NEW OBSERVATIONS AND NOVEL WAYS OF ASSIMILATING:

THE DATA ASSIMILATION RESEARCH CENTRE.

ROSS BANNISTER,

D.A.R.C., READING UNIV.

OUR TOOLS ...

  1. Met Office variational data assimilation system.
    • 3d-Var.
  2. Met Office U.M.
    • Stratosphere/troposphere, 40 levels (later 64 levels).
    • 'Climate' resolution ().
    • 'Old' dynamics.
    • Cariolle scheme.

OUR AIMS ...

(amongst other things)
  1. Assimilate new research satellite data (esp. EnviSat).

    • Retrieved products in the first instance.
    • Limb radiances.
    • (all additional to operational observations).
  2. Use dynamical ideas to improve the performance of the variational scheme.
    • Representation of .
    • Preconditioning.

MET OFFICE 3D-VAR. SYSTEM

Cost function in model space

equation

equation

equation

equation

Cost function in 'control' space

equation

equation

equation

equation

equation

DESCENT

The gradient of :

(i) WHY ARE WE INTERESTED IN NOVEL SATELLITE DATA?

  • Cal./Val. for EnviSat instruments.
  • Global fields of trace gases in U.T.L.S.
  • Investigate the value of assimilating a combination of different types of satellite data.
  • Develop new techniques.

E.g. with water vapour in the U.T.L.S. (UM simulation only):

The combination of research satellite data, an operational data assimilation system, and on-line chemistry is new.

(i) RETRIEVED PRODUCTS (STRATOSPHERIC ASSIMILATION)

Model variable Retrieved product
¬
¬
¬

Forward models (layer averaging)

+

Adjoints

(ii) VARIATIONAL CONTROL VARIABLES

  1. The background error co-variance matrix, , is:
    • too large to store ( elements),
    • multivariate,
    • not sparse,
    • ill conditioned ().

equation

equation

equation

equation

equation

equation

CURRENT SCHEME
Met. vars. Parameters equation

PROPOSED SCHEME
Met. vars. New parameters equation

( will have a similar structure)
  • Expect that the leading (balanced) control variable captures most of the flow under all flow regimes.
  • Expect that the assumption of weak correlation between variables is globally valid.

Need , and


The D.A.R.C. is an interdisciplinary collection of researchers. It will try to bridge the gap between many communities.
  • Instrumentation / remote sensing.
  • Environmental and climate science.
  • Dynamical and control theory.
  • Weather forecasting and modelling.

For more information:

http://darc.nerc.ac.uk

http://www.met.rdg.ac.uk/~ross/DARC/DataAssim.html