1
Appendix S1. Supporting online information on additional sampling protocols, multiplicative
2
diversity partitioning, characteristics of the plant-frugivore networks, species list of plants and
3
their frugivores, results of supporting analyses.
4
Monitoring of fruit abundance along transects in 2011
5
From July to October 2011 we monitored overall fruit abundance in all study sites. In each
6
study site we established one transect of 250 m length and 20 m width covering a total area of
7
0.5 ha. In some study sites ash-alder forests were continuously flooded and accessibility was
8
limited. Thus, we established transects in all study sites along pre-existing trails. As the
9
temporal turnover in fruit ripening of plants was low, we repeated transect walks twice during
10
the study period (from the 1st to 5th August, 2011 and from 1st to 5th September, 2011). During
11
the transect walks all fleshy-fruited plants bearing ripe fruits were identified within each
12
transect. For each plant the presence and number of ripe fruits were estimated on a
13
logarithmic scale and the number of fruits available per transect walk and plot was calculated.
14
For each study site we calculated the mean fruit abundance by averaging across the two
15
transect walks.
16
Estimation of consumer/resource ratios and inference about competition
17
Measuring competition for resources is not trivial. Here we used the consumer/resource ratio
18
as a rather simple surrogate for the degree of competition for fruit resources. This measure has
19
to be interpreted with caution since we have no information on the absolute effect of
20
consumer/resource ratios on the fitness of the consumers (i.e. reproductive success).
21
However, the mean consumer/resource ratios in the different forest types can be interpreted
22
relative to the distribution of the consumer/resource ratios of all observed focal plants (n =
23
98). At least, a comparison of consumer/resource ratios in a given habitat type with the
24
overall distribution provides an indication of how much the consumer/resource ratio deviates
25
from what is expected from the overall sample. In order to draw inferences about competition
26
from consumer/resource ratios we first calculated the consumer/resource ratio for each focal
1
27
plant and determined the median and the 25% and 75% quartiles of the distribution (Fig. S1).
28
Next we compared the observed distribution of consumer/resource ratios with the mean
29
consumer/resource ratios in the different habitat types (Figs. 2a and S1). In the interior of old-
30
growth forests the consumer/resource ratio (0.058 ± 0.010; mean ± SE) equalled the median
31
consumer/resource ratio of all observed focal plants (median = 0.049; 25% quartile: 0.027;
32
75% quartile: 0.11). In contrast, the consumer/resource ratios were lower (0.0096 ± 0.0089) in
33
the interior of logged forests and were higher (logged: 0.11 ± 0.010; old-growth: 0.085 ±
34
0.0086) at forest edges than the median consumer/resource ratio of all observed focal plants.
35
This suggests a higher pressure on the available fruit resources at forest edges (i.e. increased
36
competition) and a reduced pressure on fruit resources in the interior of logged forests (i.e.
37
reduced competition) compared to the interior of old-growth forests.
38
Partitioning of diversity into independent richness and evenness components
39
A recent review by Tuomisto (2012) has shown that Shannon diversity can be partitioned into
40
independent richness and evenness components in a multiplicative manner (equation 1). This
41
approach is mathematically related to the partitioning of diversity into alpha and beta
42
components (Jost 2009). According to the above mentioned concept multiplicative
43
partitioning of diversity into richness and evenness components can be written as:
44
e H i  Ei  J i
45
where e H is the exponent of the Shannon-index, Ei is the evenness component and Ji is the
46
richness component in the spectrum of frugivores on plant species i. Following the correct
47
definition of Evenness (Hill 1973):
48
Ei 
49
equation 1 can also be rewritten as follows:
50
eH 
eqn 1
i
e Hi
Ji
eH
 Ji
Ji
eqn 2
eqn 3
2
51
Note that the evenness and richness components measure two different phenomena, i.e. (1)
52
the equitability in the interaction frequency of species and (2) the number of involved species
53
(Tuomisto 2012). Further, evenness is replication invariant, i.e. does not change when a
54
dataset is replicated such that each of its species gives rise to n new species of the same
55
absolute abundance as the original one (Hill 1973). Given that both of these measures
56
quantify different phenomena they can vary independently of each other and multiplied
57
express the ’effective‘ number of species if all were equally common (Tuomisto 2012).
