Risk assessment for invasiveness differs for aquatic and terrestrial plant species

Abstract

Predictive tools for preventing introduction of new species with high probability of becoming invasive in the U.S. must effectively distinguish non-invasive from invasive species. The Australian Weed Risk Assessment system (WRA) has been demonstrated to meet this requirement for terrestrial vascular plants. However, this system weights aquatic plants heavily toward the conclusion of invasiveness. We evaluated the accuracy of the WRA for 149 non-native aquatic species in the U.S., of which 33 are major invaders, 32 are minor invaders and 84 are non-invaders. The WRA predicted that all of the major invaders would be invasive, but also predicted that 83% of the non-invaders would be invasive. Only 1% of the non-invaders were correctly identified and 16% needed further evaluation. The resulting overall accuracy was 33%, dominated by scores for invaders. While the overall accuracy increased to 57% when the points assigned to aquatic life forms were removed, 57% of the non-invaders required further evaluation rather than were identified as having low probability of naturalizing. Discrimination between non-invaders and invaders would require an increase in the threshold score from the standard of 6 for this system to 19. That higher threshold resulted in accurate identification of 89% of the non-invaders and over 75% of the major invaders. Either further testing for definition of the optimal threshold or a separate screening system will be necessary for accurately predicting which freshwater aquatic plants are high risks for becoming invasive.