Atmospheric GNSS RO 1D-Var in Use at UCAR: Description and Validation
Round 1
Reviewer 1 Report
General comments:
The paper is about the 1D-Var method to retrieve atmospheric profiles of temperature, pressure, and humidity in GNSS R0 geometry.
The work is well done, well written and complete in all its parts. Introduction and conclusion are in line with respect to the core of the paper.
All the figures are clear and complete.
Regarding the general accuracy of the 1D Var method in GNSS RO geometry, as expected, the greater errors and uncertainties are related at the lowest tangent altitudes with higher humidity values.
Moreover, the one-dimensional geometry of the RO observation (the final retrieval is always a vertical profile) do not allow to account for horizontal gradients of the atmospheric variables.
IN order increase the 2D-3D space resolution and reduce the uncertainty al the lowest altitudes even in strong humidity conditions, a different source of data could be obtained using the NDSA approach in LEO LEO satellite geometry in CO-ROTATING configuration (the meaning of NDSA that is Normalized Differential Spectral Attenuation).
The NDSA should allow the estimation of the total content of Water vapor along the radio propagation link without any a priori data. Such data could be a valuable constrain and input data for any retrieval method, 1D-var included, to reduce the Humidity errors al the lowest tropospheric altitudes.
Author Response
We thank the reviewer for encouraging and constructive comments. Our point-by-point response is given below.
The paper is about the 1D-Var method to retrieve atmospheric profiles of temperature, pressure, and humidity in GNSS R0 geometry.
The work is well done, well written and complete in all its parts. Introduction and conclusion are in line with respect to the core of the paper.
All the figures are clear and complete.
Comment: Regarding the general accuracy of the 1D Var method in GNSS RO geometry, as expected, the greater errors and uncertainties are related at the lowest tangent altitudes with higher humidity values.
Response: Agreed.
Comment: Moreover, the one-dimensional geometry of the RO observation (the final retrieval is always a vertical profile) do not allow to account for horizontal gradients of the atmospheric variables.
Response: That’s true and there were/are attempts/efforts by the authors to account for horizontal gradients – references [86-87].
Comment: IN order increase the 2D-3D space resolution and reduce the uncertainty al the lowest altitudes even in strong humidity conditions, a different source of data could be obtained using the NDSA approach in LEO LEO satellite geometry in CO-ROTATING configuration (the meaning of NDSA that is Normalized Differential Spectral Attenuation).
Response: We can’t agree more on this. Any independent, accurate information on either moisture or temperature will contribute to retrieval improvement and we don’t doubt NDSA could be a valuable source of such information. However, doing so is irrelevant to our study.
Comment: The NDSA should allow the estimation of the total content of Water vapor along the radio propagation link without any a priori data. Such data could be a valuable constrain and input data for any retrieval method, 1D-var included, to reduce the Humidity errors al the lowest tropospheric altitudes.
Response: Yes. We are very interested in utilizing such data.
Reviewer 2 Report
The authors did fantastic work on the updated GNSS RO 1D-Var method and provided thorough descriptions and validations. The paper is worthy of publication and will be beneficial for multiple applications. There are a few minor comments:
Firstly, the manuscript provides lots of figures with different altitude ranges. For example, in Figure 1, the temperature background error estimates are shown over 70 km, and in Figure 3, the temperature error correlation matrix is shown around 70 km. However, in Section 3, most of the results are shown only up to 40 km, and some of them are up to 10 km. For the data related to radiosondes, the highest altitude is limited by the balloon-burst altitude, but for the synthetic data set, it should have no such altitude limitations. Therefore, the inconsistent altitude range is quite confusing. Does the 1-D Var method provide data over 40 km? If it does, how’s its performance? If doesn’t, then the 1D-Var can be extendable to the thermosphere is questionable (statement in line from 1030 to 1032).
Line 648: It’s better to add all the necessary labels on Height-axis, for example, add “0” on Height-axis for Figures 4, 5, 7, 9, 10, 11, 13, 15.
Line 837: It’s better to increase the linewidth of the blue dashed line.
Author Response
We thank the reviewer for useful comments. We have carefully taken the comments into consideration in preparing our revision. Our point-by-point response is given below.
The authors did fantastic work on the updated GNSS RO 1D-Var method and provided thorough descriptions and validations. The paper is worthy of publication and will be beneficial for multiple applications. There are a few minor comments:
Comment: Firstly, the manuscript provides lots of figures with different altitude ranges. For example, in Figure 1, the temperature background error estimates are shown over 70 km, and in Figure 3, the temperature error correlation matrix is shown around 70 km. However, in Section 3, most of the results are shown only up to 40 km, and some of them are up to 10 km.
Response: In general, the height range of each Figure was carefully chosen to highlight the most important aspects/features (e.g., Figures 1b-c, 3b-c, 11, and 13) or to elucidate the points mentioned in the relevant text (Figures 4-8, and 10). We do not think it is important to show results above 12 km for the figures involving water vapor (Figures 9-11, 13). The temperature height range of Figure 9 is limited by the data range of OLD (39.99 kilometers).
Comment: For the data related to radiosondes, the highest altitude is limited by the balloon-burst altitude, but for the synthetic data set, it should have no such altitude limitations.
Response: The vertical range of the synthetic data experiments is limited by that of GRUAN radiosonde.
Comment: Therefore, the inconsistent altitude range is quite confusing.
Response: We understand well the reviewer’s concern. However, it would be difficult for the Figures to serve their specific purposes if a fixed height range were used for all Figures.
Comment: Does the 1-D Var method provide data over 40 km? If it does, how’s its performance?
Response: NEW covers 0-60 km and shows generally good agreements with numerical weather models.
Comment: If doesn’t, then the 1D-Var can be extendable to the thermosphere is questionable (statement in line from 1030 to 1032).
Response: Extension to the thermosphere requires a different framework, e.g., references [86] and [87].
Comment: Line 648: It’s better to add all the necessary labels on Height-axis, for example, add “0” on Height-axis for Figures 4, 5, 7, 9, 10, 11, 13, 15.
Response: The suggested change to Figures 4, 5, 7, 9, 10, 11, 13, 15 is made and these Figures now starts from 0 km. The reason that original Figures do not start from 0 km is that not all RO profiles hit the ground level. As a result, we have not included the ground level (0 km) at the time of data preparation. The lower height bounds of Figures 5 and 6 correspond to the height of GRUAN site. When the results of Figure 6 are shown from 0 km, the bottom heights in Figures 6a, b, and c-d look quite different, just because different vertical ranges are used for these Figures. After comparing the original Figure 6 and the one suggested by the reviewer side by side, we still prefer showing the results only in the actual data range.
Comment: Line 837: It’s better to increase the linewidth of the blue dashed line.
Response: Done. Thank you.
