Integrating spatial habitat and fisheries effort data to improve abundance estimates of west coast groundfish

Principal Investigator: Andrew Shelton
Co-Principal Investigators: Eric Ward, James Thorson, Marlene Bellman, Blake Feist


A comparison of predicted darkblotched rockfish density (kg ha-1) from two models.  Warmer colors indicate higher density areas. A) Predicted density from a model incor-porating habitat information. B) Mean  density in the three statistical regions from a classical stock assessment model that does not incorporate spatial information or include habitat variables. Small points show the location of survey trawls

Species-habitat associations play an important role in ecosystem management and coastal and marine spatial planning. However, species-habitat associations rarely connect with stock assessment efforts. Historically, this disconnect between habitat science and stock assessment was necessary because the fine grained spatial information needed to relate habitats to fisheries surveys and fishing effort were lacking. With the recent collation of spatially explicit habitat information, and fisheries survey and effort data from myriad sources, the links between habitat characteristics and fish abundance are becoming easier to establish, but methodological challenges remain. For example, studies of species-habitat associations are typically assessed in a purely spatial context and tend to ignore or integrate out temporal variation whereas stock assessments typically ignore or integrate across spatial variation and focus on analyzing time-series of abundance indices. Thus, a hurdle to the integration of habitat information into stock assessments is the development of methods that can incorporate the inherently spatial nature of habitat information while preserving the emphasis of stock assessments on abundance time-series.

Our research focuses on understanding how current estimates of the abundance of three west coast groundfish species (shortspine thornyhead, Sebastolobus alascanus; greenspotted rockfish, Sebastes chlorostictus; darkblotched rockfish, Sebastes crameri) can be improved by developing spatio-temporal models that explicitly incorporate information about both habitat variables and fishing effort. We assessed both how fishing may affect the habitat associations of species and looked for signatures of local depletion in the abundance of our three focal species.

We have developed and applied new methods for estimating groundfish abundance from trawl surveys, focusing on methods that directly incorporated habitat information into the estimation process to improve precision and consistency of abundance estimates for groundfish based on the trawl survey. We are particularly interested in developing methods for understanding how fishing effort may interact with habitat to affect abundance of groundfish.

We developed a set of Bayesian delta-generalized linear mixed models (delta-GLMMs) and applied them to the West Coast trawl survey. These models extend traditional stock assessment methods for estimating abundance to incorporate habitat variables but can be used in a variety of applications. We applied our models to a variety of settings to illustrate the value and power of directly incorporating habitat variables.

We completed several example projects to illustrate the potential of spatial, habitat-based approaches to a range of stock-assessment. First, we compared the results from our new approach with traditional stock assessment methods to ask if the inclusion of habitat variables improved the estimation of abundance of a darkblotched rockfish along the west coast. We showed that habitat models improve abundance estimates. Second, we used a habitat model to examine the seasonal distribution of Dover sole and asked if the abundance of this species showed evidence of localized depletion from fishing or if we could identify other mechanisms responsible for observed species distribution. Third, we applied our model to multi-species metrics of fish communities including species richness, evenness, and functional diversity to explore patterns of fish communities along the west coast. Finally, we applied models to eulachon bycatch in the pink shrimp fishery off the Oregon coast. We showed how our models can be used to identify areas of high bycatch potential and assess if these areas are persistent or variable among years.

At present, our methods are beginning the process of becoming a statistical tool available to stock assessment scientists and we hope they will be incorporated into formal stock assessments in the near future. Our work with diversity will be included in the upcoming Integrated Ecosystem Assessment. The work with eulachon will potentially help to inform the designation of critical and essential habitat for this species and inform fisheries bycatch issues in the pink shrimp fishery.