Sensitivity of Satellite-Based Skin Temperature to Different Surface Emissivity and NWP Reanalysis Sources Demonstrated Using a Single-Channel, Viewing-Angle-Corrected Retrieval Algorithm

Abstract:

Single-channel algorithms for satellite thermal-infrared- (TIR-) derived land and sea surface skin temperature (LST and SST) are advantageous in that they can be easily applied to a variety of satellite sensors. They can also accommodate decade-spanning instrument series, particularly for periods when split-window capabilities are not available. However, the benefit of one unified retrieval methodology for all sensors comes at the cost of critical sensitivity to surface emissivity (ε_s) and atmospheric transmittance estimation. It has been demonstrated that as little as 0.01 variance in ε_s can amount to more than a 0.5-K adjustment in retrieved LST values. Atmospheric transmittance requires calculations that employ vertical profiles of temperature and humidity from numerical weather prediction (NWP) models. Selection of a given NWP model can significantly affect LST and SST agreement relative to their respective validation sources. Thus, it is necessary to understand the accuracies of the retrievals for various NWP models to ensure the best LST/SST retrievals. The sensitivities of the single-channel retrievals to surface emittance and NWP profiles are investigated using NASA Langley historic land and ocean clear-sky skin temperature (T_s) values derived from high-resolution 11-μm TIR brightness temperature measured from geostationary satellites (GEOSat) and Advanced Very High Resolution Radiometers (AVHRR). It is shown that mean GEOSat-derived, anisotropy-corrected LST can vary by up to ±0.8 K depending on whether CERES or MODIS ε_s sources are used. Furthermore, the use of either NOAA Global Forecast System (GFS) or NASA Goddard Modern-Era Retrospective Analysis for Research and Applications (MERRA) for the radiative transfer model initial atmospheric state can account for more than 0.5-K variation in mean T_s. The results are compared to measurements from the Surface Radiation Budget Network (SURFRAD), an Atmospheric Radiation Measurement (ARM) Program ground station, and NOAA ESRL high-resolution Optimum Interpolation SST (OISST). Precise understanding of the influence these auxiliary inputs have on final satellite-based T_s retrievals may help guide refinement in ε_s characterization and NWP development, e.g., future Goddard Earth Observing System Data Assimilation System versions.