Assessment of Sampling Effects on Various Satellite-Derived Integrated Water Vapor Datasets Using GPS Measurements in Germany as Reference

Round 1

Reviewer 1 Report

Review: Assessment of sampling effects on various satellite-derived integrated water vapor datasets using GPS measurements in Germany as reference

General comments:

This paper presents an analysis of the quality of satellite-measured IWVs compared with IWVs from a dense network of GPS receivers in Germany.

Usually, the main limitation of such study is that satellite data have a specific spatial and time sampling that can induce systematic differences in comparisons. Therefore the authors evaluate the impact of these samplings by applying sampling of satellite data to GPS data. This allows to put the results of comparisons into perspective. Even if reduced to the German territory, this study establishes a relevant methodology that could be applied globally.

The results presented in this paper are very interesting and allow to clearly identify the impact of satellite data sampling during comparisons with more "regular" data. In particular, they should make it possible to put into perspective a large number of qualification results of satellite IWV from GPS data.

The paper is very well written, very well documented with a solid bibliography. Some figures would however deserve to be described a little more.

I recommend the editor to accept the papers with **minor revisions** according to the following specific comments and technical corrections.

Specific comments & technical corrections:

* P3 / L99-101: Maybe you could already (and shortly) explain why SEVIRI measurements will be used.

* P4 / L109-111: Could you briefly described the data that are used for troposphere delay to IWV conversion: Pressure and temperature from NMP models, ground sensors? Which model for mean temperature is used for conversion. Maybe a simple reference?

* P4 / L115: That are NRT GPS IWV? (it should be mentioned)

* P5 / L179-180: I can't figure out what "10% water fraction" means.

* P6 / Figure 2: I think you could better explain what this figure represents (it was not obvious for me!).

* P8 / Table 2: statistics are computed overall if I understand. Uncertainty for bias is computed using +/-stdev / sqrt(N)? I found slightly different values.

* P9 / Table 4: the representation of bias and standard deviations at the same time is very well found, however, can't you find another color scale for standard deviations? It's not always very readable and hard enough to compare (but I do not succeed in finding anything better!).

* P11 / L295-495: before presenting the following results, can you briefly describe the 3 types of climatology you calculated?

* P12 / L385: I'm sorry, but I don't see the connection with figure 3. Can you detail it?

* P13 / L346: delete Steink et al. (“[34]” is enough)

* P16 / Table 3 & 4: you represent differences between IWV using all values and within each periods. You also represent the relative frequency of IWV for each period wrt the complete period. I do not understand why the calculation of the weighting sum of mean difference for each period do not match with the overall mean difference.

The bibliography is relevant and properly formatted.

Author Response

General response from our side:

We would like to thank the reviewer for reading our manuscript carefully, for the positive feedback and valuable comments.

Point-by-point response to the reviewer's comments:

* P3 / L99-101: Maybe you could already (and shortly) explain why SEVIRI measurements will be used.

The line was changed from:

"The GPS measurements were collocated in space and time with information on cloud coverage obtained from measurements from the Spinning Enhanced Visible and Infrared Imager (SEVIRI) onboard the geostationary satellites MeteoSat Second Generation (MSG) [20]."

to:

"The GPS measurements were collocated in space and time with information on cloud coverage obtained from measurements from the Spinning Enhanced Visible and Infrared Imager (SEVIRI) onboard the geostationary satellites MeteoSat Second Generation (MSG) [20] to also study the impact of cloud conditions on IWV distributions."

Lines 99 - 102

* P4 / L109-111: Could you briefly described the data that are used for troposphere delay to IWV conversion: Pressure and temperature from NMP models, ground sensors? Which model for mean temperature is used for conversion. Maybe a simple reference?

The text was slightly expanded and changed from:

“The GFZ processes data of approximately 300 GPS stations within the German nation-wide network operationally and in near-real time.”

to

“The GFZ processes data of approximately 300 GPS stations within the German nation-wide network operationally and in near-real time (NRT) using the EPOS software (Gendt et al., 2004 =21) . Herein the necessary information on surface pressure and surface temperature is taken either from direct measurements at the station or interpolated from the smallest surrounding triangle of synoptic stations (see Steinke, 2017 for details). For more details on the method of IWV derivations from GNSS observations in general as well as the estimation of the IWV uncertainty, we refer to Ning  and Zhang, respectively.”

Lines 114-119

New reference added:

Steinke, Sandra, Variability of integraded water vapor: An assessment on various scales with observations and model simulations over Germany. Dissertation, University of Cologne, 2017, http://kups.ub.uni-koeln.de/id/eprint/7629

* P4 / L115: That are NRT GPS IWV? (it should be mentioned)

defined above

* P5 / L179-180: I can't figure out what "10% water fraction" means.

