Review of « Particle and front tracking in experimental and computational avalanche dynamics » by Neuhauser et al.

General comment

The paper by Neuhauser et al presents a detailed analysis of the dynamics (front position, « internal » velocities thanks to particle tracking) of one single medium size avalanche event using a combination of approaches that involve full-scale experiments with advanced instrumentation and simulations : i) particle tracking within the avalanche using the AvaNodes’ technique recently applied to snow avalanches, ii) radar measurements, and iii) specific depth-averaged simulations based on a Voellmy rheology enriched with cohesion that allow (simulated and specific) particle tracking, namely the com1DFA module of AvaFrame.

I found the paper quite difficult to digest and to follow because many statements that are quite unclear when they appear for the first time are then discussed/solved later at different locations of the paper but it is frustrating that explanations come too late. I think that the story can be improved in many places to help the reader. In parallel there is a lack of explanation on some points that may be trivial for the authors but certainly not for the reader. I will try to provide a number of detailed comments about this issue.

More generally, I’m note sure about the key objective of the paper ? Do the authors want to use AvaNodes’ measurements to better constrain numerical simulations based on a depth-averaged model with a closure Voellmy law with cohesion ? Do the authors want to better/test calibrate Avanodes’ technique based on radar measurements ? Do the authors intend to investigate the granular processes at stake within the bulk of an avalanche flow thanks to particle tracking in both experiments and simulations, having in mind that by construction Avanodes’ experiments and com1DFA simulations have their drawbacks and have fundamental differences (see detailed comments) ? I think the authors should explain more explicitly their objectives and try to propose a better structure for the paper so that the key objectives become clear.

Although I have the main concerns above regarding this initial version of the manuscript (see my details comments below) I think that the content of the paper is original and of great interest and I would be happy to read a thoroughly revised version of the manuscript.

Detailed comments

C1. line 26. At this stage of the paper, it is not clear which type of particles you are speaking of here: there is a large spectrum of particles in snow avalanches from 'tiny' particles of ice, that are relevant to aerosols (see for instance Rastello et al, J. Glaciol. 2017), to 'much larger' snow aggregates, relevant to dense snow avalanches, either dry or wet (having in mind that wet snow 'granules' are larger than dry snow 'granules')? I would suggest you to say that you deal with dense avalanches and snow granules/aggregates --made of hundred/thousands (if not more) of snow grains, right at the start of the paper (though it is obvious to you I think).

C2. Table 1 ; and lines 83-84. The density of the Avanodes: why choosing such densities of 688 kg/m3 and 415 kg/m3 ? I would expect more explanation. It seems obvious that the trajectories of isolated particles are influenced by their density. Moreover, you don’t give the size of Avanodes, which is another important input. Could you please elaborate more on this? It has something to do with the following comment (C3).

C3. lines 62-63. Not easy… as there is one more degree of freedom in reality compared to the depth-averaged simulations used, there are potential errors. For instance, segregation processes, mixing, and the existence of secondary flows in granular flowing media is well established and we may expect strong differences between trajectories of single snow granules in reality and synthetic particles in a depth-averaged simulation framework that do not take into account all this complexity.

C4. lines 70-71. There are obvious/fundamental differences between particle tracking from Avanodes’ experimental technique and particle tracking in the simulations: i) the particle are not the same, ii) the Avanodes have a size and a density (and a shape), are mixed with a snow granular assembly and subject to a number of complex granular processes (size and density segregation, geometrical trapping, mixing, secondary flows, etc.) that are not taken into in the simulations. As such trajectories shouldn’t be the same in the end. I think you should detail and explain these crucial differences earlier in the text and then explain why the cross-comparison still remains relevant? And what is your key objective of that cross-comparison ? Beyond the errors inherent to each approach (deciphering the true trajectories of the snow granules is challenging), do you expect little or huge gap between the measured and simulated velocities and trajectories ?

C5. lines 94-95: why ? Please explain here or refer to the explanation that will come later (see comment C10)

C6. line 105. Why 'a minimum'? Could you please elaborate. I think you just implement a constant yield stress that may refer to a ‘cohesion’ yield stress and add it to the total Voellmy stress.

C7. line 115: why such a huge upper value for the turbulent friction ?

C8. line 117: the range of ‘cohesion’ you use is small, finally (when compared to a typical frictional stress : \mu \rho g h (easily more than 1000 Pa based on h = 1 m, density of 300 kg/m3 and \mu=0.5). Why not using a classical Voellmy law without cohesion ? Could you please elaborate more on this point.

C9. lines 134-135: yes, this is an important (crucial) point. Could you elaborate more on the expected density of snow granules versus the one of AvaNodes? Did you have any measurements of snow granules after the event. Or at least an indication of the type of snow (dry, wet) ? Typical size of the snow granules ? I would suggest that you give more information on the survey of the specific avalanche event investigated, right at the start here. Some information is provided a bit later (see around line 325) but it is not enough I think.

C10. lines 136-137. This explanation should come earlier (see comment C5).

C11. lines 147-148. Yes, this should be emphasized earlier / clearly stated I think to help the reader to follow.

C12. lines 149-150. Yes, and much more than that: grain size segregation, grain density segregation, mixing, and secondary flows, and even changes over time of the snow granules as a direct effect of competitive aggregation and crushing of snow granules. I think you should elaborate more on the complexity of granular processes involved in snow avalanches and refer to key papers about those granular processes. I have for instance in mind the recent work by Marks and Einav (Geophys Res Lett 2015, Granular Matter 2017). This is important to highlight/state the gaps between the true snow granules’ dynamics and the one from AvaNodes. I don’t even speak of the gap between AvaNodes and simulated depth-averaged particles.

