the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 19 Jun 2025
Design and Application of Two Consistency Verification Models for Weather Radar Networks in the South China Region
Abstract. The observational consistency between ground-based weather radars significantly impacts the quality of mosaic products and severe convection identification products. The real-time monitoring of observational biases between radars can provide a basis for calibration and adjustment. This study designed a consistency verification model for weather radar networks based on the FY-3G precipitation radar (SGRCM) and a ground-based weather radar network consistency verification model (AWRCM). From January to October 2024, observational experiments were conducted in the South China region involving 19 S-band weather radars and 13 X-band phased-array weather radars. The aim was to analyze the influencing factors of the consistency verification models and the observational biases of reflectivity factors for radars with different bands and systems. For the S-band weather radars, the difference in the mean bias between the two models ranged from -1.5 dB to 1.4 dB, and the difference in the mean standard deviation ranged from -1.2 dB to 1.2 dB. For the X-band phased-array weather radars, the difference in the mean bias between the two models ranged from -6.67 dB to 0.84 dB, and the difference in the mean standard deviation ranged from -0.38 dB to 1.51 dB. The evaluation results of the two models show good consistency for weather radars with different bands.
- Preprint
(3454 KB) - Metadata XML
- BibTeX
- EndNote
Status: closed
-
RC1: 'Comment on egusphere-2025-2313', Anonymous Referee #1, 02 Jul 2025
- On pages 85 to 90 of the article, the VCP21 scanning mode is mainly used for stratiform precipitation, while VCP31 is primarily used for clear-sky conditions.It is recommended to make the changes in the text after confirmation.
- Section 3.1.2 mentions the removal of shielding caused by terrain. However, in recent years, shielding caused by buildings at low elevation angles has become increasingly common. How is this factor considered in the algorithm to ensure the reliability of the final results?
- How can the stability and accuracy of satellite data be ensured, and how can the results of satellite-ground comparisons be used to calibrate the radar?
Citation: https://doi.org/10.5194/egusphere-2025-2313-RC1 -
AC1: 'Reply on RC1', Heng Hu, 07 Jul 2025
Thank you for your valuable comments. Our detailed responses are as follows:
Comment 1: Regarding the volume scanning modes currently operated in the radar network, we will verify the operational configurations and make appropriate revisions in the updated version of the manuscript.
Comment 2: Presently, we have incorporated configurable parameters within the algorithm. For stations subject to significant obstruction—including those caused by buildings—the lowest elevation angle (0.5°) is excluded from comparative analyses. Additionally, we are developing an algorithm to systematically assess the actual shielding conditions at each station. Once refined, this will be integrated into the model to enhance overall reliability.
Comment 3: FY-3G will be cross-compared with other satellites such as GPM to ensure a robust and consistent calibration mechanism. Within this framework, we propose using FY-3G as a reference benchmark to evaluate the consistency among ground-based radar networks, rather than directly calibrating discrepancies between radar and satellite observations. This approach is adopted due to the inherent difficulty in establishing which observation—satellite or radar—more accurately represents the true characteristics of the observed targets.
Citation: https://doi.org/10.5194/egusphere-2025-2313-AC1
-
RC2: 'Comment on egusphere-2025-2313', Anonymous Referee #2, 13 Jul 2025
This paper needs a major revision. I think the material is there, but a more thorough analysis is needed. A lot of information on the radars are missing (hardware signal processing e.gI) so that the results are hard to interpret. Figures are not properly discussed (e.g Figure 17) where two modes are visible. In principle you would expect one linear dependence (ideally 1:1).
Some more specific comments;
You don’t list other options to verify the calibration and consistency of data in the network, most importantly the sun as a reference. Please include and discuss!
A technical description of the radars you investigate is missing. This is important to interpret the results:
e.g. antenna gain, beam width, transmit power, tx type, dualpol or not, signal processing, clutterfilter and so on.
you missed https://amt.copernicus.org/preprints/amt-2021-257/amt-2021-257.pdf
check out this paper
l 48: „satellite used as a reference standard”: I would disagree here. No weather radar network is using satellite radar data as a reference operationally
l.100: is an attenuation correction performed with the Ku, Ka-Band data? or do you avoid situations with attenuation?
- 135: the specific mathematical detail to get the coordinates right, should be moved to an appendix, unless there is something new here.
