Predictability and Inverse problems in chaotic Geofluid Dynamics, an application to the atmosphere
By: Carlos Pires
From: Instituto Dom Luis. Faculdade de Ciencias da Universidade de Lisboa, DEGGE –
Departamento de Engenharia Geografica, Geofisica e Energia.
At: Complexo Interdisciplinar, Anfiteatro
[2009-05-20]
Operational daily weather forecasts can be useful for lead-times between 3-12 days depending on the dynamical stability of the chaotic atmospheric flow. This means that weather predictability changes along the time. Predictability is closely related to the performance of data assimilation in the atmosphere, which is an inverse problem recovering forecasting model initial conditions from observations (satellites, measurements from ground, sea, upper-air). We will give an overview of data assimilation methods, in particular the 4D-variational method, showing characteristics of the a posteriori error for some simplified models and how they project on stable and unstable manifolds. The influence of model errors is also explored.