or bee species richness (combined pan trap and transect walk data) and land cover at different. 12 ...... Scandinavian Journal of Statistics, 6, 65-70. 202. Kline ...
1
Supporting information
2 3
Table S1. Details of the abbreviation used of field sites and the sampling dates
4
Numbers, names and codes of the field sites in the Yucatan Peninsula at which bees were
5
collected, with sampling date. Sites in which the pollination experiment was performed are
6
highlighted using a pink background. Site numbers and codes correspond to those in Figures 1,
7
S3 and S8. Site number
Sites names
Sites codes
Sampling date
1 Cepeda
Cepeda(2010)
2 Muna
Muna(2011)
22/Jul/2011
3 Tebek
Tebek(2010)
4/May/2010
4 Homun
Homun(2011)
28/Jun/2011
5 Toh
Toh(2010)
6 Motul
Motul(2010)
10/Jun/2010
7 Yobain
Yobain(2010)
11/Jun/2010
8 Buctzots
Buctzots(2011)
29/Jul/2011
9 TiziminA
TiziminA(2010)
30/May/2010
10 Tekal de Venegas A
TeVe(2011)
21/May/2011
11 Tekal de Venegas B
TeVeB(2010)
12 Moctezuma
Moctezuma(2011)
13 Rancho Alegre
RaAl(2011)
1/Jun/2011
14 Tizimin D (Santa Maria
TiziminD(2010)
5/Jun/2010
15 Tizimincen C
TiziminC(2010)
3/Jun/2010
16 TiziminB
TiziminB(2010)
1/Jun/2010
17 San Pedro Bacab
SPB(2011)
18 Tixcaltuyub
Tixcaltuyub(2011)
24/Jun/2011
19 Santa María
StaMaria(2011)
26/Jun/2011
20 Nenela C
NenelaC(2011)
14/Jun/2011
21 Timul A
TimulA(2011)
09/Jun/2011
22 Xaya
Xaya(2011)
5/Aug/2011
23 Tixmehuac
Tixmehuac(2011)
15/Jun/2011
24 Nenela A
NenelaA(2011)
9/May/2011
25 TahDziu B
TaDB(2011)
18/Jun/2011
1
19-20/May/2010
8-9/May/2010
15/Jun/2010 30/May/2011
31/May/2011
26 TahDziu A
TaDA(2011)
27 Ichmul
Ichmul(2010)
31/May/2010
28 Tekax A
TekaxA(2010)
2/May/2010
29 Tixcuytun A
TixcuytunA(2011)
30 Tekax B
TekaxB(2010)
31 Tixcuytun B
TixcuytunB(2011)
17/Jun/2011
32 Tixcuytun C
TixcuytunC(2011)
22/Jun/2011
33 Alfonso Caso
AC(2011)
34 Tzucabab C
TzucababC(2010)
9/Jun/-2010
35 Tzucabab B
TzucababB(2010)
24/May/2010
36 Yaxcopil
Yaxcopil(2011)
37 Becanchen
Becanchen(2010)
8 9
2
6/Jun/2011
7/Jun/2011 12-13/May/2010
13/May/2011
1/Aug/2011 8/Jun/2010
10
Table S2. Relationships between land cover and the bee communities across sites.
11
Spearman rank correlation coefficients of the relationship between either total bee abundance
12
or bee species richness (combined pan trap and transect walk data) and land cover at different
13
spatial scales from a site’s centre. Land cover is the proportion of land covered by:
14
agricultural fallow land, home gardens and pasture (FGP); primary or secondary growth
15
forest (Forest); cropland (Crops), comprising staples (maize, beans), cash crops (e.g. chilli)
16
and orchards; and an overall index of the diversity of land cover (Lc-diversity) of all three
17
land cover classes: FGP, Forest and Crops. The largest absolute correlation coefficient
18
(positive or negative) of a row is given in bold.
19
FGP Distance Abundance Richness (Chao-1)
200 0.04 -0.31
Distance class from the centre of a site (in m) 300 400 500 600 700 800 0.09 0.11 0.09 0.07 0.06 0.11 -0.28 -0.27 -0.28 -0.30 -0.31 -0.32
900 0.05 -0.32
1000 0.06 -0.30
900 0.03 0.21
1000 0.04 0.18
20
Forest Distance Abundance Richness (Chao-1)
200 -0.25 0.05
Distance class from the centre of a site (in m) 300 400 500 600 700 800 -0.20 -0.13 -0.06 -0.01 0.03 -0.26 0.12 0.16 0.19 0.21 0.22 0.23
21
Crop Distance Abundance Richness (Chao-1)
200 0.13 0.22
Distance class from the centre of a site (in m) 300 400 500 600 700 800 0.06 0.04 0.02 0.02 0.01 0.16 0.17 0.15 0.07 0.02 0.02 0.28
900 -0.01 0.01
1000 -0.02 0.03
22
Lc-diversity Index Distance
200
300
400
500
600
700
800
900
1000
Abundance
-0.18
-0.32
-0.11
-0.03
0.04
0.06
0.06
0.03
0.04
0.03
0.20
0.15
0.19
0.20
0.20
0.19
0.20
0.20
Richness (Chao-1)
Distance class from the centre of a site (in m)
23
3
24
Table S3. Relationships among land cover variables across sites.
