Manuscript “Moho depths beneath the European Alps: a homogeneously processed map and receiver functions database” presents an exhaustive database of P-wave receiver functions computed for teleseismic earthquakes recorded by multiple networks of seismic stations across Europe, with a central focus on the AlpArray Seismic Network. The major highlights of the manuscript are: (a) homogeneous processing scheme to compute RFs, which includes multiple quality and signal-to-noise checks on the waveforms, (b) time-to-depth migration in 3D, (c) Moho map created from the manually picking the discontinuity signal, (d) open avaliability of the RFs, codes for computation of RFs and the Moho map results. Overall, it is a well written manuscript and documents each step in fair amount of detail. I have a few observations regarding the data availability, quality checks applied and the results presented.

1. I downloaded the radial and tangential RFs from the repository, provided with the manuscript. However, the 3-component waveforms from which the RFs have been calculated are not provided. The RF waveforms have signal from 0 to 70 s, with the time marking starting ~30 s before the P-arrival time. Only 40 s of the P-to-S converted signal is provided. This is sufficient to observe crustal phases, but not upper mantle phases. Moreover, additional information about data processing e.g. Gaussian filter parameter (if any) applied to the waveforms during the RF computation is not provided. This restricts the scope of use of the data by other users. In my opinion if data is suppose to be made open and available to the community, it should be done so that (a) the RF computation can be re-done by an independent user, and (b) information/analysis, other than the one presented in this manuscript can be extractable/done. Providing the 3-component waveforms for 45 s before the P-wave arrival and 120 s after, will allow users to compute P-RFs using other algorithms, different frequency content, vary the number of iterations (other than 200) and also study structure beyond the crust.

2. An event list of all earthquakes used and the detailed information of all the stations (e.g. lat, long, elevation, instrument type and, if possible, response files) should be provided, so as to enable the user to perform time-to-depth migration for each trace using different velocity structures.

3. The CCP stacks uses only the P-to-S converted phases for ascertaining the Moho. These can have significant dependence on the velocity structure. Using the converted phases would reduce such dependence to a large extent.

4. The CCPs presented in this manuscript serve two purposes (as I see it): (a) provides confidence to the data quality and uniformity of analysis (b) re-confirms most of the Moho structure observed from previous studies. Aa this manuscript is more of a data mine article, I believe that the discussion presented on the results is acceptable. Given the quality of the analysis and the results, I would have been tempted to discuss the results (variation in the structure) in greater detail and also correlate it to the geology/other geophysical observations.

5. A few minor points:

(i) I did not entirely follow the filtering scheme of the RFs. If the data is filtered between 0.05 and 1 Hz (L130), why perform a high pass filter at 1 Hz (L125)?

(ii) L129 - What is meant by “the effect of the signal”?

(iii) Why is the time referencing of the RFs from -30 s of the P-wave arrival time and not at the point of the largest amplitude arrival?

(iv) L100 – Iterative deconvolution used for the RF calculation does not “deconvolve the vertical component seismogram”. It follows a convolution of the updated spike train with the vertical to match the radial component.

(v) the CCP is done as a 3D migration, but the 3D models are not presented. This would reveal the influence of the 3D model in the final Moho maps obtained.

(vi) L300 – why are uncertainties “difficult to assess”?

The present manuscript (and dataset) provides a consistently processed database of receiver functions as well as a crustal thickness map of the Alpine region that is a product of the AlpArray initiative. This large dataset is the result of considerable effort in data acquisition and processing, and will be a very useful and widely utilized resource for the community. Overall, the manuscript is reasonably well written, and as it is mainly intended as a data description article it makes sense that the authors refrained from going into interpreting the results. However, some parts of how the presented Moho map was obtained deserve a clearer explanation and description, and some choices on what data are made available should also maybe be reconsidered, so that moderate revisions will be necessary. I will outline my main points below, followed by more specific comments by line number.

General Comments:

1. The main product provided here is the crustal thickness map for the Alpine region, which was manually picked on a series of CCP stacked receiver function profiles. While the receiver function processing and quality checking procedure is nicely and comprehensively described, there is no detail at all on how the manual picks on the CCP profiles were retrieved. This part of the analysis is a complete blackbox at the moment. I recommend to add:

•  A description on how manual picking was performed. What guided finding the right anomaly in the profiles?  Was the pick set in the center of the anomaly or onto the maximum amplitude? What was the procedure in case of a double anomaly? Was any interpolation performed for sections where the Moho was not really visible, etc.

• These descriptions could be accompanied by one or two examples where the set picks are shown on top of the CCP profile...at the very least, they could be added to the current Figures 7 or 9 (at the moment, as far as I can see, the manual picks made in this study are not shown anywhere in the article)

•  It is also mentioned, without further explanation, that these manual picks were labeled as either certain or uncertain.  Based on what was this labeling performed, again it would be interesting to see examples

2. The other reviewer had a number of comments about the provided datasets, with which I wholeheartedly agree. It would make a lot of sense to also provide the raw, cut three-component waveforms, so that other researchers can apply different rotation (e.g. 3D, into LQT system) and/or deconvolution approaches.

