The retrieval of column abundances of atmospheric gases from atmospheric infrared spectra necessarily depends on the choice of the a priori vertical profiles of the atmospheric species involved. Most commonly a standard profile is adopted, and some additional parameter is introduced in the retrieval algorithm for improving the final result. This parameter may be unrelated to geophysical arguments. For the retrieval of ozone columns, it is proposed here to use a model of a priori O3 vertical profiles which better reflect the real state of the atmosphere at the time and location of the measurement. The model profiles depend on the daily local thermal tropopause height and include a climatological seasonality. The model has been developed for the retrieval of ozone columns from high-resolution Fourier transform infrared (FTIR) solar absorption measurements at the International Scientific Station of the Jungfraujoch in the Swiss Alps. It is demonstrated that the use of the model significantly improves the quality of the spectral fits while minimizing the need for adjusting additional nongeophysical parameters to compensate for atmospheric changes and that it enhances the accuracy of the retrieved O3 column data. Such is done by comparison of the retrieved FTIR O3 data with validated independent O3 data sets from co-located Syst¿me d'Analyze par Observations Z¿nithales and nearby Brewer instruments. Contrary to some other approaches for improving the FTIR O3 retrieval, the model concept developed here can be applied to O3 observations at any midlatitude site, and it can be implemented in an operational retrieval procedure. ¿ 1999 American Geophysical Union