Evaluation of a Random Forest Model to Identify Invasive Carp Eggs Based on Morphometric Features

Three species of invasive carp—Grass Carp Ctenopharyngodon idella, Silver Hypophthalmichthys molitrix, and Bighead H. nobilis—are rapidly spreading throughout North America. Monitoring their reproduction can help to determine establishment in new areas but is difficult due challenges associated with identifying fish eggs. Recently, random forest models provided accurate identification eggs based on morphological traits, the have not been validated using independent data. Our objective was evaluate predictive performance egg developed by Camacho et al. (2019) for classifying carp an data set. When were grouped as one category, accuracy high at following levels: family (89%), genus (90%), (91%), reduced predictor variables (94%). Invasive decreased when we only considered observations from newly sampled locations (family: 9%; genus: 22%; species: 30%; variables: 70%), suggesting potential differences characteristics among locations. Random a combination previous resulted (96–98%) class all family, genus, levels. The two most influential average membrane diameter embryo diameter; probability predicting increased these metrics. High metrics suggest that trained be used identify morphometric variables. However, suggests more research would beneficial models’ applicability larger spatial region.