Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

COMMUNITY AND ECOSYSTEM ECOLOGY

Effects of Suburbanization on Forest Bee Communities


ADRIAN L. CARPER,1,2,3 LYNN S. ADLER,4 PAIGE S. WARREN,5 AND REBECCA E. IRWIN1

Environ. Entomol. 43(2): 253Ð262 (2014); DOI: http://dx.doi.org/10.1603/EN13078


ABSTRACT Urbanization is a dominant form of land-use change driving species distributions,
abundances, and diversity. Previous research has documented the negative impacts of urbanization
on the abundance and diversity of many groups of organisms. However, some organisms, such as bees,
may beneÞt from moderate levels of development, depending on how development alters the

Downloaded from https://academic.oup.com/ee/article/43/2/253/526089 by guest on 02 May 2023


availability of foraging and nesting resources. To determine how one type of low-intensity human
development, suburbanization, affects bee abundance and diversity and the mechanisms involved, we
surveyed bees across suburban and natural forests in the Raleigh-Durham area of North Carolina. We
sampled for bees using a combination of bee bowls and hand-netting from March through July of 2008
and 2009. We found higher bee abundance in suburban than natural forests, and although observed
species richness was greater in suburban than natural forests, there were no signiÞcant differences in
rareÞed richness or evenness estimates in either year. In addition, the effects of suburbanization were
similar across bee species of varying ecological and life-history characteristics. At the local scale, bee
abundance and species richness were both positively related to the abundance and richness of
ßowering species within forests, while the proportion of surrounding developed open areas, such as
yards and roadsides, was a strong positive predictor of both bee abundance and richness at the
landscape scale. These results suggest that open habitats and the availability of ßoral resources in
suburban sites can support abundant and diverse bee communities and underscore the potential for
native bee conservation in urban habitats.

KEY WORDS bee, land use, pollinator, suburban, urban

Land-use change is a leading component of human- decline in bee populations (Biesmeijer et al. 2006,
induced environmental change (Kalnay and Cai 2003) Cameron et al. 2011, Bartomeus et al. 2013, Burkle et
that has ecological consequences at both local and al. 2013), and habitat loss because of land-use changes
global scales (Foley et al. 2005). Urbanization (the is hypothesized as a major factor (Winfree et al. 2009).
expansion of developed residential, municipal, and For example, bee abundance and diversity often de-
industrial areas) is a dominant form of land-use change cline with increasing metrics of urbanization, such as
that has drastically expanded in recent history (Irwin built landscape (Bates et al. 2011), a pattern suggested
and Bockstael 2007), reducing biodiversity through as a general trend (Hernandez et al. 2009). However,
habitat loss, fragmentation, and competition with in- the effects of urbanization likely vary with the degree
troduced, competitively dominant, and often synan- or type of anthropogenic disturbance (Winfree et al.
thropic species (Shochat et al. 2010). Declines in spe- 2009), with the potential costs and beneÞts likely de-
cies richness (Burkle et al. 2013) also raise concern pendent on a variety of factors, including the type of
over the stability of natural communities and the eco- development, the surrounding natural habitat, and the
system services they provide (Eigenbrod et al. 2011). availability of foraging and nesting resources. For ex-
Bees are an important group of organisms in urban ample, urban gardens can promote diverse bee assem-
and natural areas, given their importance as pollina- blages (Fetridge et al. 2008), and the number of gar-
tors. Approximately 87% of ßowering plants rely on dens in the surrounding landscape has been shown
pollinators, especially bees, to reproduce (Ollerton et to positively inßuence bumblebee colony survival
al. 2011). A growing body of evidence points to a (Goulson et al. 2010). Moderate levels of human dis-
turbance, such as rural and suburban development,
1 Department of Biological Sciences, Dartmouth College, 78 Col-
could represent intermediate levels of disturbance
lege St., Hanover, NH 03755.
that promote species coexistence (Connell 1978) and
2 Present address: Department of Ecology and Evolutionary Biol- therefore could hold potential for bee conservation.
ogy, University of Colorado, Ramaley N122, Boulder, CO 80309. However, more studies are needed to document pat-
3 Corresponding author, e-mail: adrian.l.carper@colorado.edu.
4 Biology Department, 221 Morrill Science Center South, 611 N.
terns and identify mechanisms driving bee abundance
Pleasant St., University of Massachusetts, Amherst, MA 01003.
and species richness.
5 Department of Environmental Conservation, University of Mas- To understand how one type of human develop-
sachusetts, 216 Holdsworth, Amherst, MA 01003. ment, suburbanization, affects bee abundance and

0046-225X/14/0253Ð0262$04.00/0 䉷 2014 Entomological Society of America


254 ENVIRONMENTAL ENTOMOLOGY Vol. 43, no. 2

richness and the mechanisms involved, we surveyed wood forest that were ⬎10 ha and primarily managed
bees across paired suburban and natural forests in a for the conservation of their natural resources, such as
metropolitan area of the southeastern United States. county and state parks. Suburban forests consisted of
We deÞned suburban development as single-family, persistent wooded patches in single-family housing
residential neighborhoods encompassed within or on developments located ⬍5 km from each correspond-
the edge of more intensively incorporated metropol- ing forested site (Supp. Table 1 [online only]). We
itan areas. We asked the following questions: 1) How established equal-sized sampling plots within each site
does suburbanization affect forest bee abundance, pair ranging from 0.22 to 0.57 ha (Supp. Table 1 [online
richness, and composition? We predicted that subur- only]). The paired approach allowed us to control for
ban sites would have higher bee abundance and rich- landscape-scale differences in site characteristics
ness given that suburbs typically have gardens and while minimizing potential overlap in bee foraging
green spaces and are often esthetically managed to within pairs.
provide ßowers throughout the growing season, which Bee Abundance, Richness, and Composition. We
could beneÞt bees. However, the response of individ- sampled bees over two ßowering seasons: May

