Next Article in Journal
Multi-Factors Synthetically Contribute to Ulva prolifera Outbreaks in the South Yellow Sea of China
Previous Article in Journal
Terrestrial vs. UAV-Based Remote Measurements in Log Volume Estimation
Previous Article in Special Issue
Fast Observation Operator for Global Navigation Satellite System Tropospheric Gradients
 
 
Article
Peer-Review Record

The Novel Copernicus Global Dataset of Atmospheric Total Water Vapour Content with Related Uncertainties from GNSS Observations

Remote Sens. 2023, 15(21), 5150; https://doi.org/10.3390/rs15215150
by Kalev Rannat 1,*, Hannes Keernik 1,2 and Fabio Madonna 3,4
Reviewer 1: Anonymous
Reviewer 2: Anonymous
Reviewer 3: Anonymous
Remote Sens. 2023, 15(21), 5150; https://doi.org/10.3390/rs15215150
Submission received: 6 September 2023 / Revised: 12 October 2023 / Accepted: 25 October 2023 / Published: 27 October 2023
(This article belongs to the Special Issue GNSS in Meteorology and Climatology)

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The manuscript “The novel Copernicus global dataset of atmospheric total water vapour content with related uncertainties from GNSS observations” presents the algorithm able to providing high-quality GNSS IWV on a web data portal. These parameters are important for climate research and the data processed in a homogeneous way are crucial to evaluate the quality of the results and to provide reliable values, especially for trend estimates, taking into account the small values we are dealing with.

The dataset, series comparison and results are presented quite clearly and the article can be published with minor edits.

However, I have a comment regarding trend estimation: an algorithm that takes into account the temporal correlated noise should be used. Some studies, such as Alshawaf et al (2017) (your reference # 13) and Klos et al (2018)* have shown that the estimates of the trend of GNSS tropospheric parameters have to take into account the noise models and autoregressive models better represent the residuals of the fit, when the linear and periodic signals are estimated. Was any noise model tested in this study?

*(Klos, A.; Hunegnaw, A.; Teferle, F.N.; Abraha, K.E.; Ahmed, F.; Bogusz, J. Statistical significance of trends in Zenith Wet Delay from re-processed GPS solutions. GPS Solut. 2018, 22, 51)

 

Some specific points are listed below

In the document different GNSS subnetworks are used for the various comparisons, so I suggest, to place alongside Figure 1, preparing a Table with all the stations involved in the analysis, maintaining the distinction between IWV difference analysis and IWV difference&trend analysis.

In Figures 3, 5, 6 and 7 there are some colors that are not so clear. For example, green and cyan are too similar, but the others can also cause confusion. I suggest using more different and bold colours.

Regarding the equations presented in Appendix A:

In eq (3) and (4): is H the ellipsoidal height or the orthometric height? Either it is the height above the ellipsoid or above mean sea level, not both.

The Π presented in eq. (7) is the inverse of the parameter presented in Bevis et al (1994) (your reference # 8). Usually your conversion parameter is called Q = 1/Π. It's just a matter of nomenclature. In general we find the formula IWV= Π ZWD and consequently in eq. (10) Π should be in the numerator. But again, it's just a formal question. The definition in eq (7) in the article is used correctly.

References #35 and #36 are the same paper, please correct.

For convenience, it may be useful to have the doi of the articles in the references.

Author Response

Dear Reviewer,

thank you for your valuable questions and comments. We have made changes to the text (marked with "blue" and "red" in *.pdf), hoping that all your questions were noted and understood correctly.

In case of any questions, we're ready to continue to find the answers and solutions. 

Best regards,

Kalev Rannat

Please find the questions/answers below:

===================================

  1. However, I have a comment regarding trend estimation: an algorithm that takes into account the temporal correlated noise should be used. Some studies, such as Alshawaf et al (2017) (your reference # 13) and Klos et al (2018)* have shown that the estimates of the trend of GNSS tropospheric parameters have to take into account the noise models and autoregressive models better represent the residuals of the fit, when the linear and periodic signals are estimated. Was any noise model tested in this study?

ANSWER 03102023

We don't actually want to estimate the trend (or validate any of the time series), we just want to assess how well the datasets agree.  In brief, Autocorrelation depends on the climate system observed using different techniques, so if we neglect it all the estimated trends will be inflated accordingly. Nevertheless, we can still use trends as a diagnostic tool for assessing the agreement among GNSS datasets (NRT and reprocessed).

