the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 14 Nov 2024
Intercomparison of bias correction methods for precipitation of multiple GCMs across six continents
Abstract. This study, conducted across six continents, evaluated and compared the effectiveness of three Quantile Mapping (QM) methods: Quantile Delta Mapping (QDM), Empirical Quantile Mapping (EQM), and Detrended Quantile Mapping (DQM) for correcting daily precipitation data from 11 CMIP6 General Circulation Models (GCMs). The performance of corrected precipitation data was evaluated using ten evaluation metrics, and the Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) was applied to calculate performance-based priorities. Bayesian Model Averaging (BMA) was used to quantify model-specific and ensemble prediction uncertainties. Subsequently, this study developed a comprehensive index by aggregating the performance scores from TOPSIS with the uncertainty metrics from BMA. The results showed that EQM performed the best on all continents, effectively managing performance and uncertainty. QDM outperformed other methods in specific regions and was selected more frequently than DQM when greater weight was given to uncertainty. It suggests that daily precipitation corrected by QDM is more stable than DQM. On the other hand, DQM effectively reproduces dry climate but shows the highest uncertainty in certain regions, suggesting potential limitations in capturing long-term climate trends. This study emphasizes that both performance and uncertainty should be considered when choosing a bias correction method to increase the reliability of climate predictions.
- Preprint
(7142 KB) - Metadata XML
-
Supplement
(3534 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
CEC1: 'Comment on gmd-2024-177', Juan Antonio Añel, 08 Dec 2024
Dear authors,
Unfortunately, after checking your manuscript, it has come to our attention that it does not comply with our "Code and Data Policy".
https://www.geoscientific-model-development.net/policies/code_and_data_policy.htmlYour "Code and Data Availability" statement does not contain the link to a permanent repository with the code and data used to produce your manuscript. I am sorry to have to be so outspoken, but this is something completely unacceptable, forbidden by our policy, and your manuscript should have never been accepted for Discussions given such flagrant violation of the policy. All the code and data must be published openly and freely to anyone in one of the repositories listed in our policy before submission of a manuscript.
Therefore, we are granting you a short time to solve this situation. You have to reply to this comment in a prompt manner with the information for the repositories containing all the models, code and data that you use to produce and replicate your manuscript. The reply must include the link and permanent identifier (e.g. DOI). Also, any future version of your manuscript must include the modified section with the new information.
Note that if you do not fix these problems as requested, we will have to reject your manuscript for publication in our journal.
Juan A. Añel
Geosci. Model Dev. Executive Editor
Citation: https://doi.org/10.5194/gmd-2024-177-CEC1 -
CC1: 'Reply on CEC1', Young Hoon Song, 08 Dec 2024
Dear Chief editor
We have identified an oversight on our part regarding the absence of a DOI for data availability. Accordingly, we have revised the Code and Data Availability section as follows:
Code availability
Codes for benchmarking the xclim of python package are available from https://xclim.readthedocs.io/en/stable/. Furthermore, the Comprehensive Index proposed in this study, along with the performance and uncertainty indices used within it, is available at https://doi.org/10.6084/m9.figshare.27987665.v2 (Song, 2024).
Data availability
The data used in this study are publicly available from multiple sources. CMIP6 General Circulation Models (GCMs) outputs were obtained from the Earth System Grid Federation (ESGF) data portal at https://esgf-node.llnl.gov/search/cmip6/. Users can select data types such as climate variables, time series, and experiment ID, which can be downloaded as NC files. Furthermore, CMIP6 GCMs output can also be accessed in Eyring et al. (2016) The ERA5 reanalysis dataset used in this study is available through the Copernicus Data Store (CDS) provided by ECMWF (https://cds.climate.copernicus.eu/cdsapp#!/dataset/reanalysis-era5-single-levels?tab=overview). ERA5 is available at https://doi.org/10.24381/cds.bd0915c6 (Hersbach et al., 2023).
