the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Quality Assessment of YUNYAO GNSS-RO Refractivity Data in the Neutral Atmosphere
Abstract. GNSS (Global Navigation Satellite System) Radio Occultation (RO) data is an important component of numerical weather prediction (NWP) systems. To incorporate more GNSS RO data into NWP systems, commercial RO data has become an excellent option. Tianjin Yunyao Aerospace Technology Co., Ltd. (YUNYAO) plans to launch a meteorological constellation of 90 satellites equipped with GNSS-RO instruments, which will significantly increase the amount of GNSS-RO data in NWP systems. This study evaluates the quality of neutral atmosphere refractivity profiles from YUNYAO satellites Y003 to Y010 during the period from May 1 to July 31, 2023. Compared with the refractivity calculated from ERA5, the absolute value of the mean bias (MB) for YUNYAO refractivity data is generally less than 1.5 % between 0 and 40 km, and close to 0 between 4 and 40 km. The standard deviation (SD) is less than 3.4 %, and there are differences in the SDs for different GNSS satellites, especially in the lower troposphere and the stratosphere. Second, the refractivity error SD of YUNYAO RO data is estimated using the "three-cornered hat" (3CH) method and multiple data sets. In the pressure range of 1000–10 hPa, the refractivity error SD of YUNYAO RO data is below 2.6 %, and the differences in refractivity error SD among different GNSS satellites do not exceed 0.5 %. Finally, compared to COSMIC-2 and Metop-C RO data, YUNYAO RO data exhibit consistent refractivity error SD and are smaller within 300–50 hPa.
- Preprint
(4945 KB) - Metadata XML
-
Supplement
(505 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on amt-2024-150', Anonymous Referee #1, 04 Nov 2024
This is a well-written paper and presents interesting results evaluating radio occultation (RO) observations from a constellation of eight small commercial satellites developed by Tianjin Yunyao Aerospace Technology. The eight Yunyao satellites produced about 12,000 RO occultations per day over the period May-June 2023. The quality of these observations are evaluated based on mean and standard deviations of Yunyao refractivities computed by comparing with radiosondes and two models, as well as the penetration rates (% of profiles reaching levels near the Earth’s surface). This paper is a potentially valuable contribution and should be published, but there are several areas where improvements are needed.
The paper says Yunyao plans to launch 90 satellites, which would provide an unprecedented number of RO profiles per day (probably greater than 150,000). The impact of this large number of RO profiles on NWP forecasts is expected to be very large, and hence this paper, which is the first to describe the quality of Yunyao RO observations, is important. The readers would be interested in the timeline for these launches and the likelihood that they will actually be launched. I know this is a hard question, but can the authors comment on this? Are the resources already obtained or are they merely a goal? Also, will all 90 satellites be launched into similar polar orbits, or will some be launched in low-inclination orbits? Some additional detail on the plans and their status would be useful.
The methodology for evaluating the Yunyao observations in this paper is sound and the statistics look comparable to those from other RO missions. However, the paper only evaluates refractivities, and most current NWP models assimilate bending angles (BA). The paper would be improved if it evaluated BA rather than refractivities. The authors should explain why they chose refractivities instead of BA. If possible, they should include at least some BA evaluations in the paper.
The paper does not describe or provide a reference on how the Yunyao data are processed. Early studies of Yunyao BA data, which were graciously provided by Yunyao for the Radio Occultation Modeling Experiment (ROMEX), showed that there were issues in the BA uncertainties around 12 km, thought to be associated with the transition from geometric to wave optics. There was another issue of large standard deviations and large negative biases below 5 km in the BA, which was apparently due to very strict quality control in Yunyao’s lower-level processing, where carrier phase measurements were removed from the excess phase data. There does seem to be large negative biases in the refractivities below 5 km in these results (Fig. 3); however, the refractivity error SD in these figures look reasonable. It would be good if the paper could summarize how the Yunyao data are processed and if the two issues above have been resolved in the current processing.
It would also be useful to include estimates of the error SD for the refractivity data up to 50-60 km, using the ERA5 and FNL model analyses as the other two datasets (extend Figs. 6-8 up to 50-60 km).
As the paper notes, the radiosonde 3CH uncertainty estimates are less than those of the RO, which is different from the results of Schreiner et al. 2020 and Rieckh et al. 2021. They attribute this difference to the fact that they used a double-differencing correction to the radiosonde data, which is plausible since Rieckh et al. and Schreiner et al. did not. They could test their proposed reason easily be redoing their Fig. 5 with uncorrected radiosonde data and present the results in a second part to Fig. 5.
The length of the paper is appropriate, and the quality of the figures overall is high, with a few exceptions (Fig. 2 and Fig. 4). Please see detailed comments on these figures.
