Abstract:The potential predictability of the seasonal mean climate anomaly has been asses sed by using the dataset of European Centre for Medium Range Weather Forecasts ( ECMWF) ensemble seasonal forecasts for the period 1979~1993.The forecasts were created with prescribed observed sea surface temperature taken from ECMWF re-an alysis and updated daily in the forecasts. By reconstruct ing the ensemble in terms of reproducible forced modes,and by defining a predict ive utility index (PUK) based on the Kolmogorov-Smirnov (KS) test,a new quant itative measure for evaluating potential predictability,i.e.the predictability i ndex (PI),which combines PUk with the local variance contribution to ensemble mean,was proposed.The quantitative analysis with PI for the 850 hPa temperature has shown that the seasonal mean anomalies over not only most of the tropical re gions but some extratropical regions are predictable.Extratropical predictable r egions are found mainly over North America,south Africa and part of Asian monsoo n region.Interestingly,the potential predictability over some extratropical regi ons like northern China,middle Asia and southern part of North America is contro lled by the non-ENSO forcing,although the ENSO forcing generally dominates most of the predictable regions.The results also indicate that the predictable regio ns have obvious seasonality.Summer possesses the largest areas with high PI valu es while winter possesses the smallest.The results have been compared with other methods,which shows that the PI analysis proposed here can extract more predicta ble information especially for the extratropical regions.