Additionally, the newly added DOI in the references is as follows:
Eyring, V., Bony, S., Meehl, G., Senior, C., Stevens, B., Stouffer, R., and Taylor, K.: Overview of the Coupled Model Intercomparison Project Phase 6 (CMIP6) experimental design and organization [Dataset]. Geoscientific Model Development, 9(5), 1937–1958. 2016. https://doi.org/10.5194/gmd-9-1937-2016
Hersbach, H., Bell, B., Berrisford, P., Biavati, G., Horányi, A., Muñoz Sabater, J., Nicolas, J., Peubey, C., Radu, R., Rozum, I., Schepers, D., Simmons, A., Soci, C., Dee, D., and Thépaut, J.-N.: ERA5 hourly data on pressure levels from 1940 to present, Copernicus Climate Change Service (C3S) Climate Data Store (CDS) [Data set], https://doi.org/10.24381/cds.bd0915c6, 2023.
Song, Y.H.: Comprehensive Index and Performance-Related Code, Figshare [Code], https://doi.org/10.6084/m9.figshare.27987665.v2
Additionally, I am attaching the revised article. The modified sections are marked in red text. Thank you.
-
CEC2: 'Reply on CC1 - compliance with policy no solved', Juan Antonio Añel, 10 Dec 2024
Dear authors,
I'm sorry, but your reply does not address the issues I have pointed out. You should read more carefully our policy. It clearly states which are the acceptable repositories to publish assets. The readthedocs.io is not acceptable (and moreover it only contains documentation, not the code). Also, generic data portals for reanalyses or datasets such as those from the CMIP6 do not provide the exact data that you use in your manuscript, that is what you have to publish. Also, papers could have DOIs, but they are not data or code repositories.
I ask you to read again our policy, and follow the rules listed there, and store your assets in adequate repositories to comply with our policy.
To be clear, currently, your submission continues to be non-compliant with our policy.
Juan A. Añel
Geosci. Model Dev. Executive Editor
Citation: https://doi.org/10.5194/gmd-2024-177-CEC2 -
CEC3: 'Reply on CEC2 - no compliance with the code policy -', Juan Antonio Añel, 10 Jan 2025
Dear authors,
I have to note that you have failed to address the problems that we have pointed out regarding the compliance with the code and data policy of the journal. It is this way despite you have addressed comments from reviewers in Discussion, which means that you have actively avoided to do it. Therefore, with this comment I request the Topical Editor to reject your manuscript for publication in Geosci. Model Dev. because of no compliance with the policy of the journal. If the editor considers it and you show interest on solving the situation, at minimum all the review process must be stalled until this issue is solved.
Juan A. Añel
Geosci. Model Dev. Executive Editor
Citation: https://doi.org/10.5194/gmd-2024-177-CEC3 -
CC7: 'Reply on CEC3', Young Hoon Song, 10 Jan 2025
Dear. Juan A. Añel,
We hope this message finds you well. We are writing to seek clarification regarding the reasons for the rejection of our manuscript due to non-compliance with the journal's "Code and Data Policy." While we have attempted to address the concerns raised in the review process, we are unable to pinpoint the exact issue that remains unresolved and would greatly appreciate your guidance.
In our revised manuscript, we have made the following modifications to the "Code and Data Availability" section to align with the journal’s policy:
Code and data availability
Codes for benchmarking the xclim of python package are available from https://doi.org/10.5281/zenodo.10685050 (Bourgault et al., 2024). Furthermore, the CI proposed in this study, along with the TOPSIS and BMA used within it, is available at https://doi.org/ 10.5281/zenodo.14351816 (Song, 2024b). The data used in this study are publicly available from multiple sources. CMIP6 General Circulation Models (GCMs) outputs were obtained from the Earth System Grid Federation (ESGF) data portal at https://esgf-node.llnl.gov/search/cmip6/. Users can select data types such as climate variables, time series, and experiment ID, which can be downloaded as NC files. Furthermore, CMIP6 GCMs output can also be accessed in Eyring et al. (2016) The ERA5 reanalysis dataset used in this study is available through the Copernicus Data Store (CDS) provided by ECMWF (https://cds.climate.copernicus.eu/cdsapp#!/dataset/reanalysis-era5-single-levels?tab=overview). ERA5 is available at https://doi.org/10.24381/cds.bd0915c6 (Hersbach et al., 2023). The daily precipitation datasets from CMIP6 GCM and ERA5 used in this study are available at https://doi.org/10.6084/m9.figshare.27999167.v5 (Song, 2024c).