58
In the following we show that this framework easily can be generalised to be used with
59
species interaction networks. To measure the effective number of dispersal vectors per plant
60
species Sq we calculated the mean across the plants in a given network where each plant was
61
weighted by its interaction strength as:
62
S q  i 1
63
where Ai is sum of interactions of plant species i and m is the sum of interactions in the
64
network. Likewise the evenness and richness components can be calculated. Finally, equation
65
1 can be generalised to the network context as follows:
66
Hi
Ai H i
I Ai e
I A
e

i1 m
i1 m J  i1 mi J i
i
67
Since equation 1 holds for all plant species I in a network and diversity as well as its
68
components for each plant species i are scaled by the same constants (i.e. the interaction
69
strength of plant species i) the assumptions of equation 1 also hold for equation 5.
70
Partitioning of Environmental and Spatial Effects on composition of frugivore
71
assemblages
72
We used a PCNM analysis (Principal Coordinates of Neighbourhood Matrix; Dray et al.
73
2006) combined with a multivariate redundancy analysis (RDA) to partition the variance in
74
the species turnover of the frugivore assemblages among study sites that was explained by
I
Ai Hi
e
m
I
eqn 4
eqn 5
3
75
environmental and spatial components. In our case the analysis required (i) a site × species
76
community matrix, (ii) a table containing spatial eigenvectors retained from PCNM analysis
77
and (iii) a table containing the environmental variables.
78
Community data
79
For the multivariate redundancy analysis (RDA) we constructed a site × species matrix. To do
80
so we first calculated the mean abundance of each frugivore species across the plant species in
81
each study site and year (i.e. the mean visitation rate of each frugivore species in each of the
82
18 networks during 18 hours). Then we calculated the mean abundance of each frugivore
83
species across the two study years for each study site (i.e. the mean abundance of each
84
frugivore species across the two networks per study site). We applied a Hellinger
85
transformation to the abundance data prior to analysis (Legendre and Gallagher 2001).
86
Hellinger transformation makes abundance data containing many zeros suitable for analysis
87
by linear methods such as redundancy analysis (Legendre 2008).
88
Spatial Variables
89
We derived the spatial variables by using principal coordinates of neighbourhood matrices
90
(PCNM), a method well suited for the detection of spatial trends across a wide range of scales
91
(Borcard and Legendre 2002; Borcard et al. 2004; Dray et al. 2006). The GPS coordinates of
92
the centre of each study site were used to construct a Euclidean distance matrix. This matrix
93
was truncated at the smallest distance that keeps all sites connected in a single network (9.8
94
km in our case). The distances above the truncation threshold were given an arbitrary value of
95
four times the threshold. Then, we used a principal coordinates analysis (PCoA) to retain the
96
eigenvectors associated with positive eigenvalues as spatial variables (PCNM variables)
97
(Borcard and Legendre 2002; Borcard et al. 2004). We used a forward selection procedure
98
based on redundancy analysis (RDA) to retain the spatial eigenvectors that explain most of
99
the variation in the species turnover among the study sites. The forward selection was based
100
on a double-stop criterion (Blanchet et al. 2008). The procedure began with performing a
4
101
global test (RDA) with all spatial eigenvectors. Afterwards, α-values (P < 0.1 based on
102
pseudo F-values after 9999 permutations) and coefficients of determination (R2) of global
103
tests were used as stopping criteria in the forward selection of variables. The spatial
104
eigenvectors that fulfilled both stopping criteria were identified as the significant spatial
105
variables influencing the variation in the species turnover among the study sites.
106
Environmental variables
107
The environmental table contained the two main factors location and logging. The main
108
factors were represented by dummy variables coded by 0 and 1. The interaction between the
109
main factors was included as the product of the two dummy variables. Moreover, we included
110
fruit abundance (ln(x) transformed, mean across the two years for each study site) as a
111
continuous predictor into the environmental table.
