To investigate the mesoscale predictability we use real data cases from both the Sierra Rotors Project and the follow up Terrain-Induced Rotor-Experiment which both took place upstream and in the lee of the California's Sierra-Nevada mountains. In particular we look at how predictions of downslope winds are sensitive to the upstream synoptic state. To accomplish this we use an 70 member ensemble generated by the Ensemble Kalman Filter (EnkF) technique. The sensitivity of downslope winds to a particular parameter is investigated by calculating the covariance between each ensemble members prediction of downslope wind in a domain downstream of the Sierra crest and the ensemble prediction of the parameter of interest. Initial results from IOP 8 of the Sierra Rotors Project indicate that that the variability of downslope wind magnitudes in the lee of the Sierra-Nevada and adjacent mountain ranges was quite large with standard deviations in excess of 8 m/s. The sensitivity of this variability to the upstream synoptic state will be presented.