the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
HUST-Grace2024: a new GRACE-only gravity field time series based on more than 20 years satellite geodesy data and a hybrid processing chain
Abstract. To improve the accuracy of monthly temporal gravity field models for Gravity Recovery and Climate Experiment (GRACE) and GRACE Follow-On (GRACE-FO) mission, a new series named HUST-Grace2024 is determined based on the updated L1B dataset (GRACE L1B RL03 & GRACE-FO L1B RL04) and the newest atmosphere and ocean de-aliasing product (AOD1B RL07). Compared to the previous HUST temporal gravity field model releases, we made some improvements on both updating background models and processing chain as follows. (1) During the satellite onboard events, the intersatellite pointing angles are calculated to pinpoint the outliers in K-band range rates (KBRRs) and accelerometer observations. To exclude outliers, the advisable threshold is respectively 50 mrad for KBRRs and 20 mrad for accelerations. (2) To relieve the impacts of KBRR noise in different frequencies, a hybrid data weighting method is proposed. Kinematic empirical parameters are used to reduce the low frequency noise, while a stochastic model is designed to relieve the impacts of random noise above 10 mHz. (3) a fully-populated scale factor matrix is used to improve the quality of accelerometer calibration. Analysis in spectral and spatial domain is then implemented, which demonstrates that HUST-Grace2024 has a noticeable reduction of 10 % to 30 % in noise level and remains consistent amplitudes over 48 basins in singal content compared with the official GRACE and GRACE-FO solutions. These evaluations confirm that our aforementioned efforts lead to a better temporal gravity field series.
- Preprint
(2497 KB) - Metadata XML
- BibTeX
- EndNote
Status: closed
-
CC1: 'Comment on essd-2024-39', Shuang Yi, 03 Apr 2024
The improvement relative to official gravity solutions is impressive. The improvement may come from 1) new accelerometer product, 2) new AOD1B product, 3) algorithm in the manuscript. I would like to see how much of the improvement comes from the new products, and how much comes from the algorithm, thus consolidating the contribution of this work. I suggest the authors add such a controlled variable experiment.
Fig. 8, the HUST result is better over the CSR result globally, with the exception in western Pacific. Is there a reason for this?
Page 13, L18, a typo in ‘shown’
Citation: https://doi.org/10.5194/essd-2024-39-CC1 - AC5: 'Reply on CC1', Lijun Zheng, 22 May 2024
-
RC1: 'Comment on essd-2024-39', Anonymous Referee #1, 05 May 2024
The manuscript outlines a data processing strategy, yielding impressive improvements in its recovered temporal gravity solutions. Specific inquiries and comments are as follows:
1) Please specify the time interval used for constructing the observation equation, considering the differing sampling rates between the kinematic orbit (10 seconds) and other L1B data (5 seconds). Additionally, please elaborate on the error assessment strategy for kinematic orbits, including the criteria for error identification, and whether interpolated epochs are included in constructing the observation equation.
2) Please quantitatively analyze the accuracy improvement of the temporal gravity recovery by kinematic orbit and GNV 1B.
3) Is the influence of the thruster accounted for in preprocessing? If so, please clarify whether THR 1B data is utilized and provide details regarding the number of epochs affected by thruster start-up time.
4) I would like to know the performance of your products in the later stage of GRACE. Please extend the time period for comparing GRACE results from 2005-2010 to 2005-2015.
5) Correct a typographical error on Page 12, Line 18, where 'equation (9)' should be amended to 'equation (6)'.
6) On Page 20, Line 7, consider rephrasing "indicating a reduction of -12.8%, -33.2%, and -34.7%" for clarity.Citation: https://doi.org/10.5194/essd-2024-39-RC1 - AC3: 'Reply on RC1', Lijun Zheng, 22 May 2024
-
RC2: 'Comment on essd-2024-39', Anonymous Referee #2, 07 May 2024
This study develops a new series of temporal gravity field models HUSTGrace2024, on the basis of a hybrid data processing chain and updated GRACE and GRACE-FO L1B datasets. The hybrid processing chain takes account of the newest atmosphere and ocean de-aliasing product (AOD1B RL07), an improved data weighting method and the updated accelerometer scale factor matrix, which leads to a better temporal gravity field series. Comparison with the current official GRACE and GRACE-FO solutions (released by CSR, GFZ and JPL) demonstrates that the HUSTGrace2024 models have a lower noise level (with a reduction of 10% to 30%) and comparable signal amplitudes in both the long-term trend and seasonal variations.
