Invasive species are well known to cause millions of dollars of economic as well as ecological damage around the world. New Zealand, as an island nation, is fortunate because it has the opportunity to regulate and monitor travel and trade to prevent the establishment of new species. Nevertheless foreign species continue to arrive at the borders and continue to cross them, thus requiring some form of management. The control and management of a new incursion of an invasive species would clearly benefit from predictive tools that might indicate where and how quickly the species is likely to spread after it has established. During the process of spread an invasing species must interact with a complex and heterogeneous environment and the suitability of the habitat in a region determines whether it survives. Many dispersal models ignore such interactions and while they may be interesting theoretical models, they are less useful for practical management of invasive species. The purpose of this study was to create and investigate the behaviour of a spatially explicit model that simulates insect dispersal over realistic landscapes. The spatially explicit model (Modular Dispersal in GIS, MDiG) was designed as am open-source modular framework for dispersal simulation integrated within a GIS. The model modules were designed to model an an approximation of local diffusion, long distance dispersal, growth, and chance population mortality based on the underlying suitability of a region for establishment of a viable population. The spatially explicit model has at its core a dispersal module to simulate long distance dispersal based an underlying probability distribution of dispersal events. This study illustrates how to extract the frequency of long distance dispersal events, as well as their distance, from time stamped occurrence data, to fit a Cauchy probability distribution that comprises the dispersal module. An investigation of the long distance dispersal modules behaviour showed that, in general, it generated predictions of the rate of spread consistent with those of analytical partial differential and integrodifference equations. However, there were some differences. Spread rate was found to be mainly dependent on the measurement technique used to determine the invasion front or boundary, therefore an alternative method to determine the boundary of a population for fat-tailed dispersal kernels is presented. The method is based on the point of greatest change in population density. While previously it was thought that number of foci rather than foci size was more important in stratified dispersal and that finer resolution simulations would spread more quickly, simulations in this study showed that there is an optimal resolution for higher spread rates and rate of area increase. Additionally, much research has suggested that the observed lag at the beginning of an invasion may be due to lack of suitable habitats or low probability of individuals striking the right combination of conditions in a highly heterogeneous environment. This study shows an alternative explanation may simply be fewer dispersal event sources. A case study is described that involved the application of the spatially explicit dispersal model to Argentine ant spread to recreate the invasion history of that species in New Zealand. Argentine ant is a global invasive pest which arrived in New Zealand in 1990 and has since spread to both main islands of New Zealand, primarily through human mediated dispersal. The spatially explicit simulation model and its prediction ability were compared to that of a uniform spread model based on equivalent total area covered. While the uniform spread model gave more accurate predictions of observed occurrences early in the invasion process it was less effective as the invasion progressed. The spatially explicit model predicted areas of high probability of establishment (hot spots) consistent with where populations have been found but accuracy varied between 40-70% depending on the year of the simulation and parameter selection. While the uniform spread model sometimes slightly outperformed or was equivalent to the simulation with respect to accuracy early in the invasion process, it did not show the relative risk of establishment and was less effective later in the invasion when stochastic random events generated by the simulation model were averaged to represent trends in the pattern of spread. Additionally, probabilistic predictions as generated by the spatially explicit model allow the uncertainty of prediction to be characterised and communicated. This thesis demonstrates that heterogeneous spread models can give more insight and detail than one dimensional or homogeneous spread models but that both can be useful at different stages of the invasion process. The importance of compiling appropriate data on dispersal and habitat suitability to aid invasion management has been highlighted. Additionally, a number of important hypotheses that need to be addressed to increase understanding of how species interact with the complex environment, have been identified and discussed.
Identifer | oai:union.ndltd.org:ADTP/211545 |
Date | January 2008 |
Creators | Pitt, Joel Peter William |
Publisher | Lincoln University |
Source Sets | Australiasian Digital Theses Program |
Language | English |
Detected Language | English |
Rights | http://purl.org/net/lulib/thesisrights |
Page generated in 0.0014 seconds