Evaluating ecological-niche factor analysis as a modelling tool for environmental weed management in island systems.

Management actions are essential for mitigating the potentially harmful changes in biodiversity, ecosystem function and crop/forest productivity caused by invasive species. Species distribution models, if reliable, could be used to design effective management strategies. Although several modelling methods well suited for studying invasive species have been developed for presence-only data, often the size of available data sets for modelling is small and results are not validated with test samples. Moreover, the impact of such methods in practical applications has been overlooked. Here, we evaluated the reliability of the modelling approach based on ecological-niche factor analysis ( ENFA) implemented in Biomapper software when applied to environmental weed data in the Azores. Presence-only data sets of two top invasive woody species ( Pittosporum undulatum and Acacia melanoxylon) were used. The continuous Boyce curve was used for validation, calculated either in Biomapper (cross-validation) or based on test samples. The species' most habitable areas that should be regarded as management targets were thus estimated from modelling and validation. By imposing size restrictions on the presence-only data sets used in modelling and validation, other habitable areas were defined and compared. The ENFA proved to be a suitable method for modelling environmental weed distributions, regardless of the presence-only dataset size. Moreover, the cross-validation of Biomapper was reliable, although its results should be interpreted with caution as they could potentially lead to statistically different management target areas. [ABSTRACT FROM AUTHOR]