Rephrased: Our primary objective is not to directly estimate the trend values themselves but to use the trends as a diagnostic tool to compare the GNSS datasets (NRT and reprocessed), using the same methodology. Autocorrelation may generally inflate trends, but the discussion of water vapour trends in this paper is beyond our scope. However, the difference between the trend for IGS NRT (daily) and trends for the other dataset will remain and this confirms the recommendation to 'avoid using non-reprocessed time series for climate research. In the discussion section, we introduced the idea of incorporating noise models (reference to Alshawaf et al. and Klos et al.) as part for future exploration of IWV climatology.

  1. In the document different GNSS subnetworks are used for the various comparisons, so I suggest, to place alongside Figure 1, preparing a Table with all the stations involved in the analysis, maintaining the distinction between IWV difference analysis and IWV difference&trend analysis.

ANSWER 03102023 We have used data from over 30 IGS stations and over 60 EPN stations. Although some of them overlap, the tables get “too long” inside the text. The distinction is already made by different colours (yellow and blue) on different subsections (IGS and EPN) in Figure 1.

ADDED a table Appendix C 10-10-2023

  1. In Figures 3, 5, 6 and 7 there are some colors that are not so clear. For example, green and cyan are too similar, but the others can also cause confusion. I suggest using more different and bold colours.

CORRECTED  A new color palette is used in those figures.

 

  1. Regarding the equations presented in Appendix A: In eq (3) and (4): is H the ellipsoidal height or the orthometric height? Either it is the height above the ellipsoid or above mean sea level, not both.

Our fault, must be written “above the geoid (above the mean sea level)”

CORRECTED 26.Sep.2023

 

  1. The Π presented in eq. (7) is the inverse of the parameter presented in Bevis et al (1994) (your reference # 8). Usually your conversion parameter is called Q = 1/Π. It's just a matter of nomenclature. In general we find the formula IWV= Π ZWD and consequently in eq. (10) Π should be in the numerator. But again, it's just a formal question. The definition in eq (7) in the article is used correctly.

ANSWER  We wanted to follow the notation found in “Springer Handbook ...” by Teunissen & Montenbruck, Part G, eq.38.7, where IWV = ZWD/Q. Will stay as it is, but avoiding using “conversion” in the “conversion coefficient”.

  1. References #35 and #36 are the same paper, please correct.

ANSWER It was a mistake, one is a journal paper, the other refers to the dataset.

REF. [35]

[1] Yuan, P., Blewitt, G., Kreemer, C., Hammond, W. C., Argus, D., Yin, X., Van Malderen, R., Mayer, M., Jiang, W., Awange, J., and Kutterer, H.: An enhanced integrated water vapour dataset from more than 10 000 global ground-based GPS stations in 2020, Earth System Science Data, 15, 723–743, https://doi.org/10.5194/essd-15-723-2023, 2023.

REF. [36] CORRECTED 26.Sep.2023

[2] Yuan, Peng, Blewitt, Geoffrey, Kreemer, Corné, Hammond, William C., Argus, Donald, Yin, Xungang, Van Malderen, Roeland, Mayer, Michael, Jiang, Weiping, Awange, Joseph, & Kutterer, Hansjörg. (2022). An enhanced integrated water vapour dataset from more than 10,000 global ground-based GPS stations in 2020 [Data set]. Zenodo. https://doi.org/10.5281/zenodo.6973528

 

  1. For convenience, it may be useful to have the doi of the articles in the references.

We have added DOIs for most of the articles. 26.Sep.2023 

Reviewer 2 Report

Comments and Suggestions for Authors

Thank you in advance for considering my comments. In this study, a new algorithm is proposed to invert GNSS IWV for tropospheric products of IGS and EPN, and ERA5, IGS daily, IGS repro3 COD, IGS repro3 TUG and IGRA dataset scheme GNSS IWV are used to evaluate RHARM, EPN repro2 and GURAN IWV, and the verification results show that The average difference of IWV between GNSS datasets reprocessed by the new algorithm is less than 0.1 mm, and the deviation of GNSS IWV values is less than 1 mm compared with RHARM and GRUAN IWV values, and the algorithm provides the first unified global GNSS IWV dataset with Copernican climate change uncertainty. Based on the above comments, I have some comments related to what is included already in your manuscript:

(1) Whether the new algorithm proposed in this paper is applicable to all GNSS site data of IGS and EPN.

