This paper introduces a new approach for continuous measurement of glacier ice melt (surface ablation). The concept is to use an accelerometer to measure the vibrational frequencies of a mass balance stake hourly to determine the amount of the stake that is exposed, which is related to surface ablation. To the best of my knowledge this concept is entirely novel and, since the cost of the instrument is low, this could be an excellent addition to the glaciological toolkit. While this shows a lot of promise, it is not quite ready for prime time. The paper shows clearly how this technology could work conceptually but the testing is not thorough and reveals issues that need to be tackled. A final test on a glacier is a real must, of course. It is good to know that this is currently in progress, but there are details yet to work through.

My recommendation would be to either A) wait to publish until there are more answers to the questions raised below (i.e. manuscript rejection/withdrawal) or B) rescope and reframe this paper to be more of a record of a proof of concept. If the authors chose the latter, the manuscript could be publishable after (1) explaining the limitations of this instrument for the goal of mass balance measurement, (2) redoing some of the tests, explaining the challenges that lie ahead for (3) field testing and (4) operational use. This would serve to anchor your contribution in the literature and provide a roadmap for the second paper.

Here are considerations and questions relating to points 1-4 above:

Lab conditions should ideal conditions when you are testing your technology. If there are issues with this setup, it is hard to interpret field tests in less than ideal conditions. If the clamping was not done correctly, it would be best to redo your work so you can be sure of the data. Likewise, if you don’t have a surface to test the sonic ranger against, then get a sheet of plywood and a bracket to create one. Is there interference with the pole itself?

Testing should be first done thoroughly in the lab. Using a lake to test is an excellent idea. It doesn’t need to be in the Arctic though. It could be near Montreal. Another option is to create a mound of ice with water. That ice can ablate over time during a test and this could be monitored closely. Close observation would be helpful before testing on a glacier.

It would be good to test both the accelerometer and sonic ranger against manual measurements. If your sonic ranger model is not good enough for a comparison, add one that is, at least for the testing. Other instruments and a timelapse camera would be great as well.

Using more than one stake at each length for field and lab testing is essential. This can be used to estimate error.

Temperature dependence should be worked out. Why is this happening? The effect needs to be isolated and removed or put in the error term.

If there is no wind, there will be no motion to measure. How often do you anticipate there not being enough wind to get a measurement of F1 and F2? Or at least can you give some indication of the wind required to induce these frequencies?

Testing different brands of the same stake (pipe) will show how sensitive the method is to differences in manufacturing. What happens if the stake is slightly bent or damaged? Will that invalidate the measurement? At what point?

How easy is it to use a different kind of stake? Can users easily recalibrate to make use of steel, wood or plastic stakes? Glaciologists will need to know this to adopt the method.

What happens when it snows? Loose snow won’t constrain the stake. Will dense snow constrain the stake vibration? At what point? Is there a way to use the sonic ranger to detect snow and account for it?

What happens when the ice surface around the stake melts a little water-filled hole such that the stake is held firmly ~15 cm below the general ice surface?

====

Specific comments:

Title: I think surface ablation is better than ice melt in the title

Abstract

Surface ablation measurements of glaciers are critical for understanding mass change over --> Surface ablation measurements are critical for understanding glacier mass change over

depends sensitively on the exposed length, --> depends sensitively on their exposed length,

Color Lake --> Colour Lake (here and elsewhere in the manuscript – proper name with Canadian spelling)

surface activity at any locations --> surface activity at any location

1. Introduction

Hulth (2010) presents a draw-wire method, although the measurement is sensitive to only surface lowering and not accumulation --- I think your method and many of the others presented also cannot measure accumulation at least directly (since for mass accumulation, snow density needs to be known)

representing annual or seasonal summer and winter balances, respectively --> representing annual or seasonal (summer and winter) balances, respectively

2 BRACHIOSAURUS design

, and a depth sensor (Figure 1 (b)). This sensor measure range. It is a sonic ranger and it measures distance. There is no quantity ‘depth’ being examined in this paper. I recommend you replace depth with range or distance.

2.2 Depth [Range] sensor

Sonic range varies according to temperature (which affects the speed of sound). Please explain how this is corrected for.

The sonic pulse will spread out with distance from the sensor. What is the angle that describes the cone of ensonification? How do you avoid having the pole interfere with this measurement?

2.5 Circuit board...

sensor ‘eyes’ -> sensor transducers

3 Vibrational frequency

Equation explanation of symbols – please include the units of measure here as well.