- 232: is this a X-Band phased array? You don’t use an attenuation correction?
this section needs be reworked in order to really provide a meaningful comparison between the two bands
- 244: the 15-35 dBZ: do you do an attenuation correction? or do you avoid any precipitation > 35 dBZ? but then 35 dBZ is probably too large for the X-Band; you will have attenuation. Please clarify.
- 12: I couldn’t find a reference to figure 9. it is not clear which radar is the Radar1 or 2. Clearly state what radar is meant! what kind of correction is shown?
l 265, fig 10: no dualpol system? no sqi, Doppler filter implemented?
- 280: describe the fuzzy logic interference removal I think you mean the left figure as the quality controlled picture?
l 310: figure 15: I don’t understand this figure. How does the ground based consistence analysis looks like? Take radar 1: what is the reference radar here? How do you come up with the bias?
Fig 16: font cannot be read. Rework the figures. X-Axis is a time axis. What time period? why not showing the times? Larger biases can be attributed to specific weather events? Are there any snow cases?
l319: reflectivity is not “strong” it is large, small I would say
Fig 17: clearly two modes are visible in each plot, they are not discussed and explained! (two linear fits with different slopes could be fitted). Two modes suggest that there is something fundamentally wrong, or?
l. 328: without discussing the quality control of the reflectivity factor from the X-Band the results are difficult to interpret: are you really sure that you can rule out attenuation effects e.g.?
l 335: so Fig 19: really doesn’t say anything about the biases. Comparing Fig 19 and 15 one would assume similar performance of the S and X-Band. Why do you show standard deviations? Doesn’t make sense to me. Please explain!
l 361: what is a SC model weather radar?
Fig 23: the result suggests that the satellite / radar has further systematic problems errors in my view. The calibration does not provide a more consistent result.
Citation: https://doi.org/10.5194/egusphere-2025-2313-RC2 -
CC1: 'Reply on RC2', Heng Hu, 25 Jul 2025
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2025/egusphere-2025-2313/egusphere-2025-2313-CC1-supplement.pdf
-
AC3: 'Reply on RC2', Heng Hu, 25 Jul 2025
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2025/egusphere-2025-2313/egusphere-2025-2313-AC3-supplement.pdf
-
RC3: 'Comment on egusphere-2025-2313', Anonymous Referee #3, 16 Jul 2025
This manuscript proposes methods for verification of weather radar networks. Not only by ground-based radars, but also space-borme radars are used for consistency verification. I have several critical queries to be solved before the final decision.
General Comments:
- The overall direction and purpose of the manuscript remain unclear. Additionally, the description of the analysis methodology is insufficient, making it impossible to reproduce the results based on the current manuscript.
- Despite the abundance of radar systems in China, the authors do not specify which radars or what time periods were used in the analysis. Furthermore, the text-only explanation of the radar locations is difficult to interpret. At minimum, a map of the radar network should be included to facilitate understanding.
- The study investigates biases through comparisons between ground-based radars and between ground-based and satellite radars. However, such comparisons merely highlight the relative biases between systems, and an independent, well-calibrated reference radar is essential. Is there no such calibrated radar within the network used in this study?
- It is also unclear what types of biases the authors are attempting to identify. Are these parameters that cannot be corrected through individual radar calibration, or are they related to factors like beam blockage or system biases that can be corrected? The manuscript lacks clarity on this point. Additionally, even if biases are identified, the manuscript does not explain how this information will be used—whether for correction or simply as observational insight.
- Although the term "model" is used, the methodology appears to be more of a data extraction and comparison approach rather than a model in the conventional sense.
- The manuscript refers to numerous parameters used in data extraction, but they are scattered throughout the text and difficult to follow. Parameters such as thresholds should be clearly summarized in a table.
- From Section 3.1.3 onward, the statistical analyses lack clarity regarding which radar(s) and what data periods were used. Without a clear listing of these, the reliability and reproducibility of the analysis cannot be ensured.
- There is insufficient explanation of the analysis methods. For example, in the paragraph starting on P.10 L.219, how was VIL calculated? Also, which radar stations correspond to Radar1 and Radar2 in Figure 7?
- In Section 3.2 and onward, only a subset of the presumably large dataset is shown. However, since the selection criteria are not explained, the reliability of the results is questionable—for example, in Figure 20.
