Page 10 - Pretto_Celesti_2012

Non-native plant species in Mediterranean islands 2567

ferry connections at the height of summer. The effect on the two response datasets (Table 2). Never-
remaining islands were further divided according to theless, when we analyzed the matrix of total non-native
the presence/absence of artificial surfaces, i.e. where species, the environmental component accounted for
human disturbance promotes the occurrence of non- most of the explained variation in species composition
native plant species. Latitude was instead found to (33.33 %) after controlling for the effects of geography
play an important role in islands on which artificial and human impact, while the geographical (28.13 %)
surfaces were absent, with the southernmost islands and human-related variables (27.15 %) yielded slightly
hosting the lowest number of introduced plants. By smaller pure effects. By considering the matrix of
contrast, in islands on which artificial surfaces were the established non-native species, the pure effect of the
present, the distance from the nearest island, which human-mediated component surpassed those of the
was closely correlated with the distance from the other two groups of variables (35.83 vs. 28.13 % and
mainland, was a major determinant of non-native plant 27.15 % respectively).
species richness, probably influencing the intensity of In both cases, we found small fractions of shared
both past and current human exploitation. variation, which were due to the joint effect of
Given the small number of casual taxa found on the different components. The majority of these shared
37 islands, the regression tree obtained by using the effects were accounted for by the joint effect of
established non-native species richness as the response geography and environment. Negative values were
variable was, as expected, very similar (Fig. 2b). The detected for the joint effect of the environmental
only noteworthy difference concerned the islands on and human-related variables. According to Legendre
which artificial surfaces were present, where the and Legendre (1998), positive values in the shared
greater proportion of volcanic lithology appears to variation component of factors indicate collinear
contribute to the establishment of a higher number of variation, whereas negative values indicate that the
non-native taxa. corresponding explanatory variables had opposite
effects (i.e. one process hindered the contribution of
Determinants of non-native species composition the other in the joint regression model).

In the preliminary DCA, the length of the first gradient
was 3.282 SD for total non-native species and 3.461 SD Discussion
for establishednon-nativespecies,therebyjustifyingthe
use of unimodal method for the variation partitioning The findings of this study highlight the predominant
analysis. Variation partitioning analysis shows that the role played by human activity, expressed both as
fraction of variance explained by the explanatory tourism pressure and extension of artificial surfaces,
variables is equal to 32.95 % for total non-native on the richness and composition of non-native flora on
species and 51.5 % for established non-native species. small Mediterranean islands.
All three groups of variables (geographical, environ- The regression tree analysis, in particular, reveals
mental and human-mediated) had a significant pure the effect exerted by human activities on differences in

Table 2 Variation partitioning of the composition of all the non-native and that of the established non-native highlighting the
independent effects of geographical (Geo), environmental (Env) and human-mediated (Man) factors, as well as their overlap
Pure components Shared components Unexplained
Geo Env Man Geo \ Geo \ Env \ Geo \ Env
Env Man Man \ Man

Total non-native 0.432 0.512 0.417 0.101 0.031 0.023 0.020 3.195
plant species (p = 0.0002) (p = 0.035) (p = 0.003)
Established non- 0.346 0.441 0.535 0.109 0.038 -0.009 0.033 2.28
native plant species (p = 0.0002) (p = 0.0008) (p = 0.0002)
Negative values for the shared components indicate synergy. P-levels for pure components as determined by Monte Carlo
permutation tests (n = 4,999) are shown in brackets

123