3. The comparison to previously existing compilations of crustal thickness in the Alpine area should be extended, at the moment there is only a quite brief section on this, and the comparisons in Figures 7 and 9 are only along selected profiles, do not show the picks of the present study, and make it difficult to appreciate the differences due to the large scale of the cross sections. I would recommend to compile a map view figure that shows absolute differences between the new crustal thickness map and one or several pre-AlpArray ones.

Specific comments:

l.1: Unnecessary first sentence; this may fit into the Introduction but not into an abstract

l.14 (and elsewhere throughout the manuscript):  why say crustal structure when you mean crustal thickness?

ll.19-32: this very basic introduction to the Moho is not necessary and makes the Intro chapter rather unstructured. Better leave out

ll.62-67: that is comparing apples and oranges. Tomography studies yield crustal velocity structure, whereas RFs give the crustal thickness (and NOT crustal structure, see above)...this means that the two methods are largely complementary

ll.79-88: quite repetitive, maybe better to present this basic network information in a small Table?

l.94: as well as having: sounds clumsy, please reformulate 

l.108: maybe mention how the orientation was determined for the ocean-bottom instruments?

l.130: this means that the last quality criterion (STA/LTA) was determined in a frequency range (f>1 Hz) that has basically no overlap with the one that is finally used for deriving the RFs (0.01-1 Hz). This seems like a rather strange choice to me.

l.146: how is this amplitude range chosen?

l.155: Unnecessary to go back to the very first RF studies here

l.158: I thought that the first uses of CCP stacking in RF analysis was by others...not sure who was first, but studies like Yuan et al.  (1997) or Kosarev et al.  (1999) already showed CCP stacked RFs

ll.170-175: circular text flow

l.174: grid spacing [...] consists of three layers: no, the model does! Please reformulate

l.175: Was some kind of half space added for the region below the EPcrust model (representing the mantle)?

l.178 (and later): ray trace paths → ray paths

l.197: look (-s)

ll. 198/199: explain in more detail or provide a reference!

l.203: I would not call the ray coverage shown in Figure 5 great. There is a gap of 90-100 degrees in southern directions, and at least one station also has very few RFs from westerly directions

l.204: maybe use amplitude instead of strength?

l.213: this was done using the EPcrust velocity model mentioned before?

l.216: maybe supply what the range of horizontal offsets from the stations is...this can then be compared with station spacing

l.224: I fail to understand what direction exactly East-East-Northeast is supposed to stand for

l.226: distinguish from what?

l.228: Hard to compare in the profiles because this study's picks are not supplied, and scale is rather large. Maybe better to plot residuals somehow? (see General Comment #3)

l.241: Needs more detail on how picking was performed, and what the criteria for certain/uncertain picks are (see General Comment #1)

l.256: just semantics, but aren't routines always systematic?

ll.260-262: This should be analyzed in much more detail, and I would appreciate some kind of map view residual plot compared to at least one previous study (see General Comment #3)

l.271: how was the presence of a double signal (overlapping Mohos) treated in the present study? Was only the shallower signal picked, were both picked, or what? The manual picking procedure needs some more explanation!

l.275: 10 km is quite substantial

l.300/301: quality and consistency of the manual Moho picks...these are a complete blackbox as is, no explanation of the procedure is given and no examples are shown       

l.307: Would a map of mists between dierent models not be a more straightforward way of identifying critical regions? Also, as your Moho picks have labels for certain/uncertain, can the spatial distribution of these labels be shown?

ll.309-311: Clumsily formulated, please change

l.312: These meetings are not relevant in the Conclusions of an article

Appendix:

In the text says that three figures are contained, then the text describes four, whereas the actual content is five

Figures:

Most figures: I do not understand why the authors use color scales with (often very few) constant colors for rather large ranges of values. Using a continuous color scale would, in many cases, give more detailed information using the same plot

Figure 2: Why plot a line for the close cut-off distance (30 degrees) but not for the far one? And why choose a color scale for depth that assigns constant color for 70 km intervals, instead of taking a continuous scale (see above)?

Figures 4/5: is there any logic according to which the sequence of the highlighted stations (a through d) was chosen? Naming according to position (e.g. start with a in the W and move E, or something similar) would seem more straightforward

Figure 7: It would also be interesting to see where the picks performed in this study are situated

Figure A2: Here you show many many dots that are all on top of each other, thus it is really hard to see much...could you maybe plot point density instead?

Figure A3b: The absolute number of discarded RFs per station is not that interesting, could you maybe display the proportion of discarded RFs per station (i.e. what percentage of all RFs for that station was discarded)? This would be more of an indicator of data quality and less of data amount