Forecast in the medium and longer range requires an accurate forecast of the scale flow. An accurate large-scale forecast requires an accurate, large-scale, initial, objective analysis. The present study examines the accuracy of large objective analyses using a state-of-the-art data assimilation scheme. Experiments with idealized uniform networks demonstrate that them are systematic deficiencies in large-scale analyses. These systematic large-scale errors, while annoying are not nearly as serious as those caused by the inhomogeneities in the operational observation network. The systematic errors can probably be corrected by improvements in analysis technique, The large-scale errors due to observation network inhomogeneities require improvements in the network.