112
Partitioning of environmental and spatial components
113
In the last step we used multivariate redundancy analysis to partition the variation in the
114
species turnover among the study sites with respect to the environmental and spatial
115
components (Borcard et al. 1992). For inference we used adjusted R² values as unbiased
116
estimators of explained variation (Peres-Neto et al. 2006). We assessed the significance of the
117
joint and independent environmental and spatial components using pseudo F-tests based on
118
9999 permutations. All analyses were conducted in R version 2.14.0 (R Development Core
119
Team 2011), using the packages vegan (Oksanen et al. 2011) and packfor (Dray et al. 2011).
120
Variation in species turnover explained by environmental and spatial components
121
The forward selection procedure identified one spatial eigenvector (PCNM4: R² = 0.05; P =
122
0.095) as predictor of species turnover. The partitioning of environmental and spatial effects
123
showed that the spatial component in our study design explained 18% of the variation in
124
species turnover among study sites (F1,4 = 2.81, P = 0.079; Table S3). After accounting for the
125
spatial component the environmental component significantly explained 36% of the variation
5
126
in the species turnover among study sites (F4,4 = 2.44, P = 0.017; Table S3). Thus, we are
127
confident that the observed patterns are not merely a spatial artefact.
6
128
Table S1. Geographical coordinates and characteristics of the 18 plant-frugivore networks quantified in Białowieża forest, Eastern Poland in the years
129
2011 and 2012.
Fruit
abundance
Site
Latitude
Longitude Logging
Location
2011
2012
Number of
plant species
Number of
frugivore
species
Number of
interaction
links
Total
number of
interactions
2011 2012
2011 2012
2011 2012
2011 2012
CV Interaction
frequencies
2011
2012
1
52.7023920044
23.6534900218 Logged
Interior
3820
4494
3
8
5
11
6
31
16
192
120.16
142.26
2
52.7048510034
23.6223939992 Logged
Interior
2303
2016
4
9
7
15
8
39
52
306
154.72
159.92
3
52.6704000402
23.6854699999 Logged
Interior
973
-
2
-
6
-
7
-
33
-
79.10
-
4
52.7421170287
23.8343590032 Old-growth
Interior
2000
6262
3
6
11
19
17
40
51
397
141.42
247.53
5
52.7892280184
23.8460719585 Old-growth
Interior
4852
5568
5
6
9
13
20
30
110
239
157.79
165.85
6
52.7162399981
23.8157899864 Logged
Edge
2809
3075
6
8
9
12
24
28
207
260
151.59
253.93
7
52.7343607090
23.7892601568 Logged
Edge
5915
3209
5
8
14
11
27
42
336
156
264.72
128.40
8
52.7307270281
23.8221490011 Old-growth
Edge
494
2464
2
5
7
9
9
17
44
111
125.03
102.76
9
52.7990119625
23.8249779772 Old-growth
Edge
9794
13739
8
8
11
16
29
51
510 1133
236.90
230.38
10
52.7795999777
23.8578750193 Old-growth
Edge
4250
-
5
-
12
-
25
-
224
149.16
-
-
1
130
Table S2. Summary of linear mixed effects models (Typ III SS) testing the effect of location
131
(forest interior vs. edge), logging (logged vs. old-growth), year and second order interactions
132
on (a) fruiting plant richness, (b) network size, (c) total number of interactions and (d)
133
coefficient of variation (CV) of interaction frequencies in the 18 plant-frugivore networks
134
quantified in Białowieża Forest, Eastern Poland in 2011 and 2012. Network size and total
135
number of interactions were ln(x) transformed prior to statistical analysis. Significant
136
predictors at a level of P < 0.05 are given in boldface type.
