Forecasting the Surface Weather Elements with a Local Dynamical-Adaptation Method Using a Variational Technique

View More View Less
  • 1 Direction de la Météorologie Nationale, Toulouse, France
© Get Permissions
Restricted access

Abstract

A simple method of dynamical adaptation of a mesoscale model has been tested to produce meteorological parameters locally adapted at meteorological stations. This method is based on the use of a soil model in association with a surface boundary-layer model (MUSCLES, modélisation uni-dimensionnelle du sol et de la couche limite en surface), coupled with the outputs of the French operational mesoscale model (PERIDOT). The meteorological station-dependent characteristic constants used for describing the soil properties are identified by comparisons with observations. This adjustment is achieved by a variational method consisting in minimizing a cost function that measures the distance between the output parameters computed by MUSCLES and the observed ones. The minimization algorithm developed for that purpose requires the computation of the gradient of this cost function, which is done in practice by using the adjoint of the MUSCLES code.

For forecasting purposes, it was found that the best way is to adjust the local constants by computing the cost function on the ten previous days. The results are encouraging; for five of the six stations considered, the quality of the gains is significant, even if they are lower than what is achieved by the operationally used statistical adaptation.

Abstract

A simple method of dynamical adaptation of a mesoscale model has been tested to produce meteorological parameters locally adapted at meteorological stations. This method is based on the use of a soil model in association with a surface boundary-layer model (MUSCLES, modélisation uni-dimensionnelle du sol et de la couche limite en surface), coupled with the outputs of the French operational mesoscale model (PERIDOT). The meteorological station-dependent characteristic constants used for describing the soil properties are identified by comparisons with observations. This adjustment is achieved by a variational method consisting in minimizing a cost function that measures the distance between the output parameters computed by MUSCLES and the observed ones. The minimization algorithm developed for that purpose requires the computation of the gradient of this cost function, which is done in practice by using the adjoint of the MUSCLES code.

For forecasting purposes, it was found that the best way is to adjust the local constants by computing the cost function on the ten previous days. The results are encouraging; for five of the six stations considered, the quality of the gains is significant, even if they are lower than what is achieved by the operationally used statistical adaptation.

Save