Assessing Invasion Risk and Associated Economic Loss of Fish Species Across Global Watersheds Using Machine Learning Algorithms

The intention of this study is to compare invasion risk and economic loss for selected fish species in both different watersheds and countries on a global scale. The potential risk of invasion for four different species was assessed by the prediction models developed from machine learning algorithms and joint dataset comprising species occurrence records and relevant environmental information. Combining invasion risk and unit economic costs derived from a comprehensive invasion cost database also enabled us to calculate the economic loss of invaders in three example countries. Results showed that the multi-layer perceptron algorithm, which was selected as the best model out of them, successfully identified areas of very-high-to-no risk for four selected fishes. The potential risk of invasion was relatively high for Perccottus glenii and low for Salmo trutta in the Republic of Korea, as compared to that of Japan and Australia. In addition, both invasion risk and unit economic costs were found to be largely responsible for the economic loss of selected fish species, specifically the species Petromyzon marinus in Australia. We expect that the methodology proposed in this study can be used to address the potential risk of invasion for non-native species along with other similar models such as species distribution models, and can be used to recommend affordable management options for controlling their populations.