Global patterns of plant invasions and the concept of invasibility

2. Patterns after controlling for site size and site descriptors (i.e., island and reserve status). - The analyses were carried out on the subset of the data analyzed in Model 5. I also excluded continents or biomes for which the number of sites was less than four: Asia (n = 2), South America (n = 2), and the wet tropics (n = 1). Comparing standardized residuals of log E from Model 5 showed that there were differences between continents [ILLUSTRATION FOR FIGURE 7A OMITTED], but considerable variation within continents meant that these were not significant ([F.sub.5,99] = 2.2, P = 0.06). By contrast, there were significant differences between biomes ([F.sub.6, 102] = 5.9, P [less than] 0.001). The relative values of these scale-controlled data were somewhat changed from the uncontrolled values. For biomes [ILLUSTRATION FOR FIGURE 7B OMITTED], the multiple biome and temperate agricultural/urban sites were all above the central trend, and deserts and savannas below, with mediterranean shrubland, temperate forest, and alpine habitats close to the trend. Three groups were recognizable by least significant differences (LSD), but with less overlap between groups than for continents [ILLUSTRATION FOR FIGURE 7B OMITTED]. Nevertheless, there was considerable variation within biomes. For example, the Australian savannas (mean residual = -1.3) were much less invaded than were the African (-0.3), whereas the Australian mediterranean habitats (0.7) were much more invaded than those of North America (-0.4) or Europe (-0.6).

Testing generalizations about plant invasions

The generalizations (a-f) can now be tested after controlling for the effect of scale dependency and, where appropriate, native richness and island or reserve status. The basic approach here will be to explore the differences in mean residuals for groups of sites, as was done for the previous biome and continent comparisons. Except where otherwise stated, I will use the residuals from Model 5 (Table 3), for the reasons previously given (see Regional and biome patterns of invasion).

TABLE 6. The fraction of the flora that is exotic for 184 sites
around the world, broken down by biome, and before controlling for
site scale and type.

                                  Mean
Biome                             (%)       1 SD     n

Temperate agricultural/urban       31         9      24
Temperate forest                   22        16      13
Multiple                           19        11      26
Mediterranean shrubland(*)         17        13      43
Alpine                             11         8      26
Savanna                             8         6      33
Wet tropics                         6                 1
Desert(**)                          6         3      18

* Of mediterranean type rather than location.

** Exclusively reserves.

[TABULAR DATA FOR TABLE 7 OMITTED]

a) Old World vs. New World sites. - From the available data, the Old World sites (n = 13) represented by Britain (a single geometric mean for all sites), Europe, Africa, and Asia can be separated from the New World sites (n = 47) represented solely by North America, and their mean standardized residuals from Model 4 then can be compared by t test. The mean residual for New World sites was close to the central trend, being 0.08 [+ or -] 0.13 (mean [+ or -] 1 SE), whereas that for Old World sites was well below at -0.45 [+ or -] 0.30, but as a consequence of the great variation about the means, the two groups did not differ significantly ([t.sub.58] = 1.8, P = 0.075). If, by contrast, we remove the effect due to native species richness from the comparison, by using the residuals from Model 6, the New World is more invaded than the Old ([t.sub.58] = 2.3, P = 0.027). This conclusion is further strengthened if we examine the residuals from the larger data set of Model 4 ([t.sub.127] = 4.3, P [less than] 0.001). Thus, there is evidence to support the generalization that the New World is more invaded than the Old, but only when we factor out the effect of site differences in native species richness.