the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 20 Mar 2024
A decomposition approach to evaluating the local performance of global streamflow reanalysis
Abstract. While global streamflow reanalysis provides valuable information for water resources management, its local performance in the time-frequency domain is yet to be investigated. This paper presents a novel decomposition approach to evaluating streamflow reanalysis by combining wavelet transform with machine learning. Specifically, the time series of streamflow reanalysis and observation are respectively decomposed and then the approximation components of reanalysis are compared to those of observed streamflow. Furthermore, the accumulated local effects are derived to showcase the influences of catchment attributes on the performance of raw reanalysis at different scales. For streamflow reanalysis generated by the Global Flood Awareness System, a case study is devised based on streamflow observations from the Catchment Attributes and Meteorology for Large-sample Studies. The results highlight that the reanalysis tends to be more effective in characterizing seasonal, annual and multi-annual features than daily, weekly and monthly features. The Kling-Gupta Efficiency (KGE) values of raw reanalysis and approximation components are primarily influenced by precipitation seasonality. That is, high values of KGE tend to be observed in catchments where there is more precipitation in winter, which can be due to low evaporation that results in reasonable simulations of soil moisture and baseflow processes. The longitude, mean precipitation and mean slope also influence the local performance of approximation components. On the other hand, attributes on geology, soils and vegetation appear to play a relatively small part in the performance of approximation components. Overall, this paper provides useful information for practical applications of global streamflow reanalysis.
- Preprint
(3431 KB) - Metadata XML
- BibTeX
- EndNote
Status: closed
-
RC1: 'Comment on hess-2024-83', Anonymous Referee #1, 13 Apr 2024
This paper presents a novel approach to investigate the performance of GloFAS streamflow reanalysis from the perspective of the time-frequency domain. The results provide interesting insights into the performance of global streamflow reanalysis datasets and attribution analysis. The paper is well-structured and well-written. Below are some comments for consideration.
- Abstract - The significance of evaluating global streamflow reanalysis in the time-frequency is in demand, which helps clarify the contribution of this paper.
- Lines 110 to 120 – Consider adding some diagnostic plots about the clustering results in the supplementary material, as the existence of outliers also indicates the variability of global streamflow reanalysis.
- (11) – Two KGE terms are used in this equation. For clarity, I suggest adding subscripts to differentiate between the observed KGE and the predicted KGE by RF.
- Results – The correspondence between approximation level (e.g., A1, A2) and time scale (e.g., daily, monthly) is mentioned in Lines 255 to 260. To improve readability, consider moving this information to Section 4.1.
- Results - The varying number of stations under investigation in this paper may raise concerns about the robustness of the results. In Line 185, it is demonstrated that there are 661 stations, while 554 stations are used in Figure 4. If the reduction is due to the use of clustering, please include a description in the paper to clarify this point.
- Results – While the Results section is well-written, it would benefit from further interpretations Overall, similar analyses are conducted for both raw reanalysis and the decomposition. Consider adding further illustrations to highlight the added value or new findings that cannot be directly found based on raw data but are derived from the novel approach.
Citation: https://doi.org/10.5194/hess-2024-83-RC1 - AC1: 'Reply on RC1', Tongtiegang Zhao, 29 May 2024
- AC2: 'Reply on RC1', Tongtiegang Zhao, 29 May 2024
- AC3: 'Reply on RC1', Tongtiegang Zhao, 29 May 2024
-
RC2: 'Comment on hess-2024-83', Anonymous Referee #2, 15 May 2024
This study presents an approach to evaluating streamflow reanalysis from a time-frequency domain using wavelet transforms. By applying the wavelet transform to both reanalysis and observation data, the authors decompose the time series into different scales and conduct a performance evaluation for each component. Additionally, they employ random forests combined with accumulated local effects (ALE) to analyze the influence of catchment attributes on reanalysis performance across various time scales. The results offer valuable insights into the understanding of reanalysis data and provide plausible explanations based on catchment characteristics, which are crucial for the practical application of reanalysis data.
Despite the findings, the manuscript requires improvements to better illustrate both the methods and the results part.
- Novelty and Justification: The manuscript does not sufficiently highlight the novelty and importance of the proposed methodology. The authors should provide a clearer rationale for choosing this method over other potential techniques, while the wavelet transform is a powerful tool. Additionally, a more succinct and cohesive summary of the methodology would enhance the manuscript. Specifically, I suggest including a part that clearly outlines the motivation for using the wavelet-based method and the connections among the main steps, including the basic inputs and outputs of each process, with clear symbols and formulas distinguishing between reanalysis and observation data.
- Robustness of Results: Could the authors discuss whether the results hold consistently across different catchment selections for training and testing and provide any relevant sensitivity analyses?
