Thermal infrared remote sensing of mineral dust over land and ocean: a spectral SVD based retrieval approach for IASI

From the high spectral resolution thermal infrared observations of the Infrared Atmospheric Sounding Interferometer (IASI) mineral dust AOD (transferred from thermal infrared to 0.5 mu m) is retrieved using a Singular Vector Decomposition of brightness temperature spectra. As infrared retrieval based on 8-12 mu m observations, dust observation with IASI is independent from solar illumination. Through the linear combinations of suitable independent singular vectors weighted by their contribution to the observed signal, and a projection of different a-priori dust spectra on the resulting signal the dust can be well distinguished from the influence of surface emissivity and gas absorption. In contrast to lookup-table based single-channel retrievals this method takes advantage of the spectral shape of dust extinction and surface and atmosphere influence over the total 8-12 mu m window band. Using different a-priori spectra for dust extinction allows also for an estimation of dust particle size in terms of effective radius based on the respective dust model size distributions. These dust models are also used for the transfer of infrared AOD to 0.5 mu m. Four months of IASI observations covering Northern Africa and Arabia are used for evaluation. Two large scale dust events, one covering the Arabian Peninsula and adjacent parts of the Indian Ocean, the other over the Atlantic Ocean off the coast of West-Africa, are analysed and compared with other satellite images. They also show the good suitability of IASI data for dust observation at day and night. Monthly means derived from IASI observations represent well the known seasonal cycles of dust activity over Northern Africa and Arabia. IASI Dust AOD(0.5 mu m) and AERONET coarse mode AOD(0.5 mu m) are reasonably well (linearly) correlated with rho = 0.623. Moreover, comparison of time series of AERONET and IASI observations shows that the evolution of dust events is very well covered by the IASI observations. Rank correlation between dust effective radius and AERONET Angstrom exponent is -0.557 indicating the general capability of (qualitative) dust particle size information being provided by this method.