We have developed a methodology for estimating weekly and monthly latent heat fluxes over the global oceans using remote observations from several satellites and the bulk formula. Our estimates are based on sea surface temperatures from the Reynolds optimal-interpolated analysis, air-humidities from the Special Sensor Microwave/Imager (SSM/I) and wind speed from both radiometers (SSM/I) and scatterometers (ERS and NSCAT). Our methods were developed by careful calibration and validation with in-situ observations during the NSCAT period (September 1996-June 1997) and our result for this period was compared with ECMWF analyses and NCEP/NCAR re-analyses. Recently we have used a 1-year version (October 1996-September 1997) of our wind and latent heat flux estimates to describe the large-scale patterns of week-to-week variability in the tropics.

In this study, we extend our satellite latent heat flux estimates to the three-year period (January 1996 - December 1998) that includes the strong 1997-1998 El Niņo and focus on seasonal and interannual time scales. The seasonal variability is extracted by fitting annual and higher-order harmonics to the data after removing a long-term trend at every grid point. The interannual variability is studied from the anomaly fields constructed by removing our estimate of the seasonal cycle from the data. The seasonal-interannual decomposition from the harmonic analysis is compared to the empirical large-scale modes of variability obtained from EOF analysis.