Models used to identify the best configuration based on Akaike information criteria weights (ΔAIC) and expected log pointwise predictive density weights (ΔELPD).

Model structure indicates the different configurations tested in the model. Note that bottom temperature (temp) and depth were added as smoothing terms (p-spline). df indicates the degree of freedom, AIC is the Akaike information criteria, and ΔAIC indicates differences among AICs. ELPD is the expected log pointwise predictive density, and differences among ELPDs are indicated by ΔELPD. The best model configuration is in bold.