Specific Comments:
- P.3 L.78: The phrase “corrected for frequency” is unclear, as reflectivity in the Rayleigh scattering regime is not wavelength-dependent. Please clarify what correction was applied and how.
- P.3 L.87: Please write out “VCP” (Volume Coverage Pattern) in full upon first use.
- P.3 Figure 1: Indicate the satellite’s direction of movement directly on the figure.
- P.4 L.88: The phrase “Evaluation results from 2024...in this study” requires a citation.
- P.4 L.96–97: The terms “PRE” and “FRE” are undefined and should be explained.
- P.4 L.102: It is unclear what the “first and second reference frames” refer to.
- P.4 L.110: Explain how the averaging and gridding were performed. These procedures can introduce bias and should be described in detail.
- P.6 L.120: The term “S-PAR” is undefined and should be clarified.
- P.7 L.138: The meaning of “Km = 4/3” is unclear and should be explained.
- P.7 L.160: The variable “Hthre” should be written with a subscript for clarity.
- P.8 L.170: If comparing a single satellite with a single ground radar, vertical resolution should not be an issue. The intent of this sentence is unclear. If multiple ground radars are being matched to one satellite, this should be clearly stated.
- P.11 L.244: Justification is needed for choosing the reflectivity range of 15–35 dBZ. If rain attenuation is a concern, then strong reflectivity along the beam path should also be considered for exclusion. Please elaborate.
- P.11 Figure 8: It is unclear which result corresponds to the S-band radar.
- P.13 Figure 10: The relative positions of the radars are not shown. Without this context, comparing the two radars is impossible for readers. At minimum, the coastline should be shown, and the map axes (latitude/longitude) should be consistent across both subplots.
- P.13 Figure 11: The figure does not indicate what parameter is being visualized. Please clarify.
- P.14 L.280: The term “Fuzzy logic” is mentioned without describing the actual algorithm or implementation used.
- P.14 Figure 12: The left and right panels may be reversed—the left appears to be quality-controlled. Also, clarify which radar (and frequency band) was used to generate these results.
- P.14 Figure 13: Specify which radar was used for these results.
- P.17 Figure 16: The axis labels are too small to read. Also, it is unclear which radar the bias was calculated from.
- P.17 Figure 17: The caption text within the figure is obscured by the data points.
- P.20 L.362: The abbreviation “SC” is undefined and should be explained.
- P.21 Figure 22: This figure would be more informative if the x-axis used a time scale.
Citation: https://doi.org/10.5194/egusphere-2025-2313-RC3 -
AC2: 'Reply on RC3', Heng Hu, 25 Jul 2025
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2025/egusphere-2025-2313/egusphere-2025-2313-AC2-supplement.pdf
Status: closed
-
RC1: 'Comment on egusphere-2025-2313', Anonymous Referee #1, 02 Jul 2025
- On pages 85 to 90 of the article, the VCP21 scanning mode is mainly used for stratiform precipitation, while VCP31 is primarily used for clear-sky conditions.It is recommended to make the changes in the text after confirmation.
- Section 3.1.2 mentions the removal of shielding caused by terrain. However, in recent years, shielding caused by buildings at low elevation angles has become increasingly common. How is this factor considered in the algorithm to ensure the reliability of the final results?
- How can the stability and accuracy of satellite data be ensured, and how can the results of satellite-ground comparisons be used to calibrate the radar?
Citation: https://doi.org/10.5194/egusphere-2025-2313-RC1 -
AC1: 'Reply on RC1', Heng Hu, 07 Jul 2025
Thank you for your valuable comments. Our detailed responses are as follows:
Comment 1: Regarding the volume scanning modes currently operated in the radar network, we will verify the operational configurations and make appropriate revisions in the updated version of the manuscript.
Comment 2: Presently, we have incorporated configurable parameters within the algorithm. For stations subject to significant obstruction—including those caused by buildings—the lowest elevation angle (0.5°) is excluded from comparative analyses. Additionally, we are developing an algorithm to systematically assess the actual shielding conditions at each station. Once refined, this will be integrated into the model to enhance overall reliability.
Comment 3: FY-3G will be cross-compared with other satellites such as GPM to ensure a robust and consistent calibration mechanism. Within this framework, we propose using FY-3G as a reference benchmark to evaluate the consistency among ground-based radar networks, rather than directly calibrating discrepancies between radar and satellite observations. This approach is adopted due to the inherent difficulty in establishing which observation—satellite or radar—more accurately represents the true characteristics of the observed targets.
