Precipitation is the result of complex physical processes, which are highly nonlinear and very sensitive to initial conditions. Under a deterministic approach is not only weather can accurately predict values of rainfall with potential applications in hydrological studies. This thesis follows a stochastic-deterministic way. 
First is a review of relevant models for simulation of daily precipitation, both purely statistical and added that relations with the meteorological variables. Below is a stochastic model for daily precipitation. Rain events are independent realizations of a Poisson process. The statistical series of simulated and observed rainfall for various stations are very similar. This model is also superior to one based on Markov chains. 
However, this model still remains a useful tool for simulation of daily precipitation is not suitable for prediction of rainfall or climate change studies. General circulation models used grids of varying degrees of latitude longitude by several much larger than the basin scale operating models related to water resources. For this reason is that presented the idea of downscaling. 
Essentially, the statistical downscaling is to translate the variations of large-scale flow in the values of local variables such as precipitation. This is done using the relationship observed between the general circulation and the variable under study. 
Thus, we propose a model based on downscaling. Used to an existing synoptic classification for the Iberian peninsula. The schedule of rates of time is simulated by a discrete Markov process. The model is successfully implemented for both areal and point prediction of daily precipitation.