the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 15 Nov 2024
Turbulent transport extraction in time and frequency and the estimation of eddy fluxes at high resolution
Abstract. We propose a framework for the estimation of eddy fluxes based on cross-scalogram smoothing. This is motivated by two main problems encountered with the standard eddy-covariance method: (1) limited temporal resolution leading to estimated fluxes unable to characterise fast dynamics (< 30 min) and with potentially large periods of data discarded after the application of quality tests; (2) limited frequency resolution leading to poor localisation of the turbulent scales and thus to potential biases in the estimations. We show that cross-scalogram smoothing can be viewed as an extension of the standard eddy-covariance approach where measurement signals are analysed in multiple frequency bands leading to a high resolution analysis of fluxes in time and frequency. A metric based on the vertical component of the Reynold's stress tensor is proposed to localise the turbulent scales in time and frequency. It conditions the estimation of any scalar flux decomposed in time and frequency. The proposed metric is similar to the u* and σw tests but it is adapted to the time-frequency setting. We also address practical issues encountered with cross-scalogram smoothing such as the choice of the wavelet family and the conservative property of the decomposition. We show application of the framework at the beech forest site FR-Hes and demonstrate its relation with standard eddy covariance calculations. The proposed method produces high temporal resolution (1 min) estimates of CO2, latent and heat fluxes that align well with estimates from the standard 30-minute eddy-covariance method. The improved localisation of turbulent scales results in higher estimates of carbon uptake during summer (+2 ± 1 µmol m-2 s-1) and a more accurate assessment of nighttime respiration compared to standard eddy-covariance estimates. The methodology is implemented in the Julia package TurbulenceFlux.jl and is readily available for use.
-
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
(14916 KB)
-
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(14916 KB) - Metadata XML
- BibTeX
- EndNote
- Final revised paper
Journal article(s) based on this preprint
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2024-3243', Anonymous Referee #1, 07 Dec 2024
- AC1: 'Reply on RC1', Gabriel Destouet, 20 Mar 2025
-
RC2: 'Comment on egusphere-2024-3243', Anonymous Referee #2, 23 Jan 2025
The manuscript "Turbulent transport extraction in time and frequency and the estimation of eddy fluxes at high resolution" by Gabriel Destouet et al. (egusphere-2024-3243) presents a method for determining if the turbulence is sufficiently developed considering time and the eddy scale.
The manuscript is clear. I appreciated the progressiveness of the introduction that guides well the reader and found that the technical parts in the methods are well explained. It is in Atmospheric Measurement Techniques’ scope and it brings clear contributions to the flux community. A few points could be clearer, please see my major concerns below:
General comments
- The manuscript seems to suggest an innovation is calculating fluxes at smaller time periods. Estimating fluxes at high resolution can be done using standard eddy-covariance. The time and frequency average are commonly the same but there is no reason why not to calculate the instantaneous deviation (c’) using the 30-min average and then averaging w’c’ every 1 min. Conserving thus lower than 1-min frequency while having the 1 min fluxes calculated. Also calculating 1 min flux using continuous wavelets has been done in Schaller et al. (2017).
The fact that Schaller et al. (2017) have used wavelets for addressing short turbulent events could be cited in the introduction when enumerating the uses of wavelets. - The authors justify choosing the Generalised Morse Wavelets to avoid making an arbitrary choice of a particular family of wavelets such as Mexican hat or Morlet wavelets. Then it presents the parameters used. It seems the Generalised Morse Wavelet is of a “more flexible use”, but once you define its parameters, it is hard to understand how different this is from choosing another particular wavelet.
- The manuscript would benefit from showing clearly an equation stating how to calculate the global flux from wavelets. The result from using eq. 15-17 together I imagine.
- The terms “smoothing” and “local smoothing” are used several times in the manuscript and seems to be one of the key concepts. Once it includes a citation “The local smoothing of cross-scalograms (Mauder et al., 2007)”, however the authors in Mauder et al. (2007) do not employ the term themselves. In the manuscript it seems to be simply a time average, so how does it differ from it and is it really relevant to employ this wording? The risk is to lose the reader with too technical terms.