Citation: https://doi.org/10.5194/egusphere-2025-2313-AC1
-
RC2: 'Comment on egusphere-2025-2313', Anonymous Referee #2, 13 Jul 2025
This paper needs a major revision. I think the material is there, but a more thorough analysis is needed. A lot of information on the radars are missing (hardware signal processing e.gI) so that the results are hard to interpret. Figures are not properly discussed (e.g Figure 17) where two modes are visible. In principle you would expect one linear dependence (ideally 1:1).
Some more specific comments;
You don’t list other options to verify the calibration and consistency of data in the network, most importantly the sun as a reference. Please include and discuss!
A technical description of the radars you investigate is missing. This is important to interpret the results:
e.g. antenna gain, beam width, transmit power, tx type, dualpol or not, signal processing, clutterfilter and so on.
you missed https://amt.copernicus.org/preprints/amt-2021-257/amt-2021-257.pdf
check out this paper
l 48: „satellite used as a reference standard”: I would disagree here. No weather radar network is using satellite radar data as a reference operationally
l.100: is an attenuation correction performed with the Ku, Ka-Band data? or do you avoid situations with attenuation?
- 135: the specific mathematical detail to get the coordinates right, should be moved to an appendix, unless there is something new here.
- 232: is this a X-Band phased array? You don’t use an attenuation correction?
this section needs be reworked in order to really provide a meaningful comparison between the two bands
- 244: the 15-35 dBZ: do you do an attenuation correction? or do you avoid any precipitation > 35 dBZ? but then 35 dBZ is probably too large for the X-Band; you will have attenuation. Please clarify.
- 12: I couldn’t find a reference to figure 9. it is not clear which radar is the Radar1 or 2. Clearly state what radar is meant! what kind of correction is shown?
l 265, fig 10: no dualpol system? no sqi, Doppler filter implemented?
- 280: describe the fuzzy logic interference removal I think you mean the left figure as the quality controlled picture?
l 310: figure 15: I don’t understand this figure. How does the ground based consistence analysis looks like? Take radar 1: what is the reference radar here? How do you come up with the bias?
Fig 16: font cannot be read. Rework the figures. X-Axis is a time axis. What time period? why not showing the times? Larger biases can be attributed to specific weather events? Are there any snow cases?
l319: reflectivity is not “strong” it is large, small I would say
Fig 17: clearly two modes are visible in each plot, they are not discussed and explained! (two linear fits with different slopes could be fitted). Two modes suggest that there is something fundamentally wrong, or?
l. 328: without discussing the quality control of the reflectivity factor from the X-Band the results are difficult to interpret: are you really sure that you can rule out attenuation effects e.g.?
l 335: so Fig 19: really doesn’t say anything about the biases. Comparing Fig 19 and 15 one would assume similar performance of the S and X-Band. Why do you show standard deviations? Doesn’t make sense to me. Please explain!
l 361: what is a SC model weather radar?
Fig 23: the result suggests that the satellite / radar has further systematic problems errors in my view. The calibration does not provide a more consistent result.
Citation: https://doi.org/10.5194/egusphere-2025-2313-RC2 -
CC1: 'Reply on RC2', Heng Hu, 25 Jul 2025
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2025/egusphere-2025-2313/egusphere-2025-2313-CC1-supplement.pdf
-
AC3: 'Reply on RC2', Heng Hu, 25 Jul 2025
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2025/egusphere-2025-2313/egusphere-2025-2313-AC3-supplement.pdf
-
RC3: 'Comment on egusphere-2025-2313', Anonymous Referee #3, 16 Jul 2025
This manuscript proposes methods for verification of weather radar networks. Not only by ground-based radars, but also space-borme radars are used for consistency verification. I have several critical queries to be solved before the final decision.
General Comments:
- The overall direction and purpose of the manuscript remain unclear. Additionally, the description of the analysis methodology is insufficient, making it impossible to reproduce the results based on the current manuscript.
- Despite the abundance of radar systems in China, the authors do not specify which radars or what time periods were used in the analysis. Furthermore, the text-only explanation of the radar locations is difficult to interpret. At minimum, a map of the radar network should be included to facilitate understanding.
