The exchange of heat fluxes across the ocean-atmosphere interface plays a significant role in the climate system variability. The determination of surface latent (LE)heat fluxes over open oceanic regions is difficult to attain because of sparse in-situ measurements of meteorological parameters. In addition to in-situ observations, LE can be estimated from satellite data. This presentation will describe our new methodology to estimate instantaneous LE from satellite observations. The approach consists in combining sea surface temperature from NCEP/NCAR reanalysis, total precipitable water, cloud liquid water and surface wind speeds from SSM/I data as input parameters to train artificial neural networks (ANN). The ANN training and validation is performed with observed in-situ estimates of LE.