A workplan to improve the stock assessments for tropical tunas was completed taking into consideration the results of the external review of yellowfin. The yellowfin review panel did not single out a particular model configuration as a replacement for the current base case model, but suggested a variety of alternatives for the staff to consider.

To encompass as many scenarios as possible, the staff developed a pragmatic risk assessment framework to apply for both species, which included the development of hypothesis, the implementation and weighting of models, and the construction of risk tables based on the combined result (SAC-11-08, SAC-11-INF-F,SAC-11-INF-J).

To solve the inconsistencies, spatiotemporal models were used to produce new purse-seine and longline indices and associated length frequencies, but the inconsistencies were not resolved (Figure B-4).

Figure B-4: Top - Relative abundance indices derived from catch per unit of effort of purse-seine and longline fisheries standardized using spatiotemporal models. Bottom – Spatial domain of the purse-seine and longline derived indices.

The mismatch was most apparent in 2001-2003, when a peak occurred earlier in the longline index and later in the purse-seine index (opposite to what was expected given the growth and selectivity assumptions of the model). By incorporating length classes in the standardization, it became clear that the differences were due mainly to the 1998 cohort (of an important El Niño year) being prominent in the longline index, while not showing in in purse-seine index, and the opposite occurring with the 1999 cohort (an equally important La Niña year). Spatial heterogeneity was considered as the most plausible explanation for the unresolved inconsistencies.

Three overarching hypotheses related to the degree of spatial mixing of the yellowfin tuna stock in the EPO where developed (SAC-11-INF-J). Of those, the high-mixing hypothesis was assumed for the benchmark assessment, with the purse-seine index assumed the most representative of the core of the exploited population (SAC-11-07). A series of lower hierarchical level hypotheses were developed regarding other major uncertainties in the previous assessment.

From those, 12 reference models were developed, which combine components that address changes in selectivity and catchability, growth, asymptotic selectivity, and density-dependence in the index catchability as shown in (Table B-1).

Table B-1: Model configurations (hypotheses) used for yellowfin tuna in the EPO (from SAC-11-08 Table A)

Each reference model was run with four steepness of the stock-recruitment relationship assumptions (0.7, 0.8, 0.9, and 1.0). A total of 48 models composed the benchmark assessment for yellowfin tuna (SAC-11-07). In addition, new fishery definitions were implemented, and spline selectivity functions were adopted for most fisheries. All EPO catches were added to the models, which were fit to a standardized purse-seine index of abundance for the EPO north of 5°N and to the length-composition data from the purse-seine fisheries that operate north of 5°N, in order to avoid contamination of the signal with that of a possible southern population. The models were diagnosed for model misspecification, lack of fit, retrospective bias, among others (SAC-11-07). Rather than choosing a base-case model, all models where used to produce management advice by combining them using relative weights determined based on several criteria, including performance on model diagnostics (SAC-11-INF-J).