Role of nesting resources in organising diverse bee
communities in a Mediterranean landscape
SIMON G. POTTS
1
, BETSY VULLIAMY
2
, STUART ROBERTS
1
,
CHRIS O’TOOLE
3
, AMOTS DAFNI
4
,GIDINE’EMAN
5
and P A T
WILLMER
2
1
Centre for Agri-Environmental Research, University of Reading, U.K.,
2
School of Biology, University of
St Andrews, U.K.,
3
Bee Systematics and Biology Unit, Hope Entomological Collections, Oxford University Museum of Natural
History, U.K.,
4
Institute of Evolution, University of Haifa, Israel and
5
Department of Biology, University of Haifa, Israel
Abstract. 1. The habitat components determining the structure of bee commu-
nities are well known when considering foraging resources; however, there is little
data with respect to the role of nesting resources.
2. As a model system this study uses 21 diverse bee communities in a Mediterra-
nean landscape comprising a variety of habitats regenerating after fire. The find-
ings clearly demonstrate that a variety of nesting substrates and nest building
materials have key roles in organising the composition of bee communitie s.
3. The availability of bare ground and potential nesting cavities were the two
primary factors influencing the structure of the entire bee community, the com-
position of guilds, and also the relative abundance of the dominant species. Othe r
nesting resources shown to be important include availability of steep and sloping
ground, abundance of plant species providing pithy stems, and the occurrence of
pre-existing burrows.
4. Nesting resource availability and guild structure varied markedly across
habitats in different stages of post-fire regeneration; however, in all cases, nest
sites and nesting resources were important determinants of bee community
structure.
Key words. Bees, community organisation, Mediterranean, nesting guilds,
resource availability.
Introduction
Organisation of bee communities is closely related to the
floral communities they forage upon, with several key char-
acters having been identified, including floral diversity (e.g.
Tepedino & Stanton, 1981; Gathmann et al., 1994), floral
abundance (e.g. Banaszak, 1996), and availability of pollen
and nectar resources (Petanidou & Vokou, 1990). However,
few studies have attempted to quantify the combined effect
of these structuring agents (but see Potts et al., 2003a).
In turn, some drivers have been identified that impact
directly upon bee communities, while others act indirectly
through modification of floral communities or other habitat
characteristics. These drivers include changing land use
practices such as agricultural intensification (Banaszak,
1995), habitat fragmentation (Jennersten, 1988) and habitat
isolation (Steffan-Dewenter & Tscharntke, 1999), grazing
(Potts et al., 2003a), and agrochemical use (O’Toole, 1993).
Other important drivers have also been recently identified:
fire (Potts et al., 2003b); disease (Watanabe, 1994) and
parasite spread (Schmid-Hempel & Durrer, 1991); climate
change (Price & Waser, 1998); introduction of non-native
plants (Brown & Mitchell, 2001; Chittka & Schu
¨
rkens,
2001); and competition with managed pollinators
(Butz-Huryn, 1997; Steffan-Dewenter & Tscharntke, 2000).
While the forage rewards provided by floral communities
are generally accepted as the primary determinants of polli-
nator community structure, there is an increasing body of
evidence suggesting that nest sites and nesting resources may
Correspondence: Simon G. Potts, Centre for Agri-Environmental
Research, PO Box 237, University of Reading, Reading RG6
6AR, U.K. E-mail: s.g.potts@reading.ac.uk
Ecological Entomology (2005) 30, 78–85
78 #
2005 The Royal Entomological Society
also play an important role for bees (e.g. Petanidou & Ellis,
1996). Several studies have investigated particular biotic and
abiotic factors influencing nesting success or nest-site selec-
tion for single species (e.g. Potts & Willmer, 1997; Wuellner,
1999), yet only two studies to date have provided quantitative
evidence showing that nesting resources may have an import-
ant role in the structuring of entire communities (Potts et al.,
2003a) or key guilds within communities (Cane, 1991). The
study of Potts et al. (2003a) examined 14 habitat character-
istics as predictors of bee community structure; floral
characters were the primary determinants, however 5% of the
entire bee community structure was explained by the diversity
of nesting resources available, and 10% of the structure when
only dominant bees were considered. These findings indicate
that nesting resources in general play a small but important
role in organising bee communities, and direct the present
study in partitioning this variation into the function of
specific resources and substrates.
