Recently, there are several surface heat flux data sets derived from satellite data. J-OFURO (Japanese Ocean Flux Data Sets with Use of Remote Sensing Observations) data (Kubota et al., 2002a) contains surface heat flux data such as shortwave radiation, logwave radiation, latent heat flux and sensible heat flux. Also, HOAPS (Hamburg Ocean Atmosphere Parameters and Fluxes from Satellite) and GSSTF (Goddard Satellite-based Surface Turbulent Fluxes) are similar data sets made in Germany and U.S.A, respectively. In addition, there are reanalysis data sets including surface heat flux data, such as NRA(NCEP/NCAR Reanalysis) and ERA(ECMWF Reanalysis). These data sets provide global and high resolution data and very important information for research on air-sea interaction. Accuracy of surface heat flux data, however, has not been discussed enough. For example, Kubota et al. (2002b) compared each latent heat flux product and mentioned that there are large differences between each latent heat flux data. In this study, we evaluate accuracy of satellite latent heat flux data in J-OFURO using several buoy data as in situ data. In equatorial region, J-OFURO latent heat flux data averaged in 3days were compared with TAO/TRITON buoy data (65 buoys) during 1992-2000. The correlation coefficient, the rms error and the bias (buoy - J-OFURO) were 0.73, 46.0 Wm-2 and -4.4 Wm-2. We found that J-OFURO latent heat flux overestimate when buoy latent heat flux shows large value. This is mainly due to the accuracy of air specific humidity derived from satellite data. J-OFURO's specific humidity (Schlussel et al., 1995) tends to overestimate the amplitude in equatorial region. If we use air specific humidity observed by buoys for estimating latent heat flux, the statistics are remarkably improved, 0.84, 27.9 Wm-2 and -0.63 Wm-2, respectively.