Bayesian Missing Value Approach Helps Explore Long Term Nonstationary Population Dynamics

Long term nonstationary population dynamics are rarely applied in population dynamics.  The reasons include both the difficulty in solving the nonstationary models and the presence of the missing historical data.  Atlantic weakfish (Cynoscion regalis) is one of the important fisheries along Atlantic coast.  As for many fisheries, stationary models have been used in its assessment and people often use the data after 1981 which is because of the missing of pre-1981 recreational catch history although commercial catch history can trace back to 1920s.  The weakfish population has been found to be very dynamic and a stationary density dependent population process model fails to capture its long term dynamics.  We used Bayesian missing data approach to model the missing recreational catch pre-1981 based on the observed commercial catch from 1929-1982 and the relationship between commercial catch and recreational catch after 1981.  We compared a nonstationary population process model with the commonly used stationary process model to assess the population dynamics with specific focus on the dynamics pattern of the productivity and the possible driving factors.  Our approach dealing with missing data through the Bayesian selection approach provided a useful way to explore long-term population dynamics.