(2) In this paper, the test method for evaluating the consistency between the two datasets is used, and whether the calculation details of the standard uncertainties u1 and u2 in the method can be explained.

(3) In this paper, can you explain why the RHARM method selects 16 pressure levels of 1000-10hPa and limits the specific humidity to 250 hPa, and the improvement of radiosonde IWV accuracy by this treatment.

(4) The abbreviation of GNSS IWV in line 104 is written as GNSS IVW, please check the abbreviation in the full text.

(5) Whether line 338 in Article 338 is "shown in Figure 1".

(6) (A) in Figure 5 is not marked x lable.

(7) ZTD's sigma uncertainty is not explained in Appendix A.

(8) Whether step 8 in Figure 8 is Calculate IPW, and if so, the English explanation corresponding to IPW does not appear in this article.

(9) Some references, such as "Multiscale Variations of Precipitable Water over China Based on 1999–2015 Ground-Based GPS Observations and Evaluations of Reanalysis Products. Journal of Climate" and "A novel regional drought monitoring method using GNSS-derived ZTD and precipitation, Remote Sensing of Environment, 2023" should be cited, because those are closely related to the processing of GNSS observation and application of GNSS, and should be introduced in Introduction.

Author Response

Dear Reviewer,

Thank you for your valuable questions and remarks. We've made several changes to the text (marked with "blue" and "red"), hoping that all your questions were noted and understood correctly. 

In case of any additional questions, we're ready to continue finding answers and solutions. 

Best regards,

Kalev Rannat

Please find our answers to your questions below:

==============================================

  1. Whether the new algorithm proposed in this paper is applicable to all GNSS site data of IGS and EPN.

ANSWER: Yes

  1. In this paper, the test method for evaluating the consistency between the two datasets is used, and whether the calculation details of the standard uncertainties u1 and u2 in the method can be explained.

29092023 ADDED after Eq. 2: where u1 for GNSS IWV is calculated according to [ning2016uncertainty] and u2 for the radiosonde according to [GRUAN_TD8].

  1. In this paper, can you explain why the RHARM method selects 16 pressure levels of 1000-10hPa and limits the specific humidity to 250 hPa, and the improvement of radiosonde IWV accuracy by this treatment.

This is not correct and this is now better clarified in the paper. According to Madonna et al. (2020), the RHARM harmonization is employed for every balloon launch, typically conducted at 00:00 and/or 12:00 UTC, using data from 16 mandatory pressure levels (10, 20, 30, 50, 70, 100, 150, 200, 250, 300, 400, 500, 700, 850, 925, 1,000 hPa). These specific levels are consistently available at the stations for each ascent. In contrast, significant level reports differ in the definition for each profile. Adjustments to humidity are confined to the 250 hPa level due to widespread sensor performance issues at higher altitudes in most commercial sondes (now mentioned in the manuscript). After the homogenization of time series at mandatory levels, the observational uncertainties are estimated for each data point, and finally, adjustments and uncertainties estimated only at mandatory pressure levels are interpolated at significant levels. On average, we had access to and utilized data from 30 levels in the analysis for each radiosounding.

Based on the high-resolution GRUAN profiles used in Section 3.4 less than 0.3% of the water vapour is located above 250 hPa level. Henceforth, we can deduce that excluding the integration of specific humidity above this level does not result in any significant bias in the RHARM IWV. Due to the uncertainty of RHARM IWV values, the estimation is made according to the latest uncertainty estimation characterisation scheme developed by GRUAN which takes into account the vertical resolution of the profile.

CORRECTED 28.Sep.2023

  1. The abbreviation of GNSS IWV in line 104 is written as GNSS IVW, please check the abbreviation in the full text.

Our mistake.

CORRECTED 26.Sep.2023

  1. Whether line 338 in Article 338 is "shown in Figure 1".

ANSWER: The question is fully understood,  and we ask a clarification by the reviewer Figure 1 (A)[7]

  1. (A) in Figure 5 is not marked x lable.

