Predicting invasive species' range expansion using models built in the native range

Background/Question/Methods

 The relationship between a species and the landscape can differ between its native and non-native ranges due to different competitors, resource availability and climates. For prediction purposes however, it is pertinent to use native range relationships to inform distribution models in the non-native range. Our study examines the landscape factors that determine the invasive mute swan’s distribution in its native and non-native ranges. The relationships between mute swan’s distribution and landscape factors were determined using generalized linear models and classification trees. We created three possible theoretical models that could explain mute swans’ distribution in its native range. We then applied the best model to the non-native range, and compared the resultant probable distribution maps with those created using non-native range data. We also used non-native range data to build a model to apply to the native range. 

Results/Conclusions

Models exhibited asymmetrical transferability as models built in the native range were more able to accurately predict mute swan distribution in the non-native range while the reciprocal transfer of the model did not exhibit high predictive accuracy. Overall, it is apparent that many of the significant landscape factors determining mute swan distribution actually remained the same between its native and non-native ranges. However, using native-range models to develop probability distribution maps resulted in a slight over-prediction of mute swan distribution in its non-native range. This was likely due to the relatively recent introduction of this invasive species to the non-native range. Establishing which landscape variables are highly correlated with invasive species’ distribution in both native and non-native ranges will help to ascertain where invasive species are likely to spread next as their population grows. This has important implications for management efforts that may require international data to supplement local distribution data.