A Walleye Pollock (Theragra Chalcogramma) Depletion Estimator For The Eastern Bering Sea

• Main Author: Battaile, Brian Charles
• Other Authors: Terrance J. Quinn, II, Kelly, Brendan, Sigler, Mike, Adkison, Milo
• Format: Doctoral or Postdoctoral Thesis
• Language: unknown
• Published: 2005
• Subjects: Aquatic sciences; Bering Sea; Fairbanks; Theragra chalcogramma; Alaska
• Online Access: http://hdl.handle.net/11122/8771

Dissertation (Ph.D.) University of Alaska Fairbanks, 2005 The decline of the Steller Sea lion in the eastern Bering Sea over the last 25 years has resulted in increased management of the pollock fishery due to requirements of the Endangered Species Act, as food competition was hypothesized to contribute to the decline. Our research focused on determining if the pollock fishery was causing significant depletion in the eastern Bering Sea, particularly in Steller sea lion critical habitat. DeLury depletion models were fitted to catch and effort data from 1995 to 1999, from the observer program, which required considerable processing to obtain a database at a temporal and spatial scale that is much finer than that used for stock assessment in the eastern Bering Sea. The catch per unit effort (CPUE) data were standardized in a unique way in that the data were stratified in space and time and standardized using separate general linear models for each stratum. A significant amount of depletion was detected in the pollock fishery from 1995--1999. Depletion estimates of fishery mortality tended to be an order of magnitude smaller than those found in traditional stock assessments. Post hoc analyses indicated that depletion is detected more easily in areas of low abundance due to the hyperstable relationship between CPUE and biomass, possibly exacerbated by a lack of search time in the model. Evidence further suggested that dispersing exploitation pressure decreases local depletion, and pollock may repopulate a depleted area within weeks. Finally, a hierarchical spatial Bayesian analysis with a conditional autoregressive model was constructed to unify the analysis. Because the data were relatively clean of outliers and not over dispersed, significant changes in the results between the frequentist and Bayesian based analyses were not found as was little evidence of spatial autocorrelation in the estimates of catchability.