The datasets developed by the authors are useful to those who investigate the global climate change and geodynamics with satellite gravimetry. Meanwhile, the hybrid data processing chain provides insights into more accurate determination of the global time-variable gravity field models. In particular, the detailed parameter setting suggestions in GRACE and GRACE-FO L1B data processing (such as the thresholds to detect outliers in range-rate observations, and to set the margin for accelerometer datasets, etc.) are beneficial to more reliable gravity field modeling. I would like to recommend minor revisions of the manuscript before publication in Earth System Science Data, according to the comments as follows.
Comments:
1. On the selection of a quiet ocean area to assess the noise level
In Section 3.1.2, the authors select an open ocean area (in the middle Pacific Ocean) to evaluate the noise level of the datasets (Page 21, Lines 4-5; Page 22, Figure 9). However, the location of the selected ocean area (as well as its latitude and longitude ranges) is not described in the text. Please clarify it.
In addition, using only parts of the ocean is probably not the best way to assess the noise level of the GRACE and GRACE-FO temporal gravity field series. I suggest the authors to add the results for the global ocean. To reduce the effect of signal leakage from the land areas, a buffer zone (with 400 or 500 km from the coastlines) can be considered when determining the global ocean coverage. The RMS comparison for the global ocean among the different products would be more convincing to evaluate the noise level.
2. On the RMSs of TWSA residuals to assess the noise level
In Section 3.1.2, the RMSs of TWSA residuals (by removing the yearly trends and annual amplitudes from the original time series) are used to assess the noise level (Page 19, Equation 13). Nevertheless, removing only the yearly trends and annual amplitudes is likely not enough to separate the signal and noise. For example, in the regions of South America (Amazon basin), East Africa (Nile basin) and South Asia (the surroundings of Bengal Bay), there are notable effects in the RMSs (see the row 3 in Figures 6 and 7), which are probably due to the inter-annual changes in the TWSA. Thus it might be better to take into account the inter-annual variations in Equation (13), with a high-pass filter (for instance, with a threshold of 1.5 or 2 years) to mitigate the inter-annual impact.
In addition, the co- and post-seismic changes of several gigantic earthquakes during the GRACE observation period may also affect the RMSs results to some extent. Since the seismic effects are local and not significant in the RMS distribution (Figure 6), adding some discussions to acknowledge this issue should be enough.
Another concern is about the semi-annual changes in the GRACE and GRACE-FO time series. Normally both the annual and semi-annual terms are included to represent the seasonal variations. So I would suggest the authors to consider the semi-annual term in Equation (13).
3. On the assessment of low degree coefficients C20 and C30
In Section 3.1.1, the authors compare the datasets in the spectral domain. From Figure 5 one can find that the low degree coefficients are notably different between the ones from HUST and those from the three official solutions. Therefore, it should be better to add a specific comparison for the low-degree coefficients (e.g., C20 and C30). By comparing the time C20 and C30 time series among the HUST, CSR (GFZ and JPL) and those from the SLR observations (i.e., TN-14; see Loomis et al., 2020), the advantage of the data product in this study may be better demonstrated.
4. On the acknowledgement of the difference in data inputs between HUST and the official solutions
The authors compare the HUST solutions and those from CSR, GFZ and JPL, and find that their solution performs better than the others. However, one should keep in mind that the data inputs are different for the solutions. The HUST solutions use the newest atmosphere and ocean de-aliasing product (AOD1B RL07), whereas the CSR (GFZ and JPL) solutions use the old version of the AOD1B product. In the near future, CSR (GFZ and JPL) will release the GRACE and GRACE-FO RL07 solutions, which are comparable (in the data inputs) to the HUST solutions. Whether the HUST datasets perform better than the RL07 official products remains a question. Please add some discussions to acknowledge this issue.
