Clear-sky outgoing far-IR spectral flux inferred from satellite observations over near globe and their usage in evaluating state-of-the-art reanalyses

The far-IR spectral flux is important to the study of earth radiation budget however no directly spectrally resolved observations of the far IR spectrum (10-600cm^-1) have been made yet. We extend the algorithm of Huang et al. (2008) to the entire globe so clear-sky outgoing spectral flux through the entire thermal-IR spectrum can be estimated from 80S to 80N from the collocated Atmospheric Infrared Sounder (AIRS) and the Clouds and the Earth's Radiant Energy System (CERES) measurements. Specifically, clear-sky spectral anisotropic distribution models (spectral ADMs) are developed for five land surface types (forest, grassland, savannas, bright desert, dark desert) as well as for the extra-tropical oceans so AIRS radiance can be converted to spectral flux over each AIRS channel. Then a multivariate linear regression scheme is developed for each surface type so the flux over spectral regions not covered by the AIRS instrument can be estimated at a 10 cm^-1 interval. ASTER Spectral Library v2.0 is used in the construction of spectral ADM for land surfaces. The spectral flux derived by this algorithm in the entire year of 2004 is then compared with collocated CERES OLR measurements. For each of six surface types, the mean difference between OLR derived using our algorithm and CERES OLR is between -0.7 and 1.7 Wm^-2 for the daytime cases and between -0.41 and 1.41 Wm ^-2 for the nighttime cases. Over the globe the mean difference is 0.96 Wm^-2 with a standard deviation of 2.02 Wm^-2 for the daytime and 0.86±1.61 Wm^-2 for the nighttime. With confidence in the derived spectral flux over the globe, we further compared far-IR spectral fluxes computed from three most recent reanalyses: ECMWF ERA-interim, NASA MERRA, and NOAA CFSR, with the far-IR portions of our derived spectral fluxes. The spectral greenhouse parameters derived from each data source are examined. Three reanalyses show some consistent discrepancies from the AIRS-inferred spectral fluxes. The reanalysis global averaged flux over 0-400cm^-1 tends to be slightly more than the AIRS-derived ones by ~ 0.4 Wm^-2 (~0.9%). The flux over 400-600cm^-1 computed from three reanalyses tends to be slightly less than the AIRS-derived ones by ~ 0.4Wm^-2 (~ 0.7%). We further composite the far-IR flux with respect to temperature and precipitable water. The largest discrepancies are seen for cold and humid scenes. The spatial distribution of the observation-reanalysis difference is also examined and diurnally dependent discrepancies in large amplitude are consistently shown over the high elevation areas such as Tibetan Plateau and Andes Mountain.