Abstract. This paper presents a comparative assessment of ultraviolet nadir-backscatter and infrared limb-emission ozone profile assimilation. The Meteorological Operational Satellite A (MetOp-A) Global Ozone Monitoring Experiment 2 (GOME-2) nadir and the ENVISAT Michelson Interferometer for Passive Atmospheric Sounding (MIPAS) limb profiles, generated by the ozone consortium of the European Space Agency Climate Change Initiative (ESA O3-CCI), were individually added to a reference set of ozone observations and assimilated in the European Centre for MediumRange Weather Forecasts (ECMWF) data assimilation system (DAS). The two sets of resulting analyses were compared with that from a control experiment, only constrained by the reference dataset, and independent, unassimilated observations.Comparisons with independent observations show that both datasets improve the stratospheric ozone distribution. The changes inferred by the limb-based observations are more localized and, in places, more important than those implied by the nadir profiles, albeit they have a much lower number of observations. A small degradation (up to 0.25 mg kg −1 for GOME-2 and 0.5 mg kg −1 for MIPAS in the mass mixing ratio) is found in the tropics between 20 and 30 hPa. In the lowermost troposphere below its vertical coverage, the limb data are found to be able to modify the ozone distribution with changes as large as 60 %. Comparisons of the ozone analyses with sonde data show that at those levels the assimilation of GOME-2 leads to about 1 Dobson Unit (DU) smaller root mean square error (RMSE) than that of MIPAS. However, the assimilation of MIPAS can still improve the quality of the ozone analyses and -with a reduction in the RMSE of up to about 2 DU -outperform the control experiment thanks to its synergistic assimilation with totalcolumn ozone data within the DAS.High vertical resolution ozone profile observations are essential to accurately monitor and forecast ozone concentrations in a DAS. This study demonstrates the potential and limitations of each dataset and instrument type, as well as the need for a balanced future availability of nadir and limb sensors and long-term plans for limb-viewing instruments.