Global patterns of plant invasions and the concept of invasibility

Exotic species are "species coming from outside the area in question," invaders sensu Williamson (1996: 58). However, it is impossible to judge from the present data set the percentages of the exotic species that actually cause pest problems, or that are naturalized but have no harmful effect.

Statistical analyses

Variables were examined for normality and were log- or arcsine-transformed as necessary. Relationships between variables were tested for significance by correlation (Snedecor and Cochran 1980: 361-363) and, where the correlation proved significant, trend lines were fitted by least squares linear regression. For multiple regressions, independent variables were first tested to ensure that they were not autocorrelated prior to their inclusion in the model. The strengths of regression relationships were measured by their adjusted [r.sup.2] values (Seber 1977). This corrects for the number of independent variables in the model. An overall indication of the accuracy with which the fitted regression predicts the dependence of y on x is given by the standard error of the estimate, expressed as a percentage of the mean of y (Zar 1996: 327-328).

Differences between groups of sites were compared using ANOVAs of standardized residuals from the regressions. The use of residuals controls for the most important explanatory variables, so that one may cautiously assume that one is comparing sites of different types on something approaching an equivalent basis. [TABULAR DATA FOR TABLE 3 OMITTED] For example, if we are interested in comparing differences between groups of sites in variable y, but we know that much of the variation in y depends on x, we can use ANOVAs of the residuals from a regression of y on x. Essentially, one is then comparing the vertical displacement of the data points from the central trend of y on x, to see whether groups differ, on average, in being above or below the trend. This avoids confounding the effects due to site factors with the effects due to x.

RESULTS

Global patterns of exotic plant richness

Across all sites, the mean value of X, the fraction of the flora that was exotic, was 16%, but there was considerable variation about this mean, the minimum value being 1.3% (Zachariashoek reserve in South Africa; Kruger et al. 1989), and the maximum being 64% (Hawaii Volcanoes; Vitousek 1988). The frequency distribution of exotic fractions was right skewed, with 18% of all sites having fractions of [less than or equal to] 5% [ILLUSTRATION FOR FIGURE 1A OMITTED]. Island sites [ILLUSTRATION FOR FIGURE 1C OMITTED] had almost three times as large an exotic fraction as mainland sites [ILLUSTRATION FOR FIGURE 1B OMITTED], whereas reserves had less than half of the exotic fraction of nonreserve sites [ILLUSTRATION FOR FIGURES 1B, C OMITTED]. Furthermore, the exotic fraction X, and the number of exotic species E, were significantly related to site area A (Models 1 and 2 in Table 3; [ILLUSTRATION FOR FIGURE 2 OMITTED]).