Bees exhibit a diverse array of nesting strategies with
respect to the part of the habitat they nest in, the type of
substrate they use, and the materials required for nest con-
struction. Indeed, bees can be partitioned into several exclu-
sive guilds on the basis of their nesting habits (O’Toole &
Raw, 1991), known as miners, masons, carpenters, and social
nesters. Miners dominate in many open habitats and exca-
vate holes in the ground that may be lined with glandular
secretions. All species of Andrenidae, Melittidae, Oxaeidae,
and Fideliidae are miners, as are most Halictidae, Colletidae,
and Anthophoridae. Masons generally use pre-existing cav-
ities in which to construct their nests, and these may be pithy
or hollow plant stems, small rock cavities, abandoned insect
burrows, or even snail shells. Masons are from the family
Megachilidae, and they line their nests with materials found
within their habitat rather than with glandular secretions.
Leaf-cutters are a subgroup of masons, also megachilids
from the genus Megachile and Creightonella;theyuse
pre-existing cavities and line their nest with freshly gathered
leaf material. Carpenters excavate their own nests in woody
substrate; this habit is found in two genera within the Apidae
(Xylocopa and Ceratina) and one from the Megachilidae
(Lithurgus). Social nesters use larger pre-existing cavities to
build large social nests; members of this guild are all from the
Apidae and include honeybees, bumblebees, and stingless
bees. Finally, the members of one guild do not construct
nests at all, but instead parasitise the nests of other bees
and are therefore referred to as cuckoo bees or kleptopar-
asites. Cuckoo bees have evolved in several families.
Potentially important habitat characteristics previously
investigated as being important resources for various nest-
ing guilds are soil texture (Cane, 1991), soil hardness
(Brockmann, 1979; Potts & Willmer, 1997), soil moisture
(Wuellner, 1999), aspect and slope (Potts & Willmer, 1997),
amount of insolation (Weaving, 1989; Jeanne & Morgan,
1992), cavity shape and size (Schmidt & Thoenes, 1992),
and diameter of pre-existing holes (Scott, 1994).
Using a hyper-diverse bee assemblage as a model system,
this study aims to: (1) quantify the importance of nesting
resources in the organisation of overall bee community
structure; (2) determine which guilds are most influenced
by nesting resources; and (3) identify the specific habitat
characteristics to which the whole community and indivi-
dual species are responding.
Methods
Sites
The National Reserve on Mt Carmel, Israel, comprises
150 km
2
of Pinus halepensis pine forest. Mount Carmel has
a typical Mediterranean climate with cool wet winters and
warm dry summers. The reserve now has a mosaic of regen-
erating post-fire patches, with freshly burnt open habitats,
intermediate-aged scrub (phrygana), mature pine forest
stands, and all intermediate habitat types. Several major
burns have occurred (two in 1998, and one in each of
1989, 1983, and 1974, giving site ages of 10, 16, and
25 years); the resultant post-fire regenerating patches are
found within large tracts of unburnt pine forest.
Within each burnt area, three 1-ha sites were selected as
being representative of that burn age and each was at least
300 m (usually > 600 m) away from another site or bound-
ary with a different vegetation type. This separation dis-
tance, though potentially within the foraging range of large
solitary and social bees (Wesserling & Tscharntke, 1995), is
unlikely to compromise the independency of the sites in this
study as forage and nesting resources were always found in
high concentrations locally within each site. Site details are
summarised in Table 1. Full descriptions of the floral com-
munities can be found in Zohary (1982), and more detailed
site information can be found in Potts et al. (2003a, b).
