Speeding up the aerosol optical thickness retrieval using analytical solutions of radiative transfer theory

We present here the aerosol retrieval technique FAR that uses radiative transfer computations in the process of retrieval rather than look-up tables (LUT). This approach provides operational satellite data processing due to the use of the accurate and extremely fast radiative transfer code RAY previously developed by authors along with approximate analytical solutions of the radiative transfer theory. The model of the stratified atmosphere is taken as two coupled layers. Both layers include aerosol scattering and absorption, molecular scattering and gas absorption. The atmosphere parameters are assumed to change from pixel to pixel in the lower atmosphere layer, but the upper stratified layer of the atmosphere over 2–3 km is supposed to be horizontally homogenous for the frame under retrieval. The model of the land spectral albedo is taken as a weighted sum of two a priory chosen basic spectra. The aerosol optical thickness (AOT), Angström exponent and the weight in the land spectral albedo are optimized in the iteration process using the least-squares technique with fast computations of the derivatives of radiative characteristics with respect to retrieved values. The aerosol model and, hence, the aerosol phase function and single scattering albedo, is predefined and does not change in the iteration process. The presented version of FAR is adjusted to process the MERIS data. But it is important that the developed technique can be adapted for processing data of various satellite instruments (including any spectral multi-angle polarization-sensitive sensors). The use of approximate analytical radiative transfer solutions considerably speeds up data processing but may lead to about 15–20% increase of AOT retrieval errors. This approach is advantageous when just the satellite data processing time rather than high accuracy of the AOT retrieval is crucial. A good example is monitoring the trans-boundary transfer of aerosol impurities, particularly in the case of emergencies such as volcano eruptions, or various industrial disasters. Beside, two important problems that determine the accuracy of the AOT retrieval are considered. The first one is the effect of the preliminary choice of the aerosol model, particularly for the retrieval from satellite instruments providing only spectral data (MERIS, MODIS). The second problem is the influence of clouds in adjacent pixels. As for our knowledge, this problem has not been given the required attention up to now and it should be properly accounted for in the AOT retrieval algorithms.