Spatiotemporal stochastic forcing of an ensemble system consisting of chaotic Lorenz cells diffusively coupled is analyzed. The nontrivial effects of time and length correlations on the ensemble mean error and spread are studied and the implications to new trends in weather forecast methodologies are discussed. A maximum for the forecast scores is observed to occur for specific values of time and length correlations. This maximum is studied in terms of an interplay between the natural scales occurring in the system and the noise parameters.