Development of algorithms for nonlinear temperature retrieval problems in atmospheric remote sensing using regularization methods

Infrared and microwave passive remote sensing of the atmosphere is used to determine the Earth's atmospheric state and surface properties. Radiance measured by the radiometers can be used to estimate atmospheric parameters such as temperature and water vapor content. These quantities are of primary importance for applications in meteorology, oceanography, and geophysical sciences.; In this research project, algorithms for atmospheric temperature retrievals based on radiometry from the High Resolution Infrared Radiation Sounder/2 (HIRS/2) and the Microwave Sounding Unit (MSU) were developed and validated. These are part of the TIROS Operational Vertical Sounder (TOVS), onboard the United States National Oceanic and Atmospheric Administration's (NOAA) operational satellites. The developed algorithms were based on the Gauss-Newton method for nonlinear least-squares, and the Tikhonov and Truncated Singular Value Decomposition (TSVD) regularization methods for linear and nonlinear inverse problems. A set of MATLAB{dollar}spcircler{dollar} functions for the retrieval algorithms was developed. Algorithms were validated by means of simulations using the GLA TOVS Code for Radiance and Jacobian Calculation: Version 1.0.; Results of retrievals for several initial values are presented and evaluated. The performance of the algorithms when supplied input data contaminated with noise was investigated. The estimated temperature profiles obtained for the retrievals of the noiseless cases were accurate. The computed root mean square (RMS) error was as low as 0.99 {dollar}spcirc{dollar}K for a tropopause initial guess at 10 km. Moreover, results of the retrievals for the noisy cases were unsatisfactory, since the estimated temperature profiles were inaccurate. The computed RMS error for these cases varied from 6.97 to 14.85 {dollar}spcirc{dollar}K, depending on the given initial guess.