Review of "Global total precipitable water variations and trends during 1958-2021" by Wan et al. 

The authors provide a valuable update in the regional changes in atmospheric water vapour content and surface temperature since 1958 based on state of the art reanalysis systems. While earlier attempts to use reanalyses demonstrated serious defects in homogeneity over time, the newer products appear more reliable and the analysis presented is complimentary to other work. Nevertheless, as noted in the discussion, the early satellite era may contain spurious variability related to changes in the observing system assimilated into the reanalysis model while the early record may be closer to a dynamically nudged "amip" atmosphere-only climate model simulation with prescribed SST and sea ice, though radiosonde information should provide some observational input in well sampled locations. The present work provides valuable information for studies interested in evaluating climate model simulations of moist processes including water vapour feedback. There is also recent interest in discrepancies between simulations and observations of low level moisture over arid and semi arid regions that could be mentioned (Simpson et al. 2023 PNAS doi:10.1073/pnas.2302480120). Despite the improvement in water vapour changes, other aspects of hydrological cycle such as precipitation show spurious global variability sincec 1979 (Allan et al. 2020 NYAS doi:10.1111/nyas.14337) so I wonder if there is any insight into this based on the water vapour evaluation? Overall, this is a well written and presented analysis that I recommend to be published with only minor modifications. A list of suggestions and comments is provided below.

1) L10 should this be water vapor responses to lower tropospheric temperature (the thermodynamic causal route)? Radiative cooling rates affected by water vapor will also feedback on temperature but I don't think this is meant here?

2) L11 "improved" is ambiguous and can be removed

3) L13 is the trend for the whole period?

4) L26 O’Gorman & Muller could be referred to on dTPW/dT

5) L28 "strengthened greenhouse effect"; could mention that changes higher in the troposphere are more important for the feedback while changes at low levels are strongly linked with precipitation

6) L35 see also the GPS network eg Douville et al. 2022 reference

7) L45 although reanalyses are much improved over earlier versions there remain some homogeneity issues eg before the mid 1990s over the tropical ocean eg Allan et al. 2022. 

8) L141 - consistent --> inconsistent since the ERA5 decline in TPW in the 1980s is not consistent with SMMR/SSM/I microwave satellite data and appears to originate in the tropical lower troposphere over the ocean.

9) L150 can the turning points be linked to phases of Pacific Decadal "Oscillation"? The jumps in TPW also seem to coincide with rapid increases in global temperature.

10) L167 the North Atlantic cooling seems to be most prominent in northern hemisphere winter (Allan & Allan 2020 JGR doi:10.1029/2019JC015379 ) while it is more a warming "hole" in the summer. Some have liniked this with changes in ocean circulation rather than heat fluxes though this is based more on modelling (e.g. Drijfhout et al. 2012 doi:10.1175/JCLI-D-12-00490.1). 

11) L190 the rapid Arctic warming is consistent with large increases in water vapour that can be menioned here (also discussed briefly in Allan et al. 2022) or signpost to the next section.

12) L195 the reduction in relative humidity related to land/sea warming contrast could also be mentioned (e.g. Byrne & O'Gorman 2019 PNAS doi:10.1073/pnas.1722312115).

13) L206 O'Gorman & Muller (2010)  would also be appropriate to cite here

14) L210 some regions are likely to exhibit more realistic trends where there is a greater density and homogeneity of data (e.g. N America, Europe)

15) L215 it could be noted that changes in atmospheric circulation will introduce regional increases or decreases in relative humidity (e.g. wet part of circulation moves from one location to another). This could be investigated based on reanalysis vertical motion fields for example.

16) Figure 5 colour bar is not very intuitive to me (red implies drier to me)?

17) L253 - I think the SSM/I satellites only began in 1987, well after 1977?

18) 267 - can the homogeneity tests be used to assess uncertainty in the computed trends (above the structural differences between reanalyses)? Trends also should be reported with at least statistical error bars relating to the linear fit

  

19) L270 - I am surprised that the radiosonde data is not assimilated in reanalyses? Is there a reason for this?

