Comparative Evaluation of SELECT and Generalized Additive Models for Estimating Gillnet Selectivity of Snow Crab (Chionoecetes opilio) in the East Sea of Korea
Comparative experimental study reveals size-selective retention across varying mesh sizes in snow crab, indicating complementary utility of SELECT and generalized additive models.
Key Points
Compare the Share Each Length’s Catch Total (SELECT) framework and generalized additive models (GAM) for estimating snow crab (Chionoecetes opilio) gillnet selectivity across various mesh sizes.
Conducted comparative fishing experiments in the East Sea of Korea evaluating four stretched mesh sizes: 210, 240, 255, and 270 mm.
Fit candidate parametric SELECT models (binormal, normal, and gamma) ranked by Akaike’s Information Criterion (AIC) and compared them with non-parametric GAM fits.
The binormal SELECT model had the lowest AIC, while normal and gamma models also showed substantial support (ΔAIC < 2).
Under an equal-fishing-power assumption, the ascending 50% retention length (L50) increased from 70.47 mm for the 210 mm mesh to 90.60 mm for the 270 mm mesh.
The final GAM model and the binormal SELECT model showed closely similar fitted selectivity curves and nearly identical in-sample agreement with observed conditional catch proportions.