollinator shifts and
loss of pollination services. Together these
sources, and many other studies (e.g. Klein
et al., 2002;Larsenetal.,2005; Williams N.M.
and Kremen, 2007), provide an overwhelming
evidence-base for marked declines of many
taxa of bees across the world.
The loss of bee species from communities
may not be a random process e.g. large-bodied
bees with higher pollination eciencies can be
more extinction prone (Larsen et al., 2005),
and so has important implications for pol-
lination services to crops and wild flowers
(Memmott et al., 2004). Drivers of local bee
extinctions can act dierentially on particular
bee traits such as size, foraging and nesting
behaviour. For instance, in Californian agro-
ecosystems, Larsen et al. (2005) reported that
large-bodied bees were more extinction-prone,
and Memmott et al. (2004) demonstrated that
larger bees (such as bumblebees) tend to be
more highly linked in plant-pollinator webs.
Similarly wildfires were found to disassem-
ble Swiss and Israeli bee communities in a
non-random fashion, with extinction probabil-
ities related to nesting guild, phenology, for-
aging behaviour and size (Potts et al., 2005;
Moretti et al., 2009). However, the sensitiv-
ity of particular bee species to the wide range
of natural and anthropogenic drivers remains
largely unknown. It is therefore essential to
understand the basic ecology of bees in order
to predict how they respond to environmental
change and how these changes can be miti-
gated against.
A general conceptual framework has been
proposed to facilitate our understanding of
how bees, and other “mobile-agent-based
ecosystem service providers”, are aected by
land-use change (Kremen et al., 2007). The
model describes interactions between bees, the
temporal and spatial availability of resources,
biotic and abiotic factors aecting bee fitness,
the delivery of pollination services and how
these are all potentially aected by land-use
change, policy and market forces (Fig. 1). By
adopting this model, i.e. incorporating individ-
ual behaviour, population biology and commu-
nity dynamics, this review aims to highlight
specific areas of bee ecology that will have the
greatest impact on the development and appli-
cation of conservation strategies.
It is now widely accepted that sensible
management begins with a sound, fundamen-
tal understanding of a species’ ecological
interactions and natural history (Simberlo,
1988; Brussard, 1991; Schultz and Hammond,
2003). Nonetheless, detailed autecological
study, like systematics, seems to lack the ca-
chet of academic respectability that ensures
practitioners of regular funding or even em-
ployment (Kim and Byrne, 2006). Unsurpris-
ingly, we are woefully short on autecological
information for rare and little-known species,
but it is also true for many of the more charis-
matic bees, such as bumblebees (e.g. Kosior
et al. 2007; Williams P.H. and Osborne, 2009).
These autecological data are of critical im-
portance for calculating a suite of parameters
which can be used to predict extinction risk
as elements in population viability analysis
(PVA) and identify traits correlated with vul-
nerability to extinction in phylogenetic com-
parative approaches (Schultz and Hammond,
2003; Fisher and Owens, 2004). Even basic
ecological knowledge gleaned from the liter-
ature can be utilized to predict the risk of
extinction and prioritise species of conserva-
tion concern. In the Lepidoptera, for example,
traits associated with vulnerability to extinc-
tion, although frequently highly taxon depen-
dent, typically include: dispersal ability, larval
specificity, adult habitat breadth and length of
flight period (Kotiaho et al., 2005).
It is important to note that autecological
studies do not need to be mundane exercises in
data gathering; rather, the autecological traits
discussed below should be used to generate
Conservation ecology of bees 213
Figure 1. Conceptual framework for studying impacts of land-use change on bee conservation; incorpo-
rating the biology of the organisms involved, in addition to ecosystem service provision and its positive
feedback into economics and conservation policy. Numbers in boxes refer to the relevant section in the text.
Modified from Kremen et al. (2007).
testable hypotheses to drive several research
projects. In the absence of detailed autecologi-
cal information and long-term population data,
we will continue to be forced to make sev-
eral assumptions when both interpreting pre-
cipitous declines in numbers and when recom-
mending appropriate management actions.
2.1. Socioecology
The primary eect of sociality on bee con-
servation biology is its impact on a species’
population biology and genetics (reviewed in:
Pamilo et al., 1997; Chapman and Bourke,
2001; Zayed, 2009). What best defines a pop-
ulation of bees is fundamentally tied to its
level of social organisation and the resul-
tant reproductive potential of each individual
(Crozier and Pamilo, 1996). Thus, a popu-
lation of a solitary species, where each in-
dividual may reproduce, would be a popu-
lation of individuals; whereas a population
of eusocial bees would be a population of
colonies, as within a colony one or few re-
lated queens reproduce and most individuals
are non-reproductive workers. Consequently,
the ratio of the eective to the census popula-
tion size can rapidly decrease with increasing
social organisation (Pamilo and Crozier, 1997;
Chapman and Bourke, 2001). Similarly, the
ease with which we can establish the popula-
tion size, and therefore the conservation status,
of a taxon typically decreases with increasing
social complexity.
Although the majority of bees are soli-
tary, all levels of social organisation occur
within the Apoidea (Michener, 1974; Crozier
and Pamilo, 1996). Some families of bees, the
sweat bees (Halictidae) in particular, are note-
worthy among insects due to their substan-
tial interspecific (Danforth, 2002; Brady et al.,
2006) and intraspecific (Soucy, 2002)vari-
ability in social behaviour. The behavioural
plasticity observed within halictid social be-
haviour is of particular relevance as an exam-
ple of how some species can exhibit more than
one type of sociality across geographic and
214 T.E. Murray et al.
climatic gradients e.g. both Lasiglossum cal-
ceatum and Halictus rubicundus are social in
lower altitudes and solitary at higher altitudes
(Sakagami and Munakata, 1972; Eickwort
et al., 1996). In the absence of detailed auteco-
logical information for many rare species, util-
ising methods such as phylogenetic compara-
tive approaches (e.g. Fisher and Owens, 2004)
in well-described groups such as the sweat
bees could highlight previously unreported
socially polymorphic taxa in other families
and stimulate further ecological research. For
those species where sociality has been con-
firmed, establishing both the level and plastic-
ity of social organisation is an important first
step in any conservation programme. As men-
tioned, the eective population size, and there-
fore the vulnerability, of a species depends
not only on the number of nests, but also on
the mode of colony foundation, number of re-
productive individuals and reproductive skew
among these individuals (see Sect. 3.2 Mating
2.2. Parasitism
An adjunct to the levels of social organisa-
tion found within bees are their varying forms
of parasitism, namely: usurpation and rob-
bing, social parasitism and cleptoparasitism
(Michener, 2007). In all cases, the parasite
benefits from the resources gathered and/or
constructed by the host, with the host presum-
ably incurring a fitness cost in the process. It
is estimated that 15–20% of all bee species are
parasites (Wcislo and Cane, 1996), with the
percentage of parasitic species tending to in-
crease with latitude (Wcislo, 1987; Petanidou
et al., 1995). Despite the relative abundance
and mostly temperate distribution of parasitic
bees, relatively little is known about their bi-
ology as their population sizes rarely become
large enough for detailed study (Scott et al.,
2000; Bogusch et al., 2006).
