Statistical models that set environmental determinism
and dispersal hypotheses against each other, while
allowing natural spatial variation in the sampled environment,
have shown that the environment appears
to be more important than dispersal at large spatial
scales (Tuomisto et al. 2003). However, at scales 1
ha and less than tens of square kilometers (the mesoscale),
the effects of the environment and dispersal are
difficult to distinguish because of a common correlation
between environmental gradients and the geographic
distribution of sampling points (Duivenvoorden et al.
2002). Gilbert and Lechowicz (2004) used a spatially
structured sampling design to break the correlation between
the environment and the geographic location,
and were able to show that environmental correlates of
plant distributions remain important at the mesoscale,
but spatial correlates do not. These studies of diverse
plants (from trees to ferns to graminoids) in both temperate
and tropical regions suggest that, at the mesoscale
and the large scale, the abiotic environment can,
in fact, be as important in affecting plant distributions
as any simple spatial effects arising from dispersal acting
independently. This contrasts with the dominant
role for stochastic factors associated with dispersal that
is predicted at larger spatial scales in neutral models
(Bell 2001, Hubbell 2001).