In addition, CSR has released the RL06.2 solutions for GRACE-FO since September 2023. But the authors do not mention this official product. Please add some necessary statements or discussions.
Corrections on writing details:
(1) Page 1, Line 25: singal -> signal
(2) Page 4, Line 1: the quiet ocean -> a selected open ocean region
For the “quiet ocean” in the rest parts of the manuscript (e.g., Page 21, Lines 4, 10 and 13; Page 22, Line 2; Page 28, Line 3; …), it should be better to change it into “open ocean”.
(3) Page 4, Line 7: background model -> background models
(4) Page 6, Line 5: Importance of renewing -> Renewing of
To be consistent with the title of Section 2.1 (Page 4, Line 7).
(5) Page 16, Lines 20-21: The result indicates that our proposed hybrid weighting approach can obtain the better temporal gravity field solutions.
This statement is not appropriate for the Methods section. Please remove the sentence and consider to make this statement in the Results or Discussion section.
(6) Page 17, Line 2: from 2005 to 2010 -> from 2005 to 2010 for GRACE solutions
(7) Page 17, Line 21: The numerical values “7.528, 9.827, 10.049, and 6.561”, and many other values in the rest parts of the text (e.g., Page 19, Lines 17 and 19; Page 20. Lines 1 and 2; …), as well as the values in the tables (e.g., Tables 3 to 5)
It does not make much practical sense to use 3 decimal places for the numerical values in the results. Using 1 decimal place (or at most 2) should be enough.
(8) Page 19, Line 4: a decorrelation filter
Which decorrelation filter? Please specify it.
(9) Page 21, Line 15: “and HUST-Grace2024” -> “and HUST-Grace2024, respectively”
(10) Page 22, Lines 5-6: the representative deserts and the east of Antarctic
Please describe the latitude and longitude ranges for these selected areas, or show their coverages in a map.
(11) Page 23, Line 13: annual amplitude -> annual amplitudes
(12) Page 26, Line 2: official solution -> official solutions
(13) Page 30, Line 27, and Page 31, Line 1:
The references Cheng & Ries (2017) and Cheng & Ries (2023) are not mentioned in the text.
(14) Page 35, Line 8: “Geo. Jou. Int.”
Please use “Geophys. J. Int.”, or use the complete journal name.
Citation: https://doi.org/10.5194/essd-2024-39-RC2 - AC1: 'Reply on RC2', Lijun Zheng, 22 May 2024
-
RC3: 'Comment on essd-2024-39', Anonymous Referee #3, 07 May 2024
The manuscript provides detailed information about their new version of the time-variable gravity field series using a new data processing strategy and new input data. However, the following are my primary comments and suggestions for major comments to the study;
-
The GRACE result section should incorporate also comparisons with previous versions of HUST-Grace to assess how they correspond with the official GRACE solutions.
- During the GRACE mission, particularly after 2010, more factors emerged, including maneuvers and the GRACE-B battery issue. Hence, it is imperative to incorporate the GRACE time period post-2010 within the results section, as it will serve as a comprehensive testing phase for the new step procedures provided in the study.
- The study's proposed improvement steps should clearly and prominently show the impact of either the new accelerometer calibration or the new AOD1B product. This can be done in the result section, comparing it with the prior version of HUST solutions.
- The metrics used to examine the models in the comparison section were used in a very subset of the significance of the magnitudes. Although the comparison is intended to emphasize that the HUST-Grace2024 model is better in a meaningful way, it may not be meaningful if the cm or mm orders are 3 digits finer after the comma. For example, Page 19, Line 17, RMSs over ocean is in cm, but the notations are well below mm.