Downloaded from https://academic.oup.com/ee/article/43/2/253/526089 by guest on 02 May 2023


ual bee species to suburbanization may vary with eco- through July of 2008 and April through July of 2009.
logical or life-history characteristics (Williams et al. Sampling occurred approximately once per month in
2010). Therefore, we predicted that suburban forests 2008 (three sampling dates) and twice per month in
would have more polylectic bees because of increased 2009 (six sampling dates). Bees were sampled using
availability of exotic ßowers in suburban gardens on pan traps and hand-netting following a standard pro-
which generalist bees could forage, and more soil and tocol (LeBuhn et al. 2003). Pan traps were made from
hive-nesting bees because of increased availability of 9-cm diameter plastic bowls (Fisherbrand Hexagonal
open areas, such as yards and roadsides, and bare Polystyrene Weighing Dishes, Pittsburgh, PA) that
ground from human activities, such as gardening and were left white or painted ßuorescent blue or yellow.
erosion. Conversely, we expected natural forests to In each site, we arranged 15 pan traps (5 of each color)
have greater abundances of oligolectic, pith, and stem- Þlled with soapy water along a single 70-m transect,
nesting bees because larger forests likely retain more with 5 m between each trap. Transects were laid hap-
native ßowers and Þne woody nesting substrates. 2) hazardly within contiguous sampling plots except in
What local and landscape factors affect bee commu- two of the suburban sites, where two smaller forest
nities? Bee abundance and richness can be driven by patches were bisected by roads or driveways (RV and
both the abundance and richness of ßoral resources SC; Supp. Table 1 [online only]). We randomly as-
(Potts et al. 2003) and the availability of suitable nest- signed the color of pan traps to their positions along
ing substrates in the landscape (Cane et al. 2006). We each transect and left the traps on the forest ßoor from
predicted bee abundance and richness would increase 0800 to 1800 hours EDT. Captured bees were strained,
with local ßoral abundance and richness. However, rinsed, and dried before sorting. Hand-netting was
bees can forage widely, and some patterns of abun- conducted for 1 h per plot per sampling period to
dance or richness may not be evident at local scales. capture bees not attracted to pan traps. We conducted
At the landscape scale, we predicted open and low- netting between 1000 and 1400 hours, during peak bee
intensity developed land covers would be strong pos- activity, and euthanized bees in ethyl acetate collect-
itive predictors of bee abundance and species richness ing jars. Suburban and natural forests were consecu-
because open surroundings likely provide superior tively sampled within each site pair, with the order
foraging and nesting resources (McFrederick and (suburban or natural forest Þrst) randomly chosen. In
LeBuhn 2006). 2008, hand-netting corresponded to pan-trapping ses-
sions by sampling one site pair per day. In 2009, pan
trapping was done simultaneously across all sites, and
Materials and Methods
hand-netting was conducted in each site pair within
Study Area. We conducted this study in Raleigh- 24 Ð 48 h of pan-trapping events. All bees were pinned,
Durham, NC (hereafter RDU). RDU is one of the labeled, identiÞed to either species or morphospecies,
fastest growing urban areas in the United States (U.S. and housed in a collection at Dartmouth College.
Census Bureau 2009) and is located in a region un- We compiled ecological and life-history informa-
dergoing some of the fastest rates of land-cover tion for each bee species from the primary literature,
change in the eastern United States (Brown et al. historical accounts, Discover Life (Ascher and Pick-
2005). The two counties surrounding RDU (Durham ering 2012), and consultation with experts (J. Ascher,
and Wake) include ⬇737,000 single-family detached personal communication). For each bee species, we
homes (62.5% of residences; NC OfÞce of State Budget characterized nesting substrate (soil, wood, pithy
and Management), creating a matrix of suburban de- stems, cavity, or hive), ßoral specialization (oligolec-
velopments in proximity to large tracts of naturally tic vs. polylectic), and sociality (solitary, subsocial,
occurring forest. Using a paired approach, we identi- eusocial, or parasitic). Only those species that foraged
Þed Þve sites within suburban forests and Þve sites in on a single plant family or a subset of ßowering genera
managed natural areas (hereafter, suburban forests were considered oligolectic. We separated bees that
and natural forests, respectively). We paired sites and nest in existing cavities in wood as cavity instead of
located each pair at least 5 km apart. Natural forests wood nesters. Lumping cavity and wood-nesting spe-
consisted of contiguous tracts of mixed pine and hard- cies had no effect on the interpretation of results. We
April 2014 CARPER ET AL.: EFFECTS OF SUBURBANIZATION ON FOREST BEE COMMUNITIES 255

also classiÞed bee species according to body size by interactions, and site as a random effect. We used the
measuring intertegular span for up to three females combined bee abundance across both years as re-
per species and calculating body mass through a sponse variables, specifying a negative binomial dis-
known relationship between intertegular span and dry tribution to reduce overdispersion. A signiÞcant in-
body mass (Cane 1987). We classiÞed species as small teraction between forest type and nesting habit, lecty,
(⬍4 mg), medium (4 Ð16 mg), or large (⬎16 mg; Win- sociality, or body size would suggest that the effects of
free et al. 2007), reßecting the quantiles of the size suburbanization on bee abundance are driven by
distribution. groups of species with particular ecological or life-
Data Analysis. All statistical analyses (throughout history characteristics.
the study) were conducted using SAS version 9.2 (SAS Factors Driving Bee Abundance and Composition.
Institute 2009) unless otherwise noted. The number of At the local scale we estimated both foraging and
bees captured using pan traps and hand-netting was nesting resources. We conducted ßower censuses
combined within each site for each sampling date. To over the entire area sampled for bees in each site,
estimate bee abundance, we calculated the mean corresponding to each bee sampling event, in 2008 and

Downloaded from https://academic.oup.com/ee/article/43/2/253/526089 by guest on 02 May 2023