"Also, GPS stations with more than 10 \% water fraction within an area of 20 km radius were not used."

was changed to:

"Also, GPS stations for which the area of 20 km radius around the station consists of more than 10% water surfaces, e.g. lakes,  were not used."

Lines 183 - 185

* P6 / Figure 2: I think you could better explain what this figure represents (it was not obvious for me!).

To clarify, a sentence has been added:

"This means the overpass times of the corresponding satellites over the region of interest for which IWV retrievals can be performed and matched to GPS measurements."

Lines 211 - 213

* P8 / Table 2: statistics are computed overall if I understand. Uncertainty for bias is computed using +/-stdev / sqrt(N)? I found slightly different values.

The uncertainty in bias was computed as follows:

Bias_unc = Err_y  * Sqrt(1/N)

Err_y = Sqrt( 1/(N-2) * Sum(dy*dy))

dy = (slope * x + intercept) - y

A very small difference might arise from taking the floor instead of round, but since the numbers  are very low, we don't consider this an issue.

* P9 / Table 4: the representation of bias and standard deviations at the same time is very well found, however, can't you find another color scale for standard deviations? It's not always very readable and hard enough to compare (but I do not succeed in finding anything better!).

In our opinion, this is the best representation also regarding the colors. For clarity we have extended the figure caption of Figure 4.

* P11 / L295-495: before presenting the following results, can you briefly describe the 3 types of climatology you calculated?

In our opinion this would mean giving the explanations two times and unnecessarily writing more text, especially since the explanation of and method behind each Figure is given right after the introduction of the Figure.

* P12 / L385: I'm sorry, but I don't see the connection with figure 3. Can you detail it?

To clarify, the sentence was extended as follows:

"Now both MODIS retrievals also show a much clearer bi-modal distribution with a minimum around 9-10 kg/m$^2$, which corresponds well with the result presented in Figure 7 for the IWV distribution under daytime and clear-sky conditions."

Lines 385 - 387

* P13 / L346: delete Steink et al. (“[34]” is enough)

Has been corrected

* P16 / Table 3 & 4: you represent differences between IWV using all values and within each periods. You also represent the relative frequency of IWV for each period wrt the complete  period. I do not understand why the calculation of the weighting sum of mean difference for each period do not match with the overall mean difference.

This is related to the fact that for the reference dataset, which consists of all 15 min. GPS time steps, the contribution of each season to the one year time period is 25% (not limited by clouds), which is not necessarily true for the subsets (sampled datasets or satellite datasets).

Let's look at the seasonal averages of certain clear-sky.  In case of certain clear-sky, a much lower number of datapoints are found in the winter months, due to large and persistent cloud coverage in Germany.  We get for all seasons average negative differences, but overall the difference is slightly positive. So per season the average of all IWV values of the sampled dataset is lower than that of the average of all GPS timesteps within the season, but not when comparing for the entire year. This is because in the reference dataset (15 min GPS timesteps) a larger number of low IWV values from the winter months are used in the calculations (25% for DFJ), reducing the annual mean value, while in the subset these winter month values are only about 13% due to high cloud coverage.

For clarity, We have added the following sentence to the captions of Table 3 and 4:

"Note that the relative frequency of IWV values from all GPS time steps and all stations with respect to the complete time period of one year is about 25\% for each season."

Reviewer 2 Report

Review for remotesensing-754739: “Assessment of sampling effects on various satellite-derived integrated water vapor datasets using GPS measurements in Germany as reference” by Henken et al.

This paper investigates the quality and sampling issues of integrated water vapor products from passive instruments on satellites by comparing with one-year-long integrated water vapor (IWV) data from ground GPS network in Germany. The paper finds that the sampling length does not affect the overall mean IWV values but the IWV frequency distributions. This paper also studies the cloud coverage impacts on the IWV distribution from both satellite and ground measurements.  I found the manuscript interesting considering the importance of water vapor in climate studies and the popularity of using satellite datasets in the research community. Therefore, evaluation of the robustness of satellite products in water vapor would be helpful. The manuscript is well-structured and clear to read. I recommend minor modifications before publication.

Minor Comments:

1. Ln 3: Please write out GNSS.
2. Ln 12: ‘is observed when limiting cases…’-----‘is’ to ‘are’
3. Ln 25: What does ‘a.o.’ mean?
4. Ln 59: ‘Roman et al. [16] propose that…’----‘proposed’
5. The IWV measurement accuracy for the GPS station is 1-2kg/m² (Ln 115). It seems the biases of the satellite products are within this measurement accuracy range (Table 2). I guess my question is does this mean the satellite data are good enough for the water vapor related studies? Maybe one sentence or two regarding this will be helpful.
6. Figure 2: Is the June duplicated on the x-axis?
7. Figure 4: The figure caption is not clear about how std is shown in the figure. It takes some time to figure out the color of the circle edge is associated with std.
8. Ln 125: Can you be clear what this atmospheric humidity quality indicator is?
9. Ln 346: Steinke et al. [34]---Duplicate reference.
10. Ln 391: ‘More specific, the low temporal…’---‘More specifically,….’

