Fluctuations in fruit abundance and synchronization of fruiting
A total of 8067 trees (0.076 tree/m2) from 97 species (41 families, 29 orders), comprising 64 woody, 15 liana and 18 herbaceous species, were observed in the route census of fruiting plants from 2005 to 2016. The number of fruiting trees was significantly different among years (annual average: 672.3 ± SD 419.1, range: 168–1367, χ2 = 2873.7, P < 0.0001, G-test). In Daphniphyllum macropodum, Sorbus gracilis, Viburnum wrightii, Vaccinium oldhamii, and Callicarpa japonica, more than 400 trees were confirmed over the 12 years. The total number of matured fruits was 10507534 (99.1 fruit/m2, annual average: 875627.8 ± 795960.7, range: 64023–2781101), and the fruit number was significantly different among years (χ2 = 7958961.0, P < 0.0001). In Zanthoxylum ailanthoides and Mallotus japonicus, more than twenty thousand of fruits were produced per tree. Additionally, these species occupied 22.9% – 88.7% of all fruits each year.
The fruit abundance of the freshy-fruited plant communities exhibited a remarkable fluctuation across years, with the number of fruiting trees and matured fruits fluctuating repeatedly every other year from 2005 to 2016 (Fig. 1A). The fruit number fluctuation corresponded to that of tree number. The plant communities could be classified into years when the fruit abundance was high and low (Supplementary Fig. 1). The CV values for the fruit numbers of 97 species were shown in Fig. 1B. They were less than 1.0 in only seven species Aralia elata, Eurya japonica, Clerodendrum trichotomum, Ampelopsis glandulosa var. heterophylla, Tripterospermum japonicum, C. japonica, and Z. ailanthoides. To confirm whether the fruiting patterns synchronise among species, the relationships between the fruit numbers were analysed in 31 dominant species, of which the number of fruiting trees was more than 20 over the 12 years. There were significant positive correlations in 133 (28.6%) of the 465 relationships among the 31 species (Spearman’s rank correlation, Supplementary Table 1). The fruit numbers of 14 species were positively correlated with those of the other 12–20 species. Specifically, a high correlation was confirmed in even three species with a CV was less than 1.0, C. trichotomum, T. japonicum and C. japonica. This means that most freshy-fruited plant species produced fruits repeatedly every other year, synchronising mutually, only four species (A, elata, E. japonica, Z. ailanthoides, and A. glandulosa var. heterophylla) constantly producing fruits every year. The synchronisation of fruit production among species leads to the periodic fluctuations in fruit abundance within plant communities.
Migration pattern of frugivorous birds and classification of the 12-year study period
From 2005 to 2016, a total 16722 individuals of 20 frugivorous species were captured and released. The number of birds captured per day was 35.9–122.2, which were significantly different among years (χ2 = 114.0, P < 0.0001, G-test, Supplementary Fig. 2). Concerning all captured birds, 76.2%-90.2% were composed of the Japanese white-eye Zosterops japonicus, pale thrush T. pallidus, and eyebrowed thrush T. obscurus. There were no significant relationships between the number of birds captured per day and the number of fruiting trees and matured fruits (tree number: P = 0.12; fruit number: P = 0.87). However, the fruit abundance was associated with the species compositions of migrant birds. The communities of migrants were classified into three groups by PCA (Fig. 2). The first and second principle components explained 27.7% and 16.8% of the total variation, respectively. The classification was associated with fruit abundance. First, in 2007, 2009, 2011, and 2013, when the fruit abundance was relatively high, the communities were characterised by Z. japonicus, Luscinia calliope, Horornis diphone, Phylloscopus xanthodryas and T. cardis (Fig. 2B). Second, in 2006, 2010, 2012 and 2014 when the fruit abundance was relatively low, the communities of frugivores were characterized by T. pallidus, T. obscurus, T. chrysolaus, T. eunomus, Ficedula mugimaki, and Tarsiger cyanurus. Third, in 2005, 2008, 2015 and 2016, the frugivore communiies were characterized by Hypsipetes amaurotis, F. narcissina, and Syrmaticus soemmerringii.
