ELE_1261_sm_SA1

advertisement
1
Supplementary Material
2
3
This material is available as part of the online article from:
4
Ozinga et al. Dispersal failure contributes to plant losses in NW Europe
5
http://www.blackwell-synergy.com/doi/full/10.1111/j.1461-0248.XXXX.XXXXX.x
6
7
Appendix S1 Check for possible confounding effects
8
Appendix S2 Historical overview of changes in dispersal infrastructure in the landscape
9
of Northwest Europe.
10
11
Please note: Blackwell Publishing is not responsible for the content or functionality of
12
any supplementary materials supplied by the authors. Any queries (other than missing
13
material) should be directed to the corresponding author for the article.
14
1
1
2
Appendix S1: Check for possible confounding effects
3
The robustness of the results was tested for the Dutch dataset, since this dataset
4
has been recorded with the highest resolution and has the smallest proportion of missing
5
values (see Table S1) and changes in frequency of occurrence are more marked at more
6
detailed spatial scales (Thomas & Abery 1995, Kunin 1998, Witte & Torfs 2003, Tamis
7
2005).
8
9
10
Correlations between variables
A potential problem in evaluating the importance of individual variables is that
11
they might be interrelated (multicollinearity). We checked for this potentially
12
confounding effect in two ways.
13
Firstly, we calculated Pearson correlations between the explanatory variables.
14
Plant characteristics were only weakly correlated between species (r < 0.25 for all
15
combinations, with the highest correlations between ‘Dispersal potential – dung’ * ‘Seed
16
longevity’: r = 0.23 and ‘Nitrogen requirements’ * ‘Seed longevity’: r = 0.22).
17
Secondly, we used a combination of conditional and marginal tests (e.g.
18
McCullagh & Nelder 1989). In marginal testing, the variable is added to the simplest
19
regression model (only including the constant) whereas in conditional testing, the
20
variable is entered in the full model (the constant and all other significant variables
21
except the variable of interest). If the contributions of the variables of interest are similar
22
in both tests, this implies a reliable estimate of the relative importance of the given
2
1
variable. Table S1 indicates that multicollinearity was not a problem. Interaction effects
2
were insignificant and did not change the effect of the dispersal vectors on the risk of
3
species decline.
4
The difference in response between species with a high capacity for dispersal by
5
the dung of large mammals and those dispersed by their fur seems surprising at first
6
glance. This finding can be understood from the fact that species dispersed in dung are
7
generally less specialized in terms of dispersal attributes than those with specialised
8
attributes for dispersal in fur (Janzen 1984; Pakeman et al. 2002; Couvreur et al. 2005),
9
and are thus less dependent on the availability of large herbivores than species with fur-
10
assisted seed dispersal. In addition the dung of cattle kept in stables is widely distributed
11
across agricultural lands, including the seeds inside. The cattle themselves, however often
12
do not leave the stables any more, and neither do the seeds attached to their fur.
13
3
1
2
Table S1 Results of marginal and conditional testing of individual variables and
3
performance of variables in the ‘environmental model’ and the ‘dispersal model’.
4
Plant characteristic
Frequency in historical species pool
Conditional testing
Environmental model
Wald χ2
Marginal testing
Sign.
R2
Wald χ2
Sign.
Wald χ2
Sign.
67.4
<0.001
0.095
56.2
<0.001
60.6
<0.001
100.7
<0.001
0.150
68.6
<0.001
95.1
<0.001
6.4
0.011
16.1
<0.001
Light requirements
12.4
<0.001
0.018
Dispersal potential – water
39.5
<0.001
0.053
30.1
Dispersal potential – wind
11.1
0.001
0.018
7.5
Dispersal potential – fur
67.0
<0.001
0.090
47.8
Dispersal potential – dung
21.8
0.006
0.031
Dispersal potential – birds
7.6
<0.001
0.013
Nitrogen requirement
Moisture
n.s.
No LDD
64.1
<0.001
<0.001
<0.001
42.1
<0.001
0.006
6.5
0.011
<0.001
40.7
<0.001
n.s.
n.s.
6.9
0.009
n.s.
9.6
0.002
55.1
<0.001
n.s.
0.089
27.7
Sign.
70.870
n.s.
n.s.
Seed longevity
Dispersal model
Wald χ2
<0.001
n.s.
5
6
Correlation with other environmental factors
7
Dispersal services for propagules may be correlated with living conditions for
8
established plants in their environment. In particular, dispersal services by water may
9
correlate with the moisture environment of the established plants, while dispersal services
10
by large mammals may correlate with the light conditions in open, grazer-dominated
11
vegetation (Ozinga et al. 2004). To evaluate the role of the environment of the
12
established plants (in terms of moisture, light and nitrogen), as compared to that of the
13
dispersal services, we calculated two models: an ‘environmental model’ (excluding
14
dispersal characteristics), and a ‘dispersal model’ (excluding the environmental variables;
15
see Table S1). Moisture and light requirements of plant species were obtained from the
16
corresponding Ellenberg indicator values. These indicator values are species-specific
4
1
scores, ranging from 1-9, for the optimal occurrence of species along environmental
2
gradients, as explained in the main document for nitrogen (Ellenberg et al. 2001). We
3
acknowledge that these indication values only represent a proxy of complex habitat
4
requirements of species. Evidence for the accuracy of Ellenberg indicator values,
5
however, has been provided by several studies reporting a close correlation between
6
average indicator values and corresponding measurements of environmental variables
7
(see Diekmann 2003 for a review).