20) L274 - why are trends reported in mm/decade here but %/decade earlier?

21) L284 - presumably the long term (1958-present) trends are mostly determined by recent rapid warming and moistening since the 1980s? It could be made clear that the values quoted here are since 2003?

22) L297 - are the decreases in the subtropical ocean cumulus transition zones explained by shifts in large-scale atmospheric circulation or changes in stability? It is noteworthy that the observed warming pattern is unlike coupled simulations, with warming more in the tropical warm pool which can increase stability in these subtropical subsidence regimes e.g. Andrews et al. 2022 JGR doi:10.1029/2022JD036675

23) L308 - the link between Arctic warming and moistening could be usefully mentioned here

24) L315 - a line of wider implications of the conclusions and future work would be welcome. Note that an intercomparison of TPW datasets is underway by Trent et al. https://doi.org/10.5194/egusphere-2023-2808. Some additional references that could also be considered are listed below:

Douville & Willett (2023) Sci. Adv. https://doi.org/10.1126/sciadv.ade6253

Patel & Kuttippurath (2023) OLA Research https://doi.org/10.34133/olar.0015

Shao et al. (2023) ACP https://doi.org/10.5194/acp-23-14187-2023

Ding et al. (2022) LNEE https://soi.org/10.1007/978-981-19-2588-7_27

Rev of Title: Global total precipitable water variations and trends during 1958-2021

Author(s): Nenghan Wan, Xiaomao Lin, Roger A. Pielke Sr., Xubin Zeng, and Amanda M. Nelson

MS No.: hess-2023-301

MS type: Research article

Rev. By Kevin E Trenberth

The topic as highlighted by the title is an important one, but one would not know from this paper that a comprehensive analysis already occurred by Allan et al. 2022.  Although the latter is referred to in a couple of spots, buried in the paper, readers would have no idea that a lot here is not new and what is new is likely suspect.   From Allan et al abstract “Global-scale changes in water vapor and responses to surface temperature variability since 1979 are evaluated across a range of satellite and ground-based observations, a reanalysis (ERA5) and coupled and atmosphere-only CMIP6 climate model simulations. Global-mean column integrated water vapor increased by 1%/decade during 1988–2014 in observations and atmosphere-only simulations.”

One should not use 1958 as a starting point for trends especially without showing time series first. Satellite data for many fields became available mainly from 1978 on, and that is highly relevant over oceans for temperatures, but for PW it was only after mid 1987 when SSM/I data began.   Allan et al also made use of SSMR data for 1979-1984.    ERA5 data are flawed prior to about 1992 when the volume of SSMI data increased substantially (see Allan et al  Fig 1b).  In general pw data are unreliable over the oceans prior to 1987.  Earlier SSMR data were not used in ERA5; see Fig 5 of Hersbach et al.   Even after 1987 changes in microwave instruments and orbits caused further discontinuities, notably in 1992 (Trenberth et al 2015).  That was evident for ERA-interim but it also holds for ERA5 as can be seen in the time series in Fig 2f of this paper for the oceans, (less so for JRA55).  However, this was missed by this paper.

Rawinsonde data have major issues through changes in the instruments and these have been noted in Trenberth et al (2005) and Zhou et al (2021), in addition to the papers included by Dai et al (2011) and Wang et al (2016). The homogenized data of Zhou et al. (2023) exhibit more spatially coherent trends and temporally consistent variations than the raw data, and lack the spurious tropospheric cooling over North China and Mongolia seen in several reanalyses and raw datasets, including ERA5.

The relationships between pw and temperature are strong over the oceans, and regression can better reveal the discontinuities in pw, see Trenberth et al 2011, 2015. 

However, even temperature data are unreliable over the oceans prior to satellite data, mainly 1979 or so, as noted about line 263.   But this reviewer does not support the procedures used to remove apparent discontinuities.