number of bees per sampling event per site for each 2009. We recorded the total number of individuals of
year. Analyzing total bees across sampling events per each ßowering species and counted the number of
site yielded similar results (data not shown). To esti- ßowers on a subset of 10 haphazardly chosen individ-
mate bee diversity, we used species richness (the uals per species. For plants in the Asteraceae, we
combined number of species captured across all sam- recorded the number of ßowering heads containing
pling events at each site across both years) and Simp- open disc ßowers on each plant. On large shrubs and
sonÕs Evenness index. To account for differences in trees with ⬎500 ßowers, we counted the number of
bee abundance on estimates of species richness, we ßowers on three separate branches and multiplied the
rareÞed samples to the lowest common abundance average number of ßowers per branch by the total
using ECOSIM software (Gotelli and Entsminger number of ßowering branches to estimate the total
2001). To account for the paired approach, we used number of ßowers per plant. We then multiplied the
one-tailed paired t-tests (JMP version 8.0.2, SAS In- mean number of ßowers per plant by the total number
stitute) to test whether suburban forests had higher of individuals to estimate the total number of ßowers
bee abundance and species richness than natural for- per species per plot, and calculated ßower density by
ests. We had no a priori prediction about the effects dividing by the area of each plot.
of suburbanization on species evenness, so we used a Nesting resources may also be important in driving
two-tailed paired t-test to assess whether suburban- bee abundance and diversity (Cane et al. 2006). To
ization affected evenness. estimate wood-nesting resources, we used three 20-m
To compare species composition between suburban line-intercept transects in each site to quantify coarse
and natural forests, we used a series of approaches. woody debris (CWD) in 2009. We recorded the length
First, we calculated MorisitaÕs C and percentage sim- and width of dead wood ⱖ7.5 cm in diameter inter-
ilarity for each site pair and for pooled captures across secting each transect (Woodall and Monleon 2008).
the Þve suburban and Þve natural sites. We used We then calculated the volume of CWD per square
MorisitaÕs C because it is less inßuenced by sample size meter using a standard formula, V ⫽ ␲2兺d2/8L, where
and species richness than other similarity indices V is the volume of woody debris per unit area, d is the
(Wolda 1981). Second, we tested for spatial autocor- diameter of each piece of wood, and L is the length of
relation in bee abundance and species richness be- the transect (Harmon et al. 2004). We observed no
tween sites by computing MoranÕs I. Third, we used bare ground along any transect and thus were unable
nonmetric multidimensional scaling (NMDS), via the to quantify it as substrate for soil nesting.
“vegan” package (Oksanen et al. 2012) in R version Lastly, we incorporated measures of other abiotic
2.14, to visualize any dissimilarity in bee community factors that could directly or indirectly affect bee
composition between suburban and forested sites. We activity or abundance. Changes in climate associated
conducted the NMDS using the Bray-Curtis dissimi- with urban development have been shown to alter
larity measure, which takes into account the number animal distributions (Parris and Hazell 2005). Differ-
of unique species between two sites relative to the ences in temperature or relative humidity (RH) could
total number of species sampled. The relative Þt of the affect bee activity and the likelihood of capture, or the
NMDS was evaluated by computing the Þnal stress of distribution or phenology of suitable host plants. To
the ordination. Stress is a measure of mismatch be- characterize temperature and RH, two iButton data
tween the rankÐ order dissimilarities and the ordina- loggers (Embedded Data Systems LLC, Lawrence-
tion distance, often referred to as a badness-of-Þt test, burg, KY) were haphazardly placed in each site to
with values ⬍0.05 indicating strong support for the record temperature and RH hourly over the ßowering
ordination distances. season in 2009. iButtons were mounted beneath low-
To determine whether the effects of suburbaniza- lying branches and shielded with a small plastic rain
tion varied with bee ecology or life-history, we used guard.
generalized linear models (PROC GLIMMIX). We At the landscape scale, we used data from the 2006
included forest type (suburban or natural forest) National Land Cover Database (Fry et al. 2011) to
along with nesting habit, lecty, sociality, and body size explore how land-use cover affected bee abundance.
as Þxed factors in four separate analyses, including The National Land Cover Database uses satellite im-
256 ENVIRONMENTAL ENTOMOLOGY Vol. 43, no. 2

Fig. 1. Suburban forests exhibited greater bee abundance (a) and observed species richness (b) than natural forests in
2008 and 2009, but no difference in rareÞed species richness (c). Bars are means ⫾ SE. Asterisks indicate signiÞcant differences
(P ⬍ 0.05).

Downloaded from https://academic.oup.com/ee/article/43/2/253/526089 by guest on 02 May 2023


agery to characterize land cover at a 30-m resolution. poeyi Lepeletier. Each of these two groupings was
We calculated the proportions of surrounding land treated as a single species in subsequent analyses. The
covers within a 1-km radius of the center of each site two most dominant bee species, Augochlorella aurata
using ArcGis (version 9.3.1, ESRI, Redlands, CA) and Smith (Halictidae) and Lasioglossum bruneri Craw-
identiÞed the four most dominant land covers: ever- ford (Halictidae), represented 17 and 14% of all bees
green (pine) forest, deciduous forest, developed open collected, respectively. Exotic bee species were rare
space (⬍20% impervious surface), and low-intensity and made up ⬍3% of the total number of captures. We
developed (20 Ð 49% impervious surface). We chose 1 captured eight honeybees (Apis mellifera L.) in sub-
km because it likely encompasses the foraging dis- urban forests and two in natural forests. We also cap-
tance of most of the species sampled while leaving tured two giant resin bees (Megachile sculpturalis
little overlap within site pairs. Smith) and Þve alfalfa leafcutting bees (Megachile
Data Analyses. At the local scale, we compared rotundata F.) in suburban forests. In addition, we
ßower species richness, ßoral density, and CWD be- captured 38 species that were each represented by a
tween forest types in each year using one-tailed paired single specimen, including 28 from suburban and 10
t-tests to test the hypothesis that suburban forests have from natural forests (Supp. Table 2 [online only]).
greater ßoral resources but fewer nesting resources Bee Abundance, Richness, and Composition. Sub-
compared with natural forests. We also used linear urban forests had 2.2 times more bees captured per
regression to determine whether ßoral density and sampling event than natural forests in 2008 (t4 ⫽ 1.99,
richness predicted bee abundance and richness. We P ⫽ 0.059) and 1.8 times more bees in 2009 (t4 ⫽ 2.88,
calculated mean, minimum, and maximum daily tem- P ⫽ 0.022, Fig. 1a). In suburban forests, bee richness
perature, as well as RH, averaging values from the two ranged from 7 to 19 species compared with natural
iButtons in each site. We compared each metric of
forests, where bee richness ranged from 6 to 16 spe-
temperature and RH between forest types using sep-
cies. Although observed species richness was signiÞ-
arate repeated measures analyses of variance.
cantly higher in suburban forests in both years (2008:
At the landscape scale, we compared land cover
t4 ⫽ 2.87, P ⫽ 0.027; 2009: t4 ⫽ 4.78, P ⫽ 0.004; Fig. 1b),
types between suburban and natural forests using mul-
rarefaction yielded no signiÞcant difference in species
tivariate analysis of variance, with site pair and forest
type as Þxed factors and the proportions of the four richness between suburban and natural forests in ei-
dominant land covers as responses. We used multiple ther year (2008: t4 ⫽ ⫺1.45, P ⫽ 0.09; 2009: t4 ⫽ ⫺0.52,
regressions (JMP version 8.0.2, SAS Institute) to de- P ⫽ 0.31; Fig. 1c). We also detected no spatial auto-
termine whether land cover predicted bee abundance correlation in either year (MoranÕs I ⫽ ⫺0.11, P ⫽
and species richness. Each proportion of land cover 0.808, MoranÕs I ⫽ ⫺0.12, P ⫽ 0.420, respectively).
was log-transformed to meet the assumptions of ho- Percentage similarity between paired suburban and
moscedasticity, and multicollinearity was assessed natural forests ranged from 21.8 to 45.2%. Pooled
from the variance inßation factors. across site pairs and years, suburban and natural for-
ests were characterized by 59.3% similarity (MorisitaÕs
C ⫽ 0.85). The most abundant genus sampled was
Results Lasioglossum, with 19 total species, 7 of which were
In 2008, we captured 258 bees (mean: 86 bees per unique to suburban forests, compared with only 3
sampling event, pooled across sites) and in 2009, 365 species unique to natural forests. Suburban forests also
bees (mean: 61 bees per sampling event), belonging to yielded nearly three times as many singletons (species
six families, 26 genera, and 71 species or morphospe- represented by a single captured individual) as natural
cies. All but Þve individuals were identiÞed to species. forests (28 vs. 10 species, respectively). However,
One Lasioglossum sp. male, one Ceratina sp., and three there was no signiÞcant difference in SimpsonÕs Even-
Nomada sp. males were identiÞed as morphospecies. ness between suburban and natural forests (t4 ⫽ 1.32,
In addition, we combined individuals in two taxo- P ⫽ 0.258). NMDS indicated overlap in bee compo-
nomic groupings currently being revised: 11 Hylaeus sition between suburban and natural forests, driven
affinis Smith/modestus Say and 19 Halictus ligatus Say/ mainly by one natural forest site (Fig. 2). The Þnal
April 2014 CARPER ET AL.: EFFECTS OF SUBURBANIZATION ON FOREST BEE COMMUNITIES 257

