[en] When using solute transport modelling to predict ground water quality, the large in situ variability of soil physical parameters is one of the most significant source of uncertainties in the reliability of the simulation results. Stochastic models were developed to take into account this phenomenon. Some of these models use the Monte Carlo approach that consists of selecting random values for each parameter from their respective statistical distributions. However, in most cases this random selection does not take into account the real spatial distribution of these parameters. With a model using the Monte Carlo approach, tests can be induced by this last fact. Theoretical relations between parameters were used to evaluate the significance of the error induced by the use of non-correlated distributions of parameters on predictive results. This theoretical approach has shown that in some cases the error can become significant. However, an application with an in situ evaluation of soil physical parameters on a sandy soil in Quebec has shown that the error levels are not significant when considering real situations. (Author) (4 refs., 6 figs., 3 tabs.)