MULTI RESOLUTION INFRARED OPTICAL PROPERTIES FOR GAUSSIAN SEA SURFACES Sandrine Fauqueux DOTA / MPSO ONERA Chemin de la Huni` ere Palaiseau , FRANCE Sandrine.Fauqueux@onera.fr Karine Caillault DOTA / MPSO ONERA Chemin de la Huni` ere Palaiseau , FRANCE Karine.Caillault@onera.fr Christophe Bourlier IREENA Ecole polytechnique de l’universit´ e de Nantes rue Christian Pauc, La Chantrerie Nantes, FRANCE Christophe.Bourlier@univ-nantes.fr Pierre Simoneau DOTA / MPSO ONERA Chemin de la Huni` ere Palaiseau , FRANCE Pierre.Simoneau@onera.fr Luc Labarre DOTA / MPSO ONERA Chemin de la Huni` ere Palaiseau , FRANCE Luc.Labarre@onera.fr ABSTRACT We are interested in the infrared radiative modelling of wind-roughened Gaussian sea for surface length from one meter to several kilometers. For the considered spectral bandwidth, the relations beween geometrical optics and global optical properties are already known for centered Gaussian sea surface. In our case, for resolutions under sea correlation length, processes have non-zero mean value and literature’s expressions are incomplete. Thus, we extend them to account for any surface length. Hiding and shadowing are included. Only single reflections are treated. The numerical method requires the mean vector- valued and covariance matrix of slopes processes for any resolution. To estimate those parameters, a new two-scale method is introduced, combining statistical approach with geometrical surface generation : the ”variable step method”. Finally, illustrations show our ability to produce good quality ocean scenes in various contextual conditions. KEY WORDS Modelling and Simulation Methodologies, Ocean Optics, Multi-resolution, 1 Introduction In the scope of the development of a new version of the in- frared background scene generator MATISSE [1], a model of sea surface optical properties will be implemented. It has to account for spatial variability ranging from 1-meter scale to large scale variability. Description of the radiative characteristics of a wind- roughened sea surface can be achieved in different ways. Models from computer graphics literature [2, 3] are based on roughness parameters which can not be expressed as a function of contextual coefficients such as wind speed or fetch. Monte-Carlo methods [4, 5] are the most accurate but ray tracing needs a thin discretization of the sea surface (smaller than one millimeter) to properly model the capillarity waves. Thus, for space range from meter to kilometers, those methods are too much time consuming. The other well-known class of methods estimates a statistical average of local optical properties which are deduced from Fresnel coefficients, for a given Gaussian slope probability density function (see refs [6] to [9]). Relation with physics is guarentee through a power spectral density defined as a function of contextual parameters, covering both gravity and capillarity waves. They also include shadowing and hiding functions. Those expressions are based on the hypothesis that the considered sea surface is flat in average. In the scope of modelling the radiation of an instantaneous sea surface, this hypothesis is only suitable for configurations where pixels footprints are large enough to include the whole sea surface variability ( for example for satellite observations) not when the pixel footprint only covers a small patch of the sea surface. The aim of this article is, first, to extend those ana- lytical expressions to non-centered processes. Then, statistical properties of the considered sea surface is determined, based on a two-scale model combining small scale statistics with large scale surface generation. Finally, radiance images of the ocean surface are shown to illustrate the quality of our model and demonstrate its feasibility for extended domains. 2 Study’s framework Effective optical properties have to be evaluated for each pixel of the sensor. We call rough facet the part of the