As I know, the CTD measurements have shown pressure dependency and hysteresis, with each probe exhibiting unique characteristics. While the authors have made efforts to correct this, such as the "linear pressure correction to the conductivity data" (L370) and offset correction for "the drift of conductivity measurement" (L177) using the WOCE/GO-SHIP dataset, there are lingering doubts about their sufficiency.

My concerns arise from the steep salinity increases close to the bottom in main figures and the substantial salinity deviations between corrected data and neighboring WOCE/GO-SHIP datasets seen in Figures S2 and S3. To address this, I guess it's crucial to compare vertical profiles among each CTD probe. Unfortunately, the main body of the manuscript lacks such profiles, with only limited data in Figures S2 and S3.

The phenomenon of increasing salinity in the hadal zone is fascinating. If confirmed, it could significantly contribute to oceanography and deep-sea biology. I remain hopeful that the data proves to be valid and reliable, considering its potential implications.

This manuscript presents analyses of CTD data taken in and over deep ocean trenches by GO-SHIP repeat hydrographic cruises (incorrectly referred to as WOCE cruises in the manuscript) and by full-ocean-depth landers.  This is a revision of an earlier manuscript, and is much improved over that earlier version. It contains interesting new information, and should be suitable for publication following revision. Specific comments follow, indexed by line number, L, where applicable.

1. Title.  Consider deleting "Examining baseline" in the title.

2. L15. Consider changing "Increases in salinity patterns" to "Salinity increases with increasing depth".

3. L20-110 and discussion section.  The potential effects of geothermal heating on halal trench mixing (e.g., van Haren, 2023, Dynam. Atm.& Oceans) as well as T-S evolution (e.g., Joyce et al., 1986, Deep-Sea Res. A) and stability should be introduced here and then incorporated into the discussion.  Even the weak bottom heating in trenches will tend to cause convective turbulence in a (possibly quite thick) bottom mixed layer, working against establishment or maintenance of stabilizing deep salinity gradients potentially caused by other mechanisms.

4. L22. Consider changing "basins" to "a few deep basins".

5. L25. Consider changing "cool" to "cold", and is light penetration really "limited" at 6000 m?  This reviewer would have thought it was effectively zero, although they are admittedly not an optical oceanographer.

6. L28-29. This sentence is confusing and needs to be rewritten, perhaps split into two.  "The 2-dimensional-area of the seafloor with depths greater than 1% (Harris et al., 2014)." is fine.  However, what is meant by the second clause?  Volume and depth are treated as somehow equivalent, which is confusing dimensionally and conceptually. At any rate, that second clause needs rethinking.

7. L50. Change "between" to "among".

8. L128. There is a grammatical error here that needs to be fixed so that readers can understand the meaning of this sentence.

9. L137. Add a comma before the last and in the series.

10. L177-181. The WOCE field program ceased circa 1998, 2000 at the latest.  Repeat Hydrographic sections collected along historical WOCE lines within ±4 years of the lander expeditions would have been completed under the auspices of GO-SHIP (e.g., Sloyan et al., 2019, Frontiers in Marine Sci.).  It would probably be useful to the reader to cite the years of the GO-SHIP sections used at each WOCE historical site.  Also, how were the offsets determined?  The optimal way would be to use conservative or potential temperature as the independent variable, and adjust the lander salinity to match the GO-SHIP CT-SA relation in a relatively stable portion of the water column (e.g., small lateral gradients and relatively slow circulation - likely the "oldest" deep waters rather than the more recently ventilated and presumably more variable bottom waters).

11. L203-15. OMP analysis typically takes advantage of a non-negativity constraint and requires the water mass fractions to add up to unity, both of which make the calculation better determined.  It is not clear from this description that this was done.  It probably should be, otherwise OMP should not be invoked.  In Matlab the function lsqnonneg in the optimization toolbox would be useful for adding the non-negativity constraint. Also, was any weighting used, as customary in OMP? If not, please note that, and if so, please note what it was.  If this is all too much, it would be fine to reframe the problem as simple end-member mixing in CT-SA space with two end-members, and not OMP at all, since it could be simplified to that if desired.

12. L227 and following.  The discussion here mostly quotes gradients from 4000 dbar to the bottom of the profiles. This practice mixes the regions above and below the sill of the trench. Readers might be more interested in gradients from the sill depth (which should be estimated for each trench) and the bottom.  This would allow a focus on trench processes and dynamics, rather than mixing trench and deep water processes and dynamics.

13. L239 (and elsewhere?).  Change "monotonously" to "monotonically". ;-)

14. Figures 2-8 and discussion. It is interesting that almost all the lander CTD profiles (with the exception of in the Japan Trench) exhibit increasing salinity with increasing pressure at high pressures (salty tails) with various amplitudes, whereas the GO-SHIP data "tails" are either absent or small (the P08 "fresh tails" are implausibly statically unstable, and are nearly within the ±0.002 PSS-78 instrumental uncertainty). All of the GO-SHIP cruise CTD data would be calibrated to bottle salinity data. In general that calibration would include a conductivity cell compressibility coefficient that was determined by least squares fitting along with other calibration coefficients for each co sensor used on that cruise (but maybe the P08 calibration didn't include that term?). So the correction would be specific to the cruise and the sensor.  It would often be different from the nominal correction (based on the compressibility of glass) that Seabird Scientific provides.  Certainly the CTDs used on Deep Argo floats have exhibited a noticeable artifact owing to this issue (Kobayashi et al., 2021, Prog. Oceanogr.)  In addition, it seems possible that under the truly extreme pressure experienced by the lander CTDs, some nonlinearity in the interaction between the glass co cells and their plastic protective jacket could come into play.  So without careful (e.g., done to GO-SHIP standards) bottle salinity analyses with multiple samples collected at a variety of pressures (from the trench sill to the bottom) this reviewer is quite skeptical regarding the salinity increases with increasing pressure reported by the lander CTDs. They could be real, but a more likely explanation is that they are an artifact owing to an incorrect coefficient, or even an inadequate model (e.g., linear when it perhaps should be non-linear), used to correct for conductivity cell compressibility.  The discussion should probably reflect this perspective.

15. L339-340. There is a repeated phrase in here.   Please edit to remove the repetition.

16. P369 and elsewhere.  Practical Salinity is reported on the dimensionless Practical Salinity Scale of 1978 (PSS-78) and Absolute salinity has "units" of g/kg.  There is no such thing as "psu".  Please revise the manuscript throughout accordingly.

17. L378. The varying rates of salinity increase with increasing pressure could easily be solely due to instrumentation.  The same co sensor used on different cruises can require different compressibility correction coefficients as it ages.