Despite these changes, we are unclear on the exact issues that remain unresolved. Specifically:
Is there an issue with how we have documented the availability of processed datasets (e.g., precipitation data derived from ERA5 and CMIP6 GCMs)?
Are there additional requirements regarding the storage and accessibility of the code used for processing and analyzing these datasets?
Does the inclusion of references to repositories like Figshare fail to meet the journal’s standards for transparency and reproducibility?
We would greatly appreciate it if you could provide detailed feedback or specify the points of non-compliance. This will allow us to address any remaining issues promptly and ensure full adherence to the journal’s policy.
We are committed to resolving this matter and making the necessary revisions to comply with the journal’s requirements. Your guidance in identifying the specific shortcomings would be invaluable. Thank you for your time and consideration. We look forward to your reply.Best regards,
Young Hoon Song -
CC8: 'Reply on CEC3', Young Hoon Song, 14 Jan 2025
Dear Editor,
While reviewing our manuscript once again, we discovered that the Zenodo URL where the code was uploaded is returning a 404 error. Therefore, we have updated the "Code and Data Availability" section as follows.
We sincerely apologize for the inconvenience caused and assure you that we will take a proactive approach to address this issue. Thank you for bringing this matter to our attention so we could resolve it.
Best regards,
Younghun Song -
CC9: 'Reply on CEC3', Young Hoon Song, 14 Jan 2025
Dear Editor,
The revised text is as follows. Please refer to it at your convenience.
Code and Data Availability
Codes for benchmarking the xclim (Version 0.48.1) Python package are available at https://doi.org/10.5281/zenodo.10685050 (Bourgault et al., 2024). Furthermore, the CI proposed in this study, along with the TOPSIS and BMA methods used within it, is available at https://doi.org/10.5281/zenodo.14351816 (Song, 2024b).The data used in this study are publicly available from multiple sources. CMIP6 General Circulation Models (GCMs) outputs were obtained from the Earth System Grid Federation (ESGF) data portal at https://esgf-node.llnl.gov/search/cmip6/. Users can select data types such as climate variables, time series, and experiment IDs, which can be downloaded as NC files. Furthermore, CMIP6 GCM outputs can also be accessed as described in Eyring et al. (2016).
The ERA5 reanalysis dataset used in this study is available through the Copernicus Data Store (CDS) provided by ECMWF (https://cds.climate.copernicus.eu/cdsapp#!/dataset/reanalysis-era5-single-levels?tab=overview). ERA5 is available at https://doi.org/10.24381/cds.bd0915c6 (Hersbach et al., 2023).
The daily precipitation datasets from CMIP6 GCMs and ERA5 used in this study are available at https://doi.org/10.6084/m9.figshare.27999167.v5 (Song, 2024c).
Best regards,
Younghun SongCitation: https://doi.org/10.5194/gmd-2024-177-CC9
-
CC7: 'Reply on CEC3', Young Hoon Song, 10 Jan 2025
-
CEC3: 'Reply on CEC2 - no compliance with the code policy -', Juan Antonio Añel, 10 Jan 2025
-
CEC2: 'Reply on CC1 - compliance with policy no solved', Juan Antonio Añel, 10 Dec 2024
-
RC1: 'Comment on gmd-2024-177', Anonymous Referee #1, 08 Dec 2024
This study proposes a Comprehensive Index that considers both performance and uncertainty, using 11 CMIP6 GCMs and compares bias correction methods across six continents. Such an approach can be evaluated as a highly significant method for GCMs and bias correction, which are essential for future predictions. The reviewer, therefore, would like to recommend that this manuscript be returned to the authors for major revisions. My comments are given below;
The authors need to address several key questions:
- Why was only the historical period used? If there is a specific reason, it must be clearly stated.
- Developing a comprehensive index is a significant strength of this study. However, there is insufficient explanation of the differences from previous methods to highlight its originality. Please address this. Additionally, the comprehensive index appears to be overly limited. Such a restrictive methodology may reduce the efficiency of the approach, which warrants further discussion.
- DQM performs worse overall compared to other methods. The reasons for this must be explained. However, as QM methods are not newly proposed in this study but adopted from existing ones, it is unnecessary to deeply analyze the cause (this could be a topic for a future study). A brief explanation in the context of specific climate phenomena should suffice.
