It is necessary to introduce consideration of fractional cloud coverage in a grid box for use in GFS, since the Ferrier scheme is designed for use in high-resolution mesoscale model and do not consider partial cloud explicitly. Using cloud cover obtained by the formulation of Sundqvist et al. (1989), each grid box is divided into the following three parts: (1) cloudy; (2) clear with precipitation from upper level; and (3) clear without precipitation from upper level. Then the Ferrier scheme is applied separately to the cloudy and clear with precipitation portion of the grid. Maximum-random cloud overlap is assumed.
Since the change in the microphysics scheme is considered a major change to the GFS, extensive testing and evaluation of the entire system of data assimilation and forecast are required before operational implementation of the scheme. We have performed extensive tests of the scheme for a two-month summer season of 2008 and are currently testing the scheme for a two-month winter season. Skill evaluation of the test will include anomaly correlation and root-mean-square errors of the synoptic fields, precipitation skill forecasts over North America, and cloud forecasts. We will present the details of the treatment of partial cloud in the Ferrier scheme and the results of the evaluation.