Data Assimilation Meetings at Reading
Date | Meeting type | Speakers |
16 Jan 2013 | Informal DARC meeting |
Alex Moodey (University of Reading) On Error Dynamics and Instability in Data Assimilation (pdf) Cycled data assimilation algorithms are a standard tool to adapt dynamical systems to measurements, usually based on a diversity of different observation systems. The data assimilation scheme produces an analysis, which is a best estimate to the state of the system. Errors in the observations, the model and the observation operator will lead to some analysis error in each time step. This analysis error can increase over time and lead to severe instabilities in the assimilation process, in particular when the analysis equation Hx=y is ill-conditioned or ill-posed. In this presentation we consider an estimate for the analysis error and derive conditions such that analysis error will be stable over time. Work will be presented for both linear and nonlinear model dynamics. |