ABSTRACT Objective Red Snapper Lutjanus campechanus is an economically and ecologically important fish for the Mississippi coast. Sustainability concerns are of importance to both stakeholders and managers. The objective of this work is to understand the environmental, habitat, and fishery characteristics that influence catch per unit effort of sublegal- and legal-sized Red Snapper in the northern Gulf of America (also known as the Gulf of Mexico). Methods We used descriptive statistics to understand patterns of habitat- and depth-specific catch per unit effort (CPUE). To explore the factors that determine catch, we employed a gradient-boosting ensemble machine learning technique on CPUE. Continuous CPUE data were modeled by converting them to a binary variable, above and below mean ln(CPUE + 1). We used data from a scientific survey of sublegal- and legal-sized Red Snapper (406.5 mm TL) in the region collected using vertical longline from 2016 to 2019. A suite of predictors (termed “features”) reflecting fishing season, habitat, and environmental characteristics was used in machine learning models. Results In the descriptive analysis of CPUE, we found that positive CPUE for both sublegal- and legal-sized fish was observed at artificial reef and oil and gas platforms in both the mid and deep strata. Conversely, CPUE in undifferentiated bottom habitat was low for all depth strata, with only a single, sublegal fish recorded in the deep depth stratum. Catch-per-unit-effort ln(CPUE + 1) was evaluated, and model output was assessed using variable importance metrics and Shapley additive explanation values for visualization and to determine the direction and magnitude of feature influence. Models were created for both sublegal- and legal-sized Red Snapper. Both models achieved accuracies above 0.90. For sublegal Red Snapper, greater-than-mean ln(CPUE + 1) was associated with high surface salinity and small hook size. For legal-sized Red Snapper, below-mean ln(CPUE + 1) was associated with small hook size. Conclusions Our descriptive analysis indicated that CPUE was greatest in structured habitats. The machine learning model indicated that the variables that predicted binary CPUE for the sublegal- and legal-sized fish were different. For both size-classes examined, our findings that hook size was an important predictor of the binary natural-log-transformed CPUE variable has implications to selectivity and perhaps gear restrictions in the fishery. These analyses can inform fisheries managers of the factors that are most influential in the catch of sublegal- and legal-sized Red Snapper in the study region.
Riedel et al. (Thu,) studied this question.