In addition, the classification was also strongly associated with the abundance of migrant birds. The first principal component was significantly correlated with the number of birds captured in one day (R2 = 0.78, P < 0.001, Fig. 3). In particular, the bird abundance in 2005, 2008, 2011, 2015 and 2016 was lower (average number: 35.9–122.2). In conclusion, the 12-year study period could be classified according to fruit and bird abundance and species composition of migratory birds as shown in Fig. 3: years when the abundances of both frugivores and fruits were high (2007, 2009 and 2013; FA group), years when the fruit abundance was low but frugivores were abundant (2006, 2010, 2012, and 2014; FP group), and years when the abundance of frugivores was low (2005, 2008, 2011, 2015, and 2016; BP group).
Seed-dispersal Networks Of Migrant Frugivorous Birds
The faeces and vomit were collected from total 6652 individuals of 15 bird species. Of the collected samples, 1671 (25.1%) included seeds from 60 plant species (Table 1). The species composition of seed dispersers and transported seeds differed among years (birds: χ2 = 356.7, P < 0.0001; seeds: χ2 = 447.4, P < 0.0001, G-test, Supplementary Fig. 3). Of the bird species, 68.4% – 97.6% were composed of T. pallidus, T. obscurus, and Z. japonicas and 48.8% – 90.1% of the seeds were composed of A. elata, Z. ailanthoides, E. japonica, C. japonica and Cornus macrophylla. In all 12 years, the average number of interactions was larger in bird species than plant species. Each bird species interacted with 2.7–9.2 plant species, whereas each plant species interacted with 1.7–3.2 bird species (Table 1). The numbers of bird and plant species, interaction numbers and interaction strengths in the networks were larger in the FP group, despite of low fruit abundance. The frequency of seed transport was also higher in the FP group, though the difference was not significant (FA: 0.19–0.34, FP group: 0.25–0.39, and BP group: 0.15–0.27, P = 0.09, ANOVA for GLM, Table 1).
Table 1
Frequency of seed removal and the characteristics of dispersal network from 2005 to 2016. The 12 years were classified with three groups by the abundance of migratory birds and fruits; FA group: years when the abundance of both frugivores and fruits were high, FP group: years when the fruit abundance was low, though frugivores were abundant, and BP: years when the abundance of frugivores was low.
|
|
|
Characteristics of interaction networks
|
|
|
|
|
|
|
Interaction number
|
|
|
Observed birds
|
|
Species number
|
|
Average ± SD
|
Year
|
Group
|
N
|
Species number
|
Seed removal
frequency
|
Birds
|
Plants
|
Total
|
Birds
|
Plants
|
2005
|
BP
|
477
|
16
|
0.15
|
12
|
19
|
48
|
4.0 ± 2.1
|
2.5 ± 1.9
|
2006
|
FP
|
1166
|
15
|
0.31
|
10
|
30
|
92
|
9.2 ± 8.0
|
3.1 ± 2.3
|
2007
|
FA
|
831
|
14
|
0.19
|
9
|
23
|
52
|
5.8 ± 6.1
|
2.3 ± 1.3
|
2008
|
BP
|
377
|
14
|
0.26
|
11
|
18
|
51
|
4.6 ± 3.6
|
2.8 ± 2.4
|
2009
|
FA
|
257
|
10
|
0.33
|
8
|
17
|
35
|
4.4 ± 5.4
|
2.1 ± 1.2
|
2010
|
FP
|
554
|
13
|
0.34
|
9
|
20
|
52
|
5.8 ± 4.7
|
2.6 ± 1.8
|
2011
|
FA
|
164
|
13
|
0.26
|
4
|
13
|
22
|
5.5 ± 3.9
|
1.7 ± 0.5
|
2012
|
FP
|
253
|
13
|
0.39
|
10
|
11
|
27
|
2.7 ± 2.3
|
2.5 ± 1.9
|
2013
|
FA
|
295
|
12
|
0.24
|
7
|
17
|
38
|
5.4 ± 4.2
|
2.2 ± 1.5
|
2014
|
FP
|
790
|
14
|
0.24
|
10
|
21
|
55
|
5.5 ± 5.3
|
2.6 ± 2.1
|
2015
