Detecting turbulent structures on single Doppler lidar large datasets: an automated classification method for horizontal scans

Cheliotis, Ioannis; Dieudonné, Elsa; Delbarre, Hervé; Sokolov, Anton; Dmitriev, Egor; Augustin, Patrick; Fourmentin, Marc

doi:https://doi.org/10.5194/amt-13-6579-2020

Articles | Volume 13, issue 12

https://doi.org/10.5194/amt-13-6579-2020

© Author(s) 2020. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/amt-13-6579-2020

© Author(s) 2020. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 13, issue 12

Research article

|

07 Dec 2020

Research article |

| 07 Dec 2020

Detecting turbulent structures on single Doppler lidar large datasets: an automated classification method for horizontal scans

Ioannis Cheliotis, Elsa Dieudonné, Hervé Delbarre, Anton Sokolov, Egor Dmitriev, Patrick Augustin, and Marc Fourmentin

Download

Final revised paper (published on 07 Dec 2020)
Preprint (discussion started on 30 Apr 2020)

Interactive discussion

Status: closed

AC: Author comment | RC: Referee comment | SC: Short comment | EC: Editor comment

- Printer-friendly version

- Supplement

RC1: 'Review of the manuscript: amt-2020-82', Anonymous Referee #1, 21 Jun 2020
- AC1: 'Authors' response to Referee 1', IOANNIS CHELIOTIS, 21 Jul 2020
RC2: 'General Comments', Anonymous Referee #2, 23 Jun 2020
- AC2: 'Authors' response to Referee 2', IOANNIS CHELIOTIS, 21 Jul 2020

Peer-review completion

AR: Author's response | RR: Referee report | ED: Editor decision

AR by Ioannis Cheliotis on behalf of the Authors (04 Aug 2020) Author's response Manuscript

ED: Referee Nomination & Report Request started (06 Sep 2020) by Marcos Portabella

RR by Anonymous Referee #1 (25 Sep 2020)

Suggestions for revision or reasons for rejection

Comment on AR1:

The altitude variations that the elevation map shows in Figure 2 in combination with the height variations of the buildings increase the heterogeneity of the terrain. I suggest explicitly stating that for the needs of the study the flow over the scanned areas is assumed to be homogeneous and avoid characterising the terrain as homogeneous.

Comment on AR2:

The root mean square error is an index on the deviations between the data sample and the corresponding predicted values of a model fit. Therefore, it is a way to evaluate the quality of the sinusoidal fit. The authors argue that instead they used the “symmetry of the radial wind field” to separate the bad cases. However, they do not explain how they have performed this step. Was it based on a visual inspection or on a statistical quantity? Please add more information in the manuscript regarding this step.

Comment on AR12:

The authors explain that: “the wind shear was estimated from the vertical profile of V_hor by subtracting the local minima from the local maxima above it, near the surface.” How consistent was the location of the local minima and the maxima in the data set used? I suggest to add one sentence stating this information. Furthermore, it should be added that the wind speed difference was divided by the corresponding difference in height between the local minima and local maxima.

Comment on AR17:

Here the authors describe their experience regarding the optimal contrast used to highlight the coherent patterns. This experience could be of value to others that would be interested in performing a similar analysis. The main points of their answer can be added as a discussion before the conclusions of this article.

Hide

ED: Publish subject to minor revisions (review by editor) (05 Oct 2020) by Marcos Portabella

AR by Ioannis Cheliotis on behalf of the Authors (13 Oct 2020) Author's response Manuscript

ED: Publish as is (25 Oct 2020) by Marcos Portabella

AR by Ioannis Cheliotis on behalf of the Authors (27 Oct 2020)

Short summary

The current study presents an automated method to classify coherent structures near the surface, based on the observations recorded by a single scanning Doppler lidar. This methodology combines texture analysis with a supervised machine-learning algorithm in order to study large datasets. The algorithm classified correctly about 91 % of cases of a training ensemble (150 scans). Furthermore the results of a 2-month classified dataset (4577 scans) by the algorithm are presented.