the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
WRF-PDAF v1.0: Implementation and Application of an Online Localized Ensemble Data Assimilation Framework
Abstract. Data assimilation is a common technique employed to estimate the state and its associated uncertainties in numerical models. Ensemble-based methods are a prevalent choice, although they can be computationally expensive due to the required ensemble integrations. In this study, we enhance the capabilities of Weather Research and Forecasting–Advanced Research WRF (WRF-ARW) model by coupling it with the Parallel Data Assimilation Framework (PDAF) in a fully online mode. Through minimal modifications to the WRF-ARW model code, we have developed an efficient data assimilation system. This system leverages parallelization and in-memory data transfers between the model and data assimilation processes, greatly reducing the need for file I/O and model restarts during assimilation. We detail the necessary program modifications in this study. One advantage of the resulting assimilation system is a clear separation of concerns between data assimilation method development and model application resulting from PDAF’s model-agnostic structure. To evaluate the assimilation system, we conduct a twin experiment simulating an idealized tropical cyclone. Cycled data assimilation experiments focus on the impact of temperature profiles. The assimilation not only significantly enhances temperature field accuracy but also improves the initial U and V fields. The assimilation process introduces only minimal overhead in run time when compared to the model without data assimilation and exhibits excellent parallel performance. Consequently, the online WRF-PDAF system emerges as an efficient framework for implementing high-resolution mesoscale forecasting and reanalysis.
-
Notice on discussion status
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
-
Preprint
(821 KB)
-
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(821 KB) - Metadata XML
- BibTeX
- EndNote
- Final revised paper
Journal article(s) based on this preprint
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2023-2311', Anonymous Referee #1, 10 Dec 2023
## Overview
The authors implemented an online data assimilation framework in the WRF-ARW weather forecasting system, and evaluated its performance utilizing a twin experiment setup of an ideal tropical cyclone case. The results indicate that this assimilation framework is effective in improving the prediction skill of the model and is computationally efficient. It would be really useful to the community if such performance can be reproduced in real-world cases, and I suggest the authors to showcase a successful real-world application to further enhance the manuscript. Anyway, I believe this development work is interesting to the community and should be tested in more real-world cases by the community.
The text is overall well-crafted and easily comprehensible, and the figures effectively convey information, although there is room for improvement in their presentation quality. I have only a few minor comments listed below. Once those have been addressed, I recommend the work be accepted for publication.
## Specific comments
- L148-149: It seems that the labels for Exp. 1 and Exp. 2 are mistakenly swapped. In addition, it's not clear what the CTL case is. My guess is that it's a single member free run without data assimilation, but the authors may consider adding some description for each case as well as the idea of the twin-experiment setup so that readers unfamiliar with the experiment can understand.
- "3.1 Subsection (as Heading 2)" seems like unfinished writing.
- L185: "in in"
- The teletype (monospace) font should be applied for routine names.
- L373: How can the RMSE of ENS be lower than that of the true state? In addition, the expression "single control run" seems odd. Perhaps the authors mean "single member control run"?
- L377: "Smaller and large" -> "Smaller and larger" ?
- L381: "in Fig. 3" --> "in Fig. 4" ?
- L387: "Figure 3b and 3c" --> "Figure 4b and 4c" ?
- L400: Can we expect that the selection of 4dx is a good choice for realistic cases? Or is the optimal selection case by case?Citation: https://doi.org/10.5194/egusphere-2023-2311-RC1 -
AC1: 'Reply on RC1', Changliang Shao, 13 Dec 2023
We greatly appreciate your very valuable comments. These are very helpful in guiding the improvement of our manuscript. A point-by-point response to the comments is provided below. We hope this could address the issues..
## Specific comments
- L148-149: It seems that the labels for Exp. 1 and Exp. 2 are mistakenly swapped. In addition, it's not clear what the CTL case is. My guess is that it's a single member free run without data assimilation, but the authors may consider adding some description for each case as well as the idea of the twin-experiment setup so that readers unfamiliar with the experiment can understand.Reply: Thanks for your comments. Now, the two labels have been swapped back to the correct positions. Moreover, to make the definition clear, descriptions of each case and basic idea of the twin-experiment setup have been added in line 150-152 as "The CTRL is derived from a single free run, which mirrors the True in all aspects, except for a 60-hour delay in its starting time. The ENS, comprising 40 ensemble members, is generated by introducing an initial perturbation to the CTRL. According to the ENS, the DA twin experiments are implemented by assimilating observations using different localization radii."
