Toward an estimation of global land surface heat fluxes from multisatellite observations

The sensitivity of a suite of satellite observations to land surface heat fluxes and the estimation of satellite-derived fluxes using a statistical model are investigated. The satellite data include visible and near-infrared reflectances (Advanced Very High Resolution Radiometer [AVHRR]), thermal infrared surface skin temperature and its diurnal cycle (International Satellite Cloud Climatology Project [ISCCP]), active microwave backscatter (European Remote-sensing Satellite [ERS] scatterometer), and passive microwave emissivities (Special Sensor Microwave/Imager [SSM/I]). Fluxes at the global scale are taken from Land Surface Models (LSM): the GSWP-2 multimodel analysis, the ISBA, and ORCHIDEE participating models, along with the National Centers for Environmental Prediction/the National Center for Atmospheric Research (NCEP/NCAR) reanalysis, on a monthly timescale from 1993 to 1995. The simulated LSM fluxes and the satellite observations are linked through a statistical model. Once calibrated, the statistical model reproduces the LSM latent and sensible fluxes for all types of snow-free environments, with global RMS errors <25 W/m(2), proving that the satellite data contain relevant information for flux estimation. The estimated fluxes have realistic spatial and seasonal patterns, although some local differences between the original and estimated fluxes are found. These differences are used to reveal potential problems in the LSMs, for instance, an anomaly in the GSWP-2 radiative forcings. Comparisons between the original and estimated fluxes and 76 tower fluxes over North America are carried out, and the differences show similar statistics. However, the largest differences between the original and estimated fluxes do not occur in these regions. Demonstrating the superiority of the proposed technique outside of these regions remains difficult in the absence of validation data sets.