Parameter estimation of a convective scheme using Ensemble Transform Kalman filter

JUAN J. RUIZ; MANUEL PULIDO

Reading

Workshop; Workshop on Representing model uncertainty and error in numerical weather and climate prediction models; 2011

European Centre for Medium-Range Weather Forecasts

In this work the model state and parameters associated with a convective scheme are simultaneously estimated using ensemble transform Kalman filter (ETKF) in a simple general circulation model (SPEEDY). The impact of parameter estimation upon initial conditions and short range forecasts are estimated using twin experiments which permits the evaluation of the technique under the assumption of the perfect knowledge of the true system evolution and the true parameters. The method is successful in the estimation of the known parameters and introduce a significant improvement in the initial conditions and in the short to medium range forecast. The method is also able to estimate the evolution of time dependant parameters. A new approach for the online determination of an optimal value for the parameter ensemble spread is provided. This approach consist in the inflation of the parameter ensemble variance in the ensemble space. Results shows that the method is able to find an optimal relationship between the spread of the parameters and the spread of the state variables. Two methods are evaluated for the inclusion of parameter estimation within a data assimilation cycle. The first one uses an ensemble with perturbations in the initial conditions and in the parameters and the second one uses two independent ensembles, one based on the perturbations of the initial conditions and other based on the perturbations of the parameters. Both methods are successful in estimating the known system parameters, however the first one is computationally more efficient.