It is not clear what this paper offers that is new and trustworthy?  The paper needs major revisions and should focus more on the time series and their relationships to highlight the observational issues and discontinuities.   Then maybe some more useful trends will emerge over times for which they are credible.

Other major comments

Introduction

There is no adequate introduction to past relevant studies.  Rather than Held and Soden 2006, Trenberth (1998) and Trenberth et al (2003) are more appropriate references for the importance and role of increasing water vapor.  Trenberth et al (2005) highlighted the issues of trends in pw in earlier years and noted the values over the ocean have no basis prior to microwave measurements from satellites that began in 1987.   Even then changes in satellite instruments have led to discontinuities and erroneous values such as the transition in 1992 (Trenberth et al. 2015) that continues in the reanalyses to this day.  The introduction is especially remiss in not recognizing relevant studies, including Trenberth et al (2005), Zhou et al. (2021) who detail sonde issues up to date, and Allan et al (2022).  The latter provides analyses of the vertical structure of water vapor changes as well as trends for periods where they are credible.  Willett (2023) may also be useful.

Surface temperature

Skin temperature: what is the value of Ts?  This is not observed except when cloud free and it should be closely related to T2m over the oceans?  In any case no observations are used, as both Ts are model variables (lines 69-74) and will critically depend on cloud which is generally poorly simulated.  The paper should first straighten out the relationship between Ts and T2m in the supplemental, then use just one of those to relate to pw.  What is the point of showing both?  I recommend deleting all the panels in the figures with Ts.

Issues with Discontinuities.

The surface temperature record is well studied and there are several versions.  Differences are mostly not large.  But pw is fraught with huge sensitivity to changes in satellites and instruments, and basically few useful observations prior to microwave over the oceans in 1987.   Even then spurious changes are readily evident.   Some of these are discussed belatedly in the supplemental material.  Evidently some adjustments are made, which is quite hazardous when one subsequently computes linear trends.   

The discussion is lines 110 to 120.  It seems the discontinuities were identified statistically, and the study did not use multiple regression (i.e. use temperature and pw together).  Nor did it adequately deal with known issues in changes in satellites and their instruments.  They did “defined the temporal coincidence when the discontinuity point positions were within a 3-month window along with observing measurement systems”.  “…the times series were adjusted using the quantile-matching (QM) procedure”.  Individual radiosondes might be corrected this way but those are likely readily dealt with by bias corrections in the data assimilation system. 

I am quite bothered by these procedures, and I have no confidence they work.  It would help to illustrate results in the supplemental.  The suspicion is that they “corrected” some things that were not warranted and missed others.   Take note of Zhou et al (2021) for example.

The issue is when the whole field is off because of a satellite change.  An example is the spurious drop in pw values over the ocean in January 1992 when SSM/I changed (Trenberth et al 2015) that is not detected here, and this affects trends.  Values after then should probably be relatively higher.  This could be properly detected using a regression with temperatures for expected pw.  i.e. assume the T record is adequate and predict the pw, then examine the discrepancies to better determine the discontinuities.   In this way ENSO effects may also be included. 

It is likely that there are no reliable trends prior to about 1992 and the maps presented are not regarded as useful. More reliable results are in Zhou et al (2021) and Allan et al (2022).

In the supplemental material, Table S1 it says for ERA5 ocean “changing assimilation data from DMSP14 SSMI to 15” as 1977-01, but this occurred in 2000, not 1977.

Some detailed comments

L 63: this (discontinuities) should not be “lastly”

L 74 regridded

L 89: why start the analysis in 1958 when there are no data over the oceans. 

L118: this adjustment procedure is not valid if one then computes trends.  The whole treatment of inhomogeneities is cavalier, given the known problems documented in the literature.

L 123: see discon discussion

L 143: finally Allan et al 2022 is acknowledged.  What is new here?

L 165 why wouldn’t they show remarkable similarities?