With this design there is inevitably a degree of pseudo-
replication imposed by the structure of the landscape which
cannot be experimentally manipulated. However, this poten-
tial shortcoming was minimised as far as possible by using
independent burns whenever available (e.g. there were two
fires in 1998, and two discrete areas of unburnt pine were
used as controls) and also separating sites as much as possible.
Control sites were selected to be close (> 600 m but < 1km)to
the burnt areas studied and were typical of the large stands of
mature pine prior to fire. In view of this a degree of caution is
therefore suggested when interpreting findings.
Bee surveys
At each of the 21 sites, bees (Hymenoptera: Apoidea)
were monitored throughout the main flight season of
February to May in 1999. Five surveys were undertaken
at each site spread evenly through the season. Each survey
consisted of a 200-m linear transect walk made over 20 min
at 08.00, 11.00, and 14.00 hours; these times were chosen to
cover the main period of bee activity. A transect started
from a random point within the site and was walked in a
random direction; all bees encountered within 1.5 m of the
observer were recorded. Those species that could be identified
Bee communities and nesting resources 79
#
2005 The Royal Entomological Society, Ecological Entomology, 30, 78–85
unambiguously while in flight were scored on a
recording sheet, while those which could not be were caught
using a hand net and retained for identification at Oxford
University Museum (U.K.).
As it is extremely difficult and time consuming to locate
nesting bees in scrub habitats, this study used surveys of
foraging bees as a surrogate. The authors consider the
assumption that bees observed foraging in any particular
habitat are representative of those nesting to be reasonable.
Family classification was in accordance with Michener
(2000), and each species was assigned to the appropriate
functional group based on nesting habit: ground nesting
miners (designated as ground); mason bees that use rock
substrate (masons), hollow stems (stem), pre-existing burrows
(old nest), or snail shells (snail); eusocial bees nesting in large
cavities (cavity); carpenters (carpenter); and kleptoparasites
(cuckoo).
Nesting resources surveys
Using existing literature (e.g. Stephen et al., 1969;
O’Toole & Raw, 1991; Potts & Willmer, 1997; Michener,
2000) in conjunction with the authors’ combined field
experience on Mt Carmel, a list of known and potentially
important nesting resources was compiled. These resources
and the methods used to measure them are listed in Table 2.
The resources available for nesting were surveyed twice at
each site during 1999, using a 200-m linear transect with 10
1-m
2
quadrats placed evenly along it. A 1-m
2
sample unit
was considered an appropriate size based on a preliminary
assessment of the distribution of resources within habitats.
Statistical approach
A variety of methods was used to explore the relationships
between nesting guilds and nesting resources and how these
varied between sites of different ages. Differences between
the species richness and abundance of bees in each guild with
respect to site age were tested using a Generalised Linear
Model, and differences between paired means with Tukey’s
honestly significant test. Changes in nesting resources
through time also used the same approach.
Ordination was used to identify associations between bee
community structure and nesting resources. Detrended
correspondence analysis of the species data indicated that
linear, rather than unimodal, ordination methods were
most appropriate, therefore redundancy analysis (RDA)
was employed. A first model included all 116 bee species;
however, given that more than half of the species in the
surveys were represented by singletons, a second model used
the 17 species that contributed > 0.5% to total abundance.
A Monte Carlo global permutation test gave the signifi-
cance of the canonical axes and the significance of the
environmental variable-axis relationships was determined
using a Monte Carlo permutation test under a reduced
model. Environmental variables were automatically
forward selected. The RDA analysis was performed using
CANOCO 4.5 (Ter Braak & Smilauer, 1999). The results of the
ordinations were used as a guide to test for the strength of
associations between specific nesting resources and the
abundance of guilds and single species.
