This manuscript examines tropical tropospheric ozone and its trends across different observational datasets including in situ (IAGOS aircraft, SHADOZ ozone sondes) measurements and six satellite products. To increase data coverage, the IAGOS and SHADOZ datasets are fused together over three regions (tropical Americas, Southeast Asia, and Malaysia/Indonesia).  A major contribution is the use of satellite products with larger sample sizes to demonstrate strong sensitivity, including in the sign, of the trends detected by sparse sampling. The tropical tropospheric ozone distributions, trends, and statistics reported in the manuscript are important benchmarks of tropical tropospheric ozone for comparison with past and future trends, and for model evaluation.  The paper should be suitable for publication in ACP once the major concerns noted below are addressed; other comments are intended to streamline the paper to ensure key messages are clearly communicated.

Major Concerns

1. This assumption underlying the correction of satellite data wth the in situ datasets from 2014-2019 at first seems in conflict with the conclusion that the in situ datasets are too sparse to reliably detect trends (for example lines 53-59; 769-771). Is the sample size sufficiently larger in this time period? Figure 8 is pointed to as the justification (lines 945-946) yet only shows trends in the satellite datasets, rather than the 2014-2019 mean in situ vs satellite data comparisons used for the correction (Figures 4 and 5) but the discussion of these figures is confusing as noted below, and it’s not clear if the satellite instrument sensitivity is accounted for (see also point 4 below and comments on Figures 4 and 5). Clarification is needed to ensure the reader can understand the assumptions involved.
2. The drift correction to OMI/MLS TCO is noted in Lines 479-480 and discussed in Section S2 but the actual documentation is quite thin if this manuscript is to serve as the sole description. If it is documented elsewhere, references are needed. Where can the reader see the evidence for the drift discussed in lines 139-141? Why is there sufficient confidence to use the ozone sondes to correct the satellite drift in light of the issue so clearly demonstrated with trend detection in sparsely sampled datasets? The ground-based total ozone comparisons probably offer some critically needed independent evidence, assuming those do not suffer from the same sparse sampling as the sondes, yet rely on a personal communication.  Why not include figures clearly demonstrating the drift in the supplement?  Is the OMI CCD product sufficiently independent from the OMI/MLS to be used to cross-check? It is important to walk the reader through the evidence here.
3. Data Availability. Will the authors provide the fused regional datasets and trends, as well as their codes for deriving them? That would help to ensure reproducibility while also lowering the barrier for their use in model evaluation. Are all values included for all days with data in each year when determining quantiles or is some effort made to normalize sample sizes across months? Section 2.5 does not provide sufficient detail to allow reproducibility of the trend analysis. The coefficients fit to equation 1 could be reported in the supplement.
4. How different are the vertical sensitivities of the various satellite instruments used to derive tropospheric column ozone? Lines 720-721 note the diminished sensitivity in the boundary layer, but it isn’t clear if these are accounted for when comparing the satellite columns with those sampled in situ prior to bias-correcting the satellite data. Has an averaging kernel been applied to the ozone sonde or IAGOS or fused data, for example as in Zhang et al. ACP 2010 (doi:10.5194/acp-10-4725-2010)? 

Specific comments

Figures 4 and 5 and surrounding discussion can be streamlined by only including the apples-to-apples comparisons in the main text that are used in the correction as a positive offset (rather than a “high bias”) is expected in Figure 4 for the satellite TCO products that retrieve through the full depth of the tropical troposphere. The alternative comparisons could be moved to the supplement to demonstrate the importance of using the correct vertical top of the column.  More generally, the use of the word “bias” should be carefully considered in the text, as its use for some of the panels in Figures 4 and 5 may mischaracterize the situation given the known difference in the tropospheric column thickness for in situ versus remotely sensed datasets.

Figure 6. Consider adding shading (2-sigma range?) to show if the fused data is statistically significantly different?  If there is sufficiently larger confidence in the fused data, why not focus on only that dataset in the text, and move the comparisons to supplement?

Lines 665-671.  This discussion is difficulty to follow.  Isn’t the MOZAIC dataset now included in IAGOS?  In any case, this could be shown in Figure 7.

Figure 7. Consider limiting the y-axis range to just the tropics to enhance legibility. Why are the months listed in the title?

Table 1 reports important information; is there a summary figure, perhaps in the style of the Figure 3 or 7 maps that could better communicate the key points, and the table could move to supplement?

Lines 897-898.  Is the range over the averages of the different products, or are spatial differences also part of this range?

