Catch me if you Stan: Bayesian estimation of size-specific selectivity and size spectrum parameters from trawl data

Background/Question/Methods

Trawl-based surveys are a critical tool for monitoring marine populations and communities. However, they suffer from a major limitation: smaller fish are less likely to be caught by trawls (i.e. trawls have low selectivity for smaller fish). Therefore the contribution of small fish to population biomass can be underestimated. The size=selectivity relationship is therefore a key function in fishery biology. It is possible to predict the expect frequency of smaller fish in a sampled area using metabolism-based size spectra models. Using trawl data, we wanted to simultaneously estimate the size-selectivity curved and species- and location-specific size spectrum parameters defining the population structure. Assuming the two primary processes describing the shape of a species distribution in trawl data are the species-specific size spectrum and a selectivity curve, a Bayesian model was constructing using the rstan package in R to estimate size spectra and selectivity parameters simultaneously. This model was then applied to 15 years of Greenland halibut (Reinhardtius hippoglossoides) survey data to determine size-specific catchabilities to Alfredo III gear and assess potential impacts of the expanding fishery on Greenland halibut population productivity.

Results/Conclusions

We found that our model converged around biologically feasible parameter estimates. A correction factor was also determined to account for the possibility of diurnal variability in catchability and to accommodate varying maximum catchability in species based on lifestyle and behaviour. Deviations from the main curve indicate this method could help detect year class strength. Preliminary analysis of multiple years of data indicate that the descending slope of the size spectra has become steeper over with time, correlated with increased harvest rates and increased energy requirements for maintenance. Although these trends don’t lend themselves to immediate concerns, it does highlight the importance of continual monitoring of these populations to ensure the stock continues to be fished sustainably. The approach proposed in this study combines theories from multiple different fields to provide a novel means of estimating otherwise difficult to measure parameters. Refining methods of estimating catchability and size spectra within an adaptive framework capable of accommodating confounding factors, such as vessel power and gear type, will allow for improved biomass and productivity estimates. Improved estimates of the catchabilities and size spectra of non-target species would furthermore be an asset to understanding community biomass and productivity of these marine ecosystems.