Humidity Profile Retrieval Using a Maximum Entropy Principle

Bernard Urban Centre National de Recherches Météorologiques, Meteo-France, Toulouse, France

Search for other papers by Bernard Urban in
Current site
Google Scholar
PubMed
Close
Restricted access

Abstract

A satellite data inversion method based on a maximum entropy principle is presented. The method is both physical since a radiative transfer model with its adjoint is needed, and also statistical since errors of the observed radiances and of a prior vertical profile are taken into account. The above errors need not be Gaussian. The technique is also intended to be applied to data assimilation and is compared to the Bayesian approach. The method can be seen as a generalization of the variational assimilation methods using strong constraints, but without explicitly using these constraints in the minimization process. In addition, the functional to minimize in the case of maximum entropy is convex and infinitely differentiable, so the computational cost of the minimization is low. The method is applied to retrieve vertical humidity profiles using TIROS-N Operational Vertical Sounder (TOVS) radiances data and a numerical model background, and to constrain the result between physically bounded values (namely zero and saturation); comparisons to radiosonde data are also presented.

Abstract

A satellite data inversion method based on a maximum entropy principle is presented. The method is both physical since a radiative transfer model with its adjoint is needed, and also statistical since errors of the observed radiances and of a prior vertical profile are taken into account. The above errors need not be Gaussian. The technique is also intended to be applied to data assimilation and is compared to the Bayesian approach. The method can be seen as a generalization of the variational assimilation methods using strong constraints, but without explicitly using these constraints in the minimization process. In addition, the functional to minimize in the case of maximum entropy is convex and infinitely differentiable, so the computational cost of the minimization is low. The method is applied to retrieve vertical humidity profiles using TIROS-N Operational Vertical Sounder (TOVS) radiances data and a numerical model background, and to constrain the result between physically bounded values (namely zero and saturation); comparisons to radiosonde data are also presented.

Save