- The figures showing the time and frequency decomposed fluxes with a mask on top are well-appreciated. They are clear and intuitive.
- In the conclusion, the authors state that the proposed method "opens up new research perspectives, in particular the analysis of ecosystem response to rapid environmental changes (< 1 hour)." This is an important point, but it could be expanded upon. The sentence seems to suggest the advantage of calculating fluxes for shorter time periods, but this is not the innovative part of the manuscript (see major comment 1). The main contribution lies in doing the analysis in the frequency domain and considering the changing conditions for turbulence flagging.
Minor comments
- In p.1 l.12, 1 min is mentioned but is not shown in the rest of the manuscript.
- In p.5 l.124-125, the sentence may not be rigorously all correct. In standard eddy covariance, although we commonly only refer to a single averaging time, in reality we do two. One to calculate the instant deviation and another to pass from 10/20Hz to 30-min. See major comment 1.
- In p.5 l.125-127, The sentence does not refer to standard eddy covariance anymore, rephrasing may help avoid misunderstanding. Such as by adding “Alternatively” in the beginning or some other alternative.
- In p.5 l.141, the term “fluxes” repeats 3 times. I suggest to reformulate as such “The advective term is decomposed into turbulent eddy fluxes and other fluxes, encompassing those generated by large-scale processes and noise.”.
- In p.8 l.211, is K defined?
- In p.8 l.213-214, is it possible that peaks at low frequencies may represent real information and thus be acceptable?
- In p.10 l. 293, “less smoothing” seems imprecise.
- In p. 14, l.357, “dotted” should be “dashed”.
The scientific approach and methods applied are valid and the results and discussion well based on it. Overall, I believe that this manuscript represents a significant contribution to the field of atmospheric measurement techniques and I recommend it for publication in Atmospheric Measurement Techniques.
Citation: https://doi.org/10.5194/egusphere-2024-3243-RC2 - AC2: 'Reply on RC2', Gabriel Destouet, 20 Mar 2025
- The manuscript seems to suggest an innovation is calculating fluxes at smaller time periods. Estimating fluxes at high resolution can be done using standard eddy-covariance. The time and frequency average are commonly the same but there is no reason why not to calculate the instantaneous deviation (c’) using the 30-min average and then averaging w’c’ every 1 min. Conserving thus lower than 1-min frequency while having the 1 min fluxes calculated. Also calculating 1 min flux using continuous wavelets has been done in Schaller et al. (2017).
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2024-3243', Anonymous Referee #1, 07 Dec 2024
- AC1: 'Reply on RC1', Gabriel Destouet, 20 Mar 2025
-
RC2: 'Comment on egusphere-2024-3243', Anonymous Referee #2, 23 Jan 2025
The manuscript "Turbulent transport extraction in time and frequency and the estimation of eddy fluxes at high resolution" by Gabriel Destouet et al. (egusphere-2024-3243) presents a method for determining if the turbulence is sufficiently developed considering time and the eddy scale.
The manuscript is clear. I appreciated the progressiveness of the introduction that guides well the reader and found that the technical parts in the methods are well explained. It is in Atmospheric Measurement Techniques’ scope and it brings clear contributions to the flux community. A few points could be clearer, please see my major concerns below:
General comments
- The manuscript seems to suggest an innovation is calculating fluxes at smaller time periods. Estimating fluxes at high resolution can be done using standard eddy-covariance. The time and frequency average are commonly the same but there is no reason why not to calculate the instantaneous deviation (c’) using the 30-min average and then averaging w’c’ every 1 min. Conserving thus lower than 1-min frequency while having the 1 min fluxes calculated. Also calculating 1 min flux using continuous wavelets has been done in Schaller et al. (2017).
