Image Sequence Assimilation in numerical models

Satellites continuously provide a huge amount of data that are still underused by numerical forecast systems. In particular, the dynamical evolution of certain meteorological or oceanic features (such as eddies, fronts, ...) that the human vision may easily detect is not optimally taken into account in realistic applications of data assimilation.

Recent work has been conducted to develop methodologies for assimilating such data into numerical models, and today two different approaches are under study:

• indirect technique:

  Indirect technique provide apparent velocity fields which are assimilated as classical observations. These measurements are obtained by certain external procedures (motion estimation) which are decoupled with the considered dynamical system.



• direct technique:

  The image sequence is considered as an observation in the assimilation scheme. This technique may use the pixel basis of the image but higher levels of interpretation (e.g. a multi-scale decomposition or a description of contours) can also be used. Adapted observation operators have to be specified in order to link model state vector to the structured information contained in the image sequence.