ANSWER: The question is unclear, Figure 5 (A) looks correctly labeled.

  1. ZTD's sigma uncertainty is not explained in Appendix A.  

ANSWER: 29092023 Please check the description at line 521: “... are the one-sigma uncertainties of … “ SigmaZTD numerical values (...) calculated by ACs and taken as is.

  1. Whether step 8 in Figure 8 is Calculate IPW, and if so, the English explanation corresponding to IPW does not appear in this article. 

Our mistake. It is now fixed in the manuscript.

CORRECTED Updated Figure 8,  27.Sep.2023

  1. Some references, such as
  1. a) "Multiscale Variations of Precipitable Water over China Based on 1999–2015 Ground-Based GPS Observations and Evaluations of Reanalysis Products. Journal of Climate" and
  2. b) "A novel regional drought monitoring method using GNSS-derived ZTD and precipitation, Remote Sensing of Environment, 2023" should be cited, because those are closely related to the processing of GNSS observation and application of GNSS, and should be reported in the introduction.

ADDED 29092023

Reviewer 3 Report

Comments and Suggestions for Authors


Comments for author File: Comments.pdf

Comments on the Quality of English Language


Author Response

Dear Reviewer,

Thank you for your valuable questions and remarks. We've made several changes to the text (marked with "blue" and "red"), hoping that all your questions were understood correctly.

In case of any additional questions, we're ready to continue finding answers and solutions.

Best regards,

Kalev Rannat

Please find our answers below:

==============================================

 Abstract

-Full description of GRUAN acronym should be made the first time it is mentioned.

 - Available datasets: it is stated that the current algorithm has been developed to support current and future CDS IWV datasets. After the first sentence, abstract is confusing on the description of which dataset is available, what will be in the future and what is advisable to have. It states that CDS GNSS datasets come primarily from reprocessed data. It says that also data in real time are available, one of these being NRT data from IGS. Checking from the portal, I only see two datasets available, one using as input EPN repro2, and the other from NRT IGS. Abstract should simply state that the algorithm is currently applied for these two datasets.

New abstract added 29092023 

================================

A novel algorithm has been designed and implemented in the frame Climate Data Store (CDS) of the Copernicus Climate Change Service (C3S) with the main goal of providing high-quality GNSS-based integrated water vapour (IWV) datasets for climate research and applications.

To this purpose, the related CDS GNSS datasets were (or better, will be?) primarily obtained from GNSS reprocessing campaigns, given their highest quality in adjusting systematic effects due to changes in instrumentation and data processing.

The algorithm is currently applied to the International GNSS Service (IGS) tropospheric products, which are consistently extended in near real-time and date back to 2000, and to the results of a reprocessing campaign conducted by the EUREF Permanent GNSS Network (EPN repro2), covering the period from 1996 to 2014.

The GNSS IWV retrieval employs ancillary meteorological data sourced from ERA5. Moreover, IWV estimates are provided with associated uncertainty, using an approach similar to that used for the Global Climate Observing System Reference Upper-Air Network (GRUAN) GNSS data product.

To assess the quality of the newly introduced GNSS IWV datasets, a comparison is made against the data from GRUAN and the Radiosounding HARMonization (RHARM) homogenized dataset, as well as with the IGS repro3, which will be the next GNSS-based extension of IWV time series at CDS. The comparison indicates that the average difference in IWV among the reprocessed GNSS datasets is less than 0.1 mm. Compared to RHARM and GRUAN IWV values, a small dry bias of less than 1 mm for the GNSS IWV is detected. Additionally, the study compares GNSS IWV trends with the corresponding values derived from RHARM at selected radiosonde sites with more than ten years of data. The trends are mostly statistically significant and in good agreement.

=========================================

 

 

  1. On a side note, it is stated that NRT IWV from IGS is updated monthly, while no data in 2023 is actually available. Why?

Thanks for this note, the (technical) reason will be clarified and data updated. ECMWF staff contacted, RESOLVED

 

  1. Page 2, line 82: Please define AC the first time it is mentioned.

CORRECTED  29092023

 

  1. Page 2, lines 73-85: Again here, I found description of the various datasets and from which sources are obtained quite confusing through the text. It is not clear what belongs already to the CDS and accessible through it and what from elsewhere ( i.e. IGS and EPS reprocessed, re-reprocessed, original time series). Please clarify in text that algorithm provides currently two CDS datasets, one from NRT IGS and the other from EPN-repro2.

