Dear Editor,



I have read and review the manuscript r.avaflow v4, a multi-purpose lansdslide simulation framework with great enthausiasm.

It represent a great innovation in the landslide modelling field, together with an open source code where everbody is able to apply these innovations for analysis and assessment. Overall a great read, and a major landmark paper in the landslide modelling field.

It should definitely be published by the journal, and I will personally use the contents in further research substantially.



It features a thorough testing of the software and model implementation, with detailed considerations, thorough testing and case studies, as well as illustrative figures. Though, some of the derivations seem to be derived somewhat idealistic.



I do have some suggestions for the authors to adapt and perhaps improve the text, which are given below:

Introduction: It might be good to get a single paragraph describing also the state of other modelling developments, and how these deal with these issues. For most, r.avaflow4 seems a unique and novel solution, but perhaps not all. There are windows-based systems with interfaces, with nicer visualization. The challenges are now perhaps a bit too focused on r.avaflow4

Line 55: with rather unsuccessful earlier attempts to create a powerful Windows-based stand-alone version of the tool

I would remove this, you are offering a great tool, and the fact this didn’t work out shouldn’t matter to the reader here.



Line 67: I would rewrite this a bit (Basically, it transports mass and momentum of up to three phases through a regular grid. -> The model applies physically-based equations to estimate and calculate numerically the motion of mass and momentum of up to three phases...)



2.2: I get the continuity derived basic advection framework of the euqations is a bit duplicate, but it might be good to explicitly define the three phases at least for the reader here., with a small figure with definitions of phases, velocities/grid



Line 105: Will there be a warning? Or is the scope of realistic parameter combinations difficult to detect (That seems very possible to me, although basic errors should be simple)



Table 1: Might I suggest putting the parameter explenations below the relevant equations, or right next to it? Now its hard to read the table and go back and forth to the small table caption.



Table 1: Can the ambient drag coefficient be spatially distrinct? Is there a parameter that can be used to depend on land cover variations for energy loss in fluid like flows (forest vs grassland for example?)



Line 140: I read this as meaning the phases are separated to form the individual layers? But that seems to come with more complications mathematically for the momentum transfer between these layers. Can individual layers be multi-phase? Now the term layer and phase are mixed and were a bit confusing to me.



Line 155 – 165: I am not completely against this approach here. Data collection and lab results would be prohibitively difficult, but currently this could use a rephrasing I think. Principle i) seems somewhat strange to me, layer surface elevation or layer absolute height? I sort of see a geometric explanation for why this increase with the angle of the interface would work to replace the general drag force normally used by the authors, but in my mind assuming the drag coefficient is 1 for a vertical interface breaks some of the original assumptions in the derivation of the drag coefficient in the earlier works of the authors.



2.5: How do the authors deal with the often observed influence of the increasing and decreasing pore pressures during the compression and spread of slow-movinglandslides?

It seems now the dynamic pore pressure is not a remaining contribution to the momentum equations? I do like the approach to accommodate more types of movements, but this limitation might be mentioned or explored?



2.6:It would be nice to see a figure here with an example already (even though its shown later in the manuscript)

Table3: I find this table not so needed in the full story. I think the point is made well by the sentences above it.

4.1: To what extent is the discretization of the layers a major influence on the final output? Now these discrete layers move separately as units, but could you remark on the balance between the influence of the modle assumptions/numerics here vs the physics of such a hypothetical slide?



Line 365: Very good visualization of the differences. I would perhaps want to ask the authors to clarify if they think the 2-layer approach would be the only valid one for such flows, or could, a single-layer approach with non-fragmented physics, potentially get equal behavior?

Line 480: Would it be possible to detail somewhat more the numerical scheme in an appendix, as its mentioned several times, and provides some of the key observations in the application cases. Besides the citation from Tai et al., these types of models nearly always require particular details for their numerical solution. I do agree with the point, that diffusion is a necessary, and unfortunate by-product of these schemes. Although approaches with SPH are, as the authors say, perhaps needlessly complex. A potential direction could be a consideration of the non-linearity/non-smoothness of the terrain and layer data. Discontinuities of a landscape in flood models require particular attention in terms of hydrostatic reconstruction and flux limitation as to prevent diffusivity of the flows.

General comment:

The paper describes the new features of a landslide simulation software and presents numerical simulations highlighting the relevance of those new functionalities. The new features seem indeed substantial enough to justify a publication. The method and results are presented clearly and concisely. The description of the numerical experiments seems precise enough to allow reproduction. 

The paper gives a rather clear description of the model, and of its new functionalities. The model is, for the most part, understandable without previous knowledge of the software. This is a very positive aspect, but some parts of the model description lack some clarity and could be improved (see next section).

The paper clearly motivates the need for the new features. The test cases are well chosen and the comparison between the different models is very well presented and easy to follow. The limitations of the model are also mentioned in the discussion, which is a good point. 

After some minor modifications in the model description, I think the paper should be accepted for publication. 

=============================================================================

Specific comments: 

Section 1.

L27. The acronym GIS is defined nowhere. You could add what the acronym stands for.

L39-58. It is not clear whether the prototypes mentioned in the list (slow-flow process and deformation control) are the new features presented in the rest of the paper, and if there are already available in previous versions of avaflow. 

L52. If the computational experiments are published results, please add the reference. 

L55. What is GRASS GIS? A software? 