The term ‘coextinction’ was coined after
Stork and Lyal (1993) drew attention to the
likelihood that many parasites may go extinct
when their hosts go extinct. In a recent re-
view, Koh et al. (2004) found that approxi-
mately 5000 insect species are likely to be en-
dangered as a direct consequence of the en-
dangerment of their hosts and that at least 100
species of beetles, lice and butterflies have
probably gone extinct in the last 200 years
due to the extinction of their hosts. Although
it is dicult to generalise about the pattern
and degree of specialisation across parasitic
bee taxa (e.g. Bogusch et al., 2006), most stud-
ies have revealed a strict concordancebetween
physiology and behaviour of parasite and host
(Wcislo, 1987; Scott et al., 2000). Therefore,
determining the degree of specialisation be-
comes increasingly important as it is proba-
bly proportional to the probability of host and
parasite becoming coextinct. Furthermore, due
to their generally lower population sizes, de-
clines in parasite populations may frequently
precede those of their host populations, as ob-
served by Rasmont et al. (2005) in Belgium
where declines of megachilid and anthophorid
cleptoparasites were found to be dispropor-
tionably higher than those of their hosts.
2.3. Floral resources and specialisation
Bees are herbivores that feed their lar-
vae with a mixture of pollen and nectar or,
rarely, plant oils (Michener, 2007). Robertson
(1925) was one of the first to recognize that
bees do not collect pollen on flowers ran-
domly but that some species demonstrate flo-
ral specificity when harvesting pollen only
on a limited number of plant taxa. He in-
troduced the terms monolecty, oligolecty and
polylecty to distinguish between pollen spe-
cialists and generalists, respectively. This clas-
sification inadequately reflects the complex
relationships between bees and their pollen
hosts and was updated by Cane and Sipes
(2006), and recently modified by Müller and
Kuhlmann (2008;Tab.I).
Even generalist bees show a restricted
range of pollen sources (Westrich, 1989;
Müller, 1996, Müller and Kuhlmann, 2008);
pollen might not be an easy to use protein
source. Praz et al. (2008) demonstrated that
several specialized bee species fail to de-
velop on non-host pollen, indicating that the
pollen of some plant taxa, e.g. Asteraceae,
Conservation ecology of bees 215
Table I. Classification of floral specificity of pollen collection by Robertson (1925), Cane and Sipes (2006),
and an updated classification proposed by Müller and Kuhlmann (2008)(dierences in terminology and
definitions in italics).
Robertson (1925) Cane and Sipes (2006) Müller and Kuhlmann (2008)
Monolecty Monolecty
Pollen collection on only one plant
species even in the presence of one
or more sympatric species of the same
Pollen collection on only one plant
species even in the presence of one
or more sympatric species of the same
genus (see also Narrow oligolecty)
Oligolecty Narrow oligolecty
Pollen collection from two to several
species belonging to one plant genus
Narrow oligolecty
Pollen collection from two to several
species belonging to one plant genus
(pollen collection on only one plant
species in the absence of co-flowering
congenerics is referred to as a special
case of narrow oligolecty)
Pollen collection from two to several
species belonging to one to four genera
belonging to one family
Broad oligolecty
Pollen collection from two to several
genera belonging to one plant tribe,
subfamily or family
Eclectic oligolecty
Pollen collection from two to four plant
genera belonging to two or three plant
Eclectic oligolecty
Pollen collection from two to four plant
genera belonging to two or three plant
Polylecty s.l.
Pollen collection from more than four
plant genera belonging to two or three
plant families or big tribes
Pollen collection from more than four
plant genera belonging to two or three
plant families
Polylecty with strong preference
Pollen collection from several plant
families, but one plant clade (family,
subfamily, tribe, genus or species)pre-
Pollen collection from various genera
belonging to four to <25% of available
plant families
Broad polylecty
Pollen collection from various genera
belonging to >25% of available plant
Polylecty s.s.
Pollen collection from various genera
belonging to at least four plant families
possesses unfavourable or protective proper-
ties that render its digestion dicult. This sug-
gests that the successful utilization of pollen
by bees in general might require special phys-
iological adaptations to cope with toxic sec-
ondary chemicals of their hosts. Selection
should favour the evolution of such adap-
tations and eventual host specificity in bees
(Praz et al., 2008) if they fit the physiological-
eciency hypothesis (Cornell and Hawkins,
2003), which predicts a physiological trade-
oin the ability to eciently utilize alterna-
tive hosts as a consequence of these adapta-
tions (Singer, 2008).
216 T.E. Murray et al.
The quantitative pollen requirements of
bees are little known. In a study of 41 bee
species Müller et al. (2006) revealed that 85%
of them require the whole pollen content of
more than 30 flowers to provision a brood cell
and some species even needed the pollen of
more than a thousand flowers to rear a single
larva. In combination with the often restricted
foraging range of small bees (see Sect. 3.1
Dispersal), this implies that tens of thousands
of flowers of a certain plant must be avail-
able within range to sustain a viable popu-
lation of an oligolectic bee species. Hence,
the loss of plant diversity and flower quantity
due to habitat destruction and fragmentation
of the landscape is assumed to be responsible
for the decline of many bee species (Müller
et al., 2006). Social bee species are typically
polylectic (Michener, 2007) and are generally
be believed to be less prone to local extinction.
However, Kleijn and Raemakers (2008)have
recently shown that bumblebee species whose
populations are in decline use a narrower spec-
trum of host plants than bumblebees with sta-
ble populations.