REPHRASED  29092023

“The usage of this algorithm enables providing in the Copernicus Climate Data Store any other new IWV time series (either global or regional) based on reprocessed GNSS tropospheric products made available by any competent Analysis Centres (ACs) in the future. Currently, the Copernicus Climate Data Store provides the EPN repro2 dataset derived from a reprocessing campaign conducted by the EUREF Permanent GNSS Network (EPN) [42] and IWV from the IGS NRT dataset (not re-processed).”

 

  1. Page 3, line 90-91: the sentence “ convince the reader about the suitability” is misleading and should be rephrased. If the dataset is suitable for climate research authors can simply state it.

REPHRASED

  1. Page 3, lines 89-94. Intercomparisons: Sentence is confusing. Please rephrase stating clearly what is compared to what.

This is handled together with the previous comment.

REPHRASING  10102023

The two GNSS IWV datasets that are available in the CDS (based on IGS NRT and EPN-repro2), the used algorithm, and its implementation in CDS infrastructure are all covered in the parts that follow. IWV data from the GNSS datasets accessible through the CDS (IGS NRT and EPN-repro2), IGS repro3 (not yet accessible at CDS), ERA5, as well as data obtained from radiosoundings that either have reference quality or have undergone bias adjustment, are all included in the intercomparison presented in this paper. This extensive intercomparison is carried out to investigate the suitability of the datasets for climate research.

  1. Page 3, section 2 data source, first sentence is very confusing, since every product, CDS or not, is called the same: i.e. GNSS IWV. Also it is not clear which comes from what. I will rephrase similarly to as follows: “ With the aim to harmonize and provide a unified global CDR GNSS IWV dataset, a novel algorithm has been created to estimate IWV from the IGS and the EPN tropospheric Zenith Total Delays (ZTD) products. The available CDS GNSS IWV consist of two dataset. The first dataset is a daily NRT IWV from the IGS network, whose initial data were retrieved from the Crustal Dynamics Data Information System (CDDIS) data portal [47]. The second dataset is a further reprocessed time series derived from the EPN repro2 dataset.”

Agreed, the sentence needs rephrasing.

REPHRASED 03102023

We want to provide standardized GNSS IWV datasets for climate research based on GNSS tropospheric outputs from various geodetic networks (global or regional). As a preliminary stage, the Crustal Dynamics Data Information System (CDDIS) data gateway is used to release a daily NRT IWV from the IGS network as well as the EPN-repro2. As geodetic reference networks, IGS and EPN are renowned for their cutting-edge technological requirements on the GNSS sites and open quality evaluation. In order to facilitate climate research, the main goal is to keep adding IWV datasets that are based on reprocessed GNSS data. The CDS platform is now implementing the IGS repro3 IWV dataset. Any additional GNSS can be added to the CDS in the future using the same algorithm.

  1. Page 3, Section 2, Data source, lines 107-108. Source of data are IGS and EPN networks. It seems that several site in Europe belong to the same network, IGS, EUREF and E-GVAP. This should be discussed in the text, especially when the same data is processed differently allows to understand uncertainty due to different algorithms.

We’re applying the same data processing algorithm to any input data source published in the CDS.

ADDED:   29092023

We are aware that in order to provide Zenith Total Delay, data processing and analysis institutions/agencies (ACs) for all networks employ their own best practices and make an effort to adhere to the strictest industry standards. We point out that different software tools, strategies, and configurations are used by GNSS analysis centers (e.g. Kacmarik et al. 2017), each of which has the potential to have a different effect on the IWV values. The precise impact of algorithms and software on IWV values cannot be quantified in this research because it is outside the scope of our investigation. Items that have been reprocessed ought to be regarded as "already harmonised" and "useable as is."

The data processed and transmitted by EPN (in EPN-repro2) is processed and integrated by EPN ACs and should be regarded and understood as a distinct product if any of the sites belong to more than one network (for example, official IGS sites in EPN).

 

  1. Page 4, Section 2.2 line 159. Algorithm is defined in the ATBD but main equations are also summarized in Appendix A. Please refer to Appendix in the text.

