Microwave Radiometric Observations of the Middle Atmosphere: Simulations and Inversions

Sammanfattning: Two well-known examples of global environmental problems are the ozone depletion and the enhanced greenhouse effect. The importance and the complexity of these questions generate a demand for improved atmospheric observations, and microwave radiometry is an approach which is increasingly important for measurements of the middle atmosphere (10-100 km). This technique is suitable for the detection of e.g. O₃, H₂O, ClO, HCl, N₂O and HNO₃, all of which are species with key roles for ozone chemistry and/or the greenhouse effect. An important step in the development of microwave radiometry will be taken by the Odin satellite, as it will carry the first sub-mm instrument for atmospheric observations from space.

Microwave radiometry needs an inversion process to convert measured emission spectra to vertical distributions of the gases studied. The inversion forms an ill-posed problem, and requires that atmospheric radiative transfer and sensor characteristics can be simulated by a forward model.

This dissertation treats forward model simulations and inversion methodology generally, and also the application of the implemented algorithms to estimate profile retrieval accuracy for both the Odin sub-mm sensor and some ground-based instruments. These preparatory studies will e.g. constitute the basis for a first operational inversion code for the Odin sub-mm limb sounding observations.

The forward model developed handles atmospheric radiative transfer, including refraction, for a non-scattering atmosphere in local thermodynamic equilibrium. The frequency region of primary concern is 10-1000 GHz. Considered sensor parts are the antenna, the mixer, the sideband filter and the spectrometer. Species abundance weighting functions are calculated by means of semi-analytical expressions.

The Optimal Estimation Method (OEM), both linear and non-linear versions, has been used for studies focusing on retrieval performance for specific observation systems. Beside OEM, the use of linear regularization of Philips-Tikhonov-Twomey type, here denoted as damped least squares (DLS), has been considered. Selection criteria to determine the DLS regularization parameter for ensemble retrievals have been designed, and considerations for OEM and DLS to obtain best possible retrieval accuracy have been compared.

Special attention has been given to inversion characterisation. Two response measures have been introduced to describe averaging kernel matrices for multiple species. These measures can further be used for a simple test of the linearity of a specific inversion problem. Ozonesonde data have been analysed to obtain statistical variables of ozone and temperature variability. These results are of direct significance for e.g. error characterisation and applications using OEM.

Detailed inversion simulations have been performed to examine retrieval limitations, accuracy and altitude coverage for the different observation modes of the Odin sub-mm sensor. In short, depending on observation mode and number of spectrometers running, 4-6 species can be detected simultaneously. The majority of the Odin profiles will be derived using a linear inversion approach, but to increase the retrieval altitude ranges downwards, also non-linear methods have been studied.