The amount of information which may be extracted from ozone measurements made with ground-based microwave receivers is examined. After reviewing the observing procedures and the sources of experimental errors, a theoretical analysis shows that there are at most 10 independent pieces of information which may be determined. Four cases have been investigated. It is found that two experimental procedures commonly used are not strictly equivalent. It is found that two experimental procedures commonly used are not strictly equivalent. The frequency-switching mode appears to be slightly better than the load-switching mode. Two methods involving transformations of the weighting function space have been examined. The first uses the differences between adjacent brightness temperatures, and the second a selection of the weighting functions. These transformations permit the application of the Chahine algorithm, but it is shown that they lead to an inherent loss of the retrieval accuracy even though the number of the independent pieces of information is preserved.

In the same way, a method based on a control of the information content is developed to define an optimal spectrometer. In the second part, an investigation with the Backus-Gilbert inversion technique is used to quantify the relationship of the retrieval accuracy against the spatial resolving power. The qualitative results derived by the previous method are completely expressed by this inversion technique. The influence of the frequency resolution and the bandwidth of the spectrometer on these parameters is examined, and the scanning interval of the microwave technique is described in terms of altitude band-pass filters. It is found that the retrieval accuracy increases when the altitude decreases and that the measurement capability vanishes rapidly above 65 km. The instrumental baseline is found to disturb the determination of the ozone concentration mainly at low altitudes. However, this effect may be regarded as a calibration term which does not affect the measurements of the ozone variability. Practical considerations lead to a preference for the load-switching technique in spite of the small theoretical advantage yielded by the frequency-switching mode. The results are presented for the 61--5→60--6 transition but can be easily transposed for the other ozone transitions. The theory and many qualitative results apply to the microwave measurements of the other minor atmospheric constituents. ¿ American Geophysical Union 1989