The fact that Schaller et al. (2017) have used wavelets for addressing short turbulent events could be cited in the introduction when enumerating the uses of wavelets. - The authors justify choosing the Generalised Morse Wavelets to avoid making an arbitrary choice of a particular family of wavelets such as Mexican hat or Morlet wavelets. Then it presents the parameters used. It seems the Generalised Morse Wavelet is of a “more flexible use”, but once you define its parameters, it is hard to understand how different this is from choosing another particular wavelet.
- The manuscript would benefit from showing clearly an equation stating how to calculate the global flux from wavelets. The result from using eq. 15-17 together I imagine.
- The terms “smoothing” and “local smoothing” are used several times in the manuscript and seems to be one of the key concepts. Once it includes a citation “The local smoothing of cross-scalograms (Mauder et al., 2007)”, however the authors in Mauder et al. (2007) do not employ the term themselves. In the manuscript it seems to be simply a time average, so how does it differ from it and is it really relevant to employ this wording? The risk is to lose the reader with too technical terms.
- The figures showing the time and frequency decomposed fluxes with a mask on top are well-appreciated. They are clear and intuitive.
- In the conclusion, the authors state that the proposed method "opens up new research perspectives, in particular the analysis of ecosystem response to rapid environmental changes (< 1 hour)." This is an important point, but it could be expanded upon. The sentence seems to suggest the advantage of calculating fluxes for shorter time periods, but this is not the innovative part of the manuscript (see major comment 1). The main contribution lies in doing the analysis in the frequency domain and considering the changing conditions for turbulence flagging.
Minor comments
- In p.1 l.12, 1 min is mentioned but is not shown in the rest of the manuscript.
- In p.5 l.124-125, the sentence may not be rigorously all correct. In standard eddy covariance, although we commonly only refer to a single averaging time, in reality we do two. One to calculate the instant deviation and another to pass from 10/20Hz to 30-min. See major comment 1.
- In p.5 l.125-127, The sentence does not refer to standard eddy covariance anymore, rephrasing may help avoid misunderstanding. Such as by adding “Alternatively” in the beginning or some other alternative.
- In p.5 l.141, the term “fluxes” repeats 3 times. I suggest to reformulate as such “The advective term is decomposed into turbulent eddy fluxes and other fluxes, encompassing those generated by large-scale processes and noise.”.
- In p.8 l.211, is K defined?
- In p.8 l.213-214, is it possible that peaks at low frequencies may represent real information and thus be acceptable?
- In p.10 l. 293, “less smoothing” seems imprecise.
- In p. 14, l.357, “dotted” should be “dashed”.
The scientific approach and methods applied are valid and the results and discussion well based on it. Overall, I believe that this manuscript represents a significant contribution to the field of atmospheric measurement techniques and I recommend it for publication in Atmospheric Measurement Techniques.
Citation: https://doi.org/10.5194/egusphere-2024-3243-RC2 - AC2: 'Reply on RC2', Gabriel Destouet, 20 Mar 2025
- The manuscript seems to suggest an innovation is calculating fluxes at smaller time periods. Estimating fluxes at high resolution can be done using standard eddy-covariance. The time and frequency average are commonly the same but there is no reason why not to calculate the instantaneous deviation (c’) using the 30-min average and then averaging w’c’ every 1 min. Conserving thus lower than 1-min frequency while having the 1 min fluxes calculated. Also calculating 1 min flux using continuous wavelets has been done in Schaller et al. (2017).
Peer review completion
Journal article(s) based on this preprint
Model code and software
TurbulenceFlux.jl Gabriel Destouet https://github.com/gabdst/TurbulenceFlux.jl
Interactive computing environment
Flux analysis of sample data with TurbulenceFlux.jl Gabriel Destouet https://github.com/gabdst/TurbulenceFlux.jl/blob/main/nb/flux_analysis.ipynb
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 290 | 92 | 20 | 402 | 13 | 32 |
- HTML: 290
- PDF: 92
- XML: 20
- Total: 402
- BibTeX: 13
- EndNote: 32
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Gabriel Destouet
Nikola Besic
Emilie Joetzjer
Matthias Cuntz
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(14916 KB) - Metadata XML