Source of Variance
(a) Fruiting plant richness
Location
Logging
Year
Location × logging
Year × location
Year × logging
Dfnum,Dfden
F
P
1,6
1,6
1,5
1,6
1,5
1,5
1.07
0.141
9.75
0.0870
2.86
4.95
0.34
0.72
0.026
0.78
0.15
0.077
(b) Network size
Location
Logging
Year
Location × logging
Year × location
Year × logging
1,6
1,6
1,5
1,6
1,5
1,5
0.469
1.34
14.5
1.14
4.68
0.612
0.52
0.29
0.013
0.33
0.083
0.47
(c) Total number of interactions
Location
Logging
Year
Location × logging
Year × location
Year × logging
1,6
1,6
1,5
1,6
1,5
1,5
2.74
0.521
14.9
0.349
7.31
0.116
0.15
0.50
0.012
0.58
0.043
0.75
(d) CV of interaction frequency
Location
Logging
Year
Location × logging
Year × location
Year × logging
1,6
1,6
1,5
1,6
1,5
1,5
0.160
0.553
1.24
1.86
1.03
0.116
0.70
0.49
0.32
0.22
0.36
0.75
1
137
Table S3. List of codes for frugivore and plant species. Mammals are marked with an
138
asterisk. Frugivores were classified into forest specialists and generalists (Jędrzejewska and
139
Jędrzejewski 1998a; Jędrzejewska and Jędrzejewski 1998b; Svensson et al. 2009). Forest
140
specialists reproduce exclusively in forest habitats, whereas generalists also reproduce in non-
141
forest habitats.
Code
Frugivores
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
Plants
P1
P2
P3
P4
P5
P6
P7
P8
P9
P10
P11
Scientific name
Vernacular name
Habitat specialisation
Sylvia atricapilla
Turdus merula
Erithacus rubecula
Turdus philomelos
Coccothraustes coccothraustes
Parus major
Sylvia borin
Poecile palustris
Sitta europaea
Muscicapa striata
Fringilla coelebs
Dendrocopos major
Luscinia luscinia
Garrulus glandarius
Ficedula hypoleuca
Tetrastes bonasia
Poecile montanus
Dendrocopos medius
Phylloscopus trochilus
Pyrrhula pyrrhula
Apodemus flavicollis*
Ficedula parva
Dendrocopos leucotos
Oriolus oriolus
Turdus pilaris
Columba palumbus
Dryocopus martius
Martes martes*
Periparus ater
Sciurus vulgaris*
Turdus iliacus
Turdus viscivorus
Eurasian Blackcap
Common Blackbird
European Robin
Song Thrush
Hawfinch
Great tit
Garden Warbler
Marsh Tit
Eurasian Nuthatch
Spotted Flycatcher
Common Chaffinch
Great Spotted Woodpecker
Thrush Nightingale
Eurasian Jay
European Pied Flycatcher
Hazel Grouse
Willow Tit
Middle Spotted Woodpecker
Willow Warbler
Eurasian Bullfinch
Yellow-necked Mouse
Red-breasted Flycatcher
White-backed Woodpecker
Eurasian Golden Oriole
Fieldfare
Common Wood Pigeon
Black Woodpecker
European Pine Marten
Coal Tit
Eurasian Red Squirrel
Redwing
Mistle Thrush
Generalist
Generalist
Generalist
Generalist
Specialist
Generalist
Generalist
Specialist
Generalist
Generalist
Generalist
Generalist
Specialist
Specialist
Specialist
Specialist
Generalist
Specialist
Generalist
Generalist
Specialist
Specialist
Specialist
Specialist
Generalist
Generalist
Specialist
Specialist
Specialist
Specialist
Generalist
Generalist
Cornus sanguinea
Euonymus europaeus
Euonymus verrucosus
Frangula alnus
Lonicera xylosteum
Prunus padus
Rhamnus cathartica
Ribes alpinum
Ribes nigrum
Ribes spicatum
Rubus idaeus
Common Dogwood
European Spindle
Spindletree
Glossy Buckthorn
Common honeysuckle
Hackberry
Common Buckthorn
Alpine Currant
Black Currant
Red Currant
Red Raspberry
-
2
P12
P13
Sorbus aucuparia
Viburnum opulus
Rowan
Guelder Rose
-
3
142
Table S4. Summary of linear mixed effects models (Typ III SS) testing the effect of fruit
143
abundance location (forest interior vs. edge), logging (logged vs. old-growth), year and
144
second order interactions between the main factors and year on (a) consumer/resource ratio
145
CRq, (b) frugivore specialisation ‹d’j›, (c) evenness Eq, and (d) redundancy Sq of the 18 plant-
146
frugivore networks quantified in Białowieża Forest, Eastern Poland in 2011 and 2012. Fruit
147
abundance was ln(x) transformed prior to statistical analysis. Given are adjusted effect sizes
148
radj according to the formula given in Rosenthal and Rosnow (1985) as the square-root of the
149
ratio: r² = dfnumerator × F / (dfnumerator × F + dfdenominator). For comparison significant effects at a
150
level of P < 0.05 and effect directions from the full and the reduced model are given in
151
boldface type. Note that inclusion of the second order interactions between year and the two
152
main factors does not affect our main conclusions qualitatively (effect direction) and that
153
interaction terms including year were not significant in any of the models.
