Satellite Aerosol Climatology Using AVHRR Channel 1 and 2 Radiances: An Update of the GACP Algorithm

The evaluation of the direct and indirect climate forcing by aerosols requires the knowledge of the global distribution of aerosol amount, size, and composition. This goal may eventually be achieved by combining satellite observations and transport models. Here we present an updated version of the algorithm for retrieving aerosol optical thickness (AOT) and Angstrom coefficient over the ocean using channel 1 and 2 AVHRR radiances developed within the framework of the Global Aerosol Climatology Project (GACP).

The main improvements include a modified aerosol model used in the retrievals, the addition of a diffuse component to the specular ocean reflectance, and changes in the retrieval approach. Compared to the initial aerosol model, the imaginary part of the aerosol refractive index was reduced in order to achieve a better balance between non-absorbing and weakly absorbing sea salt and dust aerosols. Our previous sensitivity study has shown that as the retrieved aerosol optical thickness decreases, the uncertainty in the retrieved Angstrom coefficient increases. In addition, as smaller AOTs correspond to smaller radiances for a given scattering geometry, the retrieval results become increasingly dependent on the instrumental discretisation of AVHRR radiances and the value of the "deep space" count, rendering the retrieval of the Angstrom coefficient unreliable. Analyses of the initial algorithm revealed that in many cases the range of uncertainty in the retrieval of the Angstrom coefficient exceeded the range of its natural variability. Although the effect on the average optical thickness was limited due to small respective AOT values, it was quite evident from the maps of the monthly mean Angstrom coefficient.

In the updated algorithm we address this problem in two steps. First, we retrieve the Angstrom coefficient for the points where its value does not reach the limits of our lookup table. Second, we construct asequence of 1x1 deg maps of the Angstrom coefficient by increasing the period of averaging from 1 month to 3 months to 1 year to achieve complete global coverage. We then use these fixed average values and 1st chanell radiances to retrieve the optical thickness for those pixels where the direct retrieval of the Angstrm coefficient has failed.

The new algorithm was applied to the ISCCP DX dataset corresponding tothe period of NOAA-9 observations (Feb. 1985 - Nov. 1988). An additional change compared to the original retrieval algorithm was the use of the post-launch NOAA calibration of channel 2 radiances rather than the prelaunch calibration. We will present typical results of applying the new retrieval algorithm and compare them with those obtained with the previous version of the algorithm.