C13. lines 160-165: OK you are speaking of particle distribution in your simulations. But what about the size of your depth-averaged particles ? How does it compare to the typical size of snow granules and/or the size of Avanodes? Does it make sense to compare this? Could you discuss more on this?

C14. line 189. At the first glance, I had in mind that the 'vertical' component of the velocity (=normal to the local slope) of the particles in simulations is by construction zero (depth-averaged). But as the reference frame is the absolute Cartesian one (x,y,z), I'm realising that v_{z,i}^sim is not nil. I think it would be great to show (x,y,z) frame in Fig. 1 and better state all this.

C15. lines 218-220. I think this is a nice outcome of your study. I would suggest to put more emphasis on this result if this was a key objective.

C16. line 225. the values are different from the particle velocity analysis. Could you comment on this. This is an interesting point. However, we cannot exclude the fact that as you are tracking particles that are different (avanodes or simulated one) and different from the true snow granules (see other previous comments) the particle velocities are not representative of the true snow granules in the end. As such, shall we rely more on values coming from the front (radar and simulations) rather than on the values coming from particle velocities ?

C17. line 237-238. OK you are comparing to the suggestion of Gauer (2014) for maximum velocity scaling. But what about the key difference in the curvature between the simple Gauer’s prediction and the trend of your results ? Could you comment on this ? Could it stem from complex interplay with specific topography, or would it be something else.

C18. line 264. Yes, there are key granular processes behind this observation. I’m still not convinced about the generic conclusions we can get from just three AvanNodes with two different density (that are quite high in the end). And again what is the size of the AvaNodes ? Maybe some minimum information about AvaNodes should be added (reference to key previous papers is not enough).

C19. line 295. Would it mean that taking into account processes at the grain level for snow avalanche modeling remain secondary ?

C20. lines 313-314. One single particle only cannot be representative of the whole avalanche process of course, in particular at the end when more and more particle are located at the tail. Be cautious. Again, wouldn’t it be better to rely on front position?

C21. lines 334-335. Yes, of course. We expect some frictional hysteresis between the front (inertial granular regime) and the tail (much less inertial regime), as well as complicated phase transitions in terms of densities, as known from measurements by Sovilla et al at Vallée de la Sionne.

C22. section 4.2. I’m no so convinced by the importance of initial and boundary conditions for isolated particles…

C23. lines 353-356. Not sure to get your statement. As snow granules interact through complex interactions during the avalanche I don’t really see how/why their trajectories would be primarily controlled by the initial position in the release area.

C24. lines 362-362. This kind of clear statement about the differences between simulated particles and Avanodes, and real snow granules should be said much earlier I think.

Editing issues / typos / suggestions

- title: after having read the whole paper, maybe something like this could be another relevant (if not better?) option for the title: "Particle and front tracking of one single avalanche event from inflow sensors and radar measurements backed-up with simulations".

- abstract needs a thorough revision. I think the abstract in the present state is too long. It needs to be shortened, more synthetic. I think it has something to do with major concern that we don’t really know what is the key objective of the paper.

- line 3 (abstract). I was initially surprised by “thickness integrated”. The term "depth-averaged" may refer to a more common semantics... But I've checked Tonnel et al (2023) and saw that there were some explanation about this. I think it would be pertinent to refer to Tonnel et al for this specific choice of the semantics.

- line 4 (abstract). I would invert: "the open avalanche framework, named Avaframe."

- line 25 (introduction). The start here is quite surprising (if not weird) with this sentence alone. Could you elaborate a bit more and cite some relevant literature?

- figure 1, top and bottom right plots: there are issues in the inserted legend (top plot) and x,y labels (bottom plot)... Please revise.

- line 104. Please be consistent along the whole manuscript: use a '-' between ‘thickness’ and ‘integrated’ or don’t use it but stick to one single option.

- caption of figure 6, second line: there is a typo: an empty space is missing after the comma.

- caption of Table 3, second line: there is a typo “… delta Z ) (Fig. 8) ...” Please fix it.

This is a unique and well organized study of avalanche dynamics, which uses in-flow particle sensors, FMCW radar, and avalanche dynamics simulation.  The approach is novel and provides new insight into parameterization of avalanche flow models.  This is a fundamental contribution to this field, and should be published after minor revisions.  I have some main points to consider below, and have attached an annotated PDF with detailed suggestions.

1) Sensor details.  Some additional details are needed about the sensors - where is the radar located?  What frequency range and range resolution?  How accurate do you expect the avalanche front observations to be?  This is important to help the reader interpret the results.

2) How big are the AvaNodes?  This isn't currently included.  How does the size and density compare to estimates of the snow particles in the avalanche?  Why were the two different densities chosen?  How were the density/size of the actual particles estimated?

3) Some discussion of the avalanche type (dry vs wet) and the flow regime that the sensors and simualtions are representing is needed at the beginning of the paper, as the methods are introduced.

4) How was the accumulation in the starting zone estimated?  How about the fracture depth?  It is stated that an interval board was used and estimates included an assessment of wind redistribution.  How was the wind redistribution component estimated?  Was the release volume varied in the simulations?  Seems like this would be a sensitive parameter, similar to the friction and other parameters investigated.  Some discussion is warranted here.

5) An overall summary of the uncertainties in all the estimates used for the assessment would be helpful - for example, what is the uncertainty in the vertical velocity? The avalanche front position?

6) Why was only C10 focused on in the analysis in terms of the error assessment?  Was it possible to find a set of parameters for all 3 AvaNodes that gave a low error?  I realize this might not be possible when including the avalanche front position, but what about just for all 3 AvaNodes?

Overall this is an excellent paper!