Specialized bees generally do not switch to
other host plants, even if their preferred plants
are not in flower (Strickler, 1979; Williams
N.M., 2003). Hence, selection for synchrony
of bee emergence with host plant flowering
that is positively aecting individual fitness
(Powell and Mackie, 1966) can be expected,
especially in arid and semi-arid environments
with highly variable precipitation. Anecdo-
tal observations and experimental data sug-
gest that in arid environments rainfall triggers
the emergence of oligolectic bees in particular
(reviewed in Danforth, 1999). Evidence from
western South Africa challenges this view
(Mayer and Kuhlmann, 2004) and contradicts
the assumption that bees react to the same
environmental cues for emergence as their
host plants for flowering e.g. rainfall (Linsley,
1958; Wcislo and Cane, 1996; Tauber et al.,
1998). However, generally it is hypothesized
that host plant synchrony might be a mecha-
nism for an elevated rate of speciation in desert
bees (Minckley et al., 2000; Danforth et al.,
2003) explaining the higher bee species diver-
sity in semi-arid and arid environments. Thus,
oligolectic bees that are strictly dependent on
their host plants are most species rich in desert
and Mediterranean environments, and less di-
verse in temperate biota (Moldenke, 1979).
2.4. Nesting resources
Nesting resources for bees include the sub-
strates within, or on which, they nest and also
the materials required for nest construction.
Bees are extremely diverse in their nesting
ecology and comprise a number of distinct
guilds (O’Toole and Raw, 1991): miners, car-
penters, masons, social nesters and cuckoos.
Mining bees include all Andrenidae, Melitti-
dae, Oxaeidae, Fideliinae and most Halictidae,
Colletidae and Anthophorini species. Miners
excavate tunnels in the ground or soft rocky
substrates and line their tunnels with glandu-
lar secretions. Carpenter bees also excavate
nests, but use wood as a substrate, and in-
clude species in the genera Xylocopa,Cer-
atina (Apidae) and Lithurgus (Megachilidae).
In contrast, mason bees (most Megachilidae)
utilise pre-existing holes which can be in the
form of hollow plant stems, abandoned in-
sect nest burrows in the ground or woody
substrates, small cavities or cracks in rocks
and even snail shells. Masons then line the
inside of the pre-existing hole with materi-
als such as leaves or soil. Within the ma-
son guild, the leaf-cutter bees use only freshly
gathered leaf or petal material to line their
nests and are members of the Megachile and
Creightonella genera of the Megachilidae. So-
cial nesters tend to use relatively large pre-
existing cavities to establish social colonies
and include three taxa within the Apidae: hon-
eybees (Apis), bumblebees (Bombus) and stin-
gless bees (Meliponini). One guild of bees, the
cuckoo bees or cleptoparasites, are found in
several families and do not construct their own
nests but instead parasitize the nests of other
bees by laying their eggs on larval provisions
provided by the host.
Potts et al. (2005) demonstrated that the
availability of both nesting substrates and con-
struction materials were primary determinants
of overall bee community composition. Fur-
thermore, the dominant ecological perturba-
tion, fire, resulted in a marked turnover of
Conservation ecology of bees 217
nesting resources, resulting in large shifts
in the relative proportions of nesting guilds.
Other resources shown to aect bee nest-
ing success include: the abundance, size and
species of trees in tropical forests for stingless
bees (Eltz et al., 2002; Samejima et al., 2004);
cavity shape and size for honeybees (Schmidt
and Thoenes, 1992; Oldroyd and Nanork,
2009), and the diameter of pre-existing holes
for colletid bees (Scott, 1994); soil hardness,
slope and aspect of the ground for halictid bees
(Potts and Willmer, 1997); and soil texture for
solitary bees (Cane, 1991).
The diversity of nesting strategies and the
specialisation of guilds means that the avail-
ability of the correct quantity and quality of
resources, both in space and time, are key de-
terminants for which species a landscape can
support (Tscharntke et al., 2005). Any envi-
ronmental disturbance (e.g. habitat loss, frag-
mentation, agricultural intensification, or fire)
will alter the distribution of nesting resources.
As bees are central place foragers and have
species-specific flight distances, the location
of the nest determines what floral resources are
potentially available. The nesting traits of bee
species will therefore determine their sensi-
tivity to environmental change (Moretti et al.,
2009). In order to manage the landscape for
bee conservation it is therefore essential to un-
derstand how land use change aects nesting
resources and how this interacts with the avail-
ability of other resources such as nectar and
Species are driven to extinction by both
human-mediated deterministic factors and
stochastic factors (Frankham et al., 2002). In
general, a species’ population size is initially
reduced by deterministic factors such as habi-
tat loss, fragmentation, overexploitation, intro-
duced species, pollution and climate change.
Populations may then further decline to a point
of ‘no return’ where demographic, environ-
mental and genetic stochasticity and catastro-
phes eventually drive them to extinction. Con-
sequently, as both absolute population size
and spatio-temporal variability in population
size are universally the most important pre-
dictors determining extinction risk, their accu-
rate estimation is of paramount importance to
conservation biology (Frankham et al., 2002;
O’Grady et al., 2004).
Insect populations commonly experience
large annual fluctuations in population size,
arising from natural variation in population
growth and measurement errors in population
estimates (e.g. Schultz and Hammond, 2003).
Unfortunately, bee populations and commu-
nities are no exception, displaying consider-
able spatio-temporal variation in abundance
and composition (Williams N.M. et al., 2001;
Roubik, 2001; Eltz, 2004; Tylianakis et al.,
2005; Petanidou et al., 2008). The highly vari-
able nature of these ecological data compli-
cates the determination of the conservation
status of a species, as devastating declines may
occur, but remain undetected until a sucient
period of time has elapsed after monitoringthe
population. Unfortunately, this typically leaves
managers and policy makers in a reactive pos-
ture as the decline might be demonstrated, but
only after it has already seriously weakened
the population.
Although in many cases an experimental
approach may be more appropriate, in prac-
tice, especially for rare species, the logistics,
lack of replication and spatial scales involved
have driven the proliferation of modelling ap-
proaches to test hypotheses about the causes
of decline and the response of populations to
management practices or future changes in the
environment (e.g. risk-based viable population
monitoring, Staples et al., 2005). In a review
of the ecological tools available to conserva-
tionists to intervene in the extinction process,
or even predict precipitous declines, Norris
(2004) outlines a ‘toolbox’ for the manage-
ment of threatened species: statistical mod-
els of habitat use, demographic models and
behaviour-based models. In particular, popu-
lation viability analysis (PVA) is one of the
most widely applied demographic models in
conservation (Lande et al., 2003), and as re-
cent modifications to PVA now allow for less
intensive data (e.g. shorter time series and
population surveys) more often available for
insect populations, it has been successfully
218 T.E. Murray et al.
applied to endangered insects, particularly but-
terflies (e.g. Schultz and Hammond, 2003;
Schtickzelle and Baguette, 2004).