cant. Bee abundance varied by sociality, with eusocial


bees representing 65% of bees sampled, followed by
solitary bees (21%), subsocial (9%), and parasitic bees
(5%). Suburban forests had more bees in each sociality
category compared with natural forests (Fig. 3b). Me-
dium-sized bees were the most common bees cap-
tured, representing 51% of individuals, followed by
small and large bees making up 36 and 13%, respec-
tively. Suburban forests had more bees in each size
class compared with natural forests (Fig. 3c). Finally,
soil-nesting bees were the most abundant, comprising
⬇70% of individuals, with cavity nesters making up
11%, hive and stem nesters each making up ⬇9%, and
wood nesters making up ⬍1% (Fig. 3d).

Downloaded from https://academic.oup.com/ee/article/43/2/253/526089 by guest on 02 May 2023


Factors Driving Bee Abundance and Composition.
Bee communities were affected by both local and
landscape scale factors. At the local scale, ßoral re-
sources were the strongest factor affecting bees. Flow-
ering tree species were the dominant ßoral resource in
suburban and natural forests. Oxydendrum arboreum
Fig. 2. NMDS indicated overlap in bee communities be- (L.) (Ericaceae) was the most abundant ßower re-
tween suburban and natural forests driven primarily by one source within sampling plots, constituting 47% of all
natural forest, Lake Crabtree County Park, overlapping with ßowers surveyed, followed by Ilex opaca (Alton)
suburban forests. (Aquifoliaceae) at 9% of ßowers. Aside from trees, the
three most abundant ßowering species were Erigeron
stress of the NMDS with two dimensions was 0.11, annuus (L.) (Asteraceae), Lonicera japonica (Thunb.)
indicating fair conÞdence in the ordination distances. (Caprifoliaceae), and Rubus argutus (Link) (Rosa-
Bee abundance varied as a function of ecological ceae), making up 8, 4, and 3% of surveyed ßowers,
and life-history traits and suburbanization; however, respectively. The remaining 68 species recorded made
there were no signiÞcant interactions between any of up ⬍2% of total ßower abundance each. Suburban
the ecological traits considered and suburbanization forests averaged 3.4 ⫾ 2.8 (mean ⫾ SE) and 6.2 ⫾ 3.3
(Table 1), suggesting that higher bee abundance in more ßowering species than natural forests in 2008 and
suburban sites was not driven by bees with particular 2009, respectively, but these differences were not sta-
ecological or life-history characters. Generalist bees tistically signiÞcant in 2008 (t4 ⫽ ⫺1.22, P ⫽ 0.143) or
were more abundant than specialist bees, with 2009 (t4 ⫽ ⫺1.87, P ⫽ 0.067). Suburban forests also had
polylectic species making up ⬎98% of bees captured, 14.3% greater density of ßowers than natural forests in
compared with ⬍2% that were oligolectic. Although 2008 (1.09 ⫾ 0.57 vs. 1.25 ⫾ 0.46 ßowers per square
suburban forests had 2.3 times more polylectic bees meter). However, the opposite pattern occurred in 2009
and 1.98 times more oligolectic bees (Fig. 3a), the when natural forests had 13.8% more ßowers per square
effect of suburbanization was not statistically signiÞ- meter than suburban sites (0.95 ⫾ 0.61 vs. 0.83 ⫾ 0.16
ßowers per square meter), but in neither year were these
differences in ßoral density statistically signiÞcant (2008:
Table 1. Generalized linear models of bee abundance by four t4 ⫽ ⫺0.26, P ⫽ 0.403; 2009: t4 ⫽ 0.17, P ⫽ 0.436). There
ecological characteristics (lecty, sociality, body size, and nesting was no relationship between the total number of ßowers
substrate) and forest type (suburban vs. natural)
and total bee abundance in either year (2008: r2 ⫽ 0.23,
Ecological trait df F P P ⫽ 0.164; 2009: r2 ⫽ 0.002, P ⫽ 0.905). However, ob-
served bee richness and total bee abundance increased
Lecty
Lecty 1,12 122.03 ⬍0.001 signiÞcantly with ßower richness in suburban sites in
Forest type 1,12 4.15 0.0642 2008 (richness: r2 ⫽ 0.61, P ⬍ 0.005; abundance: r2 ⫽ 0.65,
Lecty*forest type 1,12 0.05 0.8281 P ⫽ 0.005; Fig. 4a and b), though neither relationship
Sociality
Sociality 3,28 33.87 ⬍0.001
remained signiÞcant in 2009 (richness: r2 ⫽ 0.091, P ⫽
Forest type 1,28 17.57 ⬍0.001 0.396; abundance: r2 ⫽ 0.07, P ⫽ 0.46).
Sociality*forest type 3,28 1.96 0.143 The total volume of CWD ranged from 0.011 to
Body size 0.041 m3/m2 in natural forests and 0.006 Ð 0.021 m3/m2
Body size 2,20 26.92 ⬍0.001
Forest type 1,20 25.03 ⬍0.001 in suburban forests; however, natural forests showed
Body size*forest type 1,20 1.7 0.208 no signiÞcant difference in the volume of CWD from
Nesting substrate suburban forests (t4 ⫽ 0.80, P ⫽ 0.24). Given the low
Nesting 4,36 57.92 ⬍0.001 representation of wood-nesting species in our samples
Forest type 1,36 8.12 0.007
Nesting*forest type 4,36 0.99 0.424 (⬍1% of total bee captures), we did not further pursue
CWD as a mechanism driving bee abundance or rich-
*
represents an interaction between the main effects. ness. Neither mean, maximum, or minimum daily tem-
258 ENVIRONMENTAL ENTOMOLOGY Vol. 43, no. 2

