INVASIVE
Conservation Specialist
Potential distribution of Lythrum salicaria in US based on models built from coorelations between climatological variables and native range occurrences (a) or introduced occurrences (b). Actual occurrences of the species are indicated with black dots.
I address the question of whether the introduced distribution of invasive species can be predicted from its native range. I modelled the potential distribution within the United States of three alien invasive species native to Europe using the Maxent modelling technique. Using occurrence data from both the native (Europe) and introduced (US) ranges, I used reciprocal modelling to evaluate habitat discrepancies between the introduced and native ranges. The native occurrences in Europe accurately predicted the distribution within Europe; and introduced occurrences in the US accurately predicted the US distribution. However, the reciprocal models did not perform well. My model results indicate that the occupied niches are too inconsistent between the native and introduced ranges to make models useful at the scale at which early invasive species detection can occur.
The role of biotic interactions will need to play a bigger role in species distribution modelling if they are to be ecologically meaningful. Inclusion of biotic interactions such as overlap with competitor distribution and shared resources will improve model performance. Model predictions based on the native range may under-predict the potential distribution in the introduced range if biotic interactions, such as competition or parasitism, are removed when an alien species enters a new region. But accurately predicting areas of invasion in the introduced range may never truly incorporate the influence of biotic interactions because the introduced species are no longer affected by their native biotic interactions and are subject to another suite of species in the introduced range with which it may form new biotic interactions that are currently indescribable.
Another interesting avenue of research that will significantly improve the modelling of invasive species is the inclusion of mechanistic variables. Instead of relying on correlations with the environment to predict the environmental preferences of a species, a mechanistic model uses information from detailed physiological tolerance experiments to model the fundamental niche of a species. I would expect to accurately model the potential of invasive species is to model the fundamental niche and project that information onto the introduced range. This will not necessarily mean the species will be able to thrive in those locations, because new biotic interactions will be in place to limit the species range.
The methods explored in this research illustrate the potential of natural history collections and survey data have in contributing to modern biogeographical research. Although the data is not perfect and the techniques do not perfectly represent the ecology, we can still take advantage of the newly digitized historical data to answer new and fundamental questions concerning biogeography. Advances in bias reduction will no doubt occur in the next several years. Improvements will be made to modelling algorithms to better represent ecological processes. Predictor data will be enhanced by including biological meaningful and derived variables. Using any technique to model species distribution should be done with care. Too often in the literature it is apparent that the researchers plugged their data into a model, the model drew a map, and the researchers presented the map as truth. This is done with little thought to proper evaluation and noted accuracy. Researchers should understand their goal when they model and verify that their approach is appropriate for that outcome.
Areas Registry