(eng): The paper concerns the Bayesian modelling of tra c intensities measured in towns and cities. It proposes the use of a normal regressive model of low order, whose parameters (regression coe cients) are estimated using partial forgetting. This method allows to track the time varying parameters and respects the dierent variability rate of the absolute term modelling the mean value. The general theoretical concept of Bayesian modelling and parameter estimation is complemented with the theory related to the normal model in lieu.