Agriculture Reference
In-Depth Information
The two measures may be supposed to be closely related because common
species tend to be more widespread; this is supported by data from the Azores
where both range and frequency are available for the same set of species and
they are significantly correlated (F = 831.6, df 1, 908, P < 0.0001, R 2 = 47.8%).
However, they reflect different aspects of the distribution of alien plants: a spe-
cies may be present in a low number of localities but occupying a large area, or
it may be very frequent locally but with restricted overall distribution. For that
reason, the two distribution measures were analyzed separately for those
regions where data were available. Indeed, the results reported below indicate
that using range and frequency, as defined for the purpose of the present paper,
provides different results with respect to MRT.
Where information on the invasion status [35] was given or could have been
inferred from unequivocal criteria (Tab. 1), alien species were classified into
naturalized and casual, using the approach of Richardson et al. [42] and Pyˇek
et al. [35].
Statistical analysis
Where appropriate, the effect of minimum residence time (MRT) was evaluat-
ed by ANCOVAs, using a standardized measure of distribution or frequency
(Tab. 1) as the response variable, standardized MRT as a covariate, and region
or species group classified according to invasion status (Tab. 1) as factors.
For the Czech flora, where the effect of species traits together with the
effect of MRT on the occurrence of alien species was evaluated, the stan-
dardized frequency was regressed on four standardized covariates (MRT,
maximum plant height, start of flowering and propagule size) and five factors
(introduction mode with three levels: accidental and deliberate either for
ornamental or utilitary reasons; origin with three levels: America, Asia or
Europe; life history with four levels: annual, biennial, perennial or woody
plants; Grime's strategy with eight combinations; predominant dispersal
mode with four levels: no special vector, water, wind or animals; data taken
from [29]). In these analyses, minimal adequate models (MAMs) were deter-
mined, where all explanatory variables (factors and covariates) were signifi-
cantly (P < 0.05) different from zero and from one another and all non-sig-
nificant explanatory variables were removed. This was achieved by a step-
wise process of model simplification, beginning with the maximal model
(containing all factors, interactions and covariates that might be of interest),
then proceeding by the elimination of non-significant terms (using deletion
tests from the maximal model), and retention of significant terms [43]. To
prevent biases to the model structures caused by correlation between vari-
ables, model simplifications were made by backward elimination from the
maximal models by using step-wise analysis of deviance tables [44]. The
results obtained were thus not affected by the order in which the explanatory
variables were removed in the step-wise process of model simplification. The
Search WWH ::




Custom Search