Data Assimilation

The goal is to develop the theory of data assimilation, including methods to treat data and model uncertainty, to underpin applications in NCEO and partner agencies. Current research includes:

• Setting up a hierarchical parameter estimation modelling system (from a simple Lorenz system to the complex Carbon Cycle Data Assimilation System (CCDAS)) and hence developing an algorithm to quantify objectively the impact on parameter estimation of removing certain observations from the assimilation system (i.e. quantify the information content of certain observations).
• Investigating variational data assimilation techniques for multi-scale systems.
• Investigating the use of model reduction techniques for unstable systems, for use in incremental data assimilation.
• Setting up simple models for the study of data assimilation in coupled systems.
• Reviewing the methods of Bocquet's maximum entropy approach to tracer source inversion.
• Setting up Markov Chain Monte Carlo (MCMC) methods for a strongly non-Gaussian data assimilation problem, framing the problem via an infinite dimensional stochastic dynamical system.