The mass of BRACHI (assumed here to be a point mass at the end of the beam)

Fig 2 caption – make clear that numerical solution is for extended mass and analytic is for point mass

4. Data acquisition...

10 equal length chunks of 12 s each

below the noise floor – can you explain what the noise floor is in this case and why you would not just eliminate data that is below it? I think it is a matter of rewording the sentence to be clearer.

Future measurements will save and process the motion of all axes separately without squaring. - This needs to be explained (why) and moved to a better place (Future development section?)

Local maxima that lie above the noise floor – define the noise floor.

Finding a candidate F1, determined by ...highest amplitude within a frequency range of 0 to 2F1 -- this seems rather circular logically. You need F1 to look for F1? !

By numerically propagating the errors from F1 and F2 – how do you determine the errors for F1 and F2? How are they propagated? In quadrature? Other? What happens if F1 or F2 can’t be found?

Fig 3 – label the peaks adjacent to F2 as F2-F1 and F2+F1

5 Lab measurements

2.7 g/cm^3

radii were respectively measured --> measured respectively

Non-rigidity in the clamping point – please explain your set up here. Ideally you would redo it properly.

Analysis of higher-order modes, if present --- how often are they present? Under what conditions?

Estimated precision of 0.25 cm – explain how you derive this? Is it variation in ice surface? Parallax error? Other?

Although the depth sensor was tested qualitatively in the lab... I think you can do better!

6.1 BRACHI comparison against ...

was below freezing and at a local minimum – what do you mean by ‘at a local minimum’?

Surrounding the stake is solid --> replace with not actively melting

failed to detect ... degraded performance at low temperatures ... close to... operational limits – the temperature range is down to -20C but the temperature was not always close to this. Please look for another reason.

The BRACHI -derived lengths – there are 2 sensors to detect lengths. Please be specific

Temperature-dependent systematic effects – Yes, agreed you have some digging to do to explain how your system is impacted by temperature.

How are you measuring temperature? Can you measure actual ambient air temperature?

Figure 6 –

It would be good to know if you measured the length of the stake during this time. Maybe there was ablation?

How are you computing the error bars in this graph?

7 Conclusions

I think you can only characterize your field tests as partly successful

References:

axel heiberg island canadian high arctic – capitaliztion

The manuscript ”A low-cost ice melt monitoring system using wind-induced motion of mass-balance stakes” by Felix St-Amour and coauthors is an interesting read reporting on semi-automated approach for ice ablation/aggregation measurements using a low-cost “do-it-yourself” approach. The research is open source and a link to the necessary design files, codes and documents are given. The manuscript serves as a point of reference as the research is on-going and the authors describe several future updates to the system that will very likely improve the performance.

The manuscript is clearly laid out with a clear background to the research topic including appropriate references. The instrument and measurement methods are described in sufficient details and examples of data are given.

I have no objections to the manuscript being accepted after a minor revision.

Page 2: I suggest adding the Zenodo link as a reference in the manuscript and citing it in the introduction to section 2. This will make it immediately apparent to the reader where the material can be found.

Page 2: Please add the manufacturer name to the LSM6DS0X and HC-SR04 units

Page 4, batteries. I find it a little bit odd that the power draw is only “estimated” and that the authors “expect” a minimum lifetime. Surely, given the intended use case of the instrument in remote areas, some measurements of power draw and (accelerated) lifetime tests must have been done. I suggest that this issue is discussed in greater depth than just the sporadic comment on power drained by repeated wireless access attempts.

Page 6, figure 3. The spectrum shown is a good example of the acquired data. However, it is not clear to me whether this represents a typical data set or if this is data of outstanding quality. I suggest that the authors enrich the manuscript with a discussion of good, common, and bad data and signal-to-noise ratio.

Page 6, figure 3. The spectrum contains two “spurious” peaks. Are these due to motion of the sensor or artefacts from the electronic system? Please add a line to the manuscript on this.

Page 6: I understand the need for publishing intermediate results, but I can’t help feeling a little cheated in the analysis of the sensor data. It would have been great with an analysis of the individual x, y, and z components. I look forward to a follow-up paper with this analysis.

Page 6. The sentence on locating F1 in the spectrum seems a bit circular to me. Surely, if F1 is unknown, the range 0 to 2F1 is undefined.

Page 13: The formatting of the references is a little inconsistent with some references containing DOI while others do not.