Improved growth estimates from integrated analysis of direct aging and tag-recapture data: An illustration with bigeye tuna (Thunnus obesus) of the eastern Pacific Ocean with implications for management

The growth modeling framework to integrate age-at-length and tag-recapture length-increment data sources is well established in fishery science. Surprisingly, the use of the so-called Laslett-Eveson-Polacheck (LEP) statistical approach has been infrequent since its development in the early 2000s. Likely, this is due to the complex treatment of process error by the methodology and its resulting high computational demands. Using bigeye tuna in the eastern Pacific Ocean as an example, this research presents a modified LEP method to strengthen the benefits of integrating the two sources of data into a single-step modeling approach. The bigeye tuna case study is typical in the tuna world in which age estimates from direct otolith readings commonly cover the young segment of the species life-cycle only, causing growth model predictions extrapolated beyond the data to be unreasonably high and contain uncertain values of the asymptotic length parameter L-infinity. In the case of bigeye, several improvements resulted from integration of an auxiliary tag-recapture dataset containing recoveries from large (older) fish: the estimated growth curve bent over toward a much lower and more plausible value L-infinity in order to explain the tag-recapture observations from older adult bigeye and the uncertainty (precision) of most growth parameters was greatly reduced (increased). The estimate of variability of the length-at-age was lower and estimated with greater precision. It is argued that it is reasonable to relax the complex assumptions made in the original LEP approach of treating L-infinity as a random variable and assuming dependency between the lengths at tagging (l(1)) and recovery (l(2)) while treating the unknown age at release (A) for each tagged fish as a random variable. These modifications greatly reduce computational demand, potentially making the LEP approach more widely used, and simplifying the interpretation of variation of length-at-age for use in stock assessment models that fit to length composition data. In addition, opportunities are opened to integrate the methodology inside modern statistical stock assessment models. (C) 2014 Elsevier B.V. All rights reserved.