the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 29 Nov 2023
Evaluation of Total Column Water Vapour Products from Satellite Observations and Reanalyses within the GEWEX Water Vapor Assessment
Abstract. Since 2011 the Global Energy and Water cycle Exchanges (GEWEX) Water Vapor Assessment (G-VAP) has provided performance analyses for state-of-the-art reanalysis and satellite water vapour products to the GEWEX Data and Analysis Panel (GDAP) and the user community in general. A significant component of the work undertaken by G-VAP is to characterise the quality and uncertainty of these water vapour records to; i) ensure full exploitation and ii) avoid incorrect use or interpretation of results. This study presents results from the second phase of G-VAP, where we have extended and expanded our analysis of Total Column Water Vapour (TCWV) from phase 1, in conjunction with updating the G-VAP archive. For version 2 of the archive, we consider 28 freely available and mature satellite and reanalysis data products, remapped to a regular longitude-latitude grid of 2°× 2°, and on monthly time steps between January 1979 and December 2019. We first analysed all records for a 'common' short period of five years (2005–2009), focusing on variability (spatial & seasonal) and deviation from the ensemble mean. We observed that clear-sky daytime-only satellite products were generally drier than the ensemble mean, and seasonal variability/disparity in several regions up to 12 kg/m2 related to original spatial resolution and temporal sampling. For 11 of the 28 data records, further analysis was undertaken between 1988–2014. Within this 'long period', key results show i) trends between -1.18±0.68 to 3.82±3.94 kg/m2/decade and -0.39±0.27 to 1.24±0.85 kg/m2/decade were found over ice-free global oceans and land surfaces respectively, and ii) regression coefficients of TWCV against surface temperatures of 6.17±0.24 to 27.02±0.51 %/K over oceans (using sea surface temperature) and 3.00±0.17 to 7.77±0.16 %/K over land (using surface air temperature). It is important to note that trends estimated within G-VAP are used to identify issues in the data records rather than analyse climate change. Additionally, breakpoints have been identified and characterised for both land and ocean surfaces within this period. Finally, we present a spatial analysis of correlations to six climate indices within the “long period”, highlighting regional areas of significant positive and negative correlation and the level of agreement among records.
- Preprint
(5480 KB) - Metadata XML
-
Supplement
(15406 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
CC1: 'Comment on egusphere-2023-2808', Richard Allan, 09 Feb 2024
This is an important and well constructed study that should in my opinion be published. It's impact could be improved with enhanced motivation to bring out the importance and some more direction to the community in terms of more/less reliable products and strengths/weaknesses or limitations and recommending more or less suitable applications. I just have minor comments listed below.
1) L14 ice free regions change with the season and year - will this alias (slightly) into the variability? TWCV --> TCWV
2) L15 Why are % changes considered in the fit to temperature but kg/m^2 in the trends? Although ranges are shown, it would be useful to the community to have some expert judgement, such as removing obviously spurious datasets (what is the expected physical range?)
3) L23 seems to be missing text. Also Forster et al. (2021) (IPCC Chapter 7) deals more with radiative effects of water vapour. Some mention of recent updates in the field of water vapour and climate would strengthen the context and motivation of the study e.g. Colman & Soden (2021) RevModPhys doi:10.1103/RevModPhys.93.045002; Allan et al. (2022) JGR doi:10.1029/2022JD036728; Ding et al. (2022) LNEE doi:10.1007/978-981-19-2588-7_27; Douville et al. (2022) Comm. Earth Env. doi:10.1038/s43247-022-00561-z; Wu et al. (2024) GRL doi:10.1029/2023GL107909; Wan et al. (2024) HESS doi:10.5194/hess-2023-301 which build on previous assessments e.g. Trenberth et al. (2005) Clim. Dyn. doi:10.1007/s00382-005-0017-4
4) L25 water vapour feedback magnitude should be updated to the latest IPCC report chapter (Forster et al. 2021). It should also note that mid and upper tropospheric water vapour is more important to the feedback strength than lower tropospheric changes that column integrated water vapour is more closely related to. There is however an important link between column integrated water vapour and precipitation as well as downward longwave radiation and atmospheric absorption of sunlight, both of which also impact the energy-water cycle coupling (e.g. Douville et al. 2021 IPCC; Fowler et al. 2020 Nature Rev Earth Sci. doi:10.1038/s43017-020-00128-6).
