Detecting the melting layer with a micro rain radar using a neural network approach

Brast, Maren; Markmann, Piet

doi:https://doi.org/10.5194/amt-13-6645-2020

Articles | Volume 13, issue 12

https://doi.org/10.5194/amt-13-6645-2020

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

https://doi.org/10.5194/amt-13-6645-2020

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

Articles | Volume 13, issue 12

Research article

|

09 Dec 2020

Research article |

| 09 Dec 2020

Detecting the melting layer with a micro rain radar using a neural network approach

Maren Brast and Piet Markmann

Download

Final revised paper (published on 09 Dec 2020)
Preprint (discussion started on 23 Sep 2019)

Interactive discussion

Status: closed

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

- Printer-friendly version

- Supplement

RC1: 'Review of Detecting the Melting Layer with a Micro Rain Radar Using a Neural Network Approach', Anonymous Referee #1, 20 Oct 2019
- AC1: 'Answer to Comments', Piet Markmann, 17 Dec 2019
RC2: 'Review of amt-2019-248', Anonymous Referee #2, 02 Nov 2019
- AC2: 'Answer to comments', Piet Markmann, 17 Dec 2019

Peer-review completion

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

AR by Piet Markmann on behalf of the Authors (11 Mar 2020) Author's response Manuscript

ED: Referee Nomination & Report Request started (24 Mar 2020) by Laura Bianco

RR by Anonymous Referee #1 (20 Apr 2020)

Suggestions for revision or reasons for rejection

A) My key comment from my original review was not addressed “The authors have shown that a suitably trained neutral network *can* identify the melting layer in MicroRainRadar data. They have yet to show whether this is useful in the context of previous work. In my view, the paper is incomplete in its current form. “ I strongly disagree that such a comparison is outside the scope of work. The empirical algorithm used by Perry et al. (2017) is well less than 1 page of code and would be simple to implement and run on the data sets used in this paper. Without a comparison and proof that the neural network is superior method, this paper has no clear value. If someone proposed a new empirical method and did not show it was superior to previously published methods that paper would not have value either.

B) I do think it is important to mention the vertical resolutions of the C-band radars relative to the MRR since the C-band radars are being used to evaluate the algorithm in Section 3.3 and Figure 10. Please put in specific mention of the vertical resolution of the C band radar over the MRR is around 700m for the closest radar and around 1500m for the farthest radar. Please include this information in the paper.

C) Based on Figure 7, clearly the neural network algorithm does not work for short, isolated showers. I suggest that the output of the neural network be filtered to remove such short time interval data. Page 10 states that “During the training process, the NN was tuned to detect as little as possible in these cases, but it was not possible to train the NN not to detect anything without impeding the abilities of the NN for other cases. Therefore, convective cases are those cases where the NN has the most problems.” There is nothing stopping the authors from filtering the output of the NN as part of the quality control. Including data that is clearly wrong substantially weakens the case for using this new algorithm.

D) I think there is still some miscommunication regarding the application to determining risk of icing. Aircraft icing requires high concentrations of supercooled cloud water. This can occur at heights above the 0 deg C level when air temperature is < 0 deg C OR when warm layers are present aloft where the ice melts and then closer to the ground the air temperatures are < 0 deg C. The detection of the melting layer informs about the melting of snow. The MRR data cannot detect cloud-size droplets nor address the temperature of the air. While above the melting layer, the air temperature < 0 deg C an be inferred, this is not the case for below the melting layer. Whether the melting layer represents a warm layer aloft with colder air at the surface or temperature decreasing with height is not known without a temperature profile. It is unclear how this method will help to assess the risk of icing at airports (page 2 line 2, page 18 line 25)

Minor comments
Page 7, line 28 “indizes” should be “indices”

Hide

RR by Anonymous Referee #2 (21 Apr 2020)

ED: Reconsider after major revisions (19 May 2020) by Laura Bianco

AR by Piet Markmann on behalf of the Authors (30 Jun 2020) Author's response Manuscript

ED: Publish as is (18 Jul 2020) by Laura Bianco

AR by Piet Markmann on behalf of the Authors (28 Jul 2020)