This study addresses the issue of global and regional cloud cover forecasts. Using a suite of global models, the multimodel superensemble forecasts are constructed for total cloud coverage and layered (low, middle, and high) cloud coverage out to 7 days (168 hours). The experiments were conducted on a 1° latitude by 1° longitude spatial grid at a 24 hour temporal
resolution. The multimodel superensemble carries a training and a forecast phase. Several training periods were used to determine an optimal training phase. The total period of this study includes October 2006 through February 2007. Clouds from the International Satellite Cloud Climatology Project (ISCCP) were used for training and cloud validations. The metrics for the
forecast evaluation include the root mean square errors, spatial correlations, and equitable threat scores. The results show a significant improvement (~25%) of the multimodel superensemble forecasts compared to those from the member models forecasts, as well as their ensemble mean for the global and regional domains.
