The design of a camera-based fisheries-independent survey for untrawlable habitat in the Gulf of Alaska

Biomass and size composition estimates from the fishery independent biennial Alaska Fisheries Science Center Resource Assessment and Conservation Engineering Division’s Groundfish Assessment Program Gulf of Alaska bottom trawl survey are an integral component of stock assessments and management of fishes in the region. A substantial proportion of the seafloor in the Gulf of Alaska survey area consists of rocky, high-relief habitat that is not accessible to trawl survey gear and is important habitat for many harvested fish species, especially rockfishes (Sebastes spp.). Survey estimates of groundfish abundance from trawlable grounds are expanded across the entire region. Therefore, survey abundance estimates could potentially be biased as a result of sampling that does not spatially represent all of the available habitat circumscribed by the Gulf of Alaska study area. Mapping the extent of the untrawlable habitat is a critical step toward understanding the scale of these potential biases and would benefit the design of a future untrawlable habitat survey. Several previous efforts have provided trawlability maps for part or all of the Gulf of Alaska based on different data inputs. In the present work, we compile two previously published untrawlable habitat maps with three other relevant data sources to create a new untrawlable habitat map, and use this map in conjunction with density estimates from existing underwater camera studies to investigate the feasibility of an untrawlable habitat rockfish survey in the Gulf of Alaska. An analysis of 1 km2 grids sampled more than once with a lowered camera system suggested that a single-stage sampling design (one camera transect per grid) would be most effective. Post-stratification analysis showed that either a simple depth stratification or stratification based on species distribution models provided lower coefficient of variation (CV) values at a given sample size compared to random sampling. Furthermore, a target CV of 25 % could be reached for a camera survey of dusky rockfish, northern rockfish, or Pacific ocean perch with less than 200 lowered camera transects if a stratification scheme based on species distribution models was used.