PDF Archive

Easily share your PDF documents with your contacts, on the Web and Social Networks.

Share a file Manage my documents Convert Recover PDF Search Help Contact



flowering.pdf


Preview of PDF document flowering.pdf

Page 1 2 3 4 5 6 7 8 9

Text preview


B E E A N D W A S P R I C H N E S S A N D M A S S - F L O W E R I N G C R O P S 221
Table 1 Percentage areas of oilseed rape and seminatural habitats in the landscape surrounding study sites for different radii
(scale: 250–2000 m). Given are mean values for eleven study
sites and minimum (min) and maximum (max) values
Area percentage
Oilseed rape (%)

Seminatural habitats (%)

Scale (m)

Mean

Min

Max

Mean

Min

Max

250
500
750
1000
1250
1500
1750
2000

14.33
10.08
9.11
8.80
8.89
8.53
7.92
7.74

0.00
1.79
2.61
3.57
3.73
3.53
4.03
3.96

33.63
22.19
15.78
14.50
15.93
15.02
14.67
14.78

12.43
8.46
7.61
7.35
7.07
7.19
20.28
20.72

0.90
1.32
1.26
1.65
2.21
2.38
7.91
9.10

44.75
28.36
22.70
16.17
13.71
13.43
39.29
39.53

habitats (P 0.107), except between the distance to the nearest
oilseed rape field and the percentage of oilseed rape at the two
smallest scales for which, however, no variables were retained
in any model.

Trap nests
Two trap nests fixed on a wooden post 150 cm above the
ground were set up at the edge of each selected study site in
March 2008, 1 week before the flowering of oilseed rape. Each
trap nest consisted of two plastic tubes (length: 30 cm; radius:
5 cm) filled with internodes of common reed Phragmites australis Cav. (diameter: 2–10 mm). At the end of May 2008, after the
mass flowering had ceased, internodes with nests were partly
extracted from the plastic tube, individually marked with a red
permanent marker on the outside of the internode, and repositioned to enable discrimination between nests built during and
after mass flowering. The trap nests were collected in October
2008 and stored in a climate chamber (4 °C) until March 2009.
All solitary bees and wasps emerging from trap nests were
counted (‘abundance’) and determined to species level (‘species
richness’). Nests of bivoltine species’ early first generations
were vacated before the trap nests were returned to the laboratory and thus counted without further species determination or
abundance record (‘vacated nests’). Nests where eggs, larvae,
or pupae died before hatching were counted without further
species determination and abundance record. This nest number
in relation to the number of hatched nests per site constituted
the ‘mortality rate’.

Statistical analysis
Trap nests from one site were excluded from statistical analyses due to contamination by road and tire wear from a nearby
uphill motorway that prohibited proper colonization of their
reed internodes. For the remaining sites, data from both trap
nests were pooled per site. Because bee and wasp individuals
emerging from different brood cells of the same nest lack independence, we used the number of nests as sampling unit when

© 2013 Blackwell Publishing Ltd, GCB Bioenergy, 6, 219–226

testing for sampling effects on species richness (t = 1.7999,
df = 9, P = 0.105) and evaluating completeness of sampling.
Species-accumulation curves produced with R-function specaccum (method ‘random’) in the vegan package (Oksanen et al.,
2012) indicated that for various sites sampling was incomplete
(see Supporting Information). Consequently, we estimated species richness using the nonparametric estimator Chao 1 (Chao,
1984) with R-function estimateS in vegan (Oksanen et al., 2012):
SChao1 ¼ Sobs þ

F21
2F2

where Sobs = the number of species per site; F1 = the number
of observed species represented by a single nest per site (singletons); and F2 = the number of observed species represented
by two nests per site (doubletons; Magurran, 2004).
As Osmia rufa exceeded accumulated abundances of all other
trap-nesting pollinators by 6.2 times on average, it was
excluded from all dependent variables, except (i) estimated
species richness. This allowed for a detailed analysis of otherwise blurred diversity responses in the co-occurring species.
Instead, the number of nests of O. rufa entered all models as an
additional explanatory variable. As this number of nests
strongly correlated with O. rufa’s abundance (t = 19.15, df = 9,
P < 0.001), i.e., the number of provisioned cells in these nests,
we thus accounted for potential competition of this dominant
species with the remaining pollinator community not only for
nesting but also for food resources. In addition, we correlated
O. rufa abundances across the year as well as during and after
mass flowering with those of all other bees and wasps in the
community using R-function cor.test. Following the exclusion of
O. rufa nests for calculation of dependent variables, we then
determined the number of (ii) emerging bee and wasp individuals for each site; (iii) vacated nests of bivoltine species’ first
generation; and (iv) mortality rates.
Prior to testing for landscape effects, each untransformed
response variable was correlated with the scale-dependent
landscape variables percentage area of oilseed rape and percentage area of seminatural habitat across all scales with Spearman Rank Correlations using R-function cor.test (cf. SteffanDewenter et al., 2002). For each response variable, landscape
variables were selected at the scale that yielded the highest rvalue for entering the full model.
Landscape variables at these selected scales were then used
for transforming dependent variables, applying R-function
boxcox in package MASS (Venables & Ripley, 2002) to meet
normality and homoscedasticity assumptions. In addition to
these explanatory variables of main interest, the model specified
in this transformation procedure also contained the three scaleindependent explanatory variables (1) distance to nearest oilseed rape field, (2) area of study site, and (3) the number of
nests of O. rufa. Mortality rate was transformed by taking the
arcsin of the square root of the value. Each of these transformed
dependent variables was then again correlated with the percentage areas of oilseed rape and seminatural habitats at all spatial
scales to determine the decisive spatial scale to enter full models
together with the three scale-independent explanatory variables
already used in the transformation procedure. Analyses were
conducted using generalized linear models (glm) and the