Description

We propose a motion estimation method solving for the first time the complex problem of 3D winds field recovery from satellite image sequences. In order to manage incomplete observations, physical knowledge on 3D mass exchanges between atmospheric layers have been introduced within an optical flow scheme.

The estimator is based on a functional minimization. The data term relies on a model which describes the dynamics of an interacting stack of atmospheric layers. This model applies on a set of sparse pressure difference images related to the different atmospheric layers. A method is proposed to reconstruct such observations from satellite top of cloud pressure images and classification maps. To overcome the problem of sparse observations, a robust estimator is introduced in the data term. The data term is combined with a smoother that preserves the bi-dimensional divergent and vorticity structures of the three-dimensional flow and reinforces regions of homogeneous vertical winds.

Results

Retrieval of horizontal motions and ascendant winds:

ground truth
3D model
2D model
	low horizontal wind	low altitude vertical wind	intermediate horizontal wind	high altitude vertical wind	high horizontal wind

Upper line : ground truth. Horizontal velocity fields used for simulation superimposed on cloud pressure difference images for a high (left), a medium (middle) and a low (right) layer. White areas correspond to regions where no data are available. Vertical wind fields used for simulation are displayed in-between the latter images.Middle line : comparison with 3D motion estimates. Horizontal winds estimated with the 3D model are superimposed on cloud pressure difference images for the high (left), the medium (middle) and the low (right) layer and estimated vertical wind maps are displayed in-between the latter images. Bottom line : failure of a 2D model. Horizontal winds estimated with a 2D model are superimposed on cloud pressure difference images for the high (left), the medium (middle) and the low (right) layer.

Estimation of 3D winds in atmospheric convective systems :

Images over the Gulf Guinea the 5-June-2004 from 13h30 to 14h15 UTC. Cloud pressure difference images of the highest layer at 4 consecutive times. Estimated horizontal wind vectors which have been superimposed on the images range in the interval [0,10] m/s. Retrieved vertical wind maps on the highest layer lower boundary have been superimposed on the pressure difference images. Vertical winds range in the interval [-0.5,0.5] m./s.

Estimation of 3D winds in a 3-Layer atmosphere :

low horizontal wind	intermediate horizontal wind	high horizontal wind

low altitude vertical wind	high altitude vertical wind

Horizontal and vertical winds in correspondence to cloud pressure difference images (over the north Atlantic Ocean the 5-June-2004 from 12h00 to 13h00 UTC) related to the high (right), the middle (middle) and the low (left) layers. Estimated horizontal wind vectors range in the interval of [0,15] m/s. Estimated vertical winds range in the interval [-0.2,0.4] m/s.

Reference

P. Héas, E. Mémin. 3D motion estimation of atmospheric layers from image sequences . Conditionally accepted for publication in IEEE transactions on Geosciences and Remote Sensing.

Soft

P. Héas, E. Mémin, N. Papadakis. Demonstrator on layer motion estimation methods (2D and 3D estimation) . Fluid Project deliverable 4.3, 2007 (soft)