8
9
Table S1 indicates that light requirements was only significant in the marginal
testing and not significant in conditional testing and in the environmental model (due to a
10
correlation with nitrogen requirements). Moisture requirements was significant in the
11
conditional model and the environmental model but it had a limited explanatory power as
12
compared to dispersal potential by water (Wald χ2 6.4 respectively 30.1; see Table S1).
13
Moreover we found that the dispersal model performed better than the environmental
14
model (Nagelkerke’s R2 = 0.32 and 0.23, respectively; Table S1). This means that
15
relationships between species trends and dispersal services cannot be explained as
16
correlations driven by an underlying correlation between trends and the environment of
17
the established plants.
18
19
20
21
Phylogenetic non-independence
The observed patterns might be partly phylogenetically induced if related species
22
have similar characteristics and extinction risks due to their common ancestry
23
(phylogenetic conservatism, e.g. Harvey & Pagel 1991). In order to check for possible
5
1
confounding effects of such phylogenetic non-independence, we performed a post-hoc
2
test of bivariate relationships between each of the independent variables and the species
3
trend, using phylogenetically independent contrasts (Harvey & Pagel 1991).
4
Phylogenetically independent contrasts are comparisons between sister taxa, each
5
comparison describing the outcome of a separate, i.e. independent, evolutionary
6
divergence of lineages. The contrasts were calculated exclusively between extant species,
7
using the ‘Brunch’ routine of the CAIC computer program (Purvis & Rambaut 1995;
8
following Burt 1989). This method does not make any assumptions about the mode of
9
trait evolution and does not try to reconstruct ancestral states, making it suitable for
10
dichotomous variables and permitting analysis by sign tests (Purvis & Rambaut 1995;
11
Prinzing et al. 2002). The results (Table S2) largely confirmed our above analysis across
12
species as independent data points. For all variables except dispersal capacity by birds,
13
the relationship with the risk of decline was significant and in the same direction as in the
14
across-species analysis.
15
6
1
2
Table S2 Test results for bivariate comparisons across phylogenetically independent
3
contrasts. The numbers of contrasts with non-zero differences between the species were
4
compared. The table shows the percentage of cases in which these differences were
5
positive (i.e. where the declining species had a higher value than the non-declining
6
species), and the corresponding Z and P values (two-tailed). Note that only dispersal
7
potential for fur and dispersal potential for water were overrepresented among declining
8
species.
9
N contrasts
Percentage
Z-value
Sign.
Frequency in historical species pool
200
27.5
6.29
<0.0001
Nitrogen requirements
195
28.7
5.87
<0.0001
Dispersal potential – fur
65
78.5
4.47
<0.0001
Dispersal potential – water
89
73.0
4.24
<0.0001
121
24.8
5.50
<0.0001
Dispersal potential – birds
11
36.4
0.60
0.5565
Dispersal potential – wind
33
9.1
4.53
<0.0001
Dispersal potential – dung
65
32.3
2.73
0.0064
No LDD
68
32.4
2.79
0.0053
Seed longevity
10
11
7
1
References
2
Burt, A. (1989). Comparative methods using phylogenetically independent contrasts. Oxf.
3
Surv. Evol. Biol. 6, 33-53.
4
Couvreur, M., VandenBerghe, B., Verheyen, K. & Hermy, M. (2005). Complementarity
5
of epi- and endozoochory of plant seeds by free ranging donkeys. Ecography 28,
6
37-48.
7
Ellenberg, H., Weber, H.E., Düll, R., Wirth, V., & Werner, W. (2001). Zeigerwerte von
8
Pflanzen in Mitteleuropa. Scripta Geobotanica 18, 1-262. Goltze, Göttingen. [In
9
German]
10
11
12
13
14
15
16
Harvey, P.H. & Pagel, M.D. (1991). The Comparative Method in Evolutionary Biology.
Oxford University Press, Oxford.
Janzen, D.H. (1984). Dispersal of small seeds by big herbivores: foliage is the fruit. Am.
Nat. 123, 338-353.
Kunin, W.E. (1998). Extrapolating species abundance across spatial scales. Science 281,
1513-1515.
Ozinga, W.A., Bekker, R.M., Schaminée, J.H.J. & van Groenendael, J.M. (2004).
17
Dispersal potential in plant communities depends on environmental conditions. J.
18
Ecol. 92, 767-777.
19
20
21
Pakeman, R.J., Digneffe, G. & Small, J.L. (2002). Ecological correlates of endozoochory
by herbivores. Funct. Ecol. 90, 296-304.
Prinzing, A., Durka, W., Klotz, S. and Brandl, R. (2002). Geographic variability of
22
ecological niches of plant species – are competition and stress relevant?
23
Ecography, 25, 721-729.
8
1
Purvis, A. and Rambaut, A. (1995). Comparative Analysis of Independent Contrasts
2
(CAIC): A Statistical Package for the Apple Macintosh, Version 2.0 User’s Guide.
3
Oxford University Press, Oxford.
4
5
6
7
Thomas, C.D. & Abery, J.C.G. (1995). Estimating rates of butterfly decline from
distribution maps: the effects of scale. Biol. Cons. 73, 59-65.
Witte, J.P.M. & Torfs, P.J.J.F. (2003). Scale dependency and fractal dimension of rarity.
Ecography 26, 60-68.
8
9
Download