Langrangian DAta Assimilation: Method, Applications and Strategies
Stochastic Dynamical Systems and Control
March 26, 2007 02:00 PM to 03:00 PM
Speakers:
Ide, Kayo
|
 |
Abstract: |
Data assimilation is a method for estimating and forecasting the evolving
physical system along with the corresponding uncertainties by combining the
forecast model and the observations. Numerical weather prediction is an
example of atmospheric data assimilation. For geophysical flows such as
atmosphere and oceans, observations used in data assimilation have been
conventionally Eulerian because the forecast models are largely Eulerian.
For use of Lagrangian (position) observations, it has been a common practice
to transform them into Eulerian velocity observations using approximations,
which can lead to loss of precious dynamical information carried in the
position data. It is only recent that a method that directly assimilates
Lagrangian observations was developed. In this talk, we present a general
framework of the Lagrangian data assimilation (LaDA) that removes the need
for any approximation used in the past. For large-dimensional systems, the
LaDA is ensemble based. Through the oceanic applications, we demonstrate how
effectively the LaDA extracts the dynamical information from the observation
and feeds it into the forecast by the Eulerian model. Strength of the LaDA
includes its ability to incorporate Lagrangian analysis of the dynamical
systems theory. We present a new direction for the observing system design
that maximizes the impact of Lagrangian observations. |
|
|