|
BP
|
729
|
15
|
0.26
|
13
|
24
|
76
|
5.8 ± 5.4
|
3.2 ± 2.5
|
2016
|
BP
|
478
|
19
|
0.2
|
6
|
21
|
45
|
7.5 ± 4.8
|
2.1 ± 1.4
|
Figure 4 shows the networks of the three groups in which the interaction data were pooled. In all groups, Z. japonicus, T. pallidus, and T. obscurus mainly interacted with four plant species, including A, elata (ID No 3), Z. ailanthoides (ID No 47), E. japonica (ID No 22), and C. japonica (ID No 27). In all groups, Z. japonicus frequently transported seeds of two species A. elata and Z. ailanthoides that constantly produced fruits. Through 12 years, more than 40% of A. elata and Z. ailanthoides seeds were transported by Z. japonicus (average A. elata: 50.1%, Z. ailanthoides: 68.8%). The two Turdus species tended to have more interactions than Z. japonicus. In particular, in the FP group, T. pallidus and T. obscurus interacted with 29 and 33 species, respectively, whereas Z. japonicus interacted with 17 species. In the FP group, these species frequently transported not only seeds of woody species but also those of herbaceous and liana plant species; 14 of 29 (48.3%) and 14 of 33 (42.4%) were herbaceous and liana species (Fig. 4, Supplementary Table 2).
In 11 of the 12 years, the networks were significantly and highly nested, excluding 2011 (BP group) (Table 2). The NODF and WNODF values were significantly higher in the FP group than in the other two groups (NODF: F2,5.8 = 2.9, P = 0.05, WNODF: F2,7.6 = 2.9, P = 0.02, ANOVA for GLM). In the networks of the pooled data, the values were higher in the FP group than in the other two groups (Supplementary Table 3). Additionally, nestedcontribution values for T. pallidus, T. obscurus, and Z. japonicus in the FP group were 5.15, 6.0, and 3.31, respectively. This suggests that the nestedness structures were developed according to the functions of the two Turdus species as generalist dispersers. All networks through 12 years were also significantly modular (Table 2). Although the values of M and Q values were not different among years (M: P = 0.99, Q = 0.98, ANOVA for GLM), they tended to be higher in the FA group. However, when the interaction data of each group was pooled and the networks were created using the QuanBiMo algorithm, the modular structures of the groups differed. Specifically, the composition of modules in the FA group remarkably differed from that in the FP group (Supplementary Fig. 4). The number of interactions among the modules was relatively small in the FA group (average: 5.8 ± 3.9), and the modules tended to be mutually independent. In these networks, Z. japonicus, T. pallidus, and T. obscurus belonged to different modules respectively. In the FP group, 8 and 9 species of herbaceous and liana plants were included in the modules of T. pallidus and T. obscurus (Supplementary Fig. 4). In the module of Z. japonicus, A. elata and Z. ailanthoides were included in the two networks. As shown in the nested structures, there were specific interactions among these species.
Table 2
Descriptors of interaction networks between migratory birds and plants in each year. N: number of individuals transporting seeds; IN: interaction number; IS: interaction strength, WA: web asymmetry; C: connectance. Significance of nestedness and modularity was tested by comparing the values observed in real network and expected from the null model.
