This study uses a ranking method to select some good and bad ensemble members from ensemble forecasts of the 2 m temperature above the ground in East Asia provided by European Centre for Medium-range Weather Forecasts(ECMWF) and United Kingdom Met Office(UKMO).The ensemble forecast products are taken from TIGGE archive for the period from June 8, 2007 to August 31,2007.The results show that the optimal selection method may reasonably select the ensemble members with higher and lower forecast skills for 24—96 h forecasts.Case study shows that the ensemble mean forecast of the optimally selected ensemble members may predict the extreme temperature quite well,and it performs better than the ensemble mean of all ensemble members.The forecast skill of the ECMWF ensemble forecast is higher than that of UKMO forecast in terms of the root-mean-square errors of the surface air temperature forecasts.