Cumulative Results of Extended Forecast Experiments I. Model Performance for Winter Cases

A series of 2-week predictions were made with a general circulation model for 12 winter cases selected from the period 1964–69. All were January cases. The same prediction model—the most sophisticated and probably the most realistic model of those we tested in 1967—was used throughout. The model was hemispheric and had an N = 40 grid (grid size of about 270 km at mid-latitudes) with nine vertical levels. A detailed description of the model's performance is attempted by making statistical analyses of the forecast results compared with observed data. The analyses also provide useful insight into the dynamical behavior of the long waves in the middle latitude zone. The verification study reveals the practical limit of predictability with the 1967 version of the Geophysical Fluid Dynamics Laboratory model. For example, the correlation coefficient between prediction and observation of the 500-mb geopotential deviation from January normal stays above zero until the 10th day. A spectral study of the planetary and cyclone waves was also made. The behavior of the ultralong wave in this model is disappointing, but cyclone waves are reasonably well predicted until the eighth day.