|
|
|
Descriptors of interaction networks
|
|
|
|
|
|
|
|
Nestedness
|
Modularity
|
Year
|
Group
|
N
|
IN
|
IS
|
WA
|
C
|
NODF
|
WNODF
|
M
|
Q
|
M/Q
|
2005
|
BP
|
73
|
48
|
82
|
0.23
|
0.21
|
44.5**
|
19.6**
|
0.41*
|
0.34*
|
1.2
|
2006
|
FP
|
368
|
92
|
525
|
0.5
|
0.31
|
70.9**
|
48.7**
|
0.27*
|
0.19*
|
1.44
|
2007
|
FA
|
158
|
52
|
202
|
0.44
|
0.25
|
54.7**
|
38.7**
|
0.32*
|
0.26*
|
1.23
|
2008
|
BP
|
100
|
51
|
138
|
0.24
|
0.26
|
54.2**
|
35.5**
|
0.34*
|
0.27*
|
1.27
|
2009
|
FA
|
87
|
35
|
127
|
0.36
|
0.26
|
62.4**
|
39.4**
|
0.32*
|
0.25*
|
1.29
|
2010
|
FP
|
190
|
52
|
251
|
0.38
|
0.29
|
58.6**
|
36.0**
|
0.31*
|
0.24*
|
1.3
|
2011
|
BP
|
42
|
22
|
52
|
0.53
|
0.42
|
32.8
|
20.7
|
0.34*
|
0.24*
|
1.41
|
2012
|
FP
|
98
|
27
|
131
|
0.05
|
0.25
|
51.7**
|
36.3**
|
0.40*
|
0.18*
|
2.21
|
2013
|
FA
|
71
|
38
|
92
|
0.42
|
0.32
|
60.0**
|
36.0**
|
0.30*
|
0.27*
|
1.11
|
2014
|
FP
|
197
|
55
|
255
|
0.35
|
0.26
|
65.2**
|
45.4**
|
0.3*
|
0.26*
|
1.13
|
2015
|
BP
|
192
|
76
|
261
|
0.30
|
0.24
|
61.3**
|
40.0**
|
0.29*
|
0.23*
|
1.26
|
2016
|
BP
|
97
|
45
|
132
|
0.56
|
0.36
|
57.9**
|
38.1**
|
0.31*
|
0.24*
|
1.27
|
**: P < 0.01, *: P < 0.05 |
Table 3 shows the best-fit models from the GLM analysis concerning the effects on the nestedness and modular structures in the networks over 12 years. The abundance of migrant birds had a strong positive effect on the Z-NODF and Z-WNODF, though fruit abundance had also weak effects. Whereas, none of the factors had positive effects on ZM or ZQ, as the coefficients were low and seemed ineffective. The bird abundance was positively correlated with the Z-NODF and Z-WNODF (Z-NODF: R2 = 0.49, P = 0.004; Z-WNODF: R2 = 0.69, P = 0.0002, logistic regression). In particular, the abundances of T. pallidus and T. obscurus were significantly correlated with the Z-NODF (T. pallidus: R2 = 0.61, P = 0.0028; T. obscurus: R2 = 0.34, P = 0.048) and Z-WNODF (T. pallidus: R2 = 0.62, P = 0.002; T. obscurus: R2 = 0.42, P = 0.02), whereas Z. japonicus had weak correlation (Z-NODF: P = 0.2; Z-WNODF: R2 = 0.37, P = 0.03). Consequently, the nestedness structure was developed from the high abundance of the two Turdus species functioning as generalist dispersers, particularly in years when fruit abundance was low. Furthermore, the development of nestedness structures appears to induce the low levels of modular structures.
Table 3
The best-fit models from the generalized liner model (GLM) analysis concerning the effects on the nestedness and modular structures in the networks over 12 years. As predictor variables, the following values for the patterns of bird migration and fruit abundance were used: the captured number of migrant birds (bird abundance), number of fruiting trees (abundance of fruiting trees), number of mature fruits (abundance of fruits), and the first and second principal components of bird and plant communities in PCA. The models with the fewest AIC value were selected as the best model. Asterisk means significant effect.
|
Response variables
|
|
Nestedness
|
Modularity
|
Predictor variables
|
Z-NODF
|
Z-WNODF
|
ZM
|
ZQ
|
Bird abundance
|
9.8**
|
25.4**
|
27.2
|
0.4
|
PC1 of bird communities
|
-1.6*
|
-3.3
|
-1.1*
|
-0.2*
|
PC2 of bird communities
|
0.4
|
-
|
-
|
-
|
Fruit abundance
|
2.0
|
5.1*
|
8.9
|
0.2
|
PC1 of plant communities
|
-0.6
|
-1.6*
|
-0.4*
|
-0.1
|
PC2 of plant communities
|
-
|
1.1*
|
0.1
|
-0.05
|
Bird abundance × fruit abundance
|
-
|
-
|
-1.7
|
-
|
Intercept
|
-61.5*
|
-141.6*
|
-137.9
|
-7.5*
|
AIC
|
43.6
|
64.5
|
16.7
|
6.6
|
**: P < 0.005, *: P < 0.05 |