Citation: https://doi.org/10.5194/essd-2024-39-RC3 - AC2: 'Reply on RC3', Lijun Zheng, 22 May 2024
-
-
CC2: 'Comment on essd-2024-39', Zhengwen Yan, 10 May 2024
Page 20, L7, 'including (a, e, f) CSR RL06, (b, f, j) GFZ RL06, (e, g, k) JPL RL06 and (d, h, f) HUST-Grace2024' should be corrected as 'including (a, e, i) CSR RL06, (b, f, j) GFZ RL06, (c, g, k) JPL RL06 and (d, h, l) HUST-Grace2024'
Citation: https://doi.org/10.5194/essd-2024-39-CC2 - AC4: 'Reply on CC2', Lijun Zheng, 22 May 2024
Status: closed
-
CC1: 'Comment on essd-2024-39', Shuang Yi, 03 Apr 2024
The improvement relative to official gravity solutions is impressive. The improvement may come from 1) new accelerometer product, 2) new AOD1B product, 3) algorithm in the manuscript. I would like to see how much of the improvement comes from the new products, and how much comes from the algorithm, thus consolidating the contribution of this work. I suggest the authors add such a controlled variable experiment.
Fig. 8, the HUST result is better over the CSR result globally, with the exception in western Pacific. Is there a reason for this?
Page 13, L18, a typo in ‘shown’
Citation: https://doi.org/10.5194/essd-2024-39-CC1 - AC5: 'Reply on CC1', Lijun Zheng, 22 May 2024
-
RC1: 'Comment on essd-2024-39', Anonymous Referee #1, 05 May 2024
The manuscript outlines a data processing strategy, yielding impressive improvements in its recovered temporal gravity solutions. Specific inquiries and comments are as follows:
1) Please specify the time interval used for constructing the observation equation, considering the differing sampling rates between the kinematic orbit (10 seconds) and other L1B data (5 seconds). Additionally, please elaborate on the error assessment strategy for kinematic orbits, including the criteria for error identification, and whether interpolated epochs are included in constructing the observation equation.
2) Please quantitatively analyze the accuracy improvement of the temporal gravity recovery by kinematic orbit and GNV 1B.
3) Is the influence of the thruster accounted for in preprocessing? If so, please clarify whether THR 1B data is utilized and provide details regarding the number of epochs affected by thruster start-up time.
4) I would like to know the performance of your products in the later stage of GRACE. Please extend the time period for comparing GRACE results from 2005-2010 to 2005-2015.
5) Correct a typographical error on Page 12, Line 18, where 'equation (9)' should be amended to 'equation (6)'.
6) On Page 20, Line 7, consider rephrasing "indicating a reduction of -12.8%, -33.2%, and -34.7%" for clarity.Citation: https://doi.org/10.5194/essd-2024-39-RC1 - AC3: 'Reply on RC1', Lijun Zheng, 22 May 2024
-
RC2: 'Comment on essd-2024-39', Anonymous Referee #2, 07 May 2024
This study develops a new series of temporal gravity field models HUSTGrace2024, on the basis of a hybrid data processing chain and updated GRACE and GRACE-FO L1B datasets. The hybrid processing chain takes account of the newest atmosphere and ocean de-aliasing product (AOD1B RL07), an improved data weighting method and the updated accelerometer scale factor matrix, which leads to a better temporal gravity field series. Comparison with the current official GRACE and GRACE-FO solutions (released by CSR, GFZ and JPL) demonstrates that the HUSTGrace2024 models have a lower noise level (with a reduction of 10% to 30%) and comparable signal amplitudes in both the long-term trend and seasonal variations.
The datasets developed by the authors are useful to those who investigate the global climate change and geodynamics with satellite gravimetry. Meanwhile, the hybrid data processing chain provides insights into more accurate determination of the global time-variable gravity field models. In particular, the detailed parameter setting suggestions in GRACE and GRACE-FO L1B data processing (such as the thresholds to detect outliers in range-rate observations, and to set the margin for accelerometer datasets, etc.) are beneficial to more reliable gravity field modeling. I would like to recommend minor revisions of the manuscript before publication in Earth System Science Data, according to the comments as follows.
