Indirect estimation of contact selectivity for gill nets using hierarchical models

Indirect estimation of selectivities for gill nets is frequently conducted to evaluate size-related differences in vulnerability, with catches typically pooled across multiple net sets when estimating. An alternative to pooling catches would be to use hierarchical modeling to estimate selectivities, which would allow for variability in selectivity among sets to be assessed and to account for spatial or temporal autocorrelation across sets. We estimated selectivities for gill nets using several hierarchical model formulations using walleye Sander vitreus catch data from two experimental gillnet configurations in Lake Erie. Hierarchical selectivity curves were more supported from an information-theoretic perspective than nonhierarchical versions, although convergence and model complexity issues arose for some models incorporating spatial autocorrelation. Hierarchical bi-normal selectivity curves, where set-specific parameter deviations were modeled through univariate normal distributions were most supported by available data for both configurations. Given the availability of flexible software for fitting and diagnosing hierarchical models, we recommend that hierarchical modeling of contact selectivity be considered to improve understanding of set-specific variability in selectivities with multiple deployments rather than automatically pooling catch data.