Patterns in oyster natural mortality in Chesapeake Bay, Maryland using a Bayesian model

• Published in: Fisheries research 2021-04, Vol.236, p.105838, Article 105838
• Main Authors: Doering, Kathryn L., Wilberg, Michael J., Liang, Dong, Tarnowski, Mitchell
• Format: Article
• Language: English
• Subjects: Bivalve natural mortality; Box count method; Crassostrea virginica
• ISSN: 0165-7836; 1872-6763
• DOI: 10.1016/j.fishres.2020.105838

•Natural mortality estimates were derived from observations of live and dead bivalves.•A Bayesian model was used to correct biases in natural mortality estimates.•Natural mortality of eastern oysters varies spatially and temporally in Chesapeake Bay, Maryland.•Two common trends in natural mortality were identified using dynamic factor analysis.•There was a general north-south gradient in which trend contributed more to the natural mortality pattern in an area. Bivalves leave behind articulated valves when they die that can be used to estimate natural mortality. However, a common method used to estimate natural mortality in bivalves, known as the box count method, includes several assumptions that may be violated for eastern oysters Crassostrea virginica in Chesapeake Bay, Maryland. We developed a Bayesian model alternative to the box count method that included modifications to account for potential violations of assumptions. The model was applied to oysters in 32 areas in Maryland during 1991–2017 using dredge survey data, and the natural mortality estimates from the model were compared to ones derived from the box count method. The spatial and temporal trends in natural mortality from the model were summarized using dynamic factor analysis. Natural mortality showed considerable spatial and temporal variation, with median mortality rates ranging between 0.00 and 0.96 yr−1. Natural mortality spiked in most regions in 2002 and was lower in more recent years. The Bayesian model estimated slightly higher (0.02 yr−1 on average) natural mortality than the box count method, except for years following high natural mortality, after which the Bayesian model estimated lower natural mortality than the box count method. The dynamic factor analysis revealed two common trends in natural mortality and a north-south gradient in the loadings on the trends. This work improves our understanding of the variability of oyster natural mortality in Maryland and the Bayesian model could be modified for use with oysters in other regions or with other species.