A method for assimilating observations of ozone was implemented in the three-dimensional global stratospheric chemistry transport model ROSE. The model contains an extensive photochemical scheme which includes heterogeneous chemistry and uses temperature and wind fields from the UKMO (United Kingdom Meteorological Office) stratospheric analysis. Ozone measurements obtained by the Microwave Limb Sounder (MLS) on board the Upper Atmosphere Research Satellite (UARS) were assimilated in the model using the sequential statistical interpolation approach. The analysis is performed using a time invariant background error covariance matrix that only includes horizontal covariances. Results from a sixty day simulation are presented and it is shown that assimilation of the MLS observations results in improved global three-dimensional distributions of ozone as seen from comparisons with MLS data not assimilated in the model. For further validation, the stratospheric total ozone fields computed from the analysis are compared with the TOVS total ozone measurements and it is shown that they agree within the uncertainty of the data. ¿ 1998 American Geophysical Union