Predicting the future distributions of pests: Hybrid methods for current and potential distributions

We report on the development and trial of a general model, PestSpread, that predicts the future distribution of pest species, and discuss hybrid methods to provide two of its key spatial inputs: current and potential pest distributions.  The model was designed to accommodate a wide range of species and data availability, using population growth and dispersal modules, allowing different modules to be chosen to suit the requirements of different species, and new modules to be developed as needed.  It was trialled with eight plant pest species that were chosen to represent both a narrowly distributed species and a widely distributed species in each of four growth forms (tree, shrub, grass, vine).  The availability of demographic and distributional data varied widely amongst the eight species. Generally, the widely distributed species had better information on demographics and potential distribution, but current distributions were often under-reported.   Conversely, the narrowly distributed species had more complete information on current distributions, but often limited information on demographics or potential distribution.  We found that existing monitoring data could not provide robust estimates of either current or potential distributions due to lack of appropriate information and significant spatial biases.  We demonstrate hybrid methods that can incorporate data of various quality and information content to make the best estimates of pest distributions.  More effective international exchange of common pest distribution data and models would greatly facilitate the construction and use of such models for maintaining national biosecurity.