Data Assimilation Meetings at Reading

Date Meeting type Speakers
19 April 2012 DARC seminar room change: AGRIG 1U02 Neil Bowler (Met Office)
Developing 4D-En-Var as a potential replacement to 4D-Var and the ETKF
NWP data assimilation systems are using ensemble information more and more. This primarily stems from a desire to allow greater flow-dependence in the data assimilation system. However, there is also the challenge of code-scalability on future supercomputers which may make it difficult to increase the resolution of our deterministic forecasts. To address both of these areas we are developing 4D-En-Var, which is able to mimic the EnKF in a variational framework.
This talk will cover the hybrid 4D-Var system which is currently operational at the Met Office. This uses information from the ensemble at the start of the time window. Introducing the hybrid brought strong improvements to the forecasting system when it was implemented in summer 2011. Following this, the talk will cover 4D-En-Var which uses information from the ensemble throughout the time window. This means that the propagation of information through the data assimilation window is performed by the ensemble, rather than using a tangent-linear model.
Since 4D-En-Var does not use a tangent-linear model it has the potential to scale much better on future supercomputers. It also provides method to unify the ensemble generation with the data assimilation. Finally it will enable us to run a number of interesting scientific experiments.

Page navigation