Improving macroscopic maturity determination in a pre-spawning flatfish through predictive modeling and whole mount methods

• We tested a set of multivariable models to predict maturity of southern flounder.
• Both AIC and cross-validation criteria were compared.
• Each selection criteria produced a suite of well-performing models, with cross-validation selecting simpler models.
• These models, along with whole mount procedures, are an improvement over current macroscopic methods.

Accurate maturity schedules are essential for informed management of many fishery resources. Although histological methods are generally acknowledged as the best approach to correctly assign maturity status of individual fish, the methods can be expensive and time consuming. We developed and tested a set of multivariable models to predict maturity of southern flounder, a valuable flatfish occupying estuarine and coastal systems in the southeastern US. We also evaluated the potential for whole mount methods to validate maturity assignments and help discriminate transitional oocyte stages. Lastly, we used one of the better performing models to conduct retrospective analysis of variability in southern flounder maturity schedules. Several models performed well in predicting southern flounder maturity; nearly half of the models we tested achieved ≥85% prediction success. We noted that the gonadosomatic index (GSI) was included in most of the higher performing models and, by itself, was a strong predictor of maturity for southern flounder. The addition of novel quantitative predictors, such as gonad color and dimensions, pushed model success above 90% in many cases. Whole mount methods showed a high level of agreement with histological methods, and should be investigated as an inexpensive alternative for validating maturity assignments. Retrospective analysis revealed the potential for interannual fluctuations in L50 of 2–5 cm for southern flounder, which can impact yearly estimates of SSB and target harvest rates. Multivariable predictive models using routinely collected fishery biological data can provide reliable estimates of fish maturity and, when coupled with whole mount methods, should represent an improvement of traditional macroscopic maturity assignment.