The mean LHF estimated by the reanalyses tends to be overestimated (i.e. ocean loss more heat to air) over the TAO buoys array. The NCEP/NCAR reanalyses has the best estimation of the mean value of LHF but the variability of the flux, particularly on the interannual time scale, was underestimated. Despite overestimation, LHF from ECMWF reanalyses agrees well with buoys in both phase and amplitude. Using the reanalyses surface meteorological fields and COARE3.0 algorithm, LHF was recalculated and the revised flux is found to be significantly reduced in comparison with the original one. The interannual variability of the revised LHF, however, remains unchanged for most of the cases except for ECMWF reanalyses, of which the revised flux before 1996 seems to be doubtful. The revised NCEP/NCAR LHF is underestimated in both the mean value and the interannual variability, which is due to a weak wind and low variability of specific humidity, respectively.