Downloaded from https://academic.oup.com/ee/article/43/2/253/526089 by guest on 02 May 2023


Fig. 3. There were no signiÞcant interactions between forest type and ecological categories studied. Compared with
natural forests, suburban forests supported a signiÞcantly greater abundance of bees across all foraging habits (a), socialities
(b), body sizes (c), and nesting types (d). Bars are means ⫾ SE.

peratures nor RHs signiÞcantly differed between for- (WilkÕs ␭ ⫽ 0.198, F4,5 ⫽ 5.07, P ⫽ 0.052). The proportion
est types in 2009 (in all cases: F1,8 ⬍ 1.6, P ⬎ 0.24). of developed open space in the surrounding landscape
At the landscape scale, land cover varied across sites, was signiÞcantly and positively associated with both bee
with deciduous and evergreen forests, developed open abundance and species richness (Table 2). However,
space, and hay or pasture the most dominant land covers bee abundance and richness were negatively associated
(Fig. 5). Suburban forests were surrounded by four with the proportion of low-intensity development in the
times more low intensity development and nearly Þve landscape, although the relationship was only statistically
times more developed open space as natural forests signiÞcant for bee abundance (Table 2). All variance

Fig. 4. Flower richness within suburban forests in 2008 was a signiÞcant predictor of suburban bee abundance (a) and
species richness (b), though no relationship existed in natural forests.
April 2014 CARPER ET AL.: EFFECTS OF SUBURBANIZATION ON FOREST BEE COMMUNITIES 259

Downloaded from https://academic.oup.com/ee/article/43/2/253/526089 by guest on 02 May 2023


Fig. 5. The dominant land cover types surrounding suburban and natural forests were calculated using a 1-km buffer
around each site center. Land cover types in the “Other” category include cultivated crops, developed high intensity,
shrub/scrub, barren land, emergent herbaceous, and wetlands.

inßation factors were small (⬍5), suggesting multicol- port a diversity of native bees (Matteson et al. 2008,
linearity had little effect on results. Banaszak-Cibicka and Zmihorski 2012). Although we
found that observed species richness was higher in
Discussion suburban forests across both years, the lack of differ-
ence between rareÞed species richness suggests that
Suburbanization can have powerful effects on the low richness in natural forest sites could be a result of
abundance and richness of both plants and animals low bee abundance and reduced likelihood of capture.
(Shochat et al. 2006). Here, we report that forests Nonetheless, the Þnding that suburban areas harbor
within suburban development in a metropolitan area more abundant and at least as rich bee communities as
of the southeastern United States have higher bee nearby natural areas is consistent with some studies.
abundance than natural forests. Moreover, suburban For example, Winfree et al. (2007) found that both bee
forests had more bees of all ecological and life-history abundance and species richness were higher in sub-
characteristics that we sampled. At the local scale, bee urban and urban development than in extensive for-
abundance and richness were positively associated ests in New Jersey. However, other studies comparing
with ßower abundance and richness, though the re- bee richness in desert and residential areas have
lationship varied between years. At the landscape found higher richness in the natural desert habitat
level, bee abundance and richness were positively (Hostetler and McIntyre 2001, Gotlieb et al. 2011).
associated with the amount of open space, and neg- Thus, while some aspects of residential areas may
atively associated with low-intensity development. beneÞt bees, the overall landscape context is likely
Taken together, these results suggest that the combi- important in driving species richness patterns (Ahrne
nation of open habitats and availability of diverse ßoral
et al. 2009).
resources in suburban sites are capable of sustaining
Contrary to our predictions, bee community com-
abundant and diverse bee communities.
position exhibited little difference between suburban
These results add to a growing body of literature
and natural forests. We found no interactions between
demonstrating that urban and suburban areas can sup-
forest type and ecological or life-history categories on
bee abundance, implying that all types of bees in our
Table 2. Least squares multiple regressions of bee abundance study system responded similarly to suburbanization.
and observed species richness against the proportion of four dom- This result contradicts empirical work that demon-
inant land-cover types surrounding each of 10 study sites
strated guild-speciÞc responses of bees to disturbance
Land-cover type ␤ SE t P (Williams et al. 2010). Without more precise mea-
surements of characteristics, such as the distribution
Abundance
Log (deciduous forest) 11.07 12.92 0.86 0.431 and availability of nesting substrates, it is difÞcult to
Log (evergreen forest) ⫺13.59 17.27 ⫺0.79 0.467 speculate why no differences were observed. One
Log (developed, low intensity) ⫺21.33 7.85 ⫺2.72 0.042 explanation is that bees captured in the forest are
Log (developed, open space) 44.30 12.02 3.69 0.014
Species richness
primarily moving between widely distributed re-
Log (deciduous forest) 5.53 2.32 2.38 0.063 sources in the landscape, and the scale of our sampling
Log (evergreen forest) ⫺2.58 3.10 ⫺0.83 0.444 did not adequately reßect the scale at which differ-
Log (developed, low intensity) ⫺3.11 1.41 ⫺2.21 0.079 ences in community composition are evident. The 71
Log (developed, open space) 7.30 2.16 3.39 0.020
species captured are also only a subset of the regional
260 ENVIRONMENTAL ENTOMOLOGY Vol. 43, no. 2

bee fauna, and our sampling periods may have missed resources or at least a superior combination of ßower
some species, such as early and late season specialists. and nesting resources. We observed no bare ground in
Regardless, suburban bee community composition ap- our surveys, which was not surprising given that sam-
pears similar to that of natural forests, only more abun- pling took place under an extensive forest canopy.
dant. This pattern was reinforced by overlap in NMDS, Still, a greater abundance of all bees in suburban sites,
which suggests that the bee community composition including soil-nesting species, indicates that suitable
was fairly similar across forest types. However, it is nest sites may be more abundant in suburban areas,
important to note that the NMDS overlap was driven despite there being more extensive impervious surface
primarily by one natural forest, Crabtree County Park. cover in the form of low-intensity development. Win-
Interestingly, this was the only natural forest site that free et al. (2007) found a positive association between
actively planted ßowers along roadsides and park wood-nesting species and the extent of forested hab-
structures (A.C., unpublished data), and thus had itat in the landscape. Contrary to our predictions,
much higher ßoral density than the other naturally CWD did not differ between forest types even though
forested sites. If such planted resources are important wood-nesting bees were more abundant in suburban
forests; however, these made up ⬍5% of the total bees

