Improved technique for data inversion and its application to the retrieval algorithm for ADEOS/ILAS

Abstract

An improved numerical inversion scheme is discussed. The way the scheme is formulated allows one to include the iterative stabilization of inversion by solving constrained least square system using linear iterations. Therefore, the technique discussed is closely related to both matrix inversion and iterative inversion methods, which are often considered as alternative to each other. As a result, this inversion procedure provides both a statistically optimum and highly stable solution even in the case of a very large number of unknowns. Also, various a priori information can be used: non-negativity of solution (by logarithmic transformation), smoothness of retrieved characteristics, a priori estimations of solution, etc. The proposed technique is applied to provide statistical optimization of the retrieval algorithm for the Improved Limb Atmospheric Spectrometer (ILAS) aboard the Advanced Earth Observing Satellite (ADEOS). ILAS has 1024 visible channels and 44 IR channels to measure atmospheric transmittance at different atmospheric levels. More than 400 parameters (temperature, pressure, gas and aerosol profiles) should be determined from ILAS data. To realize statistical optimization for such a situation is quite a challenging problem. The main advantages and difficulties of applying this inversion technique for this optimization are discussed.