Japanese Longline CPUE Standardization (1979-2019) for striped marlin (Tetrapturus audax) in the Indian Ocean using Bayesian hierarchical spatial model

To estimate a historical trajectory of striped marlin stock abundance in the Indian Ocean, we standardized the CPUE of striped marlin caught by Japanese longliners for 1979-2019. We separated the logbook data into four areas (NW, NE, SW, SE) based on the IOTC area definition, and divided the time-period into two periods, 1979-1993 and 1994-2019. In this analysis, we applied Bayesian hierarchical spatial models. Since the catch data is countable and characterized by many zeros, we used zero-inflated Poisson generalized linear mixed model (ZIP-GLMM). All analyses were performed using R, specifically R-INLA package. The INLA procedure, in accordance with the Bayesian approach, calculated the marginal posterior distribution of all random effects and then estimated parameters involved in the model. We applied a half Cauchy distribution as a prior for the random effect. Best model was selected from multiple models using Widely Applicable Bayesian Information Criterion (WAIC) for each area in each period. Gradual annual decline trends with interannual variations were generally observed for all the standardized CPUEs. The 95% credible intervals were wider due to the inclusion of spatial effect as compared to the previous non-spatial model for 1994-2017 (Ijima 2018), while the point estimates of the standardized CPUE trends were similar. To reduce the uncertainty in the estimation of the standardized CPUEs, selecting an appropriate catchability (q), applying the state space model and/or latent variable model will be essential to improve the stock assessment of this species.