Downloaded from https://academic.oup.com/ee/article/43/2/253/526089 by guest on 02 May 2023


factors structuring bee communities, they could ex-
plain the potential overlap. Further work on the role sampled.
of planted ßoral resources is needed to determine Studies considering mechanisms driving bee abun-
whether they can affect forest bee communities. dance at landscape scales often Þnd that bee abun-
The abundance and diversity of ßoral resources are dance and richness are positively associated with the
dominant factors structuring bee communities (Potts amount of “natural” habitat in the surrounding land-
et al. 2003), and ßoral resources within urban areas can scape, which is assumed to reßect higher quality hab-
support a variety of bee species (Tommasi et al. 2004). itat (Hines and Hendrix 2005). In urban areas, green
The relationships between bees and ßoral resources in spaces have been shown to increase bee abundance
our sites were highly variable between sites and years. and diversity (Tonietto et al. 2011). At the landscape
At the local scale, bee abundance and richness were scale in our study, land cover surrounding our sites did
more strongly associated with ßower richness than predict bee abundance and diversity. The proportion
ßower abundance in 2008. Floral richness has been of developed open space was a strong, positive pre-
experimentally demonstrated to affect bee abundance dictor of both bee abundance and diversity, indicating
(Ebeling et al. 2008) and can be important in struc- that areas dominated by lawns and human structures,
turing pollinator communities in other systems such as single-family housing developments, can har-
(Moeller 2005). The lack of relationship between ei- bor bees. There were no signiÞcant effects of either
ther bee abundance or richness and ßower richness in deciduous or evergreen forest on bee abundance, al-
2009 could be because of unmeasured factors obscur- though both represent natural land covers. One ex-
ing any potential relationship, or be indicative of the planation for these results is that open space may
stochastic nature of ßoral resources. Additionally, the reßect the distribution of ßoral and nesting resources
scale of ßoral sampling may not have been adequate to outside of forested patches and serve as important
accurately depict the availability of ßoral resources for foraging and nesting areas. For example, bumblebee
bees, which are very mobile foragers. More extensive abundance in urban parks of San Francisco, CA was
ßoral sampling from roadside surveys in 2011, includ- positively associated with the openness of the sur-
ing yards, gardens, and other green spaces, indicated rounding urban matrix, suggesting open areas either
that suburban landscapes tended to have more abun- have more resources or are more easily navigated by
dant and diverse ßoral resources than forested land- foraging bees (McFrederick and LeBuhn 2006). Low-
scapes, which was not evident at local scales within intensity development had a negative effect on abun-
forest patches (A.C., data not shown). Unfortunately, dance and diversity in our study, which was not sur-
roadside surveys did not coincide with bee sampling, prising given that it has up to 50% impervious surface,
so we were unable to test for relationships between and thus may provide low foraging and nesting re-
landscape ßoral resources and local bee abundance sources for bees. The differences between bee re-
and diversity. One caveat is that the distribution of sponse to developed open space and low-intensity
nectar and pollen resources may be a better predictor development suggest that any beneÞts of human de-
of bee abundance and richness than ßoral abundance velopment decrease with increasing intensity of ur-
and richness (Potts et al. 2004). Measuring nectar and banization. In addition, the spatial scale of surveys and
pollen resources was beyond the scope of this re- landscape context likely determine the effects of ur-
search, but given the importance of conserving bee banization on bee communities (Steffan-Dewenter et
communities in urban and suburban areas and recent al. 2002). Forested landscapes, such as in this study,
interest in pollinator-friendly gardens, more studies could beneÞt from moderate levels of disturbance
are needed that assess community-wide nectar and (Winfree et al. 2007), whereas open habitats, such as
pollen resources in suburban versus natural habitats. deserts (Hostetler and McIntyre 2001, Gotlieb et al.
The abundance and distribution of nesting re- 2011), typically lose bees to disturbance through hab-
sources also play key roles in the distribution of var- itat loss. Historical land use could also be an important
ious nesting bees (Potts et al. 2005). Overall, suburban factor effecting contemporary bee communities.
areas supported more bees of all nesting categories, Comparative studies using museum records could pro-
suggesting that such areas either have more nesting vide context for temporal change in both land use and
April 2014 CARPER ET AL.: EFFECTS OF SUBURBANIZATION ON FOREST BEE COMMUNITIES 261

bee community composition, but was outside the pollinator assemblages along an urban-rural gradient.
scope of this study. Therefore, more detailed knowl- PLoS ONE 6: e23459.
edge of the endemic bee fauna, historical land use, and Biesmeijer, J. C., S.P.M. Roberts, M. Reemer, R. Ohlemuller,
land-use practices is crucial to interpreting the effects M. Edwards, T. Peeters, A. P. Schaffers, S. G. Potts, R.
of urban development on bee communities. Kleukers, C. D. Thomas, J. Settele, and W. E. Kunin.
In conclusion, this work suggests that moderate 2006. Parallel declines in pollinators and insect-polli-
nated plants in Britain and The Netherlands. Science 313:
levels of human disturbance, such as suburban devel-
351Ð354.
opment, can support abundant and diverse bee com- Brown, D. G., K. M. Johnson, T. R. Loveland, and D. M.
munities. With increasing concern over the status of Theobald. 2005. Rural land-use trends in the contermi-
pollinators and the sustainability of pollination ser- nous United States, 1950 Ð2000. Ecol. Appl. 15: 1851Ð1863.
vices (Kearns et al. 1998), such Þndings underscore Burkle, L. A., J. C. Marlin, and T. M. Knight. 2013. Plant-
the need to incorporate ecological Þndings in both pollinator interactions over 120 years: loss of species,
urban and conservation planning (Miller and Hobbs co-occurrence and function. Science 339: 1611Ð1615.
2002). Given the challenges of urban and suburban Cameron, S. A., J. D. Lozier, J. P. Strange, J. B. Koch, N.
Cordes, L. F. Solter, and T. L. Griswold. 2011. Patterns