Lines 899-901 seem misleading given the known mismatch of tropospheric column depth; why not just report the correctly matched columns?

Line 905. Need to define TTOB.

Lines 908-909.  What is being compared here?  The equator versus mid-latitudes, or the parallel tropical latitude bands in the northern vs southern hemisphere?

Line 953-954.  Some platforms have different vertical extents; isn’t it just a matter of accounting for these differences rather than needing to define a common top?

Lines 954-957.  Can this be expanded?  What is the target here, to stitch together different satellite products or something else?

Section S1 line 37.  Figure S1 --> Figure S2?

Section S1 lines 66-72.  Is it worth noting that natural variability can influence trend attribution (not just detection)? Is the statement that sampling frequency versus natural variability are typically inseparable true?  The authors seem to have succeeded in isolating sampling frequency issues by sampling the satellite data with the frequency of the in situ data.

Section S3.  It seems that something is missing here; is it possible to have high confidence that there is not a trend?  Specifically for the Americas, should there instead be moderate confidence of an insignificant trend?  Should Malaysia/Indonesia be downgraded to low confidence?  The justifications here seem a bit arbitrary.

Figure S2: blue crosses --> diamonds?

Figure S7: Caption/labels could better explain what is being plotted (i.e., why are there two panels per site; how is ND calculated). Consider plotting the 4 panels that should be compared directly on the same page.

Figure S18: Is ECCAD using GFED fire emissions or something else?  A reference or description would be helpful for readers not familiar with this inventory.

Summary

This paper presents a comprehensive analysis of tropospheric ozone in the tropics. In situ data from IAGOS and SHADOZ are presented alongside a fused in situ product and long-term satellite records. Trends are presented for multiple regions in the tropics and analyzed using techniques such as satellite+in situ comparisons and bias corrections. The authors discuss contributing factors, such as tropopause definition, ozone precursor data, and data availability. Many previous studies are cited and presented clearly throughout the Introduction and Results. This paper will be a productive contribution to ACP once a couple of comments are addressed.

Major comments

• Section 2.1.1 states “To compare with the satellite data, the profiles were averaged monthly before being converted to a tropospheric column value ranging from the surface up to 270 hPa or up to the maximum altitude (~ 200 hPa)”. A similar statement is made in Section 2.1.2 about the SHADOZ data. Did you consider applying satellite averaging kernels and priors to the IAGOS and SHADOZ profiles in order to make a more direct comparison? Why did you decide against using this method?
• Section 2.3.4 states “This produced a 1-1.5 DU difference between the earlier and latter record for stratospheric column ozone, which prevents accurate trend detection from either MERRA2 stratospheric column ozone or the derived tropospheric column ozone from OMPS/MERRA2”. However, the OMPS trend was reported in Figure 8. Why was OMPS included if an accurate trend detection is not possible?
• A bullet point in the Conclusions section should be added to address the importance of tropospheric column definition, given your results showing the impact of columns calculated up to 270, 200, 150, or 100 hPa (in addition to your mention of this topic in the paragraph beginning on line 953).

Minor comments

1. Line 277: define a.s.l.
2. Section 2.3.2 states that the tropospheric column “is constrained in the 15 ̊S-15 ̊N latitude band inherent to the CCD technique”. How can the TROPOMI CCD method reach outside of this band, from 20°S to 20°N?
3. OMI/MLS and OMPS both use the thermal tropopause, which varies seasonally. Did you consider recalculating the TTCO using a constant tropopause (at 100 hPa) to be more similar to OMI CCD? I suspect that the difference would be small, and this is not a critical edit to make, but it could be an interesting figure in the SI to present TTCO trends using OMI/MLS and OMPS up to the thermal tropopause versus up to 100 hPa.
4. Check the grammar of the sentence spanning lines 388-389.
5. Why are the CrIS and IASI/GOME2 tropopause pressures not plotted in Figure S1?
6. Define ND in Figures S7-S11.
7. Many of the figures (e.g., Figures 2 and 3) use TCO in the figure labeling, but TTCO in the captions. Should they all be TTCO?
8. Why does Figure S16 say “(nmol mol^-1 which is equivalent to nmol mol^-1)”?
9. Line 699: change “cannow” to “can now”.
10. The Conclusion is presented in a straightforward and helpful manner. For readers who may only read the Conclusion of the paper, you may consider also listing the relevant sections/figures/tables associated with each bullet point so that those readers can easily find the sections that they’re interested in learning more about.
11. Do you plan to extend this work to a global study? If so, please mention that in the Conclusion.
12. Why is Appendix A an appendix and not a section in the SI?