Flow simulations in fractured reservoirs are generally based on the dual porosity concept. The porous medium and the fracture network are modelled in the form of two juxtaposed media of parallelepiped matrix blocks separated by producing fractures. Whatever the studied medium and the observation scale, a fractured reservoir can be defined as a matrix affected by a fracture network. Parameters equivalent to the double medium, observed on an accurate scare, must be investigated and applied to the two media juxtaposed by means of discretisation, on a rough observation scale. It is difficult to estimate the equivalent parameters, in particular the permeability tenser which represents the heterogeneity and anisotropy of the medium. There are some rapid means of estimating horizontal permeability but they are simplistic and the complexity of the flow in a fracture network is not duly considered. The most reliable method is the interpretation of well tests which provides a good estimation of permeability in the two media and, in the most favourable cases, an estimation oi the vertical permeability by means of the Kv/Kh ratio. Vertical permeability is the most important parameter, although it is difficult to obtain and characterize in fractured reservoirs. This paper presents the results of a sensitivity study of the Kv/Kh ratio of the Various geological parameters of both the matrix and the fracture network. A stochastic modelling programme was developed and the modelled fracture networks were discretized with a very fine gridding. After ＇＇fractured cells＇＇ were differentiated, parameters were attributed to the two types of cells and the whole field was homogenized. One major result of the study was the theoretical curve of the evolution of permeability vertical anisotropy versus the density of fractures in the reservoir. Moreover, the study highlighted an important parameter that has not yet been studied : the fracture length/fracture height ratio. The fracture shape ratio, a notion defined in this study, determines the evolution of the Kv/Kh curve versus fracturation density. In view of these results, 3D modelling is justified, including stochastic techniques that take into account the natural reality of formation fracturing. A validation of the method used in this study was achieved by calculating the equivalent parameters in two real cases for which the permeability could be assessed by methods other than calculation.