- Terminology Clarification: In Line 124, the authors introduce ALEs and use an abbreviation. Given that ALEs might not be familiar to all readers within the hydrology community, could the authors provide a full name and a clear explanation of when the term was first introduced? Although ALEs are mentioned in the introduction, they should be explicitly defined and elaborated upon to ensure clarity throughout the manuscript.
- Figure 1. The Kling-Gupta Efficiency (KGE) is upper bounded by 1, with higher values indicating better performance. It would be more effective to use a linear color scale to represent this metric, enhancing the interpretability of the figure.
Citation: https://doi.org/10.5194/hess-2024-83-RC2 - AC4: 'Reply on RC2', Tongtiegang Zhao, 29 May 2024
Status: closed
-
RC1: 'Comment on hess-2024-83', Anonymous Referee #1, 13 Apr 2024
This paper presents a novel approach to investigate the performance of GloFAS streamflow reanalysis from the perspective of the time-frequency domain. The results provide interesting insights into the performance of global streamflow reanalysis datasets and attribution analysis. The paper is well-structured and well-written. Below are some comments for consideration.
- Abstract - The significance of evaluating global streamflow reanalysis in the time-frequency is in demand, which helps clarify the contribution of this paper.
- Lines 110 to 120 – Consider adding some diagnostic plots about the clustering results in the supplementary material, as the existence of outliers also indicates the variability of global streamflow reanalysis.
- (11) – Two KGE terms are used in this equation. For clarity, I suggest adding subscripts to differentiate between the observed KGE and the predicted KGE by RF.
- Results – The correspondence between approximation level (e.g., A1, A2) and time scale (e.g., daily, monthly) is mentioned in Lines 255 to 260. To improve readability, consider moving this information to Section 4.1.
- Results - The varying number of stations under investigation in this paper may raise concerns about the robustness of the results. In Line 185, it is demonstrated that there are 661 stations, while 554 stations are used in Figure 4. If the reduction is due to the use of clustering, please include a description in the paper to clarify this point.
- Results – While the Results section is well-written, it would benefit from further interpretations Overall, similar analyses are conducted for both raw reanalysis and the decomposition. Consider adding further illustrations to highlight the added value or new findings that cannot be directly found based on raw data but are derived from the novel approach.
Citation: https://doi.org/10.5194/hess-2024-83-RC1 - AC1: 'Reply on RC1', Tongtiegang Zhao, 29 May 2024
- AC2: 'Reply on RC1', Tongtiegang Zhao, 29 May 2024
- AC3: 'Reply on RC1', Tongtiegang Zhao, 29 May 2024
-
RC2: 'Comment on hess-2024-83', Anonymous Referee #2, 15 May 2024
This study presents an approach to evaluating streamflow reanalysis from a time-frequency domain using wavelet transforms. By applying the wavelet transform to both reanalysis and observation data, the authors decompose the time series into different scales and conduct a performance evaluation for each component. Additionally, they employ random forests combined with accumulated local effects (ALE) to analyze the influence of catchment attributes on reanalysis performance across various time scales. The results offer valuable insights into the understanding of reanalysis data and provide plausible explanations based on catchment characteristics, which are crucial for the practical application of reanalysis data.
Despite the findings, the manuscript requires improvements to better illustrate both the methods and the results part.
- Novelty and Justification: The manuscript does not sufficiently highlight the novelty and importance of the proposed methodology. The authors should provide a clearer rationale for choosing this method over other potential techniques, while the wavelet transform is a powerful tool. Additionally, a more succinct and cohesive summary of the methodology would enhance the manuscript. Specifically, I suggest including a part that clearly outlines the motivation for using the wavelet-based method and the connections among the main steps, including the basic inputs and outputs of each process, with clear symbols and formulas distinguishing between reanalysis and observation data.
- Robustness of Results: Could the authors discuss whether the results hold consistently across different catchment selections for training and testing and provide any relevant sensitivity analyses?
- Terminology Clarification: In Line 124, the authors introduce ALEs and use an abbreviation. Given that ALEs might not be familiar to all readers within the hydrology community, could the authors provide a full name and a clear explanation of when the term was first introduced? Although ALEs are mentioned in the introduction, they should be explicitly defined and elaborated upon to ensure clarity throughout the manuscript.
- Figure 1. The Kling-Gupta Efficiency (KGE) is upper bounded by 1, with higher values indicating better performance. It would be more effective to use a linear color scale to represent this metric, enhancing the interpretability of the figure.
Citation: https://doi.org/10.5194/hess-2024-83-RC2 - AC4: 'Reply on RC2', Tongtiegang Zhao, 29 May 2024
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 388 | 80 | 29 | 497 | 21 | 21 |
- HTML: 388
- PDF: 80
- XML: 29
- Total: 497
- BibTeX: 21
- EndNote: 21
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1