L 179 Fig 3.  Either panels a and d or b and e should be dropped, preferably include T2m. Also Fig 4 delete f,g,h,I,jn,o,p panels.  Trends in surface temperature have appeared in many places and should not be the focus of this paper.

L 195-198: The main reason for land ocean T change differences is surely that the land involves very shallow layers and the heat capacity differences are huge.

L 215: whoa, Fig 5 has ugly colors, and the hatching makes it hard to decipher.  Again why include 5b and 5d?

This relationship over the ocean could be used to ferret out discontinuity issues and improve the reliability of results.

L 225: these results are likely wrong because of the failure to properly treat the discon issues. 

L 229: Fasullo (2011) is highly relevant here: 

In deserts and arid regions, no amount of warming is matched by pw increases owing to absence of water.  On the other hand, in monsoon regions moisture increases are amplified by atmospheric dynamics. 

In our recent paper  Cheng et al 2024 there is an analysis of the salinity that continues to show the fresh areas getting fresher and salty areas getting saltier: i.e. over the ocean, the wet areas are wetter and the dry areas are drier, and this is without complications of dry land. So shouldn’t this also be expected this over land, subject to moisture availability?

L 241. This discussion should really have occurred earlier in the methods.

L 291 on. It is unfortunate that the trends presented are for the entire period for which they are not credible.  They would be much more useful if given for when the data are adequate to support the results.

References

Allan, R. P., Willett, K. M., John, V. O., and Trent, T.: Global Changes in Water Vapor 1979–2020, J. Geophys. Res., 127, e2022JD036728, https://doi.org/10.1029/2022JD036728, 2022.

Cheng, L. J., J. Abraham, K. E. Trenberth, T. Boyer, M. E. Mann, J. Zhu, F. Wang, F. J. Yu, R. Locarnini, J. Fasullo, F. Zheng, Y. L. Li, B. Zhang, L. Y. Wan, X. R. Chen, D. K. Wang, L. C. Feng, X. Z. Song, Y. L. Liu, F. Reseghetti, S. Simoncelli, V. Gouretski, G. Chen, A. Mishonov, J. Reagan, K. Von Schuckmann, Y. Y. Pan, Z. T. Tan, Y. J. Zhu, W. X. Wei, G. C. Li, Q. P. Ren, L. J. Cao, and Y. Y. Lu, 2024: New record ocean temperatures and related climate indicators in 2023,  Adv. Atmos. Sci., https://doi.org/10.1007/s00376-024-3378-5.

Fasullo J., 2011. A mechanism for land–ocean contrasts in global monsoon trends in a warming climate. Clim Dyn. DOI 10.1007/s00382-011-1270-3

Trenberth, K. E., 1998: Atmospheric moisture residence times and cycling: Implications for rainfall rates with climate change.  Climatic Change, 39, 667–694.

Trenberth, K. E., A. Dai, R. M. Rasmussen and D. B. Parsons, 2003:  The changing character of precipitation.  Bull. Amer. Meteor. Soc., 84, 1205-1217. https://doi.org/10.1175/BAMS-84-9-1205

Trenberth, K. E., J. Fasullo, and L. Smith, 2005: Trends and variability in column-integrated water vapor.  Clim. Dyn., 24, 741-758

Trenberth, K. E., Y. Zhang, J. T. Fasullo, and S. Taguchi, 2015: Climate variability and relationships between top-of-atmosphere radiation and temperatures on Earth.  J. Geophys. Res., 120, 3642-3659,  https://doi.org/10.1002/2014JD022887

Willett, K. M. 2023: HadISDH .extremes Part I: A Gridded Wet Bulb Temperature Extremes Index Product for Climate Monitoring, Adv. Atmos. Sci.,  doi: 10.1007/s00376-023-2347-8.

Zhou, C., Wang, J., Dai, A., & Thorne, P. W. (2021). A new approach to homogenize global subdaily radiosonde temperature data from 1958 to 2018. J.Climate, 34, 1163–1183. https://doi.org/10.1175/JCLI-D-20-0352.1