In this post Associate Editor Andre Punt discusses a paper he recently handled by James T. Thorson, Jason Jannot and Kayleigh Somers ‘Using spatio-temporal models of population growth and movement to monitor overlap between human impacts and fish populations‘

In many jurisdictions, including the U.S., conservation management of fish stocks involves comparing catches with estimates of an overfishing limit (OFL). The OFL is a catch considered to be sustainable based on a harvest control rule agreed by fishery managers that reflects their risk tolerance. Historically, evaluation of the conservation status of fish stocks was restricted to target species only. However, increasingly fishery managers consider non-target and bycatch stocks. A variety of the population dynamics models have been developed to provide the input for the harvest control rules used to compute OFLs. One of the simplest of these is the ‘surplus production model’, which tracks changes in the total biomass of a population, and ignores changes in age- and size-structure. This class of model is used in data-poor situations, in particular when few data are available on the age-structure of catches and the population, but there is an index of abundance.

Most surplus production models ignore spatial variation in the assessed population and those models that account for spatial variation ignore movement. However, harvest pressure differs spatially and fish move. Thorson et al. develop a spatial extension of the surplus production model that estimates population growth spatially and accounts for both process and observation error. This newly developed method extends previous approaches by using a computationally efficient estimation algorithm based on Template Model Builder to estimate annual movement (diffusive and advective) and spatial variation in harvest rates, along with the parameters governing a Gompertz growth function. The method is also able to make use of prior information on the parameters of the model.

Thorson et al. evaluate the new method using simulations to determine its small-sample properties, showing that with 600 samples it is possible to estimate unbiased and relatively precise estimates of the overfishing limit. However, estimates of the extent of spatial-temporal error and movement probabilities are biased.

The method is applied to big skate (Raja binoculata) in waters off the U.S. west coast. Big skate is a large and relatively long-lived species, with life history characteristics that suggest that it might be vulnerable to unsustainable catches. Information to estimate population biomass and hence the OFL is data from annual bottom trawl surveys (2003-2013) conducted each summer by NOAA, along with data on fishing effort in self-reported logbooks and data on catches inferred from observer reports.  The model predictions indicate considerable spatial variation in density in the modelled region, with the highest density in Oregon and Washington, as well as substantial year-to-year variation in density spatially. Estimates of the overfishing limit for big skate are much larger than current catches, irrespective of assumptions about movement.

The new method developed by Thorson et al. adds to the toolbox of ways fishery scientists can make better use of available data from fisheries and surveys to provide estimates of sustainable catches to managers, and should lead to lower risks to individual fish stocks, particular those that are data-poor.