Snow depth time series retrieval by time-lapse photography: Finnish and Italian case studies

Bongio, Marco; Arslan, Ali Nadir; Tanis, Cemal Melih; De Michele, Carlo

doi:https://doi.org/10.5194/tc-15-369-2021

Articles | Volume 15, issue 1

https://doi.org/10.5194/tc-15-369-2021

© Author(s) 2021. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/tc-15-369-2021

© Author(s) 2021. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 15, issue 1

Research article

|

27 Jan 2021

Research article |

| 27 Jan 2021

Snow depth time series retrieval by time-lapse photography: Finnish and Italian case studies

Marco Bongio, Ali Nadir Arslan, Cemal Melih Tanis, and Carlo De Michele

Download

Final revised paper (published on 27 Jan 2021)
Supplement to the final revised paper
Preprint (discussion started on 07 Oct 2019)
Supplement to the preprint

Interactive discussion

Status: closed

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

- Printer-friendly version

- Supplement

SC1: 'FMIPROT Webpage links notification, from authors', Cemal Melih Tanis, 07 Oct 2019
RC1: 'review', Anonymous Referee #1, 14 Nov 2019
- AC1: 'REPLY TO REVIEWER 1', Marco Bongio, 08 Apr 2020
RC2: 'review', Anonymous Referee #2, 05 Mar 2020
- AC2: 'REPLY TO REVIEWER 2', Marco Bongio, 08 Apr 2020
RC3: 'review', Anonymous Referee #3, 18 Mar 2020
- AC3: 'Reply to Reviewer 3', Marco Bongio, 08 Apr 2020

Peer-review completion

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

ED: Reconsider after major revisions (further review by editor and referees) (30 Jul 2020) by Eric Larour

AR by Marco Bongio on behalf of the Authors (31 Aug 2020) Author's response Manuscript

ED: Reconsider after major revisions (further review by editor and referees) (21 Sep 2020) by Eric Larour

ED: Referee Nomination & Report Request started (22 Sep 2020) by Eric Larour

RR by Anonymous Referee #2 (30 Sep 2020)

Suggestions for revision or reasons for rejection

General comments
I would like to thanks the authors for considering my previous comments. The novel contribution is now formulated clearly. The Introduction is still quite long and can be condensed. Maybe the list of findings presented at the end of Introduction can be, perhaps, turn to the formulation of research hypotheses of the paper (the Intro should provide motivation and research objective of the work, not the findings). Still, I have a few comments related to the clarity of presentation (please see specific comments) and improvement of the Discussion section. I like a separate discussion of results, as well as lesson learned part. What I’m still missing is the link to previous papers. What is similar, and what new findings have been obtained in the manuscript (compared to the existing literature)? And finally, I think there is still space to improve and condense the text/story/style of writing. Some information is repetitive; perhaps some native speaker can help here. Some sections are still quite long, which distract the reader from the main messages of the paper.

Specific comments
1) Eq. 1 in Introduction is likely not needed.
2) L.55: “…but they were not specifically focused on the retrieval of snow depth.” This is likely not a precise statement, as in the next sections, there is written that some studies (e.g. Parajka et al., 2012, Garvelman et . 2013, Dong et al., 2017) investigated the use of digital cameras for snow depth estimation.
3) L.116: “there is no fully automatic procedure to retrieve the snow depth in real-time…”. What is meant by fully automatic? In the next sentence, it is written as the new objective of this study, but the presented approach is not fully automatic, as you need to define several thresholds. So for an application in a new region, some setup and subjective decisions are needed.
4) L.123: I would suggest turning the formulation “we presented..” into a formulation of the research hypothesis of this paper.
5) L.185: I would suggest formulating a general procedure for a setup of the parameters in the Methods section. It should allow a reader/new user to understand and repeat the way/steps needed for estimating robust parameters in any region. Perhaps some objective calibration of these two parameters will be helpful. In its current form, it is not clear how many combinations and from which range the parameter selection should be based? Will it be possible to “calibrate it”? E.g. by minimising some RMSE or NSE? Is this procedure sensitive to the selection of test window?
6) Some figures are repeating the same observations (duplicating the same information). It will be interesting, perhaps, to see some additional evaluation of the sensitivity of the approach to the selection of the thresholds instead.

Hide

RR by Anonymous Referee #1 (04 Oct 2020)

Suggestions for revision or reasons for rejection

Abstract
Lines 22-25 the NSE is not a so diffused statistical index as the RMSE index. The not-expert reader cannot interpret numbers in the abstract without a detailed knowledge of mentioned indexes.

