Screening of algorithms for ensemble modelling of marine and terrestrial invasive species to New Zealand
Aguilar, Glenn; Gilchrist, Rowena; Blake, Jett
Citation:Aguilar, G., Gilchrist, R. , and Blake, J. (2013). Screening of algorithms for ensemble modelling of marine and terrestrial invasive species to New Zealand. NZBI National Education and Training Seminar (NETS). Shantytown, Greymouth. NOTE: ABSTRACT ONLY
Permanent link to Research Bank record:http://hdl.handle.net/10652/2850
Performance based screening of species distribution model (SDM) outputs to produce an ensemble of results was used to determine the suitability of New Zealand to highly threatening terrestrial and marine invasive species. Recorded occurrences of the yellow crazy ant (Anoplepis gracilepis) and Northern Pacific sea star (Asterias amurensis) in the nearest land masses were used as presence data for the models. The modeling process also employed terrestrial environmental variables from the Worldclim database and marine environmental layers from the BioOracle database. Performance evaluation metrics including the Area Under Curve (AUC), kappa and true skill statistic (TSS) were used as the basis for selecting the best performing algorithms used for creating ensembled outputs. The resulting maps show that the coastal waters of the country have generally higher suitability for the sea star while the terrestrial areas are not as suitable for the yellow crazy ant. Results provide an indication of the comparative performance of different algorithms, the limits of available environmental datasets and applicability of an ensemble approach to marine and terrestrial species distribution modeling.