5) L26 water vapour feedback magnitude is not "compared" to greenhouse gas forcing
6) L60 The 'long period' was also presumably chosen to commence at the start of the SSM/I record and for consistency with previous analyses e.g. Allan et al. (2020) NYAS doi:10.1111/nyas.14337
7) L75 - are the AIRS + AMSU v6 Obs4MIP data set (Tian & Hearty, 2020 Earth & Space Science, doi:10.1029/2020EA001438) version also evaluated? These were developed to remove systematic biases and allow better comparison with CMIP simulations.
8) L102 do all products vertically integrate to the top of atmosphere or are some cut off at a certain level?
9) L105 I didn't understand "lower, respectively higher spatial resolution"
10) L122 it would be useful to mention limitations of the datasets. For example ERA5 and other reanalyses are subject to a changing observing system that can introduce spurious changes, though water vapour now seems quite robust in ERA5 after the mid-1990s (e.g. Allan et al. 2022). The 20CR only assimilates SST and surface pressure so water vapour is determined by the model based on these contraints and so is for all intents and purposes an atmospheric model "amip" type simulation nudged towards realistic atmospheric circulation. For satellite datasets, degradation in sensors, orbital drifts and intercallibration present a challenge
11) L204 - is a consistent land/sea/ice mask (e.g. Fig. 2) applied to all datasets (if not, this could introduce differences in variability). Is the mask a climatology as in Figure 2 (though ice varies seasonally and interannually) or does it vary from month to month?
12) Figure 3 - is this a median across datasets? A fuller caption may help
13) Figure 4 - which of these correlations are significant or not?
14) L293 - seasonal range usually means range over the year but I think intra-seasonal range is meant?
15) L305 - large (e.g. 2-sigma deviations) could usefully be reported to suggest outliers
16) L312 ERA5 does not seem significantly wetter (e.g. probably depends on years chosen)
17) L317 IR estimates presumably sample clear-sky regions which are systematically drier than cloudy regions e.g. John et al. (2011) JGR doi:10.1029/2010JD015355 (presumably visible records are also susceptible). I think this is discussed later but could be flagged earlier.
18) L339 missing reference
19) L343 Are these annual trends? Were trends in %/decade also computed? This could remove mean bias effects (e.g. wetter datasets may vary more in absolute terms but not percent) and it would be useful to quote % changes for consistency with other analysis (e.g. sensitivity to temperature) and the literature
20) L347 is this the interannual regression or does it include the seasonal cycle (which is determined by very different processes)? Or is it the trend in TCWV divided by the trend in temperature? For example in Allan et al. (2022) the ERA5 global TCWV sensitivity to T2m is 5.76 +- 0.35%/K for 1988-2014 while the trend is 0.78+-0.08 %/decade which combined with a warming of 0.17 K/decade gives a lower sensitivity of 4.6 %/K. It was also noted that ERA5 decreases in TCWV over the ocean before the mid-1990s are at odds with the SSM/I record. Ocean and land estimates are also available in the paper.
21) Figure 9 - if microwave values are masked does this mean there are variable numbers of datasets in each grid point? MIssing reference in caption.
22) Figure 10/11 could be combined (and enlarged). It may also be useful to have a zoom in on the more homogeneous datasets since the outliers dominate somewhat
23) L395 do any of the ERA5 breakpoints coincide with the early 1990s low latitude ocean trends identified in Allan et al. (2020, 2022) and Hersbach et al. (2020) that were also in previous versions of this dataset and linked with changes in the observing system? These seem linked with decreases in surface relative humidity and 850 hPa specific humidity over the ocean in the late 1980s-early 1990s.
24) L402 - can changes in the mix of SSM/I satellites introduce spurious variability since they are observing at different overpass times. Some studies use particular SSM/I satellites with more stable or consistent overpass times to avoid this (e.g. Allan et al. 2022).