25
Relationships among different classes of land cover in the landscape: agricultural fallow land,
26
home gardens and pasture (FGP); primary or secondary growth forest (Forest); cropland
27
(Crops), comprising staples (maize, beans), cash crops (e.g. chilli) and orchards; and an
28
overall index of the diversity of land cover (Lc-diversity). In the lower left diagonal of the
29
box, correlation coefficients are shown (significant coefficients are in bold) and, in the upper
30
right diagonal, the corresponding significance (probability) after correction for multiple
31
comparisons using the False Discovery Rate (FDR) approach (Holm 1979; Fox 2005).
32 33
Lc-
FGP
Forest
Crops
-
0.01
0.41
0.51
Forest
-0.69
-
0.34
0.98
Crop
-0.21
-0.16
-
0.01
-0.14
0.01
0.75
-
FGP
Lc-diversity Index
diversity
34 35
4
36
Table S4. Statistical fit of the models of pollination success in relation to bee
37
communities and land use across sites.
38
Statistical fit of models relating pollination of chilli to bee communities and surrounding land
39
use at 11 sites; 1) combined data (abundance and species richness of bees from pan traps and
40
transect walks), 2) abundance and species richness of bees from transect walk data only, 3)
41
abundance and species richness of bees from pan trap data only and 4) abundance of
42
Lasioglossum sp. 1 and species richness of bees from all data (from pan traps and transect
43
walks) combined. Model fit was evaluated using SEM (structural equation models) in AMOS
44
v. 7.0 (Arbuckle 2006). Within each model we present the saturated model (with all possible
45
hypothesized links, in which there are as many parameters estimated as degrees of freedom),
46
and the second best fit model that corresponds to the Forest landscape variable in 1) and 2),
47
Forest and FGP in 3) and FGP in 4), as well as, the independence (null) model, which
48
assumes zero population covariance among the observed variables. For each of the three
49
models, three fit indices are provided: the Akaike Information Criterion (AIC), the Root Mean
50
Square Error of Approximation (RMSEA) and a Chi-square test (χ2). AIC†
RMSEA‡ χ2 (P value)§
1) Model using abundance of bees from combined data Independence (null) model
257.12
1.14
0.00
Best fit model FGP
256.24
0.00
0.54
Best fit model Forest
254.02
0.00
0.58
20.00
2.14
0.02
Saturated model
2) Model using bee abundance from transect walks Independence (null) model
257.41
1.23
0.00
Best fit model FGP
249.95
0.27
0.12
Best fit model Forest
250.00
0.40
0.04
42.00
0.00
0.08
Saturated model
3) Model using bee abundance from pan traps Independence (null) model
252.27
1.30
0.00
Best fit model FGP
248.97
0.00
0.66
Best fit model Forest
248.97
0.00
0.60
42.00
0.00
0.08
Saturated model
5
4) Model using Lasioglossum sp. 1 abundance from combined data Independence (null) model
261.85
1.25
0.02
Best fit model FGP
249.90
0.00
0.72
Best fit model Forest
251.52
0.00
0.70
42.00
0.00
0.02
Saturated model 51 52
† Lower AIC values indicate better fit;
53
‡ RMSEA 5%) and widespread (>10 sites) and were used in
62
further statistical analysis.
63 64
10
65
Table S6. Canonical Correspondence Analysis of the relationships between bee
66
communities and land use across sites.
67
Results of the Canonical Correspondence Analysis (CCA) presented in Fig. S2 (supporting
68
information), showing the percentage of variation explained by analysis of the relationships
69
between bee diversity and three land use variables: agricultural fallow land, home gardens and
70
pasture (FGP); primary or secondary growth forest (Forest); and crops (Crops). Inertia is a
71
mean squared coefficient, which represents the total variability in species abundance with
72
respect to the land use variables. Proportion represents the proportion of the variation explained
73
by the data, calculated by subjecting the inertia matrix (variation of the species abundance with
74
respect to the environment) to weighted regression. The constrained values represent the
75
percentage of variation explained by the axes (i.e. environmental variables), in this case 14%.
76
Here 100% represents over-fitting of the analysis and 0% represents poor explanation of the
77
variation. However, an analysis with low constrained values still provides important
78
information about the axes analyzed. ‘Unconstrained’ is the proportion of variation that has not
79
been explained by the axes (i.e. environmental variables), in this case 86%. The final
80
permutation test shows that the only variable that significantly explains the distribution of bee
81
species is FGP.
82
Inertia
Proportion
Rank
Total
6.04
1.00
Constrained
0.86
0.14
4.00
Unconstrained
5.18
0.87
32.00
Percentage of variation explained by the constrained axes. Eigenvalues for constrained axes: CCA1
CCA2
0.34
0.27
CCA3
CCA4
0.16
0.09
Eigenvalues for unconstrained axes: CA1
CA2
CA3
CA4
CA5
CA6
CA7
CA8
0.60
0.51
0.47
0.41
0.33
0.32
0.29
0.27
(we show only 8 of all 32 unconstrained eigenvalues) Permutation test for CCA under reduced model
FGP
DF
AIC
F
Pr(>F)
1
220.18
1.68
0.015 *
11
Lc-diversity
1
220.22
1.64
0.075 .
Forest
1
220.28
1.58
0.095 .
Crops
1
220.54
1.32
0.255
Significance code: ‘*’ P0.40
Significant, but not after correction by FDR
rs>0.20
Non-significant, weak correlation
** = P|t|36 DF)
Abundance (total)~ Crops
Poisson
-0.13
0.03
-1.59
0.11
FGP
Poisson
0.10
0.02
1.19
0.23
Forest
Poisson
0.43
0.03
5.54