Comments:
1. On the selection of a quiet ocean area to assess the noise level
In Section 3.1.2, the authors select an open ocean area (in the middle Pacific Ocean) to evaluate the noise level of the datasets (Page 21, Lines 4-5; Page 22, Figure 9). However, the location of the selected ocean area (as well as its latitude and longitude ranges) is not described in the text. Please clarify it.
In addition, using only parts of the ocean is probably not the best way to assess the noise level of the GRACE and GRACE-FO temporal gravity field series. I suggest the authors to add the results for the global ocean. To reduce the effect of signal leakage from the land areas, a buffer zone (with 400 or 500 km from the coastlines) can be considered when determining the global ocean coverage. The RMS comparison for the global ocean among the different products would be more convincing to evaluate the noise level.
2. On the RMSs of TWSA residuals to assess the noise level
In Section 3.1.2, the RMSs of TWSA residuals (by removing the yearly trends and annual amplitudes from the original time series) are used to assess the noise level (Page 19, Equation 13). Nevertheless, removing only the yearly trends and annual amplitudes is likely not enough to separate the signal and noise. For example, in the regions of South America (Amazon basin), East Africa (Nile basin) and South Asia (the surroundings of Bengal Bay), there are notable effects in the RMSs (see the row 3 in Figures 6 and 7), which are probably due to the inter-annual changes in the TWSA. Thus it might be better to take into account the inter-annual variations in Equation (13), with a high-pass filter (for instance, with a threshold of 1.5 or 2 years) to mitigate the inter-annual impact.
In addition, the co- and post-seismic changes of several gigantic earthquakes during the GRACE observation period may also affect the RMSs results to some extent. Since the seismic effects are local and not significant in the RMS distribution (Figure 6), adding some discussions to acknowledge this issue should be enough.
Another concern is about the semi-annual changes in the GRACE and GRACE-FO time series. Normally both the annual and semi-annual terms are included to represent the seasonal variations. So I would suggest the authors to consider the semi-annual term in Equation (13).
3. On the assessment of low degree coefficients C20 and C30
In Section 3.1.1, the authors compare the datasets in the spectral domain. From Figure 5 one can find that the low degree coefficients are notably different between the ones from HUST and those from the three official solutions. Therefore, it should be better to add a specific comparison for the low-degree coefficients (e.g., C20 and C30). By comparing the time C20 and C30 time series among the HUST, CSR (GFZ and JPL) and those from the SLR observations (i.e., TN-14; see Loomis et al., 2020), the advantage of the data product in this study may be better demonstrated.
4. On the acknowledgement of the difference in data inputs between HUST and the official solutions
The authors compare the HUST solutions and those from CSR, GFZ and JPL, and find that their solution performs better than the others. However, one should keep in mind that the data inputs are different for the solutions. The HUST solutions use the newest atmosphere and ocean de-aliasing product (AOD1B RL07), whereas the CSR (GFZ and JPL) solutions use the old version of the AOD1B product. In the near future, CSR (GFZ and JPL) will release the GRACE and GRACE-FO RL07 solutions, which are comparable (in the data inputs) to the HUST solutions. Whether the HUST datasets perform better than the RL07 official products remains a question. Please add some discussions to acknowledge this issue.
In addition, CSR has released the RL06.2 solutions for GRACE-FO since September 2023. But the authors do not mention this official product. Please add some necessary statements or discussions.
Corrections on writing details:
(1) Page 1, Line 25: singal -> signal
(2) Page 4, Line 1: the quiet ocean -> a selected open ocean region
For the “quiet ocean” in the rest parts of the manuscript (e.g., Page 21, Lines 4, 10 and 13; Page 22, Line 2; Page 28, Line 3; …), it should be better to change it into “open ocean”.
(3) Page 4, Line 7: background model -> background models
(4) Page 6, Line 5: Importance of renewing -> Renewing of
To be consistent with the title of Section 2.1 (Page 4, Line 7).
(5) Page 16, Lines 20-21: The result indicates that our proposed hybrid weighting approach can obtain the better temporal gravity field solutions.