CHANGED  29092023

  1. Page 4, Section 2.2, lines 178-179: For analysis of Tm-Ts relationship, IWP conversion from zenith wet delay and uncertainty in the Arctic please see Fionda et al., 2019.

ADDED:  29092023

Fionda, E.; Cadeddu, M.; Mattioli, V.; Pacione, R. Intercomparison of Integrated Water Vapor Measurements at High Latitudes from Co-Located and Near-Located Instruments. Remote Sens. 2019, 11, 2130. https://doi.org/10.3390/rs11182130

 

  1. Section 3, page. 5, lines 196-200. RHARM dataset is used as reference. Nevertheless, RHARM only contains 16 levels (only few contain more) and I wonder if any analysis has been performed for instance against high vertical resolution GRUAN sondes to assess that IWV is not impacted by this low resolution. It could be mentioned here. 

ANSWER: This is discussed below. Please also see the answer to reviewer #2

  1. Section 3, pag. 5 line 207. IGS repro3 from TUG and COD. IGS repro3 has been mentioned before but not really described. It should be briefly discussed in this Section. Also TUG and COD are acronyms not explained.

CORRECTED: Acronyms are now explained in the manuscript.

CORRECTED 29.Sep.2023

  1. Section 3, pag. 5 lines 209-213. I understand the reason for it when considering global validation, still the collocation criteria are quite broad when considering water vapour intercomparison. For instance 10 m difference could lead to 1 hPa pressure difference, resulting in 2.3 mm error in ZHD and about 0.35 kg/m2 in IWV. As additional comment on the results in Section 4, I would suggest investigating a bit on the altitude difference among sites. Although 100m may not seem much, it could provide some water vapour bias comparable to the results I see. It would be good to investigate this issue in my view. Biases in figure 3-6 could also be looked in function of site altitude differences between IWV sources.

ANSWER: As stated in the paper (lines 243-246 in the previous version), the water vapor data collected from radiosonde profiles underwent an upward integration starting from the altitude of the GNSS station to simplify comparison with GNSS data. The humidity value was calculated for the lowest level using either extrapolation or linear interpolation. The IWV derived from radiosonde data was adjusted for the altitude gap difference accordingly. A further constraint was added to restricting the use of measurements to locations with altitude variations under 100 meters. This restriction was applied to mitigate the potentially substantial errors associated with interpolating/extrapolating.

CORRECTION NOT NEEDED

  1. Radiosonde data were excluded in the EPN dataset analysis. I assume because of the lack of reference RAOBs, quality control and the dry bias issue. Please explain in text.

Although there are a few co-located radiosonde and GNSS sites in Europe that provide comprehensive and long-term datasets suitable for IWV climatology research, the radiosonde data were not included in this comparison in order to broaden the scope of the analysis and investigate data from a much larger number of stations at all full hours, spanning from 00 to 23 UTC, with the aim to thoroughly assess the IWV differences and trends for GNSS-based datasets and ERA5.”

CORRECTED (MODIFIED) 27.Sep.2023

  1. Section 3, page 5 lines 222-228: this section is a bit confusing. What EPN is compared against? Please explain in text as it is not mentioned. Only referring to later section does not suffice.

Modified now. Please see the comment above.

CORRECTED 27.Sep.2023

  1. Section 3, pag. 6, lines253-257: It must be pointed out that also GPS is not a point instrument, and it collects information from satellite in view of the receiving station with a cut-off angle (around 3-5 deg), so to speak, it has a large cone field of view. I do not see this mentioned anywhere in the analysis.

Correct. We have added this now.

CORRECTED 28.Sep.2023

  1. If I understand correctly, the analysis is between Ts estimated from local meteo station and from ERA5, and it is very interesting to see that there is a non-negligible difference. Then, authors compare against RAOB to validate the use of ERA5. Radiosonde are assimilated to provide the ERA5 product, so of course they agree very well. So I do not understand how you can say that Tm is overestimated by using the meteo station, and which method is better unless you have independent validation.

