<title>Hybrid inversion algorithm for nonlinear retrieval problems and its use for water vapor profiling based on microwave sounder data</title>

Abstract

It is shown that geophysical inversion problems can be solved by usage of a hybrid algorithm, which combines optimal estimation with further optimization techniques. We employ a Bayesian approach to nonlinear inversion and discuss several extensions of this method. Especially, a sensible guess of a priori information, the shape of the probability density functions, the utility of Monte Carlo methods, and the advantages of simulated annealing have been investigated. All these techniques furnish capability of retrieving state vectors, which depend on the data in a highly nonlinear manner. A combination of these powerful tools can provide solutions to questions that cannot be tackled with standard inversion methods properly. As a moderately nonlinear optimization problem, profiling of water vapor based on downlooking microwave sounder data is a typical geophysical problem that could not be treated with standard inversion algebra adequately. Based on synthetic state vector data, we show the potential and the characteristics features of all of the hybrid algorithm's components contributing to the retrieval of the state. The hybrid algorithm has been employed in a way that it is able to provide humidity profiles in a numerically stable and computationally efficient manner. Using this example application, the benefit of a hybrid approach is demonstrated.