Error analysis applied to several inversion techniques for the retrieval of limb scanning spectroscopic measurements

PULIAFITO, ENRIQUE, ; BEVILACQUA, R.M.; OLIVERO, J.

Houston, Texas

Simposio; International Geoscience and Atmospheric Remote Sensing Symposium; 1992

IEEE Remote Sensing Association

The Rodgers formal retrieval error analysis allows the quantitative determination of such important retrieval properties as measurement error sensitivity, resolution, and inversion bias. Therefore, it is an extremely powerful tool for the objective evaluation and comparison of retrieval techniques. We have applied this error analysis to five numerical inversion techniques (three of the linear matrix class, two of them constrained, and two non-linear iterative techniques) used for the retrieval of middle atmospheric constituent concentrations from limb scanning millimeter wave spectroscopic measurements. We find that the iterative methods have better vertical resolution, but are slightly more sensitive to measurement error than the constrained matrix methods. This results from the fact that (in the limit) the iterative methods converge to the exact solution, whereas two of the matrix methods under consideration have an explicit constraint, the sensitivity of the solution to the a priori profile. Tradeoffs of these retrieval characteristics are presented.