This paper is a well-written and interesting analysis of crustal Vp/Vs from receiver functions in the Corsica-Sardinia microplate, with potential implications for processes and timescales of metasomatic alteration of continental crust. The authors find low Vp/Vs < 1.75 in the Variscan batholith of most of Sardinia where crust is thicker, and higher Vp/Vs in Corsica and the southwesternmost part of Sardinia. These patterns are interpreted in terms of silica content and mafic to ultramafic, supported by thermodynamical modeling of properties, but perhaps a more fruitful way to interpret these (as detailed further below) is in terms of quartz abundance reflecting a history of metasomatism in low Vp/Vs regions and more mafic lithologies elsewhere.

One caveat worth making in the paper, and remembering in the context of interpretation, is that uncertainties in single-station estimates of Vp/Vs using H-k stacking techniques can be quite large. For example, raw one-sigma uncertainties are greater than 0.1 based on the variance at ~0 distance separation for variograms of USArray data in the automated EARS database (Crotwell & Owens, SRL 2005), see e.g. Fig. 2c of Lowry & Pérez-Gussinyé (Nature 2011). The authors appear to have done a very careful analysis here, and judging by the stacks in Figures 4-7 the younger crust of the study area is less structurally complicated than typical North American continental crust in USArray, but even so uncertainties are likely to be of order 0.07 or larger here and possible impacts of that should be addressed in the discussion.

There are a couple of other issues that it might be worthwhile for the authors to consider in a revision of the paper, described in greater detail in comments tied to §2.1.1 below. One is that there appears to be some sort of bias error in Perple_X outputs of Vp/Vs for crustal mineral assemblages, resulting in much lower modeled values than those measured in the lab for corresponding rocks. The practical significance of this is that an exotic (serpentinite, eclogite, or supersolidus) lithology is not necessary to explain higher Vp/Vs in the study area; a gabbro would be sufficient. However it also means that the absolute values of Perple_X-derived Vp/Vs are less useful for interpreting these results than how Vp/Vs changes for different chemistries and volatile contents. I would also suggest that it’s useful to recognize that Vp/Vs variation is dominated not by %-Si so much as by %-qtz, because of the unique elastic properties of quartz. This enables the use of Vp/Vs as a proxy for metasomatic history, as much of the Vp/Vs difference for dry and hydrated lithologies in Figure 9 is related to breakdown of feldspar to quartz and mica (Ma & Lowry, 2017).

§ 2.1.1 Thermodynamical (Perple_X) modeling:

The choice here to examine various bulk compositions but only use one constant (0.25wt-% H2O) volatile state unfortunately obscures one of the most significant potential takeaways for interpretation of the results. Namely, the primary factor in determining bulk crustal Vp/Vs is the abundance of the mineral quartz (Christensen, JGR 1996; Lowry & Pérez-Gussinyé, Nature, 2011). This does of course depend to some degree on SiO2 content, but it is much more sensitive to whether water is present to react with the bulk constituents, which breaks down feldspar into quartz and mica (Ma & Lowry, Tectonics, 2018). Hydration reactions that break down feldspar also presumably depend on whether CO2 is present to buffer those reactions (Yardley, J. Geol. Soc. Lond. 2009). From that perspective, it seems to me that a more useful approach to examining Vp/Vs with Perple_X is to use the bulk compositions from the rock environment of interest but vary the volatile mix, and then interpret the variations primarily in terms of hydration history.

Also, as an aside: There is a problem of some sort in the elastic parameter database of Perple_X, because it gives Vp/Vs seismic velocity ratio estimates that are consistently about 0.05 to 0.1 lower than the corresponding values from Christensen’s (JGR 1996) measurements. This becomes apparent if one compares the 1.71 to 1.86 range of Christensen’s measurements of granite to gabbro in Fig. 1a of Ma & Lowry to predictions in Figure 9 of this paper. For that reason, Ma & Lowry did not show absolute Vp/Vs in their Fig. 14, but rather the perturbations with water versus without water present in Perple_X thermodynamical modeling that used a similar database to this paper. Xiaofei Ma spent significant time and effort trying to figure out where the problem may be coming from during his dissertation studies, but we were unable to track it down. Since attenuation effects are likely to be larger for Vs than Vp, that is one candidate for the discrepancy, but I am somewhat skeptical that attenuation would be that significant for small-scale room temperature samples like those in Christensen’s (JGR 1996) database. Because of this, it is perhaps safer to use Perple_X as a tool to examine relative Vp/Vs for different choices of composition or state than for purposes that assign meaning to the absolute Vp/Vs. For purposes of this paper, the conclusions regarding likely mineral assemblages are probably still valid (within the large uncertainties that are inherent in single-station H-k stacking estimates of Vp/Vs), but note that robust measurements of whole-crustal averaged Vp/Vs less than ~1.7 are extremely rare except when errors are present due to the perturbations of amplitude stacks by other reflectivity, dipping structure and anisotropy, as this paper notes can be present. Vp/Vs exceeding 1.8 on the other hand does not require an unusual composition like that of eclogite; it simply requires lower abundance of quartz. In fact, the mean Vp/Vs for the USArray footprint in the United States is about 1.79 (Ma & Lowry, Tectonics 2017) and Vp/Vs exceeding 1.85 is possible in crust where lithologies are gabbroic.

Lines 368-369: The change in Vp/Vs due to adding water to the chemistry is not primarily because of the reduction of the solidus, as inferred here, but because hydration reactions reduce the feldspar content of the mineral assemblage and instead favor formation of quartz and mica (Ma & Lowry, Tectonics 2017). The abundance of quartz dominates crustal Vp/Vs variations because quartz has a very unusual Poisson’s ratio translating to a Vp/Vs less than 1.5 (Lowry & Pérez-Gussinyé, Nature 2011, based on measurements in Christensen, 1996).

Lines 415-418: As noted above, a high Vp/Vs coupled with high Vs does not require serpentinite or some other exotic mineralogy to explain; rather a Vp/Vs up to 1.88 and high shear velocity can be expected for a common mafic lithology. Where Vp/Vs exceeds 1.88, it can probably be attributed to the large uncertainties expected for single-station H-k stacking estimates of Vp/Vs.

§ 5 Conclusions

This section does not tie back strongly to the information presented in the previous sections. It seems to me that the modeling in the discussion section does reaffirm earlier work suggesting that hydration lowers Vp/Vs, and so– given that Vp/Vs in Corsica and southwestern Sardinia remains high– however-much hydration may contribute to regional volcanism, it does not appear to translate to substantial metasomatic modification of the crustal lithology in those particular locations (which is slightly different than what is said in lines 425-430). I also don’t know that I would strongly emphasize evidence for a thermal anomaly in Sardinia, given that Vp/Vs is insensitive to temperature. (And although it may be sensitive to partial melt, in practice the crustal averages of Vp/Vs derived from H-k stacking are not.)

Figures 2 and 3: These are a bit hard to interpret because the y-axis back-azimuth for the left and center panels is nonlinear, and must be inferred by looking back and forth to the right panel. It would make more sense to bin and sum, or if showing individual traces is preferred, to plot each trace as a linear function of back-azimuth (even if they then overlap). This would aid in identifying patterns expected for dipping layer boundaries and/or layer anisotropy (e.g., Schulte-Pelkum & Mahan, EPSL 2014).