The term ＇＇fractures＇＇ in its broadest sense can be defined as rupture surfaces in rock, whatever their geological origins. Geoscientists add to that the notion of ＇＇utility＇＇ (i.e. the ability to let fluids flow). An elementary fracture is characterized by its extension, orientation, dip and opening. Fluid flow is governed by the specific parameters of different sets of fractures : spacing, angles, relationships between sets which govern hierarchical fracture network connectivity. In oil reservoirs, fracture specific parameters at different scares are often difficult to define due to the small volume of available well data on one hand and heterogeneous resolutions and investigation depths of available tools on the other. We classically distinguish : ＇＇seismic scale＇＇ faults whose longitudinal and Vertical extensions are hecto to kilometric; sub-seismic scale fractures defined in wells by mud logging, gas seepages, fluid losses, core recoverability, production zones highlighted by production logging operations. Their spacing is decametric, often corresponding to elementary block sizes from dual porosity well-test interpretations; metric to infra metric fractures observed on cores or by high resolution logging tool images. Understanding the organization of these different scales of fractures is the major difficulty of each fractured reservoir study, due to the extrapolation of precise data. This paper, after a review of the different state of the art fracture scale modelling, presents the methodology used in the definition of the open fracture scheme and gridded reservoir model building of the Meillon - Saint-Faust field, located in the Aquitanian Basin, South-West France, under the city of Pau. Hecto to kilometric fractures have been characterized by a 3D seismic structural interpretation compared with outcrop observations, based on remote sensing techniques, of the Mailh-Arrouy massif. Sub-seismic fracturing has been studied from gas seepages and fluid losses while drilling, open-hale injections and production logging operations as well as outcrop observations. Small fractures have been studied on cores and high resolution logging images have been compared with outcrop observations. Three different families of fractures have been noted on each scale. These results together with dynamic data (water breakthrough in the wells, transmissivities from well test interpretations, productivity and the injectivity index) have enabled us to determine the fundamental sets and scales of open fractures which govern fluid flow in the field. Large scale open fractures have been plotted explicitly on the gridded reservoir model (i.e. open seismic scale faults have been characterized by the attribution of equivalent properties depending on the structural context). The results of this study have led us to substantially modify the image of flows in the field inherited from former studies. The finest results from a dynamic simulation show the good accuracy of the gridded reservoir model in reproducing the flowing scheme defined in the field.