Source of Variance
(a) Consumer/resource ratio
Fruit abundance
Location
Logging
Year
Location × logging
Year × location
Year × logging
Dfnum,Dfden
F
P
radj full
radj reduced
1,4
1,6
1,6
1,4
1,6
1,4
1,4
28.7
1.20
7.94
4.21
9.97
6.45
0.757
0.0058
0.32
0.031
0.11
0.020
0.064
0.43
(–) 0.94
0.41
(–) 0.75
0.72
(+) 0.79
0.79
0.40
(–) 0.92
0.55
(–) 0.77
0.47
(+) 0.79
-
(b) frugivore specialisation
Fruit abundance
Location
Logging
Year
Location × logging
Year × location
Year × logging
1,4
1,6
1,6
1,4
1,6
1,4
1,4
8.92
0.008
9.30
2.36
8.97
1.93
0.197
0.041
0.93
0.023
0.20
0.024
0.24
0.68
(+) 0.83
0.04
(+) 0.78
0.61
(–) 0.77
0.57
0.22
(+) 0.83
0.48
(+) 0.88
0.49
(–) 0.85
-
(c) Evenness
Fruit abundance
Location
Logging
Year
Location × logging
Year × location
Year × logging
1,4
1,6
1,6
1,4
1,6
1,4
1,4
27.0
12.9
0.0785
2.89
2.85
3.71
0.596
0.0066
0.011
0.79
0.16
0.14
0.13
0.48
(–) 0.93
(–) 0.83
0.11
0.65
0.57
0.69
0.36
(–) 0.92
(–) 0.81
0.47
0.12
0.59
-
(d) Redundancy
Fruit abundance
Location
Logging
Year
Location × logging
Year × location
Year × logging
1,4
1,6
1,6
1,4
1,6
1,4
1,4
8.81
2.53
8.73
2.36
6.85
0.410
0.0942
0.041
0.16
0.026
0.20
0.040
0.56
0.77
(–) 0.83
(–) 0.54
(–) 0.77
0.61
(+) 0.73
0.30
0.15
(–) 0.81
(–) 0.75
(–) 0.85
0.69
(+) 0.78
-
4
154
Table S5. Partitioning of variation in the composition of frugivore assemblages in the local
155
networks that was explained by environmental and spatial components. The spatially
156
structured environment fraction is not testable in a separate model. Note that the negative
157
variance component for spatially structured environment indicates that environmental and
158
spatial components are not completely uncorrelated, i.e. they are not orthogonal. Dfnum and
159
Dfden give numerator and denominator degrees of freedom, respectively. Significant predictors
160
at a level of P < 0.05 are given in boldface type.
Source of Variance
Dfnum,Dfden
R²adj
F
P
Environment + spatially structured environment
4,5
0.31
2.04
0.019
Space + spatially structured environment
1,8
0.14
2.44
0.10
Environment + space + spatially structured
5,4
0.50
2.79 0.0047
Environment
4,4
0.36
2.44
0.017
Space
1,4
0.18
2.81
0.079
Spatially structured environment
-
-0.04
-
-
Residual
-
0.50
environment
5
161
162
Fig S1. Distribution of observed consumer/resource ratios [visits fruit-1 18h-1] of all observed
163
focal plants (n = 98). The vertical lines represent the median (solid line) and the 25% and 75%
164
quartiles (dashed lines).