Downloaded from https://academic.oup.com/ee/article/43/2/253/526089 by guest on 02 May 2023


areas for bees (Cane 2005), insights from such work
could have broad implications for the conservation of of widespread decline in North American bumble bees.
native pollinators and the sustainability of pollination Proc. Natl. Acad. Sci. USA 108: 662Ð 667.
services in human-dominated environments. For ex- Cane, J. 2005. Bees, pollination, and the challenges of
sprawl, pp. 109 Ð124. In E. Johnson and M. Klemens (eds.),
ample, planned development in forested areas could
Nature in Fragments: The Legacy of Sprawl. Columbia
promote abundant and diverse bee communities University Press, New York.
through the establishment of ßower-rich open areas in Cane, J. H. 1987. Estimation of bee size using intertegular
concordance with the retention of natural forest hab- span (Apoidea). J. Kans. Entomol. Soc. 60: 145Ð147.
itat. Ultimately, comparative studies integrating both Cane, J. H., R. L. Minckley, L. J. Kervin, T. H. Roulston, and
planned development and conservation strategies are N. M. Williams. 2006. Complex responses within a des-
needed to determine how to best manage for and ert bee guild (Hymenoptera: Apiformes) to urban habitat
conserve native pollinators. fragmentation. Ecol. Appl. 16: 632Ð 644.
Connell, J. H. 1978. Diversity in tropical rain forests and
coral reefs - high diversity of trees and corals is main-
Acknowledgments tained only in a non-equilibrium state. Science 199: 1302Ð
1310.
This research was made possible by cooperation from Ebeling, A., A. Klein, J. Schumacher, W. W. Weisser, and T.
landowners, homeownersÕ associations, private citizens, the Tscharntke. 2008. How does plant richness affect polli-
North Carolina Division of Parks and Recreation, the Wake nator richness and temporal stability of ßower visits?
County Department of Parks, Recreation and Open Space, Oikos 117: 1808 Ð1815.
and North Carolina State University. We thank S. Droege for Eigenbrod, F., V. A. Bell, H. N. Davies, A. Heinemeyer, P. R.
assistance in bee identiÞcation; J. Ascher for help with bee Armsworth, and K. J. Gaston. 2011. The impact of pro-
biology and life-history; and D. Bolger, Z. Gezon, R. Schaef- jected increases in urbanization on ecosystem services.
fer, L. Richardson, and C. Urbanowicz for comments on the Proc. R. Soc. B. 278: 3201Ð3208.
manuscript. Funding was provided by the National Science Fetridge, E. D., J. S. Ascher, and G. A. Langellotto. 2008. The
Foundation (DEB-0743535) and Dartmouth College (the R. bee fauna of residential gardens in a suburb of New York
Melville Cramer Fund and a Rockefeller Center Urban Stud- city (Hymenoptera: Apoidea). Ann. Entomol. Soc. Am.
ies Research Grant). Any opinions, Þndings, and conclusions 101: 1067Ð1077.
or recommendations expressed in this material are those of Foley, J. A., R. DeFries, G. P. Asner, C. Barford, G. Bonan,
the authors and do not necessarily reßect the views of the S. R. Carpenter, F. S. Chapin, M. T. Coe, G. C. Daily, H. K.
funding sources. Gibbs, et al. 2005. Global consequences of land use. Sci-
ence 309: 570 Ð574.
Fry, J., G. Xian, S. Jin, J. Dewitz, C. Homer, L. Yang, C.
References Cited Barnes, N. Herold, and J. Wickham. 2011. Completion
Ahrne, K., J. Bengtsson, and T. Elmqvist. 2009. Bumble bees of the 2006 National Land Cover Database for the Con-
(Bombus spp) along a gradient of increasing urbanization. terminous United States. Photogramm. Eng. Remote
PLoS ONE 4: e5574. Sens. 77: 858 Ð 864.
Ascher, J. S., and J. Pickering. 2012. Discover Life bee species Gotelli, N. J., and G. L. Entsminger. 2001. EcoSim: null models
guide and world checklist (Hymenoptera: Apoidea: Antho- software for ecology, Version 7.0. Acquired Intelligence Inc.
phila). (http://www.discoverlife.org/mp/20q?guide⫽ & Kesey-Bear, Jericho, VT. (http://homepages.together.
Apoidea_species). net/⬃gentsmin/ecosim.htm).
Banaszak-Cibicka, W., and M. Zmihorski. 2012. Wild bees Gotlieb, A., Y. Hollender, and Y. Mandelik. 2011. Gardening
along an urban gradient: winners and losers. J. Insect in the desert changes bee communities and pollination
Conserv. 16: 331Ð343. network characteristics. Basic Appl. Ecol. 12: 310 Ð320.
Bartomeus, I., J. S. Ascher, J. Gibbs, B. N. Danforth, D. L. Goulson, D., O. Lepais, S. O’Connor, J. L. Osborne, R. A.
Wagner, S. M. Hedtke, and R. Winfree. 2013. Historical Sanderson, J. Cussans, L. Goffe, and B. Darvill. 2010.
changes in northeastern US bee pollinators related to Effects of land use at a landscape scale on bumblebee nest
shared ecological traits. Proc. Natl. Acad. Sci. USA 12: density and survival. J. Appl. Ecol. 47: 1207Ð1215.
4656 Ð 4660. Harmon, M. E., J. F. Franklin, F. J. Swanson, P. Sollins, S. V.
Bates, A. J., J. P. Sadler, A. J. Fairbrass, S. J. Falk, J. D. Hale, Gregory, J. D. Lattin, N. H. Anderson, S. P. Cline, N. G.
and T. J. Matthews. 2011. Changing bee and hoverßy Aumen, J. R. Sedell, et al. 2004. Ecology of coarse woody
262 ENVIRONMENTAL ENTOMOLOGY Vol. 43, no. 2