- The authors used entropy as a criterion weight for TOPSIS. Is there a reason for this choice? Since MCDA is widely used in other fields, incorporating more complex weights might be more appropriate.
- Model performance can vary significantly depending on the climate zone, but in most cases, the performance is reported to be very high (e.g., EVS: 0.98 (DQM)). Such results indicate exceptional performance. This should be mentioned in the Discussion section.
- Why was only ERA5 selected for comparison? It may not pose a significant issue without a valid reason, but considering multiple reference datasets is essential for evaluating performance.
- Why does DQM yield higher estimates for extreme precipitation? Could this be because DQM emphasizes variability in the data due to detrending?
This study provides valuable insights into climate change research by comparing bias correction methods and incorporating an integrated uncertainty evaluation. Therefore, sincerely addressing the above comments will significantly enhance the scientific value of this study. Additionally, the findings are anticipated to be actively applicable to new stages of GCMs, or models developed after CMIP7.
https://onlinelibrary.wiley.com/doi/10.1029/2021EF002240
https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2022EF002830
Citation: https://doi.org/10.5194/gmd-2024-177-RC1 -
CC4: 'Reply on RC1', Young Hoon Song, 10 Jan 2025
Dear Ullah,
Thank you for showing great interest in our research. Your insightful comments have further enhanced our study, and we sincerely appreciate them once again. Please find attached our response to your comments and the revised article.-
CC5: 'Reply on CC4', Young Hoon Song, 10 Jan 2025
Dear Dr. Irfan Ullah,
We sincerely apologize for omitting significant details in our earlier communication. To address this, we have revised and rewritten the content for your reference. We regret any inconvenience this may have caused. Thank you for your deep interest in our research.Attached are our responses to your comments and the revised article.
Best regards,
Younghoon Song
-
CC5: 'Reply on CC4', Young Hoon Song, 10 Jan 2025
-
CC2: 'Reply on CEC2', Young Hoon Song, 11 Dec 2024
Dear Editor,
We have actively utilized Zenodo to address your feedback. For the code, we were able to provide example datasets on Zenodo due to their small size. However, the datasets used in this study, such as CMIP6 GCM and ERA5, exceed 50 GB in size. Therefore, we utilized Springer Nature's Figshare to make these datasets available. Additionally, we have updated the version of xclim we used to reflect it on Zenodo. Accordingly, we have revised the Code and Data Availability sections as follows.
Code availability
Codes for benchmarking the xclim of python package are available from https://doi.org/10.5281/zenodo.10685050 (Bourgault et al., 2024). Furthermore, the CI proposed in this study, along with the TOPSIS and BMA used within it, is available at https://doi.org/ 10.5281/zenodo.14351816 (Song, 2024b).Data availability
The data used in this study are publicly available from multiple sources. CMIP6 General Circulation Models (GCMs) outputs were obtained from the Earth System Grid Federation (ESGF) data portal at https://esgf-node.llnl.gov/search/cmip6/. Users can select data types such as climate variables, time series, and experiment ID, which can be downloaded as NC files. Furthermore, CMIP6 GCMs output can also be accessed in Eyring et al. (2016) The ERA5 reanalysis dataset used in this study is available through the Copernicus Data Store (CDS) provided by ECMWF (https://cds.climate.copernicus.eu/cdsapp#!/dataset/reanalysis-era5-single-levels?tab=overview). ERA5 is available at https://doi.org/10.24381/cds.bd0915c6 (Hersbach et al., 2023). The daily precipitation datasets from CMIP6 GCM and ERA5 used in this study are available at https://doi.org/10.6084/m9.figshare.27999167.v5 (Song, 2024c).The additional and revised references are as follows:
Bourgault, P., Huard, D., Smith, T.J., Logan, T., Aoun, A., Lavoie, J., Dupuis, É., Rondeau-Genesse, G., Alegre, R., Barnes, C., Beaupré Laperrière, A., Biner, S., Caron, D., Ehbrecht, C., Fyke, J., Keel, T., Labonté, M.P., Lierhammer, L., Low, J.F., Quinn, J., Roy, P., Squire, D., Stephens, Ag., Tanguy, M., Whelan, C., Braun, M., Castro, D.: xclim: xarray-based climate data analytics (0.48.1). Zenodo [Code], https://doi.org/10.5281/zenodo.10685050, 2024.