25) Fig. 12 - the stippling seems to show discontinuity south of Alaska and south of India?
26) L435 but only some of the breakpoints are matched to phyiscal causes?
27) L460 although clear-sky sampling introduces dry biases, it only affects moisture variability if the clear-sky regions vary in a different way to the cloudy regions (e.g. Allan et al. 2003 QJRMetS doi:10.1256/qj.02.217)
28) Conclusion - the impact of this considerable work could be enhanced with some recommendations to the community with regard to better and worse datasets for particular applications (e.g. climatological, regional variability, interannual variability and long-term trends).
Richard Allan
Citation: https://doi.org/10.5194/egusphere-2023-2808-CC1 -
AC3: 'Reply on CC1', Tim Trent, 26 Apr 2024
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2023/egusphere-2023-2808/egusphere-2023-2808-AC3-supplement.pdf
-
AC3: 'Reply on CC1', Tim Trent, 26 Apr 2024
-
RC1: 'Comment on egusphere-2023-2808', Eric Fetzer, 20 Feb 2024
This study does an impressive job of consolidating the information in a wide variety of data sets, and manages to nicely balance depth of analysis with breadth of topic. The results are relevant and interesting, and reveal several important insights. For these reasons, the study is appropriate for publication. I recommend some minor changes to the text.
The first recommended change is to revamp the discussion of diurnal cycle in around line 465 and elsewhere. The text mentions potential biases from a lack of ‘full day’ observations without being specific about what this means. Sun-synchronous satellite measurements of the true diurnal cycle (exactly 24 hour period) taken every 12 hours will average to the diurnal mean. Consequently, the long-term mean will be unbiased. In contrast, sampling once daily will introduce a bias since the average is over only a single phase of the diurnal cycle. This is the case for the daytime-only satellite observations. The distinction between these two situations isn’t clear from the text, but some minor editing should fix this. (A semidiurnal cycle in TCWV or the effects of diurnally varying clouds on sampling further complicate the picture. This is worth mentioning, but beyond the scope of the study.) Because of the potential for the diurnal cycle to introduce a bias, the study should mention the times of day of all satellite data sets and whether the observations are obtained only during day or night. Some information about orbit local times and day-only sampling is provided in the current text, but not for every data set. Consistent statements for all satellite data sets would be helpful. A table may even be appropriate, though that is at the authors’ discretion.
A discussion of HOAPS data set is needed, especially since it is the reference data set over ocean. (The first reference on p. 9 does not even spell out the acronym.) Even a short paragraph would be very helpful.
The text would also benefit from a brief discussion of the use of ocean surface temperature and near-surface air temperature over land in calculating the temperature-TCWV regression. The use of different reference temperatures has implications for interpreting the temperature-humidity relationship over land versus over ocean, and intercomparison of land and ocean relationships.
Here are more specific comments:
Line 38 forward. How are these goals different from the first phase of the GVAP assessment?
Line 70. It’s not obvious that Aqua is included in this study. Words like “now includes both Terra and the Aqua” would be helpful.
Section 2.1.1. The local time of the Aqua spacecraft is needed here. Also, the reference to Manning et al. on line 83 should be Susskind et al. Here is the first link on a web search of the title: https://docserver.gesdisc.eosdis.nasa.gov/public/project/AIRS/L2_ATBD.pdf
Sections 2.1.2. Please mention the local time of the satellite observations, and/or the drifting orbits, as appropriate.
Section 2.1.3. Are the GOME observations obtained day and night, or 10:00 daytime only and not 22:00?
Section 2.1.4. Some mention is needed of which satellite data sets considered in the study are also assimilated into ERA5. Even a statement like “nearly all” would be helpful.
Section 2.1.5. The local times of Terra and Aqua overpasses are needed, and it should be stated that the retrievals are available both day and night (if so).
Section 2.1.6. Are MPIC daytime-only observations?
Line 175 forward. Is the lack of SSM/I etc., data different from ERA5? That seems to be the major distinction, but it’s not stated.
Line 195. Is this bilinear interpolation in latitude and longitude. This is worth mentioning because bilinear interpolation is actually a quadratic function, so has a specific (if confusing) definition.
