The ECMWF operational implementation of four dimensional variational assimilation. Part. II: experimental results with improved physics
L'implémentation opérationnelle au CEPMT de l'assimilitation variationnelle quadri-dimensionnelle. Partie II : résultats expérimentaux avec des contrôles physiques
Mahfouf, J.-F. ; Rabier, F.
Année de publication
2000
A comprehensive set of physical parametrizations has been linearized for<br> use in the European Centre for Medium-Range Weather Forecasts (ECMWF's)<br> incremental four-dimensional variational (4D-Var) system described in <br>Part I. The following processes are represented: vertical diffusion, <br>subgrid-scale orographic effects, large-scale precipitation, deep moist <br>convection and long-wave radiation. The tangent-linear approximation is <br>examined for finite-size perturbations. Significant improvements are <br>illustrated for surface wind and specific humidity with respect to a <br>simplified vertical diffusion scheme. Singular vectors computed over 6 <br>hours (compatible with the 4D-Var assimilation window) have lower <br>amplification rates when the improved physical package is included, due <br>to a more realistic description of dissipative processes, even though <br>latent-heat release contributes to amplify the potential energy of <br>perturbations in rainy areas. A direct consequence is a larger value of <br>the observation term of the cost-function at the end of the minimization<br> process when improved physics is included in 4D-Var. However, the <br>larger departure of the analysis state from observations in the <br>lower-resolution inner-loop is in better agreement with the behaviour of<br> the full nonlinear model at high resolution. More precisely, the <br>improved physics produces smaller discontinuities in the value of the <br>cost-function when going from low to high resolution. In order to reduce<br> the computational cost of the linear physics, a new configuration of <br>the incremental 4D-Var system using two outer-loops is defined. In a <br>first outer-loop, a minimization is performed at low resolution with <br>simplified physics (50 iterations), while in the second loop a second <br>minimization is performed with improved physics (20 iterations) after an<br> update of the model trajectory at high resolution. In this <br>configuration the extra cost of the physics is only 25%, and results <br>from a 2-week assimilation period show positive impacts in terms of <br>quality of the forecasts in the Tropics (reduced spin-down of <br>precipitation, lower root-mean-square errors in wind scores). This <br>4D-Var configuration with improved physics and two outer-loops was <br>implemented operationally at ECMWF in November 1997.</div>
Accès à la notice sur le site du portail documentaire de Météo-France