There are innumerable parameters that may
be discussed in relation to studying the pop-
ulation biology of a single species. Below we
highlight three keys areas of study that can sig-
nificantly influence the estimation of popula-
tion size, growth rate and persistence of bee
3.1. Dispersal
Female bees forage from a single loca-
tion, their nest, and as such are central place
foragers (Schoener, 1979). Theoretical stud-
ies frequently emphasise the role of disper-
sal ability and emigration between patches in
predictions of minimum viable population size
(Hanski and Pöyry, 2006), thus information
of their flight range is vital for bee conserva-
tion to make sure their habitat requirementsare
met within their range of activity (Westrich,
1996;Cresswelletal.,2000). Known forag-
ing distances of bees range from 0.1 km to
a maximum of 45.5 km in Eufriesea surina-
mensis, with values for most species below
1 km (summarized in Greenleaf et al., 2007)
and a strong correlation exists between the in-
tertegular span (ITS) and foraging range that
can be used as a valuable tool to predict for-
aging ranges based on a simple measurement
of ITS. In addition, knowledge about the for-
aging range is also important for estimation of
area requirements of bees for providing opti-
mal pollination service in agriculture (Kremen
et al., 2004). For conservation measures it is
vital to take into account that species with a
small foraging range require more diverse re-
sources per unit area than species with simi-
lar needs but greater range (Cresswell et al.,
2000). Thus, for many species local habitat
structure appears to be more important than
large-scale landscape composition (Gathmann
and Tscharntke, 2002).
However, considering the abundance of
data on the relationship between body size
and foraging distance (e.g. Greenleaf et al.,
2007), little data exist on how species’ forag-
ing ranges relate to male- or female-mediated
gene flow between populations. In the Eu-
glossini, long-distance gene flow across the
Andes was related to body size, with larger
body sizes potentially conferring a disper-
sal advantage in terms of thermal tolerance
and energy reserves for long-distance disper-
sal (Dick et al., 2004). In contrast, the level
of population genetic subdivision (e.g. Fst)
within the honeybee Apis mellifera was almost
twice as high in South Africa (0.105) com-
pared to Germany (0.064), despite the fact that
the sampled subpopulations in South Africa
were geographically closer together (Moritz
et al., 2007). Generalisations regarding gene
flow and body size may be spurious, as a
species’ dispersal ability may equally be de-
termined by a range of other socio-ecological
factors, such as nest-site philopatry, lecty and
tolerance to inbreeding depression (Packer and
Owen, 2001; Packer et al., 2005). Resolving
the determinants of gene flow between popu-
lations is of critical importance to all species-
based conservationprogrammes and should be
a fruitful avenue for future research in bee con-
Regardless of the lack of data regarding
gene flow, population viability and minimum
habitat requirements of bees (e.g. Larsen et al.,
2005), small and isolated bee populations can
persist for relatively long periods of time
provided that the habitat remains unchanged
(Kratochwil and Klatt, 1989). Basically bees
are physiologically highly capable of flying
long distances increasing the likelihood that
a habitat is (re)colonized. Solitary bees have
been found on lightships in the North Sea and
Baltic Sea up to 10 km othe coast, bum-
blebees even up to 30 km distance from the
nearest land, but most of the specimens were
close to the end of their reproductive phase
and thus not of significance for colonization
(Haeseler, 1974). Capacity for (re)colonization
is correlated to distance and habitat frag-
mentation (Stean-Dewenter and Tscharntke,
1999;Stean-Dewenter and Westphal, 2008)
but even areas neighbouring ancient suitable
bee habitats are slowly recolonized (Forup
et al., 2008) and isolated habitats like islands
(Haeseler, 1976,1978) are hardly reached
by most bee species, limiting the potential
success of ecological restoration measures
Conservation ecology of bees 219
(Kuhlmann, 2000; Franzén et al., 2009). De-
spite the fact that the colonization ability of
bees seems to be generally low, massive and
short-term range extensions are known from
a few species like Andrena fulva and Bom-
bus hypnorum in NW Germany about a cen-
tury ago (Wagner, 1938) and, more recently,
Andrena cineraria (Haeseler, 1973), Halic-
tus scabiosae (Frommer and Flügel, 2005)
and Colletes hederae (Kuhlmann et al., 2007;
Frommer, 2008).
3.2. Mating systems
In the context of population biology, the
eective population size, population growth
rate and variance in population growth rate
of a species depend not only on the num-
ber of individuals and nests, but also on the
level polygamy and inbreeding present within
the population (Pamilo and Crozier, 1997;
Chapman and Bourke, 2001). Due to the hap-
lodiploid nature of bees, the eect of inbreed-
ingoneective population size is a function
of the sex ratio and level of polygamy (see
Zayed, 2009). Briefly, as females are diploid,
they contribute twice the genetic diversity to
a population than haploid males, thus mul-
tiple mating by females increases the eec-
tive population size, but multiple mating by
males decreases it, as several females in a
population would carry the sperm of a single
male (Crozier and Pamilo, 1996; Hedrick and
Parker, 1997).
The majority of the 20–30000 estimated
bee species are solitary (Michener, 2007),
yet knowledge of solitary bee mating sys-
tems is still quite limited compared to social
species; the assumption that females of most
species are monogamous remains (Eickwort
and Ginsberg, 1980). Currently, there is in-
sucient genetic pedigree (genetic analysis of
mother and ospring) data or behavioural ob-
servations to support this assumption (Paxton,
2005). For example, Blanchetot’s (1992)ge-
netic pedigree study of the solitary leafcutter
bee Megachile rotundata supports monoandry
in this bee, whereas in the primitively euso-
cial sweat bee Lasiglossum malachurum both
field observations (Knerer, 1992) and genetic
pedigree data (Paxton et al., 2002) indicate that
polyandry is common in this species. More
data on other solitary and primitively social
species are necessary before any generalisa-
tions over female mating systems can be made.
In contrast, better data on female mating
systems exist for the eusocial bumblebees,
honeybees and stingless bees. Single mating
by females appears to be the norm for most
bumblebees (Schmid-Hempel and Schmid-
Hempel, 2000), but there are exceptions (Bom-
bus hypnorum; Paxton et al., 2001). Honey-
bees are highly polyandrous (Schlüns et al.,
2005), whereas stingless bees are typically
monoandrous (Paxton et al., 1999; Peterset al.,
1999). Given this disparity in female mating
frequency, it is notable that both honeybees
and stingless bees are exceptional in thatmales
only mate once (Roubik, 1989; Koeniger,
1991); whereas, for most species, males prob-
ably mate repeatedly (Paxton, 2005).