Results
Nesting guild structure in habitats of different post-fire age
Total bee species richness shows a curvilinear relationship
with site age (Fig. 1a). The number of species is highest
in the freshly burnt sites, lowest in the intermediate-aged
habitats, and mature habitats show a slightly higher
species richness than the intermediate sites. The most
speciose guild, the ground nesters, mirrors this pattern with
significant differences between mean species numbers
(F
4,16
¼ 4.58, P ¼ 0.012): freshly burnt sites have signifi-
cantly more species than the 16- and 25-year-old sites
(P < 0.05). The mean number of mason and stem-nesting
species varies with site age (F
4,16
¼ 3.71, P ¼ 0.025 and
F
4,16
¼ 2.91, P ¼ 0.055 respectively). Masons are more
speciose in 10-year-old sites (P < 0.05) than in freshly
burnt or mature sites. Similarly stem nesters are repre-
sented by more species in 10-year-old sites than mature
sites (P < 0.05). No other guilds show significant changes
in species numbers with site age.
The total number of bees is highest in freshly burnt sites
and lowest in the 16-year-old sites with intermediate numbers
in the mature sites (Fig. 1b). Ground nesting bees dominate
the community at all sites, but they decrease steadily in
abundance with site age, being least common in the 25-year-
old site, and then increase in abundance again in mature sites.
Table 1. Summary of site locations, burn ages, and habitat area used in the 1999 surveys on Mt Carmel, Israel.
Site code Location Year burnt
Coordinates
(latitude, longitude) Approximate area (ha)
Den98 Wadi Denia 1998 32
45.8
0
,35
00.1
0
100
Hod98 En Hod 1998 32
41.9
0
,34
58.6
0
450
Hai89 Hai Bar Reserve 1989 32
44.9
0
,35
01.2
0
300
Mit83 Mitla 1983 32
44.2
0
,34
59.6
0
400
Etz74 Etzbah 1974 32
42.4
0
,34
58.9
0
80
EtzNB Etzbah <1950 32
42.4
0
,34
58.8
0
240
DenNB Wadi Denia <1950 32
45.7
0
,35
00.4
0
190
80 Simon G. Potts et al.
#
2005 The Royal Entomological Society, Ecological Entomology, 30, 78–85
Eusocial cavity nesters are also abundant in all sites, while the
other guilds are represented by relatively few individuals.
There are, however, no statistically significant differences in
the abundance of bees through time for any guild.
Nesting resources in habitats of different post-fire age
The availability of nesting resources varies with site age
and mean values are given in Table 3. Not surprisingly, the
availability of soil decreases with post-fire age, with freshly
burnt sites having significantly more bare ground available
(F
4,16
¼ 7.25, P ¼ 0.002) than all older sites (P < 0.05). The
freshly burnt sites have 58.6% coverage whereas all other
sites have less than 23% coverage. The number of pithy
stems available is greater in the 10-year-old site (2.0 per m
2
)
than all other ages (< 1.2 per m
2
), though the difference is
not statistically different.
Relationship between nesting guilds and resources
The redundancy analysis using the entire bee community
indicates that bee community structure is significantly cor-
related with nesting resources (Table 4: sum of all canonical
variables ¼ 0.667; F ¼ 1.34, P ¼ 0.004 for all canonical
axes). The first and second axis explain 12.8% and 10.8%
of the variance respectively, in the model (Fig. 2a) and four
variables are significant: soil (7% of variation explained,
F ¼ 1.55, P ¼ 0.024); cavities (12%, F ¼ 2.50, P ¼ 0.001);
slope (7%, F ¼ 1.70, P ¼ 0.009); and flat (6%, F ¼ 1.51,
P ¼ 0.035).