- The study investigates biases through comparisons between ground-based radars and between ground-based and satellite radars. However, such comparisons merely highlight the relative biases between systems, and an independent, well-calibrated reference radar is essential. Is there no such calibrated radar within the network used in this study?
- It is also unclear what types of biases the authors are attempting to identify. Are these parameters that cannot be corrected through individual radar calibration, or are they related to factors like beam blockage or system biases that can be corrected? The manuscript lacks clarity on this point. Additionally, even if biases are identified, the manuscript does not explain how this information will be used—whether for correction or simply as observational insight.
- Although the term "model" is used, the methodology appears to be more of a data extraction and comparison approach rather than a model in the conventional sense.
- The manuscript refers to numerous parameters used in data extraction, but they are scattered throughout the text and difficult to follow. Parameters such as thresholds should be clearly summarized in a table.
- From Section 3.1.3 onward, the statistical analyses lack clarity regarding which radar(s) and what data periods were used. Without a clear listing of these, the reliability and reproducibility of the analysis cannot be ensured.
- There is insufficient explanation of the analysis methods. For example, in the paragraph starting on P.10 L.219, how was VIL calculated? Also, which radar stations correspond to Radar1 and Radar2 in Figure 7?
- In Section 3.2 and onward, only a subset of the presumably large dataset is shown. However, since the selection criteria are not explained, the reliability of the results is questionable—for example, in Figure 20.
Specific Comments:
- P.3 L.78: The phrase “corrected for frequency” is unclear, as reflectivity in the Rayleigh scattering regime is not wavelength-dependent. Please clarify what correction was applied and how.
- P.3 L.87: Please write out “VCP” (Volume Coverage Pattern) in full upon first use.
- P.3 Figure 1: Indicate the satellite’s direction of movement directly on the figure.
- P.4 L.88: The phrase “Evaluation results from 2024...in this study” requires a citation.
- P.4 L.96–97: The terms “PRE” and “FRE” are undefined and should be explained.
- P.4 L.102: It is unclear what the “first and second reference frames” refer to.
- P.4 L.110: Explain how the averaging and gridding were performed. These procedures can introduce bias and should be described in detail.
- P.6 L.120: The term “S-PAR” is undefined and should be clarified.
- P.7 L.138: The meaning of “Km = 4/3” is unclear and should be explained.
- P.7 L.160: The variable “Hthre” should be written with a subscript for clarity.
- P.8 L.170: If comparing a single satellite with a single ground radar, vertical resolution should not be an issue. The intent of this sentence is unclear. If multiple ground radars are being matched to one satellite, this should be clearly stated.
- P.11 L.244: Justification is needed for choosing the reflectivity range of 15–35 dBZ. If rain attenuation is a concern, then strong reflectivity along the beam path should also be considered for exclusion. Please elaborate.
- P.11 Figure 8: It is unclear which result corresponds to the S-band radar.
- P.13 Figure 10: The relative positions of the radars are not shown. Without this context, comparing the two radars is impossible for readers. At minimum, the coastline should be shown, and the map axes (latitude/longitude) should be consistent across both subplots.
- P.13 Figure 11: The figure does not indicate what parameter is being visualized. Please clarify.
- P.14 L.280: The term “Fuzzy logic” is mentioned without describing the actual algorithm or implementation used.
- P.14 Figure 12: The left and right panels may be reversed—the left appears to be quality-controlled. Also, clarify which radar (and frequency band) was used to generate these results.
- P.14 Figure 13: Specify which radar was used for these results.
- P.17 Figure 16: The axis labels are too small to read. Also, it is unclear which radar the bias was calculated from.
- P.17 Figure 17: The caption text within the figure is obscured by the data points.
- P.20 L.362: The abbreviation “SC” is undefined and should be explained.
- P.21 Figure 22: This figure would be more informative if the x-axis used a time scale.
Citation: https://doi.org/10.5194/egusphere-2025-2313-RC3 -
AC2: 'Reply on RC3', Heng Hu, 25 Jul 2025
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2025/egusphere-2025-2313/egusphere-2025-2313-AC2-supplement.pdf
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 349 | 35 | 16 | 400 | 10 | 21 |
- HTML: 349
- PDF: 35
- XML: 16
- Total: 400
- BibTeX: 10
- EndNote: 21
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1