Despite diculties in systematically
recording mating behaviour, there is a clear
need to confirm where, when and how often
mating occurs, especially as the reproductive
rates of solitary bees can be surprisingly low
(Minckley et al., 1994). Furthermore, the
presence of distinct intraspecific variability in
male mating behaviour, associated with (e.g.
Amegilla dawsoni, Alcock, 1997) or without
(e.g. Andrena agilissima, Paxton et al., 1999)
size dierences, represents a unique opportu-
nity to further advance our understanding of
how mating behaviour can influence repro-
ductive success at a population level, and how
this influences our estimation of population
size and growth rate for conservation.
3.3. Predators, parasites and pathogens
Major factors limiting bee populations such
as nest site or nest material (e.g. Potts
and Willmer, 1997), climatic conditions (e.g.
Pekkarinen, 1997) and pollen availability (e.g.
Minckley et al., 1994) have received rela-
tively detailed study compared to how preda-
tors, parasites and pathogens aect bee mor-
tality, population dynamics and community
composition. The potential for parasites and
pathogens to limit, or regulate, bee populations
is now glaringly apparent considering that
220 T.E. Murray et al.
Varroa destructor parasitic mites, for exam-
ple, destroyed 25–80% of managed honey-
bee colonies, and nearly all feral colonies, in
parts of the United States during the mid-
1990s (Sammataro et al., 2000). The epidemic
of ‘Colony Collapse Disorder’ destroyed 50–
90% of US colonies in aected apiaries in
2006/07, and also has been suggested to
involve a contagious pathogen (Cox-Foster
et al., 2007). It is possible that such marked
declines may also occur in many populations
of non-managed bees over similar geographic
scales (e.g. Andrena scotica;Paxtonetal.,
1997), but such declines are frequently un-
recorded or may be obscured by the consider-
able spatiotemporal variation in abundance ob-
served in many species (Williams N.M. et al.,
2001; Roubik, 2001; Tylianakis et al., 2005).
Furthermore, even sub-lethal eects of in-
fection may alter plant-pollinator interactions
and, therefore, ecosystem functioning (Eviner
and Likens, 2008). For example, the ability
of Bombus terrestris foragers to discriminate
rewarding flowers based on either colour or
odour decreased after being infected by the
protozoan parasite Crithidia bombi (Gegear
et al., 2006).
Better data on the potential of parasites and
pathogens to aect ecological interactions be-
tween species are available for the eusocial
bumblebees. At the regional scale (50 km2),
parasite diversity is proportional to regional
bumblebee host distribution and local abun-
dance; parasite load (the average number of
parasite species per individual worker) is in-
versely proportional to host species diversity
(Durrer and Schmid-Hempel, 1995;Schmid-
Hempel, 2001). These data support the hy-
pothesis that widely distributed (common)
host species may be adapted to a wider vari-
ety of parasites, pathogens or strains of both,
whereas locally occurring (rare) host species
may be adapted to a subset of the parasites
present (Price et al., 1986,1988). When both
rare and common host species coexist, the
common species is expected to have a com-
petitive advantage due to both its adaptation
to a wider variety of parasites, and by main-
taining a greater diversity of potentially more
virulent parasites within regions. However,
aside from cases of pathogen spillover from
introduced species (see Stout and Morales,
2009); there is currently little evidence re-
garding the role of parasites and pathogens
as drivers of decline in bumblebees (Goulson,
2003; Williams P.H. and Osborne, 2009). Us-
ing a novel comparative approach to inves-
tigate the role of pathogens in eusocial lin-
eages, Boomsma et al. (2005) propose a series
of testable hypotheses regarding eusocial bee
pathogens in comparison to eusocial wasps,
ants and termites: orally transmitted diseases
should be more common, more virulent and
endemic than those found in ants and termites;
the incidence of macroparasites (e.g. mites,
nematodes, parasitoids) should be intermedi-
ate between wasps and termites; and verti-
cal transmission of parasites should be more
common in bees and wasps, particularly in
colonies with multiple queens.
Other than phenomenological reports, lit-
tle data exist on the impact of parasites or
pathogens on the population dynamics in soli-
tary bees. For example, the Stylopidae have the
highest diversity of any strepsipteran family
(ca. 160 species) and are exclusive endopar-
asites of bees (Pohl and Beutel, 2008), yet
aside from their ability to eectively neuter
their host (Wülker, 1964;Askew,1971), rel-
atively little is known about their impact on
host populations (Kathirithamby, 1989). In the
context of social evolution, predators, particu-
larly ants, cause significant levels of mortality
in solitary versus social ground-nesting bees
and are, therefore, constraints on independent
nesting. Over a five-week period, all nests con-
taining only one female of the facultatively
social halicitid bee Megalopta genalis failed
to survive brood predation by ants, whereas
nests containing multiple females all survived
(Smith et al., 2007). Similarly, Zammit et al.
(2008) recorded colony survival and brood
production rates in predator-excluded versus
control nests of the cooperatively nesting al-
lodapine bee Exoneura nigrescens: 77.8% of
nests protected from ants survived, with an
average 3.55 brood per nest, compared to
41% survival and 1.56 brood per nest in un-
protected nests. Likewise, for eusocial bees,
high predation rates have been recorded by
predators such as the bee wolf, Philanthus
spp.; typically-sized aggregations of these
Conservation ecology of bees 221
crabronid wasps have been estimated to con-
sume 1015 bumblebees per hour (Dukas,
2005) and around 30000 honeybees per day
(Simonthomas and Simonthomas, 1980). It is,
therefore, unsurprising that in a recent meta-
analysis on the impact of predation on ecosys-
tem functioning, Knight et al. (2006) found
that predators had a substantial negative im-
pact on pollinator visitation rate and reproduc-
tive success of plants. However, the relative
importance of top-down (predators, parasites
and pathogens) versus bottom-up factors (flo-
ral and nesting resources) in regulating bee
populations is still largely unknown.
The response of individuals, populations
and communities of bees to changes in land-
use is primarily driven by the spatial and tem-
poral distribution of floral, nesting and over-
wintering resources in relation to the foraging
and dispersal abilities of bees (Kearns et al.,
1998;Kremenetal.,2007). Using concepts
from network theory, recent advances in the
study of plant-pollinator networks have con-
siderably improved our ability to define and
predict interactions between species at a com-
munity level and how the number, strength
and symmetry of these interactions influence
community tolerance to extinction (Memmott
et al., 2004;Larsenetal.,2005; Bascompte
and Jordano, 2007), especially in the context
of global environmental change (Tylianakis
et al., 2008).