Eyring, V., Bony, S., Meehl, G., Senior, C., Stevens, B., Stouffer, R., and Taylor, K.: Overview of the Coupled Model Intercomparison Project Phase 6 (CMIP6) experimental design and organization. Geoscientific Model Development, 9(5), 1937–1958. 2016. https://doi.org/10.5194/gmd-9-1937-2016
Hersbach, H., Bell, B., Berrisford, P., Biavati, G., Horányi, A., Muñoz Sabater, J., Nicolas, J., Peubey, C., Radu, R., Rozum, I., Schepers, D., Simmons, A., Soci, C., Dee, D., and Thépaut, J.-N.: ERA5 hourly data on pressure levels from 1940 to present, Copernicus Climate Change Service (C3S) Climate Data Store (CDS), https://doi.org/10.24381/cds.bd0915c6, 2023.
Song, Y.H.: Comprehensive Index and Performance-Related Code, Zenodo [Code], https://zenodo.org/records/14351816. 2024b
Song, Y.H.: Historical Daily Precipitation Data of CMIP6 GCMs and ERA5, Figshare [Dataset], https://doi.org/10.6084/m9.figshare.27999167.v5. 2024c
-
CC3: 'Comment on gmd-2024-177', Shamsuddin Shahid, 13 Dec 2024
This study presents an innovative approach by systematically comparing the performance and uncertainty of Quantile Mapping (QM) methods across six continents using 11 CMIP6 GCMs. Specifically, it makes an academically significant contribution by quantitatively evaluating the performance of three major QM methods—QDM, EQM, and DQM—and developing a Comprehensive Index (CI) that integrates performance and uncertainty using TOPSIS and Bayesian Model Averaging (BMA).
In climate change research, bias correction of GCMs is a critical step in determining the reliability of climate projections and impact assessments. While previous bias correction studies have often been limited to single regions or small datasets, this study attempts to compare bias correction methodologies on a global scale, considering continental differences and regional characteristics. Additionally, the development of CI, which integrates uncertainty quantification and performance evaluation, is expected to improve the reliability of climate models and provide substantial contributions to the formulation of adaptation and mitigation policies for climate change.
I have the following major questions:
Why did this study only use the historical period? For example, do the authors believe that bias correction performance during historical periods sufficiently accounts for uncertainty in future projections?
What are the strengths of the Comprehensive Index? The study claims to have developed this index, but could you provide a detailed explanation of its strengths and how it differs from previous methods?
Why were TOPSIS and BMA specifically used for performance and uncertainty in the Comprehensive Index? Could other methods have been equally applicable? Please discuss whether alternatives might be feasible.
Why does DQM perform worse than other methods? Could the authors explain why this method shows lower performance compared to QDM and EQM?
Why was ERA5 chosen as the reference dataset without comparing it to other reanalysis datasets (e.g., CHIRPS, GPCP)? What criteria led the authors to determine ERA5 as the most suitable for the study’s regions and precipitation characteristics?
While unifying data resolution to 1°x1° is advantageous, it might obscure detailed regional characteristics that can be captured with higher resolutions. Why was this resolution chosen, and do the authors believe that bias correction performance could differ at higher resolutions?
Entropy theory was clearly used to determine the weights, but could the authors explain the impact and significance of low weights for certain metrics (e.g., EVS, NSE)?EQM showed the lowest uncertainty. Is this result due to the characteristics of the methodology itself, or is it influenced by specific regional or data traits?
Each QM method shows strengths and weaknesses in specific regions. Was there an attempt to develop a hybrid approach that combines these methods? For instance, could DQM’s ability to remove long-term trends and EQM’s stable performance be integrated?
The authors claim that CI considers both performance and uncertainty. However, there is insufficient explanation on whether region-specific weights were applied. Was CI calculated for each grid? For example, regions with extreme precipitation distributions (e.g., the Sahara Desert) differ significantly from moderate regions (e.g., Northern Europe). If CI was calculated at a larger scale, would that be appropriate? Alternatively, is it reasonable to apply uniform weights?
