Traffic road modelling in the dense network of a wide area needs a large amount of data. It renders a keen modelling unmanageable in practice. To simplify the problem, an idea is to aggregate the network links as continuous medium where traffic road flows as a fluid on a surface. This modelling is called two-dimensional. Even if the detailed geometric structure of the network is lost, such a modelling avoids the traffic keen description on a network where measure points are not numerous enough to allow an exhaustive evaluation of traffic state. A series of articles started in the 80's have highlighted relevant concepts. Nevertheless, these previous works have not solved fundamental modelling issues: how to deduce and model global behaviours basing on local behaviours (flow on an axe, interactions at intersections). Two models are developed. The first model is static. Traffic road flows on privileged directions (orthogonal). The model takes into account the balance between the network trip supply and the users' elastic trip demand. The principal outputs are, for each point of the network and each destination, the directional flows and the trip costs. The second model is dynamical. The description of traffic flows stands at the level of elementary traffic cells, where concepts of supply and demand are defined. With a behaviour law deduced from a discrete network, we establish the conservation dynamic equation of road traffic at each point of an anisotropic network
