The manuscript Stratopause trends observed by satellite limb instruments by Dubé et al. examines trends in temperature and stratopause height over the 2005-2021 period using OSIRIS observations. The results are compared with SABER measurements and chemistry-climate model simulations. The authors apply a newly developed OSIRIS temperature retrieval, allowing a latitude-resolved observational assessment of stratopause height variability and trends. Comparison of these trends with chemistry-climate models provides useful context for model representations of the middle atmosphere. Together, these aspects make the study a valuable contribution to middle atmosphere research. I recommend publication in Atmospheric Chemistry and Physics after the comments below are addressed.

Page 1, line 15: The statement “little attention has been given to the stratosphere”  is bit misleading, as it is immediately followed by a discussion of the importance of stratospheric temperature trends, which have been investigated in several earlier studies. The authors should clarify that the lack of attention refers specifically to observational studies of stratospheric height rather than to stratospheric trends more generally.

Page 4, line 101: The stratopause is identified as the local maximum using cubic interpolation of monthly zonal means. Although cubic interpolation is reasonable it can introduce small oscillations, especially near boundaries or when gradients are weak. Have the authors tested whether the diagnosed stratopause height and trends are sensitive to the choice of interpolation method? Brief discussion or comparison with different method would help demonstrate robustness. 

Page 4, Section 2.5:

• Please state in Section 2.5 that LOTUS model was used (and version), and list the source of proxies implemented.
• It would be helpful to state explicitly that the MLR is applied to the deseasonalized stratopause height and temperature anomalies described in Section 3.1.
• The MLR equation includes QBO and ENSO as regression terms with constant coefficients which means they are treated as season-independent. Since the impacts of QBO and ENSO are seasonally modulated could the authors clarify this choice and comment on how sensitive the stratopause height and temperature trends are to this assumption?

Page 8, section 3.3: The model results would benefit from a short comparison with the stratopause height trends reported in previous modelling studies cited in the manuscript, i.e., Pisoft et al. 2021.

Page 7&8, Figure 3&4: The authors note that the OSIRIS sampling pattern affects the SABER stratopause height trends, particularly near 50°. As differences at other latitudes are relatively small, could the authors provide additional insight into why this latitude appears to be especially sensitive to the sampling choice?

Page 9, Figure 5: In section 2.3 authots states that 11 CCMI-2022 models are used, but Figure 5 shows distributions based on only seven models. Please clarify the discrepancy.

Page 9, line 186:  Figure 5 presents a multimodel distribution of stratopause trends but each model contributes only a single free-running ensemble member. As a result, the spread reflects both structural model differences and internal variability based on single realization. One sample only is not representative, especially with the large variability in the tropics.This represents a limitation that should be mentioned by the authors. 

Page 9, line 197: The authors describe this study as the first comprehensive observational analysis of stratopause height trends, while also noting in the Introduction that some observational studies have previously examined stratopause heights. These statements needs some modification (e.g., emphasizing the latitude-resolved nature of the analysis or the use of the new OSIRIS temperature retrieval) to avoid overstating complete lack of work in this area.