[7-4]

Parameterisation Estimation using Data Assimilation

M. Lang (University of Reading) , P.J. van Leeuwen (University of Reading) , P. Browne (University of Reading)

In Numerical Weather Prediction (NWP), parameterisations are used to compensate for errors in the model. Errors in NWP models can be due to a lack of scientific understanding or a lack of computing power available to address all the known physical processes. Parameterisations are sources of large uncertainty in a model as parameter values used in these parameterisations are often not well known and/or unmeasurable quantities.

We propose a new method for estimating parameters, Derivative based Parameter Estimation (DerPE), that examines the differences between an analysis forecast state to an analysed data assimilation state to estimate the parameters.

Whilst there are many efficient and effective methods for state/parameter estimation in data assimilation, there are few methods that deduce the structure of model errors and hence be extended to parameterisation estimation. DerPE has the potential to utilise these structures and thus estimate parameterisations.

We have applied DerPE to estimate parameters and simple parameterisations in an advection model. This is used as it is understood how changes in parameters affect the dynamics of the state. Analysis of the method provides a criterion on the quality of the data assimilation to ensure an improved parameter estimate.