the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
A comprehensive 22-year global GNSS climate data record from 5085 stations
Abstract. This work presents a comprehensive global GNSS climate data record derived from 5085 stations, spanning a 22-year time period 2000–2021. Generated through the GPAC-Repro campaign, the dataset utilises state-of-the-art processing methodologies and precise products from the International GNSS Service (IGS) Repro-3 initiative. The dataset includes high-quality hourly estimates of Zenith Total Delay (ZTD) and Precipitable Water Vapour (PWV), offering improved accuracy and spatiotemporal coverage. A rigorous data screening and quality assessment framework was implemented, including formal error detection, offset identification, and extensive cross-validation with ERA5 reanalysis dataset, radiosonde profiles, and Very Long Baseline Interferometry (VLBI) measurements. Collectively, these efforts ensured the consistency, accuracy, and homogeneity of the dataset. In addition, diurnal, monthly, and annual variations in ZTD and PWV have been analysed to evaluate and demonstrate its feasibility for monitoring climate variability, atmospheric circulation, and weather extremes. The insights provided by the dataset address critical data gaps in global climate observing systems and provide a robust foundation for advancing climate research and applications. Representing a significant milestone in GNSS climatology, this dataset serves as a vital resource for the scientific community, supporting improved understanding of atmospheric processes and more effective responses to climate-related challenges.
- Preprint
(11759 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 25 Aug 2025)
-
RC1: 'Comment on essd-2025-283', Anonymous Referee #1, 25 Jul 2025
reply
General Comments
Overall I find that the dataset is of general interest to the community and should be published.
The first four sections are of the highest priority, whereas the section on "Further Analysis" more describes applications for future users of the dataset. Therefore, my attention is focused on the first sections to have a clear and understandable description in order to increase the possibility of many users of (and citations to) the dataset.
I have a small problem with the title because it gave me the impression that all 5085 stations have 22 year long time series. In fact the majority of the stations have time series that are significantly shorter. What about something like: " A global GPS climate data record for 5085 stations covering up to 22 years"?
Early in the manuscript you make clear that only GPS data are used. Later it is however very often you refer to GNSS data. I think it is okey only when you refer to GNSS data in general terms, not when you discuss your dataset. Furthermore, in the summary section you may have some ideas to discuss the future use of GNSS (rather than just GPS), pros, and cons?
Specific commentsline 79: You state that a shortcoming of previous studies is that the length of the time series is only around 10 years. This is certainly true and climate scientists often use 30 year averages which means that the time series in your dataset suffer from the same shortcoming. In fact many of the sites in the dataset have time series of about 10 years or less. I think it is fair to state this. The decision to use 30-year averaging periods was taken by IMO (now WMO) at a congress in 1935. Also at line 106 you give the false impression that all time series are 22 years.
line 89: later in the manuscript you also mention changes of hardware, such as antennas and radomes. It can be meaningful to mention these also here.
line 127: "Following a rigorous data screening process, 95 sites were excluded due to identified issues with the atmospheric results, leading to a final dataset comprising 5085 GNSS stations." I will be helpful to list these 95 sites, not in the manuscript but in the data archive, together with the reason for excluding the site. I searched for a specific IGS site, with a long time series, but could not find it among the 5085 sites included. Perhaps it is among the 95 sites? In any case it will be helpful for future users of GNSS PWV to be aware of problem sites.
line 162: Please motivate why you chose such a low cutoff angle. For example, when searching for trends varying systematic errors are very important and should be reduced. Multipath is an effect which get worse closer to the horizon and this is especially true if the horizon mask change over long time.
line 165: "a 27-hour time window was adopted," How was these 27 hours defined? In the paper by Dousa et al. (which you refer to) I do not find this specific value, only that three days were combined and thereafter the atmospheric estimates from the day in the middle were selected.
line 175+: Why did you chose this rather complicated procedure to derive the ZHD. Other studies I have seen use the Saastamoinen model for the gravity with the ground pressure and the site position as input parameters which I have assumed is sufficiently accurate for the ZWD retrieval. Did you examine if there were ay differences that motivate your choice?
Section 3.3: Why do you wait until this stage by removing unrealistic negative values of the PWV. You do not need reference data for this action.
When you remove outliers by comparing with reference data you assume that the reference data are correct. I think that needs to be discussed. From my point of view, one application of GPS/GNSS PWV is to use it as an independent dataset in order to identify problems in other datasets, such as the ERA5.
Table 2: The observing periods of the VLBI stations are much longer. I understand that you do not pick up data before the start of the GPS time series but there should be data after 2018? Please explain.
line 420: How do you define a robust agreement? For which value of the STD is it no longer robust?
Section 4.3. Is not clear to me what was the action taken after finding these changepoints. Did you modify the PWV time series in the data archive, or not? Also in this case, are you sure that some detections of changepoints are not due to problems with ERA5?
A related question is if you searched for changepoints in the ZHD. If such exist they will indirectly cause a corresponding junp/offset in the PWV.
Technical CorrectionsI find that the font size in all figures is unnecessarily small. The size could in general be say 50-100 % larger in order to improve the readability.
line 11: GPAC is not explained
line 34: please explain "data gaps". Are the gaps temporal or spatial or both?
line 53: changes in signals --> changes in the arrival time of the signals
line 57: With used together --> When used together
Figure 2: use more different colours of the different symbols, both i the a and in the b graph. The present version is useless to see the differences, only the distribution of sites on the globe is clear. Also I wonder if "data integrity" is identical to "data completeness" in the summary file downloaded from the archive? If so use the the same expression at both places.
lines 137, 141: Because you use British English spelling for "vapour" it will be consistent to write "colour".
lines 139, 303, 309, 406, 420, 427, 468, 492, 499, 607: there shall be a space between value and unit according to SI standards.
Table 1: If you include the citations in the strategy column you can shorten the running text significantly.
lines 176, 190, 328: Units shall not be in italics
Eq.(4): will be more clear if it is split into two equations. Furthermore, the cosine function shall not be written in italic font.
line 205: improve the contrast between font and background colour.
Figure 4: Also in this figure it is difficult to see the different colours.
line 239: These stations are referred to as MDO1, MDO2, and MDO3 in Figure 5 ( not MGxx). Furthermore, when you show examples in the manuscript I think it will be informative if you added where the stations are located. It will save the work of going into the data archive.
line 292: "with 95 problematic sites" indicates that these are included among the 5085 sites. Please rewrite.
Figure 11: Also this Figure is difficult to get any useful information from Are all these radiosonde stations used? I mean is there a GPS site close enough. Perhaps increase the size of the graph and make the VLBI symbols larger? Or delete the figure?
Figure 14: he quality should be improved so that it is clear where the GPS sites are located. Perhaps these graphs are not needed as well? Everyone interested for sure knows the topography of Hawaii and the Andes.
lines 406-407: Is this true solar time, local time, or UT?
linne 464: You can add "radomes" here.
line 491: Equator --> equator
Figure 18: Improve the contrast between the symbols, e.g., light green for daily values and black for monthly values.
Figures 20 and 21: Colour only areas around the GPS sites, say with a radius of 10-20 km?
line 571: Northern Hemisphere --> northern hemisphere
Citation: https://doi.org/10.5194/essd-2025-283-RC1
Data sets
A comprehensive 22-year global GNSS climate data record from 5085 stations Xiaoming Wang et al. https://www.pangaea.de/tok/3945654965e0ab80bb82b695dda9426b3e7b597c
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
382 | 60 | 26 | 468 | 16 | 22 |
- HTML: 382
- PDF: 60
- XML: 26
- Total: 468
- BibTeX: 16
- EndNote: 22
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1