debris in temperate ecosystems. Adv. Ecol. Res. 34: 59 Ð communities structure pollinator communities? Ecology
234. 84: 2628 Ð2642.
Hernandez, J. L., G. W. Frankie, and R. W. Thorp. 2009. Potts, S. G., B. Vulliamy, S. Roberts, C. O’Toole, A. Dafni, G.
Ecology of urban bees: a review of current knowledge Ne’eman, and P. G. Willmer. 2004. Nectar resource di-
and directions for future study. Cities Environ. 2: 1Ð15. versity organises ßower-visitor community structure. En-
Hines, H. M., and S. D. Hendrix. 2005. Bumble bee (Hy- tomol. Exp. Appl. 113: 103Ð107.
menoptera: Apidae) diversity and abundance in tallgrass Potts, S. G., B. Vulliamy, S. Roberts, C. O’Toole, A. Dafni, G.
prairie patches: effects of local and landscape ßoral re- Ne’eman, and P. Willmer. 2005. Role of nesting re-
sources. Environ. Entomol. 34: 1477Ð1484. sources in organising diverse bee communities in a Med-
Hostetler, N. E., and M. E. McIntyre. 2001. Effects of urban iterranean landscape. Ecol. Entomol. 30: 78 Ð 85.
land use on pollinator (Hymenoptera: Apoidea) commu- SAS Institute. 2009. SAS/STAT 9.2 userÕs guide, 2nd ed. SAS
nities in a desert metropolis. Basic Appl. Ecol. 2: 209 Ð218. Institute, Cary, NC.
Irwin, E. G., and N. E. Bockstael. 2007. The evolution of Shochat, E., P. S. Warren, S. H. Faeth, N. E. McIntyre, and
urban sprawl: evidence of spatial heterogeneity and in- D. Hope. 2006. From patterns to emerging processes in
creasing land fragmentation. Proc. Natl. Acad. Sci. USA mechanistic urban ecology. Trends Ecol. Evol. 21: 186 Ð
104: 20672Ð20677. 191.

Downloaded from https://academic.oup.com/ee/article/43/2/253/526089 by guest on 02 May 2023


Kalnay, E., and M. Cai. 2003. Impact of urbanization and Shochat, E., S. B. Lerman, J. M. Anderies, P. S. Warren, S. H.
land-use change on climate. Nature 423: 528 Ð531. Faeth, and C. H. Nilon. 2010. Invasion, competition, and
Kearns, C. A., D. W. Inouye, and N. M. Waser. 1998. En- biodiversity loss in urban ecosystems. Bioscience 60: 199 Ð
dangered mutualisms: the conservation of plant-pollina- 208.
tor interactions. Annu. Rev. Ecol. Syst. 29: 83Ð112. Steffan-Dewenter, I., U. Munzenberg, C. Burger, C. Thies,
LeBuhn, G., T. Griswold, R. Minckley, S. Droege, T. Roulston, and T. Tscharntke. 2002. Scale-dependent effects of
landscape context on three pollinator guilds. Ecology 83:
J. Cane, F. Parker, S. Buchmann, V. Tepedino, N. Williams,
1421Ð1432.
C. Kremen, and O. J. Messinger. 2003. A standardized
Tommasi, D., A. Miro, H. A. Higo, and M. L. Winston. 2004.
method of monitoring bee populationsÑThe Bee Inventory
Bee diversity and abundance in an urban setting. Can.
(BI) Plot. (http://online.sfsu.edu/beeplot/).
Entomol. 136: 851Ð 869.
Matteson, K. C., J. S. Ascher, and G. A. Langellotto. 2008.
Tonietto, R., J. Fant, J. Ascher, K. Ellis, and D. Larkin. 2011.
Bee richness and abundance in New York City urban
A comparison of bee communities of Chicago green roofs,
gardens. Ann. Entomol. Soc. Am. 101: 140 Ð150. parks and prairies. Landscape Urban Plan. 103: 102Ð108.
McFrederick, Q. S., and G. LeBuhn. 2006. Are urban parks U.S. Census Bureau. 2009. Estimates of resident population
refuges for bumble bees Bombus spp. (Hymenoptera: change for incorporated places over 100,000, ranked by
Apidae)? Biol. Conserv. 129: 372Ð382. percent change: July 1, 2008 to July 1, 2009 (SUB-EST2009-
Miller, J. R., and R. J. Hobbs. 2002. Conservation where 03). (http://www.census.gov/popest/cities/cities.html).
people live and work. Conserv. Biol. 16: 330 Ð337. Williams, N. M., E. E. Crone, T. H. Roulston, R. L. Minckley,
Moeller, D. 2005. Pollinator community structure and L. Packer, and S. G. Potts. 2010. Ecological and life-
sources of spatial variation in plantÐpollinator interac- history traits predict bee species responses to environ-
tions in Clarkia xantiana ssp. xantiana. Oecologia 142: mental disturbances. Biol. Conserv. 143: 2280 Ð2291.
28 Ð37. Winfree, R., T. Griswold, and C. Kremen. 2007. Effect of
Oksanen, J., F. G. Blanchet, R. Kindt, P. Legendre, P. R. human disturbance on bee communities in a forested
Minchin, R. B. O’Hara, G. L. Simpson, P. Solymos, ecosystem. Conserv. Biol. 21: 213Ð223.
M.H.H. Stevens, and H. Wagner. 2012. Vegan: Commu- Winfree, R., R. Aguilar, D. P. Vazquez, G. LeBuhn, and M. A.
nity Ecology Package. R package version 2.0-4. (http:// Aizen. 2009. A meta-analysis of beesÕ responses to an-
CRAN.R-project.org/package⫽vegan). thropogenic disturbance. Ecology 90: 2068 Ð2076.
Ollerton, J., R. Winfree, and S. Tarrant. 2011. How many Wolda, H. 1981. Similarity indices, sample size and diver-
ßowering plants are pollinated by animals? Oikos 120: sity. Oecologia 50: 296 Ð302.
321Ð326. Woodall, C. W., and V. J. Monleon. 2008. Sampling proto-
Parris, K. M., and D. L. Hazell. 2005. Biotic effects of climate col, estimation, and analysis procedures for the down
change in urban environments: the case of the grey- woody materials indicator of the FIA program. U.S. De-
headed ßying-fox (Pteropus poliocephalus) in Melbourne, partment of Agriculture, Forest Service, Northern Re-
Australia. Biol. Conserv. 124: 267Ð276. search Station, Newtown Square, PA.
Potts, S. G., B. Vulliamy, A. Dafni, G. Ne’eman, and P.
Willmer. 2003. Linking bees and ßowers: how do ßoral Received 25 March 2013; accepted 18 November 2013.

You might also like