the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Generalized spatiotemporally-decoupled framework for reconstructing the source of non-constant atmospheric radionuclide releases
Abstract. Determining the source location and release rate are critical in assessing the environmental consequences of atmospheric radionuclide releases, but remain challenging because of the huge multi-dimensional solution space. We propose a generalized spatiotemporally-decoupled two-step framework to reduce the dimension of the solution space in each step and improves the reconstruction accuracy, which is applicable to non-constant releases. The decoupling process is conducted by applying a temporal sliding-window average filter to the observations, thereby reducing the influence of temporal variations in the release rate and ensuring that the features of the filtered data are dominated by the source location. A machine learning model is trained to link these features to the source location, enabling independent source localization. Then the release rate is determined using projected alternating minimization with the L1-norm and total variation regularization algorithm. Validation using SCK-CEN 41Ar experimental data demonstrates that the localization error is less than 1 %, and the temporal variations, peak release rate, and total release are reconstructed accurately. The proposed method exhibits higher accuracy and a smaller uncertainty range than the correlation-based and Bayesian methods. Furthermore, it achieves stable performance with different hyperparameters and produces low error levels even with only a single observation site.
- Preprint
(2505 KB) - Metadata XML
-
Supplement
(1276 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on gmd-2023-173', Anonymous Referee #1, 25 Dec 2023
The paper presented a source reconstructing procedure by first locating the source location before estimating the emission rates. A machine learning method has been used in the first step to locate the source location. The overall results are quite interesting and encouraging.
However, there are several shortcomings in this manuscript. Some of the statements are not accurate and some terminology uses are also questionable. The presentation of the machine learning method is not easy to follow for those who are not quite familiar with the same method and software. In addition, the dispersion model errors affect the results but are not sufficiently considered or discussed. It is also a concern that the method is only tested with a single set of experimental data. More test cases are probably needed.
Major points:
Title: Both “Generalized” and “Spatiotemporally–decoupled” are not accurately reflecting the current two-step method. The word “non-constant” in the title does not sound appropriate either. In reality, there are rarely constant releases. The author should reconsider the title.
Abstract, lines 12-14: This statement is not accurate. The temporal variation of the release rates may be reflected on the plume shape, not only on the temporal variations of the observations. In theory, some problems cannot be decoupled. So the proposed method cannot be a real general framework. The limitation of the method has to be pointed out in the paper.
Abstract, line 15: Locating a source location is not “localization”. This needs to be corrected throughout the paper.
Abstract, line 18: A relative error of about 50% for the Oct. 4 total release is probably not deemed “accurate”. It is better to present the results more objectively with the actual number listed in Table 3.
Line 94: The authors seem to suggest that the correlation-based method only applies when constant-release assumption is made. This is not accurate. Constant release is only one assumption that reduces the complexity of the problem. If the release starting time or duration is not known. Such assumption may not be enough to guarantee a unique solution of the source location. On the other hand, if a source is not constant, but the release time period and temporal profile are known, it is probably easy to get the source location even without the constant-release assumption.
Line 108: It is wrong to assume a square matrix. The dimensions of the observation and source vectors are independent and rarely the same.
Lines 122-124: The statement is not correct. The emissions combined with the meteorological conditions together determine the concentrations at any given measurement site, including the peak values and its timing.
Line 228: It is very confusing to use “sample” for the different candidate source locations.
Lines 287-288: The hyper-parameters used in the 50 runs should be given in the supplementary document.
Figure S3: What is the sliding window applied here? It does seem to be a sided window rather than centered one. Please explain this in the paper.
Minor points:
Line 46, T3-10: Please explain what T3-10 distributions are.
Line 60: What does “deterministic assumption” mean? It is quite confusing.
Figure 1: What do the different shapes and colors in the diagram mean?
Equation (7): Please explain all the parameters here.
Line 160: Why is the amplitude quantity called “wave rate”?
Lines 160-161: The median value is not a central moment.
Line 232: If it is 40th percentile, the number of samples for Oct.3 and Oct. 4 should be 1200 and 600.
Line 237: The authors probably mean 80th, 60th, 50th, 40th, 20th, and 0th.
Line 238: “A lower percentile” should be “a higher percentile”.
Figure 8: No shade appears for the Bayesian inversion results in the lower left panel.
Line 416: What do the various pre-screening ranges refer to?
Figure S1: Should it be 20% instead of 10% for the five-fold cross-validation?
Table S1: Brief descriptions of the hyperparamters should be provided.
Citation: https://doi.org/10.5194/gmd-2023-173-RC1 -
AC1: 'Reply on RC1', Yuhan Xu, 23 Jan 2024
Dear Referee,
Thank you for taking the time to provide such a constructive and thorough review of our manuscript (GMD-2023-173). According to the suggestive comments, we have made some modifications to the manuscript, and the comments have been uploaded in the form of a supplement file.We are looking forward to your reply.