4.1. Local scale: availability of critical
At the individual site or local scale, man-
agement and land-use practices determine the
community composition of both pollinators
and plants, and the extent to which biotic
and abiotic factors aect both groups (Kremen
et al., 2007;Fig.1). Whether clumped in
discreet patches or dispersed throughout the
landscape, the distribution of floral and nest-
ing resources largely dictates the structure and
composition of bee communities. In relation to
floral resources, including oils and resins, bees
are obligate pollen-foragers and generally both
bee and flower abundance and species richness
are positively associated (Wcislo and Cane,
1996;Stean-Dewenter and Tscharnke, 2001;
Potts et al., 2003; Holzschuh et al., 2007).
For example, flower species richness, nectar
resource diversity and nectar energy content
explained 23% of the variation in bee com-
munity structure across six distinct Mediter-
ranean habitats (Potts et al., 2006). Ultimately,
increasing floral diversity provides a wider
array of foraging niches for dierent func-
tional groups of flower visitors (Fenster et al.,
2004). Agri-environment schemes that alter
the spatial and temporal distribution of floral
resources frequently have the greatest impact
on pollinator community composition. For ex-
ample, in a study of 42 wheat fields in Ger-
many, organic fields had over twice the species
richness and twenty times the percentage cover
of flowering plants compared to conventional
fields, resulting in three times the species rich-
ness and almost eight times the abundance
of bees in organic versus conventional fields
(Holzschuh et al., 2007).
The availability of nesting resources also
plays a key role in structuring bee commu-
nities (Cane, 1991; Eltz et al., 2002; Potts
et al., 2005). In parallel with floral resources,
the temporal and spatial distribution of nest-
ing resources may determine the bee com-
munity composition in a given location. Eltz
et al. (2002) found that the abundance, size and
species of trees in tropical forests of Southeast
Asia influenced the density of stingless bee
nests. Similarly, in a hyperdiverse Mediter-
ranean bee assemblage, the amount of ex-
posed soil, the number of sloped surfaces and
the number of cavities available as nest sites
accounted for 26% of the variation in commu-
nity composition (Potts et al., 2005). However,
compared to floral resource use, relatively few
data exist on the nesting requirements for
many species. This may be due to the vari-
ety and often cryptic nature of bee nesting
habits, ranging from burrows in the soil, in
small pre-existing cavities in wood or stone,
to nests constructed of excreted wax in larger
cavities found in trees, rocks or rodent nests
(see Sect. 2.4). Further studies are needed to
assess whether nest sites are limiting resources
222 T.E. Murray et al.
and what factors within a bee’s flight rangede-
termine nest site selection.
In addition, indiscriminate use of pesticides
and herbicides can both increase mortality
rates (Johansen, 1977) and considerably re-
duce the availability flowering plants in agri-
cultural areas (Firbank et al., 2003; Morandin
and Winston, 2005). However, the intensity
of pesticide and herbicide use is often asso-
ciated with increased agricultural intensifica-
tion and subsequent decline in floral and nest-
ing resources (Kremen et al., 2002; Schweiger
et al., 2005; Williams N.M. and Kremen,
2007). Therefore, separating the relative ef-
fects of each factor on bee community com-
position will improve future decisions about
eective management of pollinators in agricul-
tural habitats.
4.2. Landscape scale: habitat loss,
fragmentation and land-use
Native habitat loss and fragmentation re-
sulting from human activity are two of the pri-
mary factors driving declines of native species
worldwide (Pimm et al., 2001). The synergis-
tic eect of loss and fragmentation reduces
gene flow and recolonisation between patches,
lowering persistence of both subpopulations
and networks of meta-populations (Hanski,
1998). Currently, there is no consensus on how
bee communities respond to habitat fragmen-
tation as empirical studies reveal a range of re-
sponses to fragment size, from positive (Cane
and Tepedino, 2001;Donaldsonetal.,2002)
to negative (Stean-Dewenter et al., 2002;
Klein et al., 2003; Ricketts, 2004;Tab.II).
This variability parallels that found in other
animal groups (e.g. Vandergast and Gillespe,
2004) and indicates that responses may dier
depending on life history and other species-
specific attributes, such as dispersal ability and
floral specificity (Cane and Tepedino, 2001;
Zayed et al., 2005). A further complication oc-
curs when the response of a taxon is obscured
by the composition of the surrounding matrix
of habitats and its influence on the availabil-
ity of floral and nesting resources (Eltz et al.,
2002;Caneetal.,2006; Williams N.M. and
Kremen, 2007). Therefore, our limited knowl-
edge of dispersal (see Sect. 3.1) and popula-
tion structure (see Zayed, 2009) prevent re-
liable estimation of the carrying capacity of
dierent habitats and habitat mosaics, further
complicating the assessment of their conserva-
tion value.
Globally, conversion of native habitats to
agriculture is the primary form of land-use
change and the largest cause of native habi-
tat loss and fragmentation (Tilman et al.,
2001; DeFries et al., 2004). The dominance
of agro-ecosystems worldwide means that in-
creasingly bee populations exist at the in-
terface of agricultural and natural habitats
or within agricultural areas. Although mass-
flowering crops can be beneficial in some
cases (e.g. Westphal et al., 2003,2009), de-
clines in both bee abundance and species
richness with increasing agricultural inten-
sity have been reported from a wide va-
riety of agro-ecosystems (Steen-Dewenter
and Tscharntke, 1999;Kremenetal.,2002;
Ricketts, 2004; Chacoand Aizen, 2006).
Generally, agro-ecosystems that contain a mo-
saic of semi-natural habitats throughout the
landscape can maintain significant levels of
bee diversity and abundance (Tscharntke et al.,
2005;Winfreeetal.,2007), even at re-
gional scales (Tylianakis et al., 2005). Results
from the recently completed European Union
Greenveins project demonstrated that, for tem-
perate European agro-ecosystems, once re-
gional eects (i.e. country) were removed,
variables describing land-use intensity and the
spatial distribution of semi-natural habitats at
the landscape scale (4 km2) were superior to
local scale variables (0.008 km2) in explaining
bee community composition (Schweiger et al.,
2005). In a separate analyses of the same data,
Hendrickx et al. (2007)arm that land-use
intensity and proximity to semi-natural habi-
tats best explained bee species richness across
landscapes, but loss of bee species richness
was not solely the result of declines within
habitats, but was also due to increased homog-
enization of community composition between
4.3. Plant-pollination networks
As there is a distinct lack of data
on how pollinator communities disassemble;
Conservation ecology of bees 223
Table II. Attributes of habitat fragmentation studies for native bee communities. Modified from Cane (2001).
