A Calibrated GPT3 (CGPT3) Model for the Site-Specific Zenith Hydrostatic Delay Estimation in the Chinese Mainland and Its Surrounding Areas
, Feijuan Li 1,2, Lilong Liu 1,2, Liangke Huang 1,2, Lv Zhou 1,2 and Hongchang He 1,2Round 1
Reviewer 1 Report
Review
The manuscript describes a calibrated GPT3 (CGPT3) model that adds the ZHD residuals to the GPT3-derived ZHD products. The temporal variations of residuals between GPT3-based ZHD and radiosonde-based ZHD data are thoroughly analyzed using the radiosonde data in China and its surrounding areas. Then, the ZHD residuals are modeled using a periodic function. The authors compare the CGPT3 model with the GPT3 model and newly GTrop model, and the results show that the proposed model is superior to the other two models. Overall, this is an interesting study for refining the GPT3 model. However, some important clarifications are needed (see below for specific comments), and the English language also needs to be improved. Therefore, I recommend that the manuscript can be accepted after minor revisions.
Specific Remarks
1. Section 2.2 Dataset: Although the radiosonde instruments can provide the most realistic meteorological profile observations, the quality control for radiosonde data is still necessary.
2. Line 123: This sentence lacks a full stop. Please check the full text.
3. An important issue: According to the title of the manuscript, the study should be focused on the China region, but the whole experiment is performed based on the radiosonde data in China and its surrounding areas. Therefore, please revise the title or the experiment. Moreover, in Figure 2, the power spectrums of the ZHD residuals at the three radiosonde stations are shown. However, only JIUQUAN station is located in China. Please select representative stations in China.
4. Figure 3: It's very nice to see the authors analyze and classify the periodic characteristics of the ZHD residuals from different stations. Concerning the modeling of the ZHD residuals, whether the proposed method uses the same periodic function (Equation 7) or different functions for all stations? Please clarify it. Based on the classification results in Figure 3, it is better to select the appropriate periodic function according to the periodic characteristics of different stations.
5. Furthermore, do the authors consider how to apply the proposed CGPT3 model to some areas without radiosonde stations?
6. Figure 9: Please use the same color bar for the first two subfigures.
7. Section 4.3: the manuscript evaluates the impact of the CGPT3 model on retrieving PWV observations using the radiosonde data. In terms of the applicability of the CGPT3 model, this assessment is meaningless. As mentioned in the manuscript, accurate ZHD observations play an essential role in the retrieval of GNSS PWV products, so it is better to evaluate the contribution of the proposed model in GNSS PWV retrieval.
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
The paper is interesting and fairly straightforward. It analyses two widely accepted global models used for the determination of zenith hydrostatic delays (ZHD) in China and proposed improvements, which could be used in China for VLBI and GNSS. Apart from some minor suggestions/corrections listed below for authors' considerations, two main comments are:
1. Since RS data are used as a ground truth here, the errors in RS derived ZHDs should also be given/discussed
2. Since GNSS and VLBI fix ZHDs and estimate zenith wet delays (ZWDs) and not the zenith tropo delay (ZTD), implied here, a ZHD error causes significant mapping separation errors also affecting ZWD estimates. This is due to the fact that hydrostatic and wet mapping functions, used for slant tropospheric delays are different. This also needs to be considered and discussed.
---- Some detail suggestions/corrections -------
l. 18-20: are confusing, not sure what is here meant ...
l. 35, 81: trajectory ? you mean baseline?
l. 36: not only PWV, also GNSS/VLBI precise solutions (positions, orbits, clocks/baselines) require accurate ZHD/ZWD separations
l. 37: GNSS/VLBI typically solve for ZWD, not ZTD!
l. 49-50: vertical decline rate => vertical lapse rate?
l.57: with negligible attention to => neglecting?
l.72: received approval from all across the world => been adopted worldwide ?
l. 123 : selected sites (marked with dots) is shown in Figure 1? Not sure here if all the sites in Fig, 1 are the selected ones, or all the RS sites used in this study?
l.139: parameter ( please list the GPT3 parameters here).
l. 159: what is the grid spacing for the GTrop coefficients? this should be listed here
l. 179-180: are confusing, you mean : with annual or semi-annual .. ?
l. 185-: .. diurnal signals ..? Note due to the 12h RS sampling, the FFT diurnal signals represent the sum of all the signals with frequencies <= 1 day, including the diurnal one. This should be mentioned here, I think
l. 202: semi-diurnal variations ... challenging? Impossible due to 12h RS sampling!
Fig. 2: Doy= ? you mean Day= ?
Eq. 7: the height variation, included in GPT3 is not considered here? Why? (remember CGPT3 is intended for the grid, not just for the RS sites)
Tab. 2, 3 and Fig. 4 legends: should include ZHD as well as the dissuasion related to the results should also include an estimate of the ground truth errors, i.e., RS errors, which by no means are negligible
l.271: bias % improvements makes a little or no sense, I think , only MAE and RMS ones are meaningful
Figs 5-7 legends should include ZHD
Fig. 8: how much of the RS variation here is caused by RS errors? Are RS errors negligible?
l. 361: where dPWV and dZWD are the partial derivatives of .. with respect to ZHD
Eqs 13, 14: in addition to ZHD errors should also consider hydrostatic/wet mapping separation errors, caused by ZHD errors. This is caused by an incorrect separation in GNSS and VLBI analyses of ZHD (which fixed) and ZWD, which is estimated and the fact that ZHD and ZWD mapping functions used for slant tropo delays are different. (ZWD not ZTD is estimated in GNSS and VLBI). The mapping separation error can reach up to 10% of ZHD errors, See e.g.
Boehm J, Werl B, Schuh H (2006a) Troposphere mapping functions for GPS and very long baseline
interferometry from European Centre for Medium-Range Weather Forecasts operational analysis data, J.
Geophys. Res. 111, B02406, doi:10.1029/2005JB003629
l. 395: see l. 271
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