In summary, this paper could be acceptable for publication after the authors consider the above comments. I look forward to seeing a revised version.
Detailed comments
- Lines 45-46—The statement that CMA-GFS incorporates about 20,000 RO profiles per day is interesting, and it would be useful to summarize the impact of these observations and refer to a study that shows this impact if one exists. By coincidence, this number is the number that the WMO International Working Group on Radio Occultation (IROWG) has been recommending for many years now for an operational RO backbone. Are there any published studies or reports that discuss the impact of these observations on the CMA forecasts?
- Lines 69-70—I am sure the three references here do not demonstrate that the Yunyao receivers are “significantly smaller and lighter” than the COSMIC and Metop/GRAS receivers. I don’t doubt that they are, but some numbers should be given, i.e.how much smaller and lighter?
- The penetration depths (minimum heights above ground reached by the RO profiles) shown in Fig. 2 are not very clear. For example, the color difference between 500 m and 2 km (a large difference in penetration depth) is very small, and it is difficult to distinguish between them. I suggest contours of penetration depth with a contour interval of 250 or 500 m for these figures, probably after the data in the grid boxes are smoothed. Alternatively, plot cumulative % profiles of penetration depths of zonal averages of the profiles for several representative latitudes for the GPS, GLONASS and BDS, similar to Fig. 2 of Schreiner et al. (2020).
- Line 183—Give the value of the mean bias between 4-40 km, and also 10-30 km (the sweet spot for RO).
- 4—Similar to the comment for Fig. 2, the colors for the middle panel (SD) do not show the variations clearly (everything is dark blue) Again, contours of the smoothed data would be clearer.
- 4, right panel: Why is the latitude distribution of numbers not symmetric? The are many more between 25-40°S than 25-40°N.
- Line 192: These are fairly small differences, and may not be important for DA. Also, most models assimilate BA not N.
- Line 211: Fig. 4 should be Fig. 5.
- Line 227: “is” should be “are.”
- Line 253: Shouldn’t 8a be 7a?
- Line 314: Earlier you said the double differencing was the likely reason for the radiosondes SD being less than the RO SD, and I think this is probably the main reason. You should mention this here. Or, since this is not a major conclusion for this study, you could delete reference to it in the Summary and Conclusions section.
Reference
Rieckh, T., J. Sjoberg and R. Anthes, 2021: The three-cornered hat method for estimating error variances if three or more atmospheric data sets-Part II: Evaluating recent radio occultation and radiosonde observations, global model forecasts, and reanalyses. J. Atmos. and Ocean. Technol., 35, https://doi.org/10.1175/JTECH-D-20-0209.1
Citation: https://doi.org/10.5194/amt-2024-150-RC1 -
RC2: 'Comment on amt-2024-150', Anonymous Referee #2, 28 Nov 2024
This manuscript by Xu et al assess the mean difference, standard deviation and error for eight different Yunyao satellites by comparing them to radiosondes, COSMIC-2 and Metop B/C observations. The main measure is retrieved refractivity. The study is very detailed and the figures are in a good shape. This paper demonstrates nicely the quality of this new dataset. However, I miss the assessment of bending angles instead of refractivity as this measure is most often used for the assimilation in NWP models. I accept the publication after addressing this issue. Further, I have the following minor issues:
-page 1, l.30: I wouldn’t phrase that refractivity is a function of liquid and frozen water – of course if you have a polarized signal than the polarisation would be affected but not the bending of the ray path. Of course, super refraction can occur in stratocumulus regions.
-page 2, l.46. It would be good to know what data goes already into CMA-GFS. 20.0000 daily profiles is impressive. Maybe a small table would help.
-page 2, l.47: Does GeoOptics still provides data in year 2024?
-page 6, l.126/127: This phrase summarizes very shortly how one derives e.g. refractivity or also physical measures, like temperature. It is good to mention that one has to make certain assumptions to get there. Probably good to add this here.
-page 7, l.142: How is the interpolation done in the vertical? Linearly or doing a spline interpolation?
-page7, l.146/7: Here, I am slightly confused. Which method to you use to calculate MB and SD? The method by Lanzante or eq 2,3 and 4. I guess you use Lanzante to get rid off outliers and then use this cleaned sample to compute MB, SD according to the given equations. Probably rephrasing this sentence, makes this clearer.
-p.12, l.216: instead of will be write is
Citation: https://doi.org/10.5194/amt-2024-150-RC2
Viewed
HTML | XML | Total | Supplement | BibTeX | EndNote | |
---|---|---|---|---|---|---|
106 | 26 | 89 | 221 | 10 | 4 | 5 |
- HTML: 106
- PDF: 26
- XML: 89
- Total: 221
- Supplement: 10
- BibTeX: 4
- EndNote: 5
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1