- "3.1 Subsection (as Heading 2)" seems like unfinished writing.Reply: Thanks for your comments. "3.1 Subsection (as Heading 2)" has been corrected to "Description of PDAF"
- L185: "in in"Reply: Thanks for your comments. "in in" has been corrected to "in"
- The teletype (monospace) font should be applied for routine names.Reply: Thanks for your comments. We rechecked the content, and the font of all routine names has been changed to 'consolas', including the main text and figures.
- L373: How can the RMSE of ENS be lower than that of the true state? In addition, the expression "single control run" seems odd. Perhaps the authors mean "single member control run"?Reply: Thanks for your comments. In line 373, "and the true state" has been changed to "compared to the true state". Furthermore, the odd expression "single control run" has been changed to "control run" to make the context consistent. As the currently added description in line 150-152, the control run is defined as a single run, actually the same meaning with "single member control run".
- L377: "Smaller and large" -> "Smaller and larger" ?Reply: Thanks for your comments. In order to avoid mistake and ambiguity, "Smaller and large" has been changed to "Compared to ANA4, either smaller or larger"
- L381: "in Fig. 3" --> "in Fig. 4" ?Reply: Thanks for your comments. Corrected.
- L387: "Figure 3b and 3c" --> "Figure 4b and 4c" ?Reply: Thanks for your comments. Corrected. in addition, the label (a),(b),...,(f) have been added to the subfigures.
- L400: Can we expect that the selection of 4dx is a good choice for realistic cases? Or is the optimal selection case by case?Reply: Thanks for your comments. Actually, the optimal selection is case by case. As we noted in line 405-407, "It is important to emphasize that the experiments conducted based on different localization radii serve as fundamental demonstrations of the functionality of the DA program, with in-depth analysis not being the primary focus of this study." Even in the same case, different assimilation observation (i.e. U or V) may have different optimal selection. To avoid misleading, we added a clarification in line 407 as "In addition, the optimal selection is case dependent."
Citation: https://doi.org/10.5194/egusphere-2023-2311-AC1
-
AC1: 'Reply on RC1', Changliang Shao, 13 Dec 2023
-
RC2: 'Comment on egusphere-2023-2311', Anonymous Referee #2, 10 Mar 2024
- AC2: 'Reply on RC2', Changliang Shao, 14 Mar 2024
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2023-2311', Anonymous Referee #1, 10 Dec 2023
## Overview
The authors implemented an online data assimilation framework in the WRF-ARW weather forecasting system, and evaluated its performance utilizing a twin experiment setup of an ideal tropical cyclone case. The results indicate that this assimilation framework is effective in improving the prediction skill of the model and is computationally efficient. It would be really useful to the community if such performance can be reproduced in real-world cases, and I suggest the authors to showcase a successful real-world application to further enhance the manuscript. Anyway, I believe this development work is interesting to the community and should be tested in more real-world cases by the community.
The text is overall well-crafted and easily comprehensible, and the figures effectively convey information, although there is room for improvement in their presentation quality. I have only a few minor comments listed below. Once those have been addressed, I recommend the work be accepted for publication.
## Specific comments
- L148-149: It seems that the labels for Exp. 1 and Exp. 2 are mistakenly swapped. In addition, it's not clear what the CTL case is. My guess is that it's a single member free run without data assimilation, but the authors may consider adding some description for each case as well as the idea of the twin-experiment setup so that readers unfamiliar with the experiment can understand.
- "3.1 Subsection (as Heading 2)" seems like unfinished writing.
- L185: "in in"
- The teletype (monospace) font should be applied for routine names.
- L373: How can the RMSE of ENS be lower than that of the true state? In addition, the expression "single control run" seems odd. Perhaps the authors mean "single member control run"?
- L377: "Smaller and large" -> "Smaller and larger" ?
- L381: "in Fig. 3" --> "in Fig. 4" ?
- L387: "Figure 3b and 3c" --> "Figure 4b and 4c" ?
