SCIENTIFIC COMMITTEE SEVENTH REGULAR SESSION 9-17 August 2011 Pohnpei, Federated States of Micronesia SPATIAL SIZE DATA STRATIFICATION FOR LENGTH-BASED STOCK ASSESSMENTS

Introduction Length data from the Japanese distant-water and offshore longline fleets are available aggregated in spatial strata of 10 degrees of latitude by 20 degrees of longitude, and more recently 5 degrees by 10 degrees (Figure 1). In previous yellowfin assessments, quarterly length frequency distributions were derived for the principal longline fisheries weighted by the spatial distribution of the quarterly catch from the individual fishery. However, there is considerable spatial variation in the sizes of longline-caught yellowfin and bigeye tuna within individual regions of WCPO stock assessments. There have also been large shifts through time in the spatial distribution of both longline catch and size sampling. These changes have influenced the composite regional-specific length compositions. Length sample locations aggregated by decade for yellowfin tuna are illustrated in Figure 1. Locations are even more variable at the year-quarter scale used in stock assessments, but it is the long-term shifts in sample locations that have the greatest effects on stock assessment outcomes (Harley et al. 2010). For the current 2011 bigeye and yellowfin assessments, a new approach was applied to re-stratify the size frequency data according to our best estimate of the distribution of the population, to reduce the influence of spatial changes in the distribution of catch and sample collection. The objective of this approach was to generate size frequency distributions that were more consistent with the underlying size distribution of the population within a region (mediated by the long-term average selectivity of the fishery). This change was necessary because in a catch at length model such as MULTIFAN-CL (Fournier et al. 1998), the predicted catch is removed rather than the observed catch. According to the separability assumption used in MULTIFAN-CL, the length distribution of the catch is a function of the population length distribution and the selectivity. i.e. , and therefore (Quinn & Deriso 1999). Selectivity does not change from year to year in the model, so the model interprets changing observed lengths as changing population lengths. If length distributions change due to fleet movements coupled with spatial variation in fish sizes, rather than population length changes, the model results will be affected and potentially biased, depending on the relative weighting assigned to the size data. Methods The following procedure was applied to generate an aggregated year/quarter length composition for a specific region longline fishery from the Japanese length frequency data. i. In order to define the average long-term spatial …