The introduction is overall very clear and well structured. However, there is no reference or comparison to other mass flow simulation tools (e.g. Shaltop). Putting some context here would help to situate r.avaflow. 

Section 2. 

L75. Please write what the TVD-NOC acronym stands for.

L76. The information that the grid is moved half a cell does not seem relevant here. Either add more details or remove the sentence. 

L79. Please specify what does a hydrograph record in your context. 

L90. It is not clear whether the indices x,y,z denote derivation with respect to x, y, t, or if it is just the name of your variables. 

L91. The flux are expressed as the derivatives of which wuantity? A word seems to be missing here. 

L91. I don't understand how you can neglect the time derivative since all your experiments are time-dependent. Or do you mean that you won't describe the term F_t in this section? For the sake of completeness, the term F_t should be included in Table 1. 

L108. I think it should be $g_x^*$ instead of $g_x$. 

Table 1.

- For the fluxes in x and in y directions, please add parentheses to show clearly which terms are differentiated. 

- For the internal deformation in the fluid, the "+" sign is not necessary.

Table 2. 

- The expression "mechanically controlled" is defined nowhere in the text. Unless it is a classical notion in the target community, please explain what it is. 

L144. If I understood correctly, the layered model is described by the governing momentum Equation (1) and Table 1. When you describe the controlled components (gravity, flow height, drag), you could refer to their notation, either in Equation (1) or in Table 1. It would help a lot to understand how the equation is actually used.

L149, 150. Similar remark: the notation h_e appears nowhere in the equations before, so it is difficult to understand what role the effective flow height plays in your model. 

L155. Please specify the value of C_{DX} when the conditions are not met. 

L155, 156, Figure 1. The equation is written for the phases P1 and P2, but the Figure 1.(b) shows results for P1 and P3. It is not clear to me whether P2 necessarily represents the mixture layer, and if its properties are fundamentally different from P3. 

L175. Same remark as before, the deformation factors f_dx, f_dy do not appear in the governing equations.

L176. If you have only one parameter f_d* controlling both deformation factors, it means that you don't have complete freedom over the choice of those deformation factors. Does this choice of parametrization have a physical meaning? Could you also give practical cases, for example: which value should be given to f_d* to completely lock the transformation? 

L182. The choice of vocabulary is a bit confusing: "deformation control" and "sliding model" refer to the same functionality, but seem to describe very different physical behaviors. 

L182. Please be more precise: which combination are not possible, and why? 

L184. The gravitational acceleration of the entire landslide seems to be a new functionality that has not been introduced earlier in the paper. What motivates this new feature? Is this case still described by the momentum equation (1)? Again, the parameters g_x, g_y, g_x^g, g_y^g, f_g do not appear in the equations before.  

L192. Since you are giving the order of magnitude for the velocity of slow landslides, you could also give it for extremely rapid flows. 

L200. There is a notation conflict, since F is already the flux in Equation (1). 

L213,214. The viscosity term and the ratio between kinematic viscosity and the square of the basal layer are denoted by the same letter \xi. 

L215. Which equations are you finally solving? The model described in Section 2.5 seems very different from the previous sections, so I assume you are not using the Equation (1). Can you write the governing equation? Can the slow-flow model be combined with the 3-phase model or with other new features from r.avaflow? 

Section 3.

As a general comment, I found the description of the test cases very clear and well written. 

Figure 2. 

- I don't understand why the two other layers are not visible in this view. Could you choose another point of view or add a figure to make all layers visible?

- The colors are very dark, and it makes this figure difficult to read (in color, and even more when printed in black and white) 

- If possible, please add the axes in the figure

L278. Is there a motivation to take two layers in this test case, apart from testing the new feature? Does each layer represent a different type of rock, for example? 

Figure 6. 

- Is this figure an output from the software? Please add the scales like you did in Figure 3(d) or 4(c) and the axes.

Table 4. 

- For SD1, please specify which values correspond to P1, P2, P3. 

Section 4. 

Overall, the figures of this section look good and are very readable. 

A general comment for the Section 4.3: the notations FO1, FO2, FO3 and FOC1, FOC2 are quite confusing. Can you use another notation for the control points?

L384. "Displacement wave" is not very precise, replace by "water wave"

L390. "only slightly decreasing thereafter": did you mean decreasing or increasing? The curve does both. 

L398. It took me some time to understand that you show the results of two different simulations on each profile (the dry simulation and the fluid and solid simulation). This is indeed mentioned in the end of Section 3.3, but you should recall it in the description of Figure 9. 

L400. I agree that the landslide becomes less mobile for FO1, but for FO2 the landslide profiles almost overlap, how is it less mobile?

The figure 10 is very nice, but is described nowhere in the text. You could maybe put it in Appendix. 

=============================================================================

Technical corrections:

L95. "This implies that in sum the terms do not induce...": the formulation sounds weird. Add a comma before and after "in sum", or rephrase. 

Table 1. When printed in black and white, it is almost impossible to tell the difference between the blue and the green letters. Please consider using colors with a stronger contrast (also for color blindness). 

L135. "The way how gradients are computed,": I don't think "how" is necessary here, please check. Also, there should be no comma after "computed".

L136,137. The whole sentence is clumsy, please reformulate it. 

L200. Please separate more clearly the two equations, for example with \quad. 

Figure 3. "View direction in A" and "Profile in D": Replace "A" and "D" by "(a)" and "(d)".