Adaptation of a particle filtering method for data assimilation in a 1D numerical model used for fog forecasting
Adaptation d'une méthode de filtrage de particules pour l'assimilation des données dans un modèle numérique 1D utilisé pour la prévision du brouillard
Rémy, Samuel ; Pannekoucke, Olivier ; Bergot, Thierry ; Baehr, Christophe
COBEL-ISBA, a boundary-layer 1D numerical model, has been developed for the very-short-term forecasting of fog and low clouds. This forecast system assimilates local observations to produce initial profiles of temperature, specific humidity and liquid water content. As fog forecasting is a threshold problem, the model is strongly nonlinear.
A new assimilation method based on a genetic selection particle filter was tested to produce the initial conditions. The particle filter was adapted for a deterministic forecast and to take into account the time dimension by minimizing the error on a time window. A simplified particle filter was also used to determine the initial conditions in the soil. The filter was tested with two sets of simulated observations. In all cases, the initial conditions produced by this algorithm were of considerably better quality than the ones obtained with a Best Linear Unbiased Estimator (BLUE). The forecast of the control variables and of fog events was also improved. When comparing scores with the ones obtained with an ensemble Kalman filter (EnKF), the particle filter showed better performances for most of the cases. The ensemble size impacted the frequency of filter collapse and the quality of the initial temperature and specific humidity profiles in the lower part of the domain. Copyright © 2011 Royal Meteorological Society
Accès à la notice sur le site du portail documentaire de Météo-France