Partitioning the bee community into nesting guilds also
shows that these are organised by nesting resources
(Table 5: sum of all canonical variables ¼ 0.722; F ¼ 1.73,
P ¼ 0.025 for all canonical axes). This model had two sig-
nificant terms, cavities (9% of variation explained, F ¼ 3.31,
P ¼ 0.009) and old nests (6%, F ¼ 3.09, P ¼ 0.006), and the
first two axes explain 44.3% of the variation (Fig. 2b). The
abundance of cavity nesting bees is positively correlated
with the number of cavities (r ¼ 0.482, n ¼ 21, P ¼ 0.027)
and with steep ground (r ¼ 0.682, n ¼ 21, P ¼ 0.001) avail-
able in the habitat. For ground nesters, the guild with the
greatest number of species and individuals, species richness
is positively correlated with area of bare ground (r ¼ 0.523,
n ¼ 21, P ¼ 0.015) and abundance is correlated with avail-
ability of hard soils (r ¼ 0.378, n ¼ 21, P ¼ 0.091) though the
latter is only marginally significant.
The ordination focusing on only the most numerically
dominant species indicates that the structure of this partial
community is not as strongly related to nesting variables
Table 2. Definition, code, and methods for measuring nesting resources in Mt Carmel National Reserve, Israel.
Resource code Method
Soil % of exposed ground in a quadrat free of vegetation and litter
Flat % of ground in a quadrat with slope < 30
Slope % of ground in a quadrat with slope 30–60
Steep % of ground in a quadrat with slope > 60
Wood Amount of dead woody substrate in a quadrat, scored on a categorical scale 0–2
Stems Number of exposed pithy or hollow plant stems per quadrat
Soft Number of soil penetrometer (ELE International, Leighton Buzzard, U.K.) readings out of 10 with values < 2.0 kgf cm
2
Medium Number of soil penetrometer readings out of 10 with values 2.0–4.0 kgf cm
2
Hard Number of soil penetrometer readings out of 10 with values > 4.0 kgf cm
2
Cavities Number of large (> 2 cm diameter) cavities in rocks, trees, and rodent holes per quadrat
Old nests Number of pre-existing insect burrows in the ground or wood per quadrat
Snail shells Number of empty snail shells per quadrat
kgf, Kilograms of force.
0
5
10
15
20
25
Species richness (mean no.species/site)
Carpenter
Cuckoo
Snail
Mason
Old nest
Cavity
Stem
Ground
(a)
0
110162550
10
20
30
40
50
60
70
A
g
e (years post-fire)
Abundance (mean no. bees/site)
(b)
Fig. 1. (a) Species richness, and (b) abundance within bee nesting
guilds as a function of site age (time since burnt) for the bee
communities of Mt Carmel, Israel. Mature pine sites were burnt
before 1950 and labelled as age 50.
Bee communities and nesting resources 81
#
2005 The Royal Entomological Society, Ecological Entomology, 30, 78–85
(Table 6: sum of all canonical variables ¼ 0.580; F ¼ 0.92,
P ¼ 0.675 for all canonical axes). However, there are two
significant nesting resources entered into the model
(Fig. 2c): soil (12% of variation explained, F ¼ 2.53,
P ¼ 0.005) and cavities (8%, F ¼ 1.75, P ¼ 0.041). The abun-
dance of four dominant ground-nesting species are posi-
tively correlated with the availability of soil: Lasioglossum
transitorium (r ¼ 0.732, n ¼ 21, P < 0.001); Lasioglossum
caspicum (r ¼ 0.427, n ¼ 21, P ¼ 0.054); Andrena rufomacu-
lata (r ¼ 0.648, n ¼ 21, P ¼ 0.002); and Eucera transversa
(r ¼ 0.444, n ¼ 21, P ¼ 0.044). The single most common bee
in the surveys, the honeybee (32.4% of total bee abundance)
is positively correlated with both the number of cavities
(r ¼ 0.486, n ¼ 21, P ¼ 0.025) and availability of steep
ground (r ¼ 0.687, n ¼ 21, P ¼ 0.001). Cavities and steep
ground are also autocorrelated (r ¼ 0.693, n ¼ 21,
P ¼ 0.001).