Why was TOPSIS chosen over other MCDA techniques like AHP (Analytic Hierarchy Process) or VIKOR? A discussion on this decision is necessary.
Lastly, some sentences could benefit from structural refinement. To enhance the clarity and flow of the text, it would be helpful to revise the sentences throughout the manuscript.Citation: https://doi.org/10.5194/gmd-2024-177-CC3 -
CC6: 'Reply on CC3', Young Hoon Song, 10 Jan 2025
Dear Dr. Shamsuddin Shahid,
Thank you for your deep interest in our research. Your insightful comments have greatly contributed to the advancement of our study, and we sincerely appreciate them once again. Attached are our responses to your comments and the revised article.
Best regards,
Younghoon Song
-
CC6: 'Reply on CC3', Young Hoon Song, 10 Jan 2025
-
CC10: 'Comment on gmd-2024-177', Brian Ayugi, 01 Apr 2025
This work is interesting and relevant to the diverse regions, especially for Africa, which faces a dearth of datasets for reliable estimates. I have provided a few suggestions that could help the authors improve the quality of the manuscript. Thank you.
- CC11: 'Reply on CC10', Young Hoon Song, 08 Apr 2025
-
RC2: 'Comment on gmd-2024-177', Anonymous Referee #2, 15 May 2025
This paper uses historical CMIP6 simulations from 11 climate models to intercompare different forms of quantile mapping bias correction. Mean statistics of daily precipitation values are the focus of the effort, though there is a short discussion of extreme daily precipitation. The novel part of the paper appears to be the joint assessment of bias correction using uncertainty (via BMA-derived weights among GCMs) and performance (quantified using a closeness index), combining them into a single comprehensive index. The paper could be much shorter to make its point. Far too much time is spent on individual results with mostly qualitative interpretations that do not adequately support the conclusions. A more streamlined paper focused on demonstrating the methods and showing how they can inform decision-making would be a valuable contribution.
Specific comments:
1. Abstract, lines 19-26, the results should focus not on how these three methods ranked based on daily precipitation but on what this application revealed about the method. How sensitive is the ranking to the selection of evaluation metrics? To GCM selection? To weighting of uncertainty vs. performance? To different climatological regions (as mentioned briefly at Lines 666 and 722)? Three QM methods is too small a pool to draw useful conclusions, and the vague language leaves too much unanswered (“QDM outperformed other methods in specific regions…” and “DQM…shows the highest uncertainty in certain regions”).
2. Lines 142-144, the three QM methods are outlined. Were these selected because they represent very different approaches to bias correction? Many other bias correction methods exist, from simple delta change to multivariate. Would a broader selection of methods test your ranking methods more robustly?
3. Line 148-149, the “frequency-adaptation technique” is applied to address potential biases. I did not see this explained further in the methods section. Did this add something beyond what is described by the definitions of the QM methods?
4. Table 2 lists 10 evaluation metrics (though the table caption and line 193 state “seven”). It would help to add columns for the range of values each can take, and what value would indicate a ‘perfect’ fit. An obvious question is why so many somewhat redundant metrics are used (for example RMSE, MAE, MdAE). Could the method be employed with 2 or 3 metrics and perform as well? That would be a useful detail to explore.
5. For clarity, variables should not be used for different quantities. For example, α is used in equation 7 as the scale factor in the GEV distribution, αw is in equation 10 (and doesn’t appear to be defined), and α is used in equation 16 as a performance weight.
6. Figure 1, Since all points in all panels are in the first quadrant, just include that in the figure. That will help readers see the individual points better. That is a pretty standard way to present Taylor diagrams. Also, I would assume this is for the validation period (1997-2014) – it should be noted in the caption.
7. Line 319 (the beginning of section 3.1.2). From here to line 421 offer very little to the aim of the paper. The interpretation of the figures is all qualitative (two examples: for South America, Fig 2 “EQM demonstrated lower JSD values, as well as higher EVS and KGE values, compared to other methods.” and “QDM and DQM also performed well but exhibited slightly larger errors in some regions than EQM.”). First, statistical tests are needed to determine if any differences are statistically significant. Second, a discussion of whether the differences are physically meaningful is needed, such as EVS varying from 0.95 to 0.98 across the region and across methods. Again, if the effort is to rank QM methods, this is far from adequate. If the aim of the study is to demonstrate the application of the method and its sensitivity to methodological choices, then most of this section is not needed.
