Biological Data Assimilation
The plankton ecosystem provides the foundation of the marine food chain. A central goal of marine ecology is to understand the spatial and temporal variability of these lower trophic levels, and ultimately to make quantitative predictions of the current and future state of the ocean. This sub-project is concerned with developing statistical methods for biological data assimilation, i.e. estimating the marine ecosystem state using nonlinear ecological models and available marine observations of various types.
In the last decade, modern ocean observing systems have transformed ocean data collection. A variety of new data types are now available such as optical time series, satellite imagery, and autonomous profiling platforms (underwater gliders and drifting underwater floats). These marine ecological data are characterized by high sampling rates, and very complex spatial and temporal dependence. Interpretation these data is facilitated by "biogeochemical" models describing the interactions between components of the ocean ecosystem. They are typically posed as nonlinear (stochastic) differential equations of varying complexity and dimension.
The statistical problem thus considered is how to hindcast, nowcast and forecast the ocean ecosystem state, and estimate key parameters, relying on both dynamical ecosystem models and observations. Our current work has identified sequential Monte Carlo approaches as a promising means to make predictions using non-Gaussian measurements, and complex nonlinear ecological models.