This poster proposes a new procedure in agriculture supply modelling by the positive mathematical programming (PMP) approach. This approach is now widely used in last CAP reform simulations. However, simulation behaviour and performances of PMP procedures depend of the way parameters of the non linear total cost function in the objective function are recovered. We propose a new specification of the total cost function where land is explicitly considered as a fixed input. By using relative parts of land of the different activities this new PMP procedure permits to better capture production behaviour when economic conditions. It also permits to avoid a drawback of the early procedures concerning marginal activities