This statement is not appropriate for the Methods section. Please remove the sentence and consider to make this statement in the Results or Discussion section.
(6) Page 17, Line 2: from 2005 to 2010 -> from 2005 to 2010 for GRACE solutions
(7) Page 17, Line 21: The numerical values “7.528, 9.827, 10.049, and 6.561”, and many other values in the rest parts of the text (e.g., Page 19, Lines 17 and 19; Page 20. Lines 1 and 2; …), as well as the values in the tables (e.g., Tables 3 to 5)
It does not make much practical sense to use 3 decimal places for the numerical values in the results. Using 1 decimal place (or at most 2) should be enough.
(8) Page 19, Line 4: a decorrelation filter
Which decorrelation filter? Please specify it.
(9) Page 21, Line 15: “and HUST-Grace2024” -> “and HUST-Grace2024, respectively”
(10) Page 22, Lines 5-6: the representative deserts and the east of Antarctic
Please describe the latitude and longitude ranges for these selected areas, or show their coverages in a map.
(11) Page 23, Line 13: annual amplitude -> annual amplitudes
(12) Page 26, Line 2: official solution -> official solutions
(13) Page 30, Line 27, and Page 31, Line 1:
The references Cheng & Ries (2017) and Cheng & Ries (2023) are not mentioned in the text.
(14) Page 35, Line 8: “Geo. Jou. Int.”
Please use “Geophys. J. Int.”, or use the complete journal name.
Citation: https://doi.org/10.5194/essd-2024-39-RC2 - AC1: 'Reply on RC2', Lijun Zheng, 22 May 2024
-
RC3: 'Comment on essd-2024-39', Anonymous Referee #3, 07 May 2024
The manuscript provides detailed information about their new version of the time-variable gravity field series using a new data processing strategy and new input data. However, the following are my primary comments and suggestions for major comments to the study;
-
The GRACE result section should incorporate also comparisons with previous versions of HUST-Grace to assess how they correspond with the official GRACE solutions.
- During the GRACE mission, particularly after 2010, more factors emerged, including maneuvers and the GRACE-B battery issue. Hence, it is imperative to incorporate the GRACE time period post-2010 within the results section, as it will serve as a comprehensive testing phase for the new step procedures provided in the study.
- The study's proposed improvement steps should clearly and prominently show the impact of either the new accelerometer calibration or the new AOD1B product. This can be done in the result section, comparing it with the prior version of HUST solutions.
- The metrics used to examine the models in the comparison section were used in a very subset of the significance of the magnitudes. Although the comparison is intended to emphasize that the HUST-Grace2024 model is better in a meaningful way, it may not be meaningful if the cm or mm orders are 3 digits finer after the comma. For example, Page 19, Line 17, RMSs over ocean is in cm, but the notations are well below mm.
Citation: https://doi.org/10.5194/essd-2024-39-RC3 - AC2: 'Reply on RC3', Lijun Zheng, 22 May 2024
-
-
CC2: 'Comment on essd-2024-39', Zhengwen Yan, 10 May 2024
Page 20, L7, 'including (a, e, f) CSR RL06, (b, f, j) GFZ RL06, (e, g, k) JPL RL06 and (d, h, f) HUST-Grace2024' should be corrected as 'including (a, e, i) CSR RL06, (b, f, j) GFZ RL06, (c, g, k) JPL RL06 and (d, h, l) HUST-Grace2024'
Citation: https://doi.org/10.5194/essd-2024-39-CC2 - AC4: 'Reply on CC2', Lijun Zheng, 22 May 2024
Data sets
HUST-Grace2024: GRACE and GRACE Follow-On monthly gravity field solution Hao Zhou et al. https://dataservices.gfz-potsdam.de/panmetaworks/review/d8ed23eb2f84c503cf70494fc827800197693e1100b3a4b8976e9f9a1eb0d323-icgem/
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
530 | 100 | 47 | 677 | 27 | 29 |
- HTML: 530
- PDF: 100
- XML: 47
- Total: 677
- BibTeX: 27
- EndNote: 29
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1