- L400: Can we expect that the selection of 4dx is a good choice for realistic cases? Or is the optimal selection case by case?Citation: https://doi.org/10.5194/egusphere-2023-2311-RC1 -
AC1: 'Reply on RC1', Changliang Shao, 13 Dec 2023
We greatly appreciate your very valuable comments. These are very helpful in guiding the improvement of our manuscript. A point-by-point response to the comments is provided below. We hope this could address the issues..
## Specific comments
- L148-149: It seems that the labels for Exp. 1 and Exp. 2 are mistakenly swapped. In addition, it's not clear what the CTL case is. My guess is that it's a single member free run without data assimilation, but the authors may consider adding some description for each case as well as the idea of the twin-experiment setup so that readers unfamiliar with the experiment can understand.Reply: Thanks for your comments. Now, the two labels have been swapped back to the correct positions. Moreover, to make the definition clear, descriptions of each case and basic idea of the twin-experiment setup have been added in line 150-152 as "The CTRL is derived from a single free run, which mirrors the True in all aspects, except for a 60-hour delay in its starting time. The ENS, comprising 40 ensemble members, is generated by introducing an initial perturbation to the CTRL. According to the ENS, the DA twin experiments are implemented by assimilating observations using different localization radii."
- "3.1 Subsection (as Heading 2)" seems like unfinished writing.Reply: Thanks for your comments. "3.1 Subsection (as Heading 2)" has been corrected to "Description of PDAF"
- L185: "in in"Reply: Thanks for your comments. "in in" has been corrected to "in"
- The teletype (monospace) font should be applied for routine names.Reply: Thanks for your comments. We rechecked the content, and the font of all routine names has been changed to 'consolas', including the main text and figures.
- L373: How can the RMSE of ENS be lower than that of the true state? In addition, the expression "single control run" seems odd. Perhaps the authors mean "single member control run"?Reply: Thanks for your comments. In line 373, "and the true state" has been changed to "compared to the true state". Furthermore, the odd expression "single control run" has been changed to "control run" to make the context consistent. As the currently added description in line 150-152, the control run is defined as a single run, actually the same meaning with "single member control run".
- L377: "Smaller and large" -> "Smaller and larger" ?Reply: Thanks for your comments. In order to avoid mistake and ambiguity, "Smaller and large" has been changed to "Compared to ANA4, either smaller or larger"
- L381: "in Fig. 3" --> "in Fig. 4" ?Reply: Thanks for your comments. Corrected.
- L387: "Figure 3b and 3c" --> "Figure 4b and 4c" ?Reply: Thanks for your comments. Corrected. in addition, the label (a),(b),...,(f) have been added to the subfigures.
- L400: Can we expect that the selection of 4dx is a good choice for realistic cases? Or is the optimal selection case by case?Reply: Thanks for your comments. Actually, the optimal selection is case by case. As we noted in line 405-407, "It is important to emphasize that the experiments conducted based on different localization radii serve as fundamental demonstrations of the functionality of the DA program, with in-depth analysis not being the primary focus of this study." Even in the same case, different assimilation observation (i.e. U or V) may have different optimal selection. To avoid misleading, we added a clarification in line 407 as "In addition, the optimal selection is case dependent."
Citation: https://doi.org/10.5194/egusphere-2023-2311-AC1
-
AC1: 'Reply on RC1', Changliang Shao, 13 Dec 2023
-
RC2: 'Comment on egusphere-2023-2311', Anonymous Referee #2, 10 Mar 2024
- AC2: 'Reply on RC2', Changliang Shao, 14 Mar 2024
Peer review completion
Journal article(s) based on this preprint
Data sets
Observation assimilated by WRF-PDAF Changliang Shao https://doi.org/10.5281/zenodo.10083810
Model code and software
WRF-PDAF v1.0 Changliang Shao https://doi.org/10.5281/zenodo.8367112
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
192 | 63 | 23 | 278 | 14 | 12 |
- HTML: 192
- PDF: 63
- XML: 23
- Total: 278
- BibTeX: 14
- EndNote: 12
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1
Cited
1 citations as recorded by crossref.
Lars Nerger
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(821 KB) - Metadata XML