The computation of the weighted mean temperature of the atmosphere, Tm, can be directly carried out using temperature and humidity profiles obtained from radiosonde data, which serves as the primary in situ dataset available for atmospheric temperature and is regarded as the reference method. We can assess the extent of the systematic deviation introduced by estimating Tm using temperature solely from the lowest level of the sounding (as was described in lines 292-294, but in principle, it can also be based on near-surface air temperature measurements from ground-based meteorological stations). Our study emphasises that such simplification leads to a more significant systematic deviation compared to using ERA5 profiles. Similarly, we can estimate IWV by using Tm which is derived from a) the lowest level of ERA5 data, and b) ERA5 temperature and humidity profiles (presented in Fig. 2). These results indicate that the simplification by a) introduces a noticeable diurnal dependency in IWV that has to be avoided. We have updated Section 3.1 and tried to clarify it. In addition, some information on that is added to Section 2.1.2.

CORRECTED 27.Sep.2023

  1. Section 4, page 8, Results, Intercomparisons against RHARM. RHARM dataset is used as reference. Nevertheless, RHARM only contains 16 levels and I wonder if any analysis has been performed for instance against high vertical resolution GRUAN sondes to assess that IWV is not impacted by this low resolution. It could be discussed here.

RHARM provides temperature and relative humidity time series that have been homogenized, as well as an evaluation of the observational uncertainty for each individual observation at each pressure level (there is a higher vertical resolution in the lower troposphere where humidity exhibits higher variability). According to Madonna et al. (2020), there is better agreement between RHARM and GRUAN in terms of temperature and humidity than there is between IGRA and GRUAN.  Based on this paper, there is no significant systematic deviation in RHARM’s temperature and humidity. The uncertainties linked to RHARM levels, along with the respective level thicknesses, have been factored into the uncertainty calculation for RHARM's IWV.  It is accurate to say that there may occasionally be differences in IWV values obtained from individual sounding data by RHARM and GRUAN. These differences can be attributable to RHARM's inherent restriction in capturing very fine-scale vertical humidity variations. When taking into account long time periods, however (assuming a linear absolute humidity gradient between RHARM's pressure levels when averaged over e.g. several years), there is no known physical cause that would lead to a systematic bias in RHARM's IWV. For each radiosounding, we had access to and used information from 30 levels on average. Please view the reply to the remark #3 from Rev2.

ADDITIONAL INFORMATION ADDED TO SECTION 2.2

  1. Section 4, Results, Intercomparisons. IGS daily and ERA5 seem to agree well among each other when compared against RHARM in Figure 3. Conversely, they seem to provide larger biases when compared against EPN repro2 in Figure 5. Can you explain?

The mean nighttime and daytime differences between IGS daily and ERA5 based on Figure 3 are 0.08 and 0 kg/m2. In the case of Figure 5, the respective differences are 0.15 and 0.14 kg/m2. It's essential to note that these two figures utilize data from distinct sets of sites. These sets have unique geographical coverage and time span, which probably accounts for most of the variations in the observed differences between IGS daily and ERA5 in Fig 3 and Fig 5.

CORRECTION NOT NEEDED

  1. Figure 7, please add bias and std of the comparisons in the figure

ADDED 28.Sep.2023

  1. It would be also interesting to add in the Section a discussion on the consistency among the results in the paper and those found in literature in terms of IWV comparison between ERA5, RAOB and IWV from GNSS.

The chosen study period, geographic scope, and particular collection of monitoring sites are some of the variables that affect the results of intercomparisons with GNSS, radiosonde, and reanalysis IWV time series.

Furthermore, the extent to which harmonization is applied, or whether it is absent altogether, significantly affects the systematic discrepancies and trends observed in IWV. To the best of our knowledge, this study represents the first attempt to compare reprocessed GNSS-derived IWV against values obtained from post-processed radiosoundings and the most up-to-date reanalysis, using a standardized approach to estimate IWV uncertainties. As a result, the findings of this research cannot be directly compared to previously published studies. Nonetheless, the magnitude of the differences observed in reprocessed GNSS IWV is comparable to what has been reported when using unhomogenized radiosonde data (Yuan et al., 2023) or earlier-generation reanalysis datasets (Pacione et al. 2017).

 

References: please, cross-check reference 19 for typo.

CORRECTED 26.Sep.2023

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

I have no further questions, thanks for the authors' response.

Reviewer 3 Report

Comments and Suggestions for Authors

I am happy with the answers for my comments and I do not have further requests.

Back to TopTop