Country Habitat Cause of
Parameter No. of
No. of
bee taxa
Abundance of
non-Apis bees
Trend Ref.
Argentina Dry thorn
Fragment size 8 43 481 Species richness
declined with
decreasing fragment size.
Aizen and
Brazil Rainforest Experimental
Fragment size 4 16 1092 Abundance decreased
with fragment size.
Powell and
Powell (1980)
Brazil Rainforest Experimental
Fragment size 7 16 290 Abundance increased
with fragment size.
Becker et al.
Brazil Atlantic
Fragment size
and level of disturbance
9 21 3653 No eect of
fragmentation due to
high variability of
species composition and
abundance between sites.
Tonhasca et al.
Costa Rica Agricultural
Distance to
forest patch
16 40 618 Species richness
was significantly higher
in farms within 100 m
from forest patches.
Ricketts (2004)
Costa Rica Tropical
Fragment size, shape,
isolation and context
22 117 1537 Fragment size,
shape, isolation and
context aected
community composition, but not
abundance or species richness.
Brosi et al.
Europe (7)* Agricultural
Land-use intensity;
habitat diversity;
distance to
semi-natural habitat
24 115 >14529 ∗∗ Across landscapes,
bee species richness
increased with habitat
diversity and proximity
of semi-natural habitat,
but decreased with increasing
land-use intensity.
Hendrickx et al.
Germany Agricultural
Distance to
semi-natural grassland
4023 212 Species richness
and abundance decrease
with increasing isolation.
and Tscharntke
Germany Agricultural
% semi-natural
grasslands within 3 km
15 36 1340 Abundance and diversity
of solitary bees were
correlated with %
semi-natural areas up
to 750 m, no eect found
for Bombus or Apis.
Dewenter et al.
224 T.E. Murray et al.
Table II. Continued.
Country Habitat Cause of
Parameter No. of
No. of
bee taxa
Abundance of
non-Apis bees
Trend Ref.
Indonesia Rainforest Agricultural
12 22 401 Solitary bee abundance,
not species richness,
increased; social bee
abundance and species
richness decreased.
Klein et al.
Indonesia Agricultural
Amount of shade,
distance to forest
24 29 >895Solitary bee diversity
increased with less shade;
social bee diversity
decreased with distance
to forest patch.
Klein et al.
South Africa Renosterveld
Fragment size 24 19 - Vegetation cover had
a greater eect than
fragment size on bee
species richness and composition.
et al. (2002)
U.S.A. Scrub desert Urbanisation Fragment size
and age
59 62 2512 Species richness
decreases and density
increases with smaller
fragment size. Fragment
size and age had
greatest eect on
ground-nesting specialists.
Cane et al.
U.S.A. Agricultural
% of semi-natural
habitat within 3 km
16 33 5732 Species richness
and abundance increased
with increasing %
semi-natural habitat.
and Kremen
U.S.A. Various Agricultural
% of forest habitat
within 1.6 km
40 130 2551 Species richness and
abundance decreased with
increasing forest cover,
but increased with agriculture
and urbanisation.
Winfree et al.
U.S.A. Agricultural
melon, pepper)
% of forest habitat
within 0.5–3 km
29 54 4592 No eect of % of
forest habitat and
species richness on crops.
Winfree et al.
* Belgium, Czech Republic, Estonia, France, Germany, Netherlands, Switzerland.
** The number of non-Apis bees was not indicated; therefore bees of 8.5–12 mm in length identified by Schweiger et al. (2005) were excluded.
40 habitat ‘islands’ consisting of 4 mustard and 4 radish plants.
The number of non-Apis eusocial bees was not indicated.
Conservation ecology of bees 225
predictions arising from the recent prolifera-
tion of simulation studies based on networks
of plant-pollinator interactions may be a valu-
able source of testable hypotheses (reviewed
in Bascompte and Jordano, 2007). At the com-
munity level, interactions between species are
the “glue of biodiversity” (Thompson, 2005)
and mutualistic networks provide well-defined
and predictable patterns of interdependence
between species. Specifically, plant-pollinator
networks can be described by two proper-
ties: they are very heterogeneous, with a few
species much more connected than by chance;
and they are highly nested, whereby special-
ists interact with distinct subsets of the species
interacting with generalists (Bascompte and
Jordano, 2007). As a result of this asymme-
try in specialisation, plant-pollinator networks
are reciprocally redundant and predicted to be
relatively tolerant to extinction, as only the mi-
nority of plant species are likely to lose all
their pollinator species as pollinator communi-
ties disassemble (Memmott et al., 2004). No-
tably, community resilience in plant-pollinator
networks is predicted to be enhanced by in-
creased species diversity and the number of
species interactions (Okuyama and Holland,
2008). Furthermore, understanding the nested
structure of plant-pollinator networks can ex-
plain why the reproductiveoutput of both spe-
cialist and generalist plant species is similarly
aected by habitat fragmentation (Ashworth
et al., 2004); how invasive species rapidly be-
come integrated into existing plant-pollinator
networks (Memmott and Waser, 2002;see
Stout and Morales, 2009); the rate of com-
munity disassembly with habitat loss (Fortuna
and Bascompte, 2006); and how the structure
of plant-pollinator networks remain largely
stable, despite considerable temporal variation
in the number, strength and symmetry of plant-
pollinator interactions (Petanidou et al., 2008).
Currently, the only empirical study of bee
community disassembly and its eect on pol-
lination is that of Larsen et al. (2005). Us-
ing wild bee data from Kremen et al. (2002),
the study found that the relationship between
species richness and pollination is approxi-
mately concave up, indicating that the first
species extinction could lead to rapid reduc-
tion in pollination. Two main factors corre-
lated with the loss of pollination function
in bee communities: the non-random loss of
species and the absence of strong density de-
pendence following species loss. Bee species
did not disappear randomly from sites; in-
stead species were lost in an ordered fash-
ion, with the largest, more ecient pollinators
more likely to experience local extinction. Ac-
cordingly, 86% of sites experienced a greater
loss of pollination function than would be ex-
pected by random species loss. This study af-
firms the sentiment expressed by Memmott
et al. (2004), that plant-pollinator networks
may be tolerant, but not immune, to extinction.
