Data Assimilation for Ocean Biology
Mike Dowd, Dalhousie University
New marine observation technologies and dynamical models are improving our understanding of the ocean. A major challenge is identifying and developing statistical approaches that can efficiently and effectively combine the large-scale, nonlinear, spatio-temporal numerical ocean models (that encapsulate our mechanistic understanding of the system) with the wide variety of available data types (e.g. time series, spatial imagery, time-space transects). This problem is termed data assimilation in the ocean and atmospheric sciences. In this talk, I explore the data assimilation in the context of ocean biology. My focus is mainly on lower trophic levels (the planktonic ecosystem or marine biogeochemistry), but I will also discuss some work with higher trophic levels (fisheries and marine mammals). I argue that Bayesian approaches and state space models provide a unifying framework for state and parameter estimation for such systems, including the treatment of model identification and sampling design. Challenges and potential new statistical directions are emphasized.