The aim of this paper is to present a statistical model which could exploit further dynamical information, capable to give a better forecast of daily hydric precipitation. The two variables, amount of water (q) and number of dry days which precede the wet day (d), appeared reasonably independent. The statistical distribution of q is well fitted by a Gamma distribution. Each day is considered wet or dry according to the result of a drawing from an urn. The composition of the urn changes day to day as in the contagious models.