Résumé: The identi .cation of appropriate spatial models requires the previous knowledge of the process exploratory properties such as the degree of homogeneity in local and global effects, and the existence of a directional component which determines an anisotropic behaviour of the phenomenon. This will allow the assumption of some form of stationarity that reduces the number of parameters involved in the model so that it becomes manageable. The methods for exploringstationarity both from global and local levels are varied and well- known. Techniques used for analysing anisotropy are, on the contrary, limited and they are only devoted to examining ‘‘by eye ’’experimental variograms in different directions. In this paper, we present a method based on second- order bivariate circular statistics that allows us to determine the anisotropy existence in digital images and the quanti .cation of the direction in which this appears. A study about the .ow of seawater through the Strait of Gibraltar, based on a remotely sense image, is presented as an example of the potential use of the proposed method.

Langue: Anglais