Author Response

General response from our side:

We would like to thank the reviewer for reading our manuscript carefully, for the positive feedback and valuable comments.

Point-by-point response to the reviewer's comments:

1.                 Ln 3: Please write out GNSS. 

Has been corrected

2.                 Ln 12: ‘is observed when limiting cases…’-----‘is’ to ‘are’  

Has been corrected

3.                 Ln 25: What does ‘a.o.’ mean?  

Was changed to i.a.

4.                 Ln 59: ‘Roman et al. [16] propose that…’----‘proposed’

Has been corrected

5.                 The IWV measurement accuracy for the GPS station is 1-2kg/m2 (Ln 115). It seems the biases of the satellite products are within this measurement accuracy range (Table 2). I guess my question is does this mean the satellite data are good enough for the water vapor related studies? Maybe one sentence or two regarding this will be helpful.

The biases for all instruments are indeed within the 1-2kg/m2 range. These biases were computed for all available GPS measurements from the German network for a one year time period, where match-ups with satellite measurements were found. In that sense, the satellite IWV datasets show good accuracy.  Results later on show that temporal sampling and retrieval limitations, such as clear-sky conditions, can have a larger impact on the statistics and especially on the IWV distribution. In water vapor related studies where satellite IWV datasets are used, these kind of issues should be taken into account for correct interpretations.

Section 2.1 on GPS data has been extended.

6.                 Figure 2: Is the June duplicated on the x-axis?

No it is not,  but for clarity the labels along the x-axis have been adjusted.

Line 174 "The data sets were collected for a time period of one year, from May 2012 to June 2013…" 

was corrected:

"The data sets were collected for a time period of one year, from June 2012 to May 2013.."

7.                 Figure 4: The figure caption is not clear about how std is shown in the figure. It takes some time to figure out the color of the circle edge is associated with std.

The caption of Figure 4 has been changed to:

Bias (inner circle colors ranging from blue to red) and standard deviation (std; outer circle colors in greyish colors) at each GPS station for all matched IWV observations within one year. Only the GPS stations are included with at least 20 valid satellite-GPS data pairs.

8.                 Ln 125: Can you be clear what this atmospheric humidity quality indicator is?

The sentence was extended:

"In this study, the IASI IWV data is used only if the atmospheric humidity quality indicator (average uncertainties along the profile in dew point temperature) is below 2 (personal communication Thomas August, EUMETSAT)"

Lines 130 - 132

9.                 Ln 346: Steinke et al. [34]---Duplicate reference.

Has been corrected

10.             Ln 391: ‘More specific, the low temporal…’---‘More specifically,….’

 Has been corrected

Reviewer 3 Report

The manuscript “Assessment of sampling effects on various satellite-derived integrated water vapor datasets using GPS measurements in Germany as reference” by Henken et al. presents  a verification of IWV products derived from satellite-born instruments against quality checked GPS-based measurements, used as reference, for time period of one year (May 2012 – June 2013) over Germany.

In order to assess quality and sampling related problems of satellite-based IWV 3 analysis have been carried out: direct comparison of matched satellite – GNSS data; comparison of IWV climatologies; investigation of the impact of temporal, spatial and retrieval-specific sampling on IWV.

Generally speaking, this is an interesting and relevant study that adheres to best scientific practice so I believe it is worth publishing.

I have only a minor comment:

Since GNSS IWV products are used as reference, more details on their reliability and uncertainty sources should be presented (e.g.  Ning et al. 2016 and Zhang et al. 2016 may be cited).

Ning, T., Wang, J., Elgered, G., Dick, G., Wickert, J., Bradke, M., & Sommer, M. (2016). The uncertainty of the atmospheric integrated water vapour estimated from GNSS observations. Atmospheric Measurement Techniques, 9, 79-92.

Zhang, D., Guo, J., Chen, M., Shi, J., & Zhou, L. (2016). Quantitative assessment of meteorological and tropospheric Zenith Hydrostatic Delay models. Advances in Space Research, 58(6), 1033-1043.

Author Response

General response from our side:

We would like to thank the reviewer for reading and reviewing our manuscript and for the positive feedback.

The two suggested references have been added in Section 2.1 on GPS data.  Line 119