A new stochastic approach for modeling the internal geometry of deltaic sand bodies is presented that incorporates both genetic (or process-based) and geometric aspects. The genetic aspect takes geologic concepts into account, whereas the geometric aspect reduces the number of parameters for use in applications. Sand-body models resulting from this method honor the well data. This approach is applied to model the internal architecture of an Eocene deltaic sand-body outcropping near Roda, Spain. The simulations using a few wells are compared with the geologic reference model obtained using all the outcrop and log data. The sensitivity of the simulation results is also checked with respect to the fluctuation of the input parameters.