Development of a Global Validation Package for Satellite Oceanic Aerosol Optical Thickness Retrieval Based on AERONET Observations and Its Application to NOAA/NESDIS Operational Aerosol Retrievals

In this paper, a global validation package for satellite aerosol optical thickness retrieval using the Aerosol Robotic Network (AERONET) observations as ground truth is described. To standardize the validation procedure, the optimum time/space match-up window, the ensemble statistical analysis method, the best selection of AERONET channels and the numerical scheme used to interpolate/extrapolate these observations to satellite channels have been identified through sensitivity studies. The package is shown to be a unique tool for more objective validation and inter-comparison of satellite aerosol retrievals, helping to satisfy an increasingly important requirement of the satellite aerosol remote sensing community. Results of applying the package to the 2 generation operational aerosol observational data (AEROBS) from NOAA-14/AVHRR in 1998 and to the same year aerosol observation data (CERES-SSF4) from TRMM/VIRS are presented as examples of global validation. The usefulness of the package for identifying improvements to the aerosol optical thickness (τ) retrieval algorithm is also demonstrated. The principal causes of systematic errors in the current NOAA/NESDIS operational aerosol optical thickness retrieval algorithm have been identified and can be reduced significantly, if the correction and adjustment suggested from the global validation are transfer model parameters that reduce systematic errors in τ retrievals are suggested for consideration in our next generation algorithm. Basic features that should be included in the next generation algorithm to reduce random error in τ retrievals and the resulting error in the effective Angstrom wavelength exponent have also been discussed. Compared to the AERONET observation, the NOAA-14/AVHRR (AEROBS) τ values for mean conditions are biased high by 0.05 and 0.08, with random errors of 0.08 and 0.05, at 0.63μm and 0.83μm, respectively. Correspondingly, the TRMM/VIRS (CERES-SSF4) values for mean conditions are biased high by 0.06 and 0.02, with random errors of 0.06 and 0.04, at 0.63μm and 1.61μm, respectively. After corrections and adjustments to the retrieval algorithm, the biases in both channels of AVHRR and VIRS are reduced significantly to values close to zero, although random error is almost unchanged. The effective Angstrom wavelength exponent (α) derived directly from the aerosol optical thicknesses (τs) has been shown to be poorly correlated both before and after adjustments, indicating that random error in the τ measurement (possibly related to aerosol model parameter variations or cloud/surface reflectance contamination) needs to be reduced.