Best regards,
Yours sincerely
Sheng Fang
-
AC1: 'Reply on RC1', Yuhan Xu, 23 Jan 2024
-
RC2: 'Comment on gmd-2023-173', Anonymous Referee #2, 05 Feb 2024
General comments A new approach to source reconstruction of atmospheric radionuclide emissions in non-stationary emission scenarios is proposed in this study. By relaxing the unrealistic assumption of constant emissions and developing a spatiotemporally separated source reconstruction method, it effectively exploits the fact that temporal variations in emission rates dominate the impact on observations. The approach involves the training of machine learning models using the XGBoost algorithm and the determination of detailed temporal variations in emission rates using the PAMILT algorithm. The paper provides an important contribution in source reconstruction of atmospheric radionuclide emissions. The proposed methodology provides an effective approach to the accurate localisation of sources and estimation of emission rates in non-stationary emission scenarios and presents a promising framework for future research and practice. [1] The application of a temporal sliding-window average filter should be particularly appreciated. However, a more detailed description of the feature selection criteria used and the impact of different combinations of observation sites on source estimation would be beneficial. [2] Validation of the proposed method against SCK-CEN 41Ar field experiment data is an important step in demonstrating its effectiveness and practicality. However, additional validation studies under different scenarios and conditions would provide a better understanding of the method's scope and limitations. Additional case studies on different types of releases and different weather conditions are recommended to further assess the efficiency and versatility of the proposed method. Specific comments L215 Figure 2 The vertical and horizontal axes are distances, so the respective scales need to be the same. L222 What about vertical information? L302 Figure 3 The vertical and horizontal scales need to be the same. L340 Why is it possible to estimate with greater accuracy than the mesh size? And is such accuracy meaningful? Please discuss. L346 Figure 5 The vertical and horizontal axes are distances, so the respective scales need to be the same. L374 Figure 7 The vertical and horizontal axes are distances, so the respective scales need to be the same. These graphs are complex, please explain the figure in the caption.ChatGPTChatGPT
General Comments:
This study proposes a novel approach to the source reconstruction of atmospheric radionuclide emissions in non-stationary emission scenarios. By moving away from the unrealistic assumption of constant emissions and developing a method for spatiotemporally decoupled source reconstruction, it effectively leverages the fact that variations in emission rates significantly impact observations. The methodology involves training machine learning models with the XGBoost algorithm and determining detailed temporal variations in emission rates using the PAMILT algorithm. The paper makes a significant contribution to the field of atmospheric radionuclide emission source reconstruction. The proposed methodology offers an effective means for accurately localizing sources and estimating emission rates in non-stationary scenarios, presenting a promising framework for future research and applications.[1] The utilization of a temporal sliding-window average filter is commendable. However, elucidating the criteria for feature selection and the impact of varying combinations of observation sites on source estimation would enhance the paper.
[2] Validating the proposed method against the SCK-CEN 41Ar field experiment data underscores its efficacy and applicability. Nonetheless, conducting further validation studies under diverse scenarios and conditions would enrich our understanding of the method's applicability and limitations. It is recommended to include additional case studies involving different types of releases and weather conditions to assess the method's efficiency and adaptability more comprehensively.
Specific Comments:
L215 - Figure 2: The axes represent distances and should therefore have identical scales for clarity and accuracy.
L222 - Consideration of vertical information could provide a more comprehensive understanding of the dispersion patterns. How does the model account for vertical dispersion?
L302 - Figure 3: To ensure clarity and accuracy in data representation, the scales on the vertical and horizontal axes must be consistent.
L340 - The capability to estimate with greater accuracy than the grid size warrants a discussion. What implications does this have for the model's precision and its practical significance?
L346 - Figure 5: As these axes represent distances, maintaining identical scales on both axes is crucial for accurate data interpretation.
L374 - Figure 7: Given that both axes represent distances, their scales should be identical. The complexity of the graphs necessitates a detailed explanation within the figure caption to aid in interpretation.
Citation: https://doi.org/10.5194/gmd-2023-173-RC2 -
AC2: 'Reply on RC2', Yuhan Xu, 18 Feb 2024
Dear Referee,
Thank you for taking the time to provide such a constructive and comprehensive review of our manuscript (GMD-2023-173). In response to your insightful comments, we have made several modifications to the manuscript, and the comments have been uploaded in the form of a supplement file.
We are looking forward to your reply.
Best regards,
Yours sincerely
Sheng Fang
-
AC2: 'Reply on RC2', Yuhan Xu, 18 Feb 2024
Viewed
HTML | XML | Total | Supplement | BibTeX | EndNote | |
---|---|---|---|---|---|---|
185 | 49 | 24 | 258 | 16 | 13 | 9 |
- HTML: 185
- PDF: 49
- XML: 24
- Total: 258
- Supplement: 16
- BibTeX: 13
- EndNote: 9
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1