Evaluation of ecological risk assessment protocols and application to non-native fishes in trade in the Laurentian Great Lakes

Background/Question/Methods

A global increase in international trade has led to an increase in both intentional and unintentional transport of nonindigenous species. Because of this, there is an increased likelihood that more species will become established and/or invasive outside of their native geographic ranges. We evaluated the performance of five ecological risk assessment tools for predicting establishment and impact of fishes using a validation data set of species with a known invasion history in the Great Lakes. Following evaluation, one statistically-based and two questionnaire-based tools were selected based upon their ability to best predict both establishment and impact. For each of these tools, thresholds distinguishing either established from failed or high impact from low impact species were identified by plotting the classification accuracy by score for each of the analysis types. We identified fishes in live trade (aquarium, live food, biological supply, and water garden trades) in the US and Canada, but not yet present in the Great Lakes, as a priority for screening. The three risk assessment tools were subsequently applied to the 12 freshwater fishes in live trade that were identified as having an environmental match to the Great Lakes.

Results/Conclusions

All five of the evaluated tools performed well in distinguishing either established from failed invaders or high impact from low impact species. We performed ROC analyses, with resulting Area Under the Curve (AUC) values above 0.77 for all protocols (an AUC value greater than 0.7 indicates that a model is adequate at distinguishing two groups (e.g., established vs. not established or invaders vs. failed invaders)). The selected questionnaire-based protocols have AUC values above 0.9, while the statistical tool has a value above 0.8. While the statistical tool has a lower performance metric than the other tools, it has other features that could make it preferable. It requires the user to gather ecological trait data, so is less subjective than the questionnaire tools, which require more subjective input. It usually takes less time to gather data. The questionnaire tools provide more detailed species information. The statistical tool predicted establishment of all species in trade and impact for most of the species, while the questionnaire tools had lower completion and prediction rates. The results of the assessments of species in live trade provide a species list that allows for better prioritization of resource allocation and pre-border screening.