Discussion
For bee communities, habitat quality depends upon two
sets of environmental characteristics: those related to fora-
ging requirements and those related to nesting require-
ments. The idea of partial habitats (Westrich, 1996)
highlights the need for both these complementary resources
to be available locally to bee species, even if spatially segre-
gated within the habitat. The role of foraging resources
(pollen and nectar sources) has been clearly demonstrated
for bee communities (e.g. Petanidou & Vokou, 1990), but
the relative importance of nesting resources has received
relatively little attention (but see Cane, 1991). Given that
an earlier study indicates that 5–10% of bee community
organisation is influenced by the availability of nesting
resources (Potts et al., 2003a), this study clearly demon-
strates that several specific resources are key factors influ-
encing community structure.
For the entire bee community and the dominant species
within the community, approximately 40% of the variation
in species-abundance pattern is explained by the availability
of nesting resources (Tables 4 and 6); and when the com-
munity is analysed in terms of guild structure 61% of the
variation is accounted for (Table 5). Two nesting resources
consistently contribute to the models: the availability of
bare ground and the occurrence of suitable nesting cavities.
The area of exposed earth for ground-nesting bees to nest in
has been suggested as a possible limiting factor for Medi-
terranean bee communities (Petanidou & Ellis, 1996); how-
ever, this is the first study to quantify the effect for an entire
community. The abundance of ground nesting bees varies
markedly across sites and is highest in freshly burnt areas.
One of the most obvious impacts of burning is the clearance
of vegetation and consequent exposure of bare ground,
which is highest immediately post-fire and then decreases
as the plant community regenerates. The absolute abun-
dance of four dominant ground-nesting species is positively
correlated with the area of soil available within a site. The
analysis indicates that it may not be only the quantity but
Table 4. Redundancy analysis for the entire bee community of 116 species.
Axis 1 Axis 2 Axis 3 Axis 4 Total variance
Eigenvalues 0.128 0.108 0.073 0.069
Species–environment correlation 0.988 0.963 0.979 0.944
Cumulative percentage variance of species data 12.8 23.6 30.9 37.8
Cumulative percentage variance of species–environment data 19.2 35.4 46.3 56.7
Sum of all eigenvalues 1.00
Sum of all canonical eigenvalues 0.667
Table 3. Mean ( SE) availability of nesting resources at sites of different ages.
Site age (years since burnt)
Characteristic 1 10 16 25 > 50
Soil 58.6 9.3 12.2 6.2 22.5 8.6 15.9 5.8 16.5 5.0
Flat 70.4 1.6 100.0 0.0 87.9 2.7 93.7 0.9 85.5 1.6
Slope 20.6 7.2 0.0 0.0 12.1 12.1 1.6 1.6 14.5 9.1
Steep 9.0 3.7 0.0 0.0 0.0 0.0 4.7 2.6 0.0 0.0
Wood 12.1 5.1 5.2. 0.7 20.5 2.0 6.7 2.6 9.3 3.0
Stems 0.2 0.2 2.0 1.0 0.3 0.3 1.2 0.6 1.2 0.6
Soft 5.4 0.4 7.3 0.2 6.2 0.2 6.3 1.2 6.9 0.4
Medium 3.0 0.3 1.7 0.2 2.6 0.2 2.6 0.9 1.7 0.3
Hard 1.6 0.6 1.0 0.3 1.2 0.2 1.1 0.6 1.4 0.7
Cavities 6.7 2.9 0.8 0.6 0.5 0.3 1.7 0.9 1.6 1.1
Old nests 0.2 0.2 1.0 0.6 0.0 0.0 0.3 0.3 0.0 0.0
Snail shells 0.8 0.2 0.2 0.2 0.7 0.4 0.6 0.3 1.0 0.2
82 Simon G. Potts et al.
#
2005 The Royal Entomological Society, Ecological Entomology, 30, 78–85