Line 224. This paragraph on comparison temperatures is worthy of its own short subsection.
Line 237. As discussed above, this is the first mention of HOAPS.
Line 284. The region described as California is almost entirely along the coast of Mexico. (The Baja California Peninsula is in the northern part of the region of interest, but is within Mexico and mostly outside of the California climate region.)
Line 305. Is this the first use of Delta-TCWV? It does not appear to be defined earlier (or else I can’t find it).
Line 317. We noted the wet biases in stratiform regions in Fetzer et al. (2006) and attributed them to a combination of subsidence-induced extensive cloudiness and low TCWV, while clearer conditions have higher TCWV. Interestingly, that study considered only a few weeks of observations.
Line 339. Fix Figure ??.
Line 347. The ‘also’ can be deleted since it’s redundant with ‘In addition’. A similar argument can be made for ‘also’ and its relationship to ‘include’ in the next sentence.
Line 352. The sentence starting ‘The UWHIRS…’ should start a new paragraph since the HIRS data sets obviously deviate from the other data sets. This is worth be highlighting.
Line 358. Change “range of regression coefficients are” to “…coefficients is” since ‘range’ is the subject.
Line 361. Is the ‘theoretical range’ the one set by the Clausius-Clapeyron relationship? It’s not obvious from the text.
Line 365. Should “noise” be “large, and likely anomalous, high variability” or something like that? A noise mechanism hasn’t been clearly established.
Line 388. Change to “Also noteworthy are…”
Line 405. Change the sentence ending slightly to “…trend estimates discussed earlier.”
Line 411. That should be “at least 50% and 100%” since ‘or’ implies either one or the other, but not both.
Line 415. Are the “expected regions” those mentioned in Table B1? The text is ambiguous here.
Line 416. Changing “observed” to “observe” will made this statement present tense, like others in the section.
Line 425. Delete “partly” or else explain the differences more fully.
Line 430. State the reference data sets used to generated the metrics since they are fundamental to the results.
Line 440. Are ‘expected ranges’ determined by the Clausius-Clapeyron relationship? In any case, the reason should be given.
Line 444. Shi (2018) should be Shi et al.
Line 445. Not clear what is meant by ‘that time.’
Line 465. See earlier comments about diurnal sampling with sun-synchronous satellites.
Thank you for the all the careful work!
Eric Fetzer
Citation: https://doi.org/10.5194/egusphere-2023-2808-RC1 -
AC1: 'Reply on RC1', Tim Trent, 26 Apr 2024
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2023/egusphere-2023-2808/egusphere-2023-2808-AC1-supplement.pdf
-
AC1: 'Reply on RC1', Tim Trent, 26 Apr 2024
-
RC2: 'Comment on egusphere-2023-2808', Anonymous Referee #2, 26 Feb 2024
I thoroughly enjoyed reviewing this manuscript. The authors work with an interesting dataset of global total column water vapour measurements from satellite observations and reanalysis within the GEWEX water vapour assessment. The evaluation was conducted through multiple approaches, some discrepancies and biases are observed, particularly in regions with complex topography or under certain meteorological conditions.
I believe that the manuscript addresses a relevant topic and includes a timely discussion. This is a well-written manuscript that only needs to undergo a few minor changes in addition to the other reviewers' comments:
- L193: please justify the reason for conducting this evaluation based on the monthly mean at 2° x 2°, since the coarse resolution may overlook some discrepancies among the datasets.
- L296: please briefly explain the reasons that there is a “significant disagreement” between the datasets for dry atmosphere.
- L339: missing cross-reference for the figure.
Citation: https://doi.org/10.5194/egusphere-2023-2808-RC2 -
AC2: 'Reply on RC2', Tim Trent, 26 Apr 2024
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2023/egusphere-2023-2808/egusphere-2023-2808-AC2-supplement.pdf
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 316 | 98 | 21 | 435 | 27 | 11 | 13 |
- HTML: 316
- PDF: 98
- XML: 21
- Total: 435
- Supplement: 27
- BibTeX: 11
- EndNote: 13
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1