8. Following on comment 7, Figures 2-7 show the results for different metrics for different continents (again, presumably for the validation period, but that should be clearly stated in figure captions). Figures 2-7 all suffer from the same issues, but I will mostly focus on Figure 2. The color bars are non-linear – while each color segment is the same length, the interval they represent varies widely. For example, the row for Pbias the yellow represents a range of less than 2% while the purple represents over 20%, so a 3% negative bias is indistinguishable from a 20% negative bias. The scales themselves are confusing: for JSD the red colors are the worst skill, while for EVS red is the best skill. Every row is different in this regard. Some indices are unitless and others have units, and that should be represented. Some rows show a wide range of values (Pbias in Fig 2) while others show virtually identical values (NSE in Fig 3, where the colors vary only from 0.98 to 1.0), so where some differences are shown they may essentially be all the same value.
9. The subset of indices shown in Figures 2-7 varies. While the supplemental material may complete the set, it should be consistent.
10. Figure 8 summarizes on a continental scale the performance using each metric. Without any statistical test, the apparent differences cannot be claimed to represent anything. Also, with wide variability (in some metrics) across each continent, a single continent-wide average may not be very meaningful.
11. Section 3.1.3 looks at extreme precipitation. This is not well integrated into the rest of the paper, and only includes a cursory look at continent-aggregated values. It does not fit into any of the rest of Section 3. Line 445 claims differences are “relatively significant” and that distributions “vary significantly”. Since there is no mention of statistical tests, these terms are inappropriate.
12. Figure 9, Why do the scales between the top and bottom rows vary so widely (max of 0.15 vs 0.00024)? Does this figure only represent the 95th percentile like Fig 10?
13. Figure 10, what are the values and units for the x-axis? Again, while these lines appear to be different the densities are extremely close (the y-axis scales cover a very small range) and no statistical test results are presented, so drawing conclusions is limited.
14. Line 464, the weights are “calculated by applying entropy theory”, but that does not appear to have been discussed in the methods section of the paper. This should be mentioned there with appropriate citations.
15. Line 464, While this seems like the more important part of the effort, higher and lower weights are only discussed qualitatively, and while “significant importance” is mentioned (Line 471) and differences are claimed to be “significant” (Line 499) no assessment of significance is shown.
16. Figs. 12, 14, and 16 have many of the same issues as Figure 2-7, noted in comment 8 above.
17. Lines 557-563 discuss Fig. 15. It is claimed that “the EQM ensemble showed the lowest standard deviation across all continents.” The box and whisker plots clearly shows nearly identical results for all methods, and I would be shocked if any of the differences showed any statistical significance. The conclusions here are just not supported.
18. Fig 17 shows the final comprehensive index across all continents. That some apparent differences emerge is due to weighting the uncertainty (as in Fig 15) less. Statistical tests are needed here too, and then it could be explored why the closeness index changes the results so strongly.Minor comments and typos:
1. Line 42, rather than saying the bias corrections differ in “physical approaches”, “statistical approaches” would be more accurate.
2. Line 47, Correct spelling of Maraun.
3. Line 61, Elaborate on “higher performance.” Is this the same as “better skill”?
4. Line 136, the model resolution “was provided by the institution for research availability.” This is confusing. Is there a citation to add?
5. Line 212, “JSD” is used, where prior to this JS-D is used. The term should be consistent throughout.
6. Line 361, The sentence begins with “In this study, the daily precipitation in Africa was corrected using three QM methods.” This sort of recap appears at the beginning of many sub-sections, and is not needed. Search for “This study,” and rephrase.
7. Line 438-439, ”adjusted by the biased bias correction methods” must be an errorCitation: https://doi.org/10.5194/gmd-2024-177-RC2
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 411 | 136 | 54 | 601 | 37 | 17 | 15 |
- HTML: 411
- PDF: 136
- XML: 54
- Total: 601
- Supplement: 37
- BibTeX: 17
- EndNote: 15
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
