Data Assimilation with Numerical model error
Seminar Room 2, Newton Institute Gatehouse
Data assimilation addresses the inverse problem of, given a set of uncertain observations and a numerical approximation to a physical system, what set of parameters, especially the initial condition, leads to a forward computation (the 'analysis') which best solves this problem. Data assimilation is widely used in meteorological applications. In this talk I will briefly describe the 4D-VAR method for data assimilation, and then show how its results are influenced by using a variety of different numerical schemes with associated numerical model error.
Joint work with Sian Jenkins, Melina Freitag and Nathan Smith (Bath)