Olivier PANNEKOUCKE, INPT-ENM, CNRM, Météo-France/CERFACS
Parametric Covariance Propagation in the non-linear diffusive Burger equation
Coauthors
Richard MENARD, Marc BOCQUETAbstract:
The forecast error covariance propagation at a reduced cost is proposed in place of the ensemble methods. The reduction relies on the parametric description of the forecast error covariance matrix featured only by its variance and its correlation length-scale fields. The situation is illustrated for the non-linear diffusive Burger equation. We detail how each processes contribute to the forecast error statistics. In particular we theoretically discuss the very difficult non-linear interaction between the variance and the length-scale that results from the diffusive process.