We studied the invasion of a New Zealand mountain beech (Nothofagus so
landri var. cliffortioides) forest by the exotic perennial herb, Hiera
cium lepidulum. We used data from 250 randomly located permanent plots
(400 m(2)) established in 1970 that sampled 9000 ha of forest. Freque
ncy of H. lepidulum was 11%, 43%, and 57% in 1970, 1985, and 1993, res
pectively. For each year of measurement, invasion patterns were relate
d to (a) distance to the forest margin as a measure of dispersal limit
ation, (b) community structure, (c) Stem biomass dynamics indicating d
isturbance history, and (d) environmental characteristics. In 1970, in
vaded plots had more species and lower potential solar radiation, and
they were closer to the forest margin; however, invaded plots were onl
y weakly predicted by these site variables. H. lepidulum also invaded
relatively species-rich subplots (0.75 m(2)) showing that community st
ructure was also significant at a microsite scale. Using the same sets
of variables, the ability to predict which plots were invaded in any
year increased from 1970 to 1993. This supports our hypothesis that in
early invasion stages, with dispersal limitation, an invader may occu
r in only a subset of suitable sites giving a weak relationship with s
ite variables. By 1993, distance to the forest margin was no longer re
lated to which plots were invaded, and invaded plots had more species,
occurred at lower elevations on more sheltered topographic positions,
and had more fertile soils than uninvaded plots. Even though site var
iables were not independent (e.g., plots on fertile soils tend to have
more species), multiple logistic regression showed that, all else bei
ng equal, invaded plots still tended to have more species than those n
ot invaded. Our study therefore questions the hypothesis that, all els
e being equal, species-poor habitats are more prone to invasion by exo
tic species.