6
165
References
166
Blanchet, F., Legendre, P. & Borcard, D. (2008) Forward selection of explanatory variables.
167
Ecology, 89, 2623-2632.
168
Borcard, D. & Legendre, P. (2002) All-scale spatial analysis of ecological data by means of
169
principal coordinates of neighbour matrices. Ecological Modelling, 153, 51-68.
170
Borcard, D., Legendre, P., Avois-Jacquet, C. & Tuomisto, H. (2004) Dissecting the spatial
171
172
173
174
175
176
structure of ecological data at multiple scales. Ecology, 85, 1826-1832.
Borcard, D., Legendre, P. & Drapeau, P. (1992) Partialling out the spatial component of
ecological Vvariation. Ecology, 73, 1045-1055.
Dray, S., Legendre, P., & Blanchet, G. (2011) packfor: Forward selection with permutation
(Canoco p. 46). [R package version 0.0-8/r100].
Dray, S., Legendre, P. & Peres-Neto, P.R. (2006) Spatial modelling: a comprehensive
177
framework for principal coordinate analysis of neighbour matrices (PCNM). Ecological
178
Modelling, 196, 483-493.
179
180
181
Hill, M.O. (1973) Diversity and evenness: A unifying notation and its consequences. Ecology,
54, 427-432.
Jędrzejewska, B. & Jędrzejewski, W. (1998a) Abundance of prey. Predation in vertebrate
182
communities: the Białowieża Primeval Forest as a case study. Ecological Studies 135
183
(eds M. M. Caldwell, G. Heldmaier, O. L. Lange, H. A. Mooney, E.-D. Schulze & U.
184
Sommer), pp. 28-92. Springer, Heidelberg.
185
Jędrzejewska, B. & Jędrzejewski, W. (1998b) Numbers, distributions and home ranges of
186
predators. Predation in vertebrate communities: the Białowieża Primeval Forest as a
187
case study. Ecological Studies 135 (eds M. M. Caldwell, G. Heldmaier, O. L. Lange, H.
188
A. Mooney, E.-D. Schulze & U. Sommer), pp. 97-173. Springer, Heidelberg.
189
190
Jost, L. (2009) Partitioning diversity into independent alpha and beta components (vol 88, p.
2427, 2007). Ecology, 90, 3593.
7
191
192
193
194
Legendre, P. & Gallagher, E.D. (2001) Ecologically meaningful transformations for
ordination of species data. Oecologia, 129, 271-280.
Legendre, P. (2008) Studying beta diversity: ecological variation partitioning by multiple
regression and canonical analysis. Journal of Plant Ecology, 1, 3-8.
195
Oksanen, J., Blanchet, G, Kindt, R., Legendre, P., O'Hara, R.B., Simpson, G.L., Solymos, P.,
196
Henry, M., Stevens, H., & Wagner, H. (2011) vegan: Community Ecology Package. [R
197
package version 1.17-6].
198
Peres-Neto, P.R., Legendre, P., Dray, S. & Borcard, D. (2006) Variation partitioning of
199
species data matrices: Estimation and comparison of fractions. Ecology, 87, 2614-2625.
200
R Development Core Team. (2011) R: A language and environment for statistical computing.
201
202
203
204
205
206
207
[R package version 2.14.1]. Vienna, Austria, R Foundation for Statistical Computing.
Rosenthal, R. & Rosnow, R.L. (1985) Contrast analysis: focused comparisons in the analysis
of variance. Cambridge University Press, Cambridge.
Svensson, L., Mullarney, K. & Zetterström, D. (2009) Collins Bird Guide: The Most
Complete Guide to the Birds of Britain and Europe. HarperCollins Publishers, London.
Tuomisto, H. (2012) An updated consumer's guide to evenness and related indices. Oikos,
121, 1203-1218.
8