Biology Reference
In-Depth Information
and D'Antonio 1998; Lonsdale 1999). These factors include disturbance, propagule
pressure, and the species pool of potential invaders (Rouget and Richardson 2003).
Invasions can be facilitated by biological interactions such as pollination and
seed dispersal. Theoretically, incorporating these processes into predictive models
could be very useful, but in most instances it would be extremely difficult to do in
a meaningful way (Araujo and Luoto 2007). Lack of data on the processes, decid-
ing what metric to use in the models, and matching the scale of the process to the
scale where species and environmental data have been collected would be problem-
atic. The issue of matching scales where predictor variables and species data are
collected is a general issue that confronts even models found to have reasonable
predictive value (Underwood et al. 2004).
In developing a useful predictive model, it is essential to only include predictor
variables that are available in the management unit's database, especially in the
case of spatial data. Although other predictors may be very important, if spatially
explicit information is not available for the management unit the model cannot be
used to predict areas of the unit that should be searched for invasives. It may be
possible to include some important predictors, such as propagule pressure, through
the use of available surrogates such as vectors and pathways.
2.5.4 Predicting Risk of Occurrence Using Multispecies Models
Information that can be used for modeling species assemblages is essentially the same
as that for individual species. The main difference is the statistical methods used to
develop the models, not the data themselves (Guisan et al. 1999; Underwood et al.
2004; Ferrier and Guisan 2006). Most landscapes have been invaded by multiple spe-
cies, so an approach focused on assemblages may be very efficient (Underwood et al.
2004). Because of computerized databases, the time required to collect information
on species assemblages is not much greater than for a single species. Nevertheless,
care must be taken with assemblage-based models. Because species tend to respond
individualistically to environmental gradients, predictions of distribution patterns
could either be narrower or broader depending upon the shape of the species response
curves (Austin 2002). In an early detection program, this could result in areas not
being monitored where invasive species do occur, or spending time searching areas
where few if any occur. An additional consideration is that within an assemblage only
one or a few species are truly prone to be problematic. In these instances, it is more
useful to predict where the problem species occur rather than the entire assemblage
(Zimmerman and Kienast 1999; Ferrier and Guisan 2006).
Multiple species models assume that species within an assemblage respond simi-
larly to environmental gradients. Numerous studies have shown this assumption is
tenuous, so great care needs to be used when using these models. Careful analysis
of species distribution data is needed before developing models to determine
whether the assumption of similar niche responses among species is justified. Even
if the assumption appears justified, the results need to be interpreted cautiously.
