A review of the data availability, model configuration and catch estimation for the 2017 blue shark (Prionace glauca) stock assessment in the Indian Ocean.

This paper presents a review of the 2017 stock assessment of blue shark in the Indian Ocean using Stock Synthesis (version 3.24f http://nft.nefsc.noaa.gov/Download.html).  This paper is largely based on the assessment document (IOTC–2017–WPEB13–33 Rev_1), as well as the catch estimation document (IOTC-2017-WPEB13-23). Herein the “assessment” refers The blue shark assessment model is an age structured (25 years), spatially aggregated (1 region) and two sex model. The catch, effort, and size composition of catch, are grouped into 8 fisheries covering the time period from 1950 through 2015. Seven indices of abundance, all from longline fisheries, were available as well as three alternative time series of total catch. The base case model is parameterized using indices of abundance from the Portugal (2000-2015), Reunion (2007-2015) and the Japanese late (1992-2015) series, along with estimates of catch generated via a generalized additive model. The estimated abundance trend is decreasing throughout the time frame of the model, and spawning stock abundance has decreased to approximately 1.503 times SSBMSY, (80% CI is 1.33-1.63). The fishing mortality has increased steadily over the model time frame with F2015/FMSY= 0.904 (80% CI =0.68 to 1.13).

Blue sharks are most often caught as bycatch in the Indian Ocean tuna fisheries, though some directed mixed species (sharks and tunas/billfish) fisheries do exist. Commercial reporting of landings has been minimal, as has information regarding the targeting and fate of blue sharks encountered in the fisheries. Useful data on catch and effort is mostly limited to recent years, and time series of historical catches have been estimated based on reported and observed catch rates, as well as observed ratios of blue shar to target species.

This analysis was developed as an assessment model that included the Portuguese, EU-France (Reunion) and Japanese late CPUE series, with the estimates of total catch based on generalized additive model based GAM series as the reference case, as it is referred to in the main text when presenting the model parametrization and diagnostics.  The 13^th meeting of the Indian Ocean Tuna Commission Working Party on Ecosystems and Bycatch (WPEB13) recommend this parameterization as a base case model for the provision of stock status. A grid of sensitivity runs using the alternative CPUE series is presented to characterize one of the major axes of uncertainty. These models vary in their groupings of CPUE series for inclusion in the model, these groupings were determined by hierarchal cluster analysis and expert opinion during the WPEB13 meeting. Although the alternative catch series available were considered unreliable for use in the base case model, sensitivities using the nominal, EUPOA estimated catches and trade base catches were analysed to represent assumptions of higher and lower total catch. As such the estimated stock status differs between combinations of the catch datasets and CPUE series.

The results of the assessment are compared across different groupings of CPUE series and show the  reference case parameterization resulting in estimates of  SB₂₀₁₅ /SB_MSY  =1.503%  and F₂₀₁₅/F_MSY  = 0.904  though the range of uncertainty, based on alternative model runs considered covers   1.49-2.36 and 0.29-0.904 for SB₂₀₁₅ /SB_MSY and F₂₀₁₅/F_MSY  respectively.  Stock status is reported in relation to MSY based reference points, however, the authors note that the IOTC has not yet adopted reference points for sharks.  Due to the inherent unreliability of recruitment estimates in the terminal year  this study defines ‘current’  as the  average of the first four of the last five years (i.e. 2011-2014), and reports ratios of SB and F as current as well as with respect to 2015.

The main conclusions of this assessment are:

• The stock status is highly dependent on the CPUE series used to fit the model. Among the candidate CPUE models in this assessment no CPUE series runs through the entire time series.
• The estimates of catch are highly influential in the model, but mostly in terms of scale, as the current depletion and fishing mortality indicators are approximately equal across all  catch estimates for a given CPUE series.
• The scale of the assessment is influenced by the CPUE series chosen, across these estimates the estimates of B0 range from approximately 1million mt to approximately 1.9 million mt.

When considering which model(s) to use for the provision of management advice, it is recommend that advice be based upon multiple model runs that consider the major axes of uncertainty.