Introduction
There are a lot of references but there is also a critical inconsistency: the manucript is focused on snow depth (looking at the title), good to say that SD is important for estimating SWE from eq.1 (maybe “not rather than” as stated in lines 35-36) and how that it is possible to measure it (lines 40-49). Authors define accuracy for different reference methods focused on SD in lines 45-49 (probably from Pirazzini et al 2018). This is the most important statement in the introduction since the reader should understand which is the state of the art about measuring the SD. From line 50 to 55 authors summarize the state of the art about time-lapse photography (good to do that in this case) stating “In these studies...but they were not specifically focused on the retrieval of snow depth”. From line 56 to 114 they try to extract information about estimating SD from some of those papers: no information about SD in lines 56-70 and 82-85; 10cm accuracy in Parajka et al 2012 (lines 71-81); 10cm accuracy and 7cm RMSE? Against visual inspection in Garvelmann et al 2013 (lines 86-91); 92-103 a full description without any statistics; 104-114 RMSE 1-2cm in Dong and Menzel 2017 (please fix the citation in line 104). Finally, authors state “In conclusions, some works about the snow depth retrieval from time-lapse photography are available…”. I would remove references and sentences without relevant details and I would find some comparison in the discussion section between the presented results and those available in literature.

Authors declare in the abstract (line 18-21) that “we present new possibilities and advantages in the retrieval of snow depth...which can be summarized as follows: 1) retrieving the snow depth at high temporal resolution with an accuracy comparable to the most common method (manual measurements); 2) visual identification of errors or misclassifications; 3) estimating the spatial variability of snow depth; 4) correcting the under catch problem of pluviometer when this instrument is used.

But they present the following results in the introduction (lines 124-126): “how the accuracy of snow depth estimations can be increased using stakes with 1 cm spacing markers; proper geometric and parameter configurations of the camera-stake system; sources of errors and post processing procedure corrections.

Methods
lines 267-271 the distance between the camera and the stake is still missing as well as the position in relation to the ultrasonic measurements.

Line 325 there is some text to be fixed close to the fig5 citation
Lines 325-334. Authors considered only the 2018-2019 period but describe a 2013-2109 starting interval adding also one figure (fig.2) that is wrongly numbered. I suggest to mention only the considered period, the whole dataset and the considerations about problems for the visual observation create confusion to the reader. I suggest also to remove fig.5 since it is redundant with fig.11 and 12 where the same data are plotted again. I suggest also to merge fig 11 and 12 evidencing the calibration and the validation periods.

Line 358-371 I suggest to mention only the considered period, the whole dataset and the considerations about problems for the visual observation create confusion to the reader. I suggest also to remove fig.6 since it is redundant with fig.13 where the same data are plotted again. I suggest to evidence the calibration and the validation periods as for the Gressoney case study (there is no declaration in this case: 10 days as in Sodankyla?).

Results
Line 383 The distance between the camera and the stake is ideal but there are no info about that in the methods section

Line 398 be careful declaring RMSE and NSE as accuracy estimations. Do Authors used the term accuracy differently in the introduction (precision, lowest limit of detection…..)?

Lines 402-404 no distances or maps have been provided to the reader

Lines 437-445 is the parallax effect only a local effect?

Lines 446-457 Why the melting near the stake is an issue only in this case study? How can you train your procedure without an AINEVA operator? This could be moved to the discussion section describing the difference between the AINEVA correction and the “standard” visual inspection.

Discussion
There is a list of potential problems and of suggestions for the right configuration of observations. From my point of view, I would know:

1) what kind of corrections are still missing (parallax or heating-related issues for example). This could be firstly assessed looking at differences between visual observations on the stake and the AINEVA reading procedure.

2) what kind of differences are detected between ultrasonic measurements, photography-derived observations and other studies (Garvelmann et al 2013; Dong and Menzel 2017) that described similar measurements.

Referring to the abstract statements, I don’t see discussion about: retrieving the snow depth at high temporal resolution with an accuracy comparable to the most common method (manual measurements); estimating the spatial variability of snow depth; correcting the under catch problem of pluviometer when this instrument is used. It is preferable to use the points cited in the introduction (lines 124-126): “how the accuracy of snow depth estimations can be increased using stakes with 1 cm spacing markers; proper geometric and parameter configurations of the camera-stake system; sources of errors and post processing procedure corrections.

Hide

ED: Publish subject to minor revisions (review by editor) (28 Oct 2020) by Eric Larour

AR by Marco Bongio on behalf of the Authors (14 Nov 2020) Author's response Manuscript

ED: Publish as is (01 Dec 2020) by Eric Larour

AR by Marco Bongio on behalf of the Authors (01 Dec 2020)

Short summary

The capability of time-lapse photography to retrieve snow depth time series was tested. We demonstrated that this method can be efficiently used in three different case studies: two in the Italian Alps and one in a forested region of Finland, with an accuracy comparable to the most common methods such as ultrasonic sensors or manual measurements. We hope that this simple method based only on a camera and a graduated stake can enable snow depth measurements in dangerous and inaccessible sites.