Clearly, there is a dire need for empirical data
regarding critical thresholds of species extinc-
tion and collapse of plant-pollination networks
(Fortuna and Bascompte, 2006).
4.4. Climate change
Thomas et al. (2004) predicted that by 2050
climate change, even in the absence of other
drivers of extinction, would doom 15–37% of
all species to eventual extinction. The predic-
tion is based on the application of an estab-
lished ecological pattern, the species-area re-
lationship, to data on the current distributions
and climatic envelopes of 1103 species. Al-
though their approach has been heavily crit-
icised (e.g. Lewis, 2006), the study did raise
critical questions regarding the response of
species to climate change, such as: how many
species have distributions primarily governed
by climate; and to what extent do current dis-
tributions truly reflect limits of climate toler-
Although the uncertainty in their models
preclude any firm generalisation, Dormann
et al. (2008) found that climate accounted
for 64.1% of the variation in the pattern of
species richness at the landscape scale for wild
bees in temperate Europe, compared to land-
scape structure (27.7%), soil (7.1%) and land-
use intensity (1.2%). Furthermore, in a re-
cent study simulating phenological shifts with
a real community of plants and pollinators,
Memmott et al. (2007) estimate that, over the
past 100 years, global warming has advanced
the first flowering date of plants and the sea-
sonal flight activity of some pollinating insects
226 T.E. Murray et al.
(mostly butterflies) by, on average, 4 days per
degree C in temperate zones, resulting in be-
tween 17–50% of all pollinator species experi-
encing a disruption in food supply.
Broad trends regarding the impact of cli-
mate change on plant-pollinator networks are
emerging (Tylianakis et al., 2008), but as
any alteration in climate will be superim-
posed upon other, multiple interacting drivers
of global change, significant challenges in pre-
dicting future responses remain.
Knowledge of the basic ecology of bees
is essential for underpinning the development
and implementation of conservation strategies
(Byrne and Fitzpatrick, 2009). Understanding
the factors that regulate bee populations and
communities, and the sensitivity of bee traits
to these factors, allows specific management
options to be identified for individual species
and entire assemblages. A first step is to iden-
tify the resource requirements of the target
bee taxa to ensure that the appropriate quan-
tity and quality of these are provided spa-
tially and temporally. These include both for-
age (pollen, nectar) and nesting (substrates and
construction materials) resources and the pro-
vision of suitable abiotic conditions (microcli-
mate and local topography). The distribution
of resources must fall within the forage and/or
dispersal ranges of the bee species considered,
as dierent parts of the landscape often pro-
vide complementary resources, leading to the
concept of partial habitats (Westrich, 1989).
Assessing the resource distributions within the
landscape may reveal that a single limiting
resource is missing which can then be sup-
plemented through management practices (e.g.
provision of flower-rich field margins) to sup-
port bee conservation (see below).
A supply of optimal resources is not enough
on its own; it is also important to understand
how bees respond to landscape-scale changes
in their environment. The geographic scale
over which negative pressures aect bees, cou-
pled with individual species’ ability to dis-
perse throughout the landscape, will determine
whether bees can avoid irreversible popula-
tion decline. For instance, habitat loss and
fragmentation can result in resource deple-
tion, and thresholds for foraging and nest-
ing resources will determine the carrying ca-
pacity of remaining habitats. Habitat area is
known to impact on bee community compo-
sition (Stean-Dewenter, 2003) and minimum
patch sizes are important for the persistence of
these communities (Hanski and Pöyry, 2006;
Kremen et al., 2004). The configuration of
the landscape, and how bees are able to dis-
perse through the landscape, will determine
whether spatially fragmented resources are
available (Stean-Dewenter and Tscharntke,
2002; Tscharntke and Brandl, 2004). The pro-
tection of key habitats and connectivity within
the landscape (Gilbert et al., 1998;Tewksbury
et al., 2002) therefore represents an important
tool for bee conservation (see below). Other
pressures such as grazing (Vulliamy et al.,
2006) and fire (Potts et al., 2003) modulate the
availability of resources in the landscape and
managing grazing regimes and burning prac-
tices can fundamentally alter habitat quality
for bees. Daily activity patterns and seasonal
phenology may also determinethe level of risk
of bee exposure to pesticides and other agro-
chemicals (Brittain et al., unpubl. data). Al-
tering the timing of pesticide application from
midday to early morning or late evening can
ensure that the window of maximum toxicity
does not overlap with the times when bees are
foraging on crops (Kevan, 1975). Similarly,
modifying the types and application regime
of herbicides can facilitate the maintenance of
diverse flower communities within intensive
agro-ecosystems (Westbury et al., 2008).
There are two broad and complementary
categories of bee conservation approaches,
both of which can be more eective when
informed by bee ecology: species-targeted
approaches and habitat-targeted approaches.
Species-targeted approaches focus on individ-
ual species, or broader taxa, and often pro-
vide some kind of legal protection which aims
to prevent activities causing direct mortal-
ity to individuals (including collection) (see
Byrne and Fitzpatrick, 2009). One example
is the Biodiversity Action Plan (BAP) in the
UK which includes 17 species of bees (www. BAP is the UK Government’s
Conservation ecology of bees 227
response to the Convention on Biological
Diversity and is a detailed plan for the pro-
tection of the UK’s biological resources. Each
UK BAP priority species action plan has an
overview of the status of a species, infor-
mation on the threats facing it, actions to
achieve the action plan, and broad policies
developed to conserve it. The core of the
bee BAPs is based on detailed autecology
of the target species or, if this is not avail-
able, then this is an initial priority step in de-
veloping a conservation plan. Another form
of species protection is the IUCN Red Lists
which include bees in several countries (e.g.
Finland, Germany, Hungary, Ireland, Nether-
lands, Slovenia, Sweden, Switzerland and UK;
see Byrne and Fitzpatrick, 2009). These prior-
ity species lists can clearly direct prioritization
of conservation actions at national levels, but
action should not be exclusive to listed species
(Fitzpatrick et al., 2007).
Complementing the species-targeted ac-
tions are habitat-based approaches which fo-
cus on conserving and restoring important bee
habitats. In some cases these will be specif-
ically for bees, but in many cases they are
aimed at enhancing wider biodiversity, which
may also include bees. Many national pro-
tected areas and networks are natural or semi-
natural areas designated on the basis of priority
habitats (e.g. Special Areas of Conservation
in Europe, EC Habitats Directive 92/43/EEC)
which in many terrestrial systems often in-
clude diverse floral resources which can sup-
port bees. In addition to protected areas, more
ubiquitous habitats can be