Estimating model error with the ensemble Kalman filter

JUAN RUIZ

Buenos Aires

Workshop; WCRP/SPARC Workshop with focus on the Southern Hemisphere and South America; 2012

SPARC/WCRP

Data assimilation is used operationally, for the optimum estimation of the state of the atmosphere. Most data assimilation methods can be extended to estimate model error. Several techniques have been proposed and implemented for the estimation of model biases. The estimated biases, can be used to partially correct model error. This improves the accuracy of the analysis and the forecast. There are several sources of model error: one of these sources are the assumptions made in model parametrizations of subgrid scale processes and the tuning of the parameters associated with these parametrizations. Most data assimilation schemes can also be extended to find the optimal value for these parameters (whose optimal values are unknown a priori). In this talk the extension of ensemble based data assimilation schemes to include model bias and model parameters will be discussed and some examples of their application will be shown.