Standardizing us blue marlin longline cpue using habitat covariates

The relative value of species habitat at the location of a longline set can be estimated with a species distribution model (SDM). These data can then be used as covariates in GLM CPUE standardizations to replace intra-annual spatiotemporal strata (area, season). We compared the two approaches using blue marlin (Makaira nigricans) data from US longline logbooks. GLM covariates included 1) gear features, month and area, or 2) gear features and either a habitat coefficient (w) or habitat relative density (H). Habitat relative densities were obtained from either of two SDMs, and habitat coefficients were estimated from H using hook depths of individual gears. All GLM standardized abundance predictions differed from the trend in nominal catch rates. The trends predicted with the habitat-based covariates for the baseline SDM were essentially the same as those from the standard approach using month and area (r=0.98, n=30). Those with covariates from the alternative SDM were also very similar (r=0.95-0.96). SDMderived habitat covariates could obviate problems with statistical imbalance and improve standardizations in many situations.