Review Reports

Round 1

Reviewer 1 Report

Review of manuscript “The 2019 eruptive activity at Stromboli volcano: a multidisciplinary approach to reveal hidden features of the “unexpected” July 3, paroxyxm” by M. Mattia et al. submitted for publication in Remote Sensing.

The manuscript uses a multidisciplinary approach to search for precursory signals for two “unexpected” explosive events at Stromboli volcano that occurred in July and August 2019. Routine monitoring had not detected or called out precursory signals. The analysis includes geodetic data, satellite and camera derived thermal data, and a discussion of very long period (VLP) seismic events. After analysis (and modeling of the volcanic source) the authors present a model for how the intermittent Strombolian explosions might happen and suggest a to have presented a forecasting tool for monitoring operations.

The manuscript falls short on many levels. I list the main shortcomings is followed by a huge, basically unedited list of items that need clarification or changes.

(1) Several papers have been published (some co-authored by some of the current co-authors) that already tackle aspects of the short-term geodetic precursors and the overall seismic behavior during build up of the event (like RSAM etc.) (references 3 and 4). Results from these other studies (in particular the seismic study by Giudicepietro et al.) are not adequately explained and included in the overall discussion. Certainly the model presented here could draw from the other studies or at least be consistent with results found there; this is not clear to me from the current manuscript.

(2) The introduction contains an entire paragraph of summary results and interpretation. Why in the introduction and without any evidence? The paragraph is pretty strongly worded that ‘classic paradigm of magma … is no longer able …’ I don’t see that and sentences like that should be used after strong evidence for their validity have been presented.

(3) On the other hand, since the manuscript presents a model for explosive activity and suggests to provide guidelines for a forecasting tool, it completely lacks bringing the 2019 events in a broader context. What is the history of these events? How often do they occur? How similar are the 2019 events to any previous events? (reading other papers that deal with the 2019 events, it is clear that they are not very common on Stromboli; but why do I or any potential reader need to consult other papers to get that information?). If you want to present a model and ‘tools’ then much more context is needed. Generating a model and suggesting to be able to forecast further occurrences based on a single (2019) episode seems implausible to me. Lack of overall context, lack of general question, and no framing of the activity with ‘normal’ Strombolian activity or possibly bringing it in a more global context.

(4) What is the purpose of the model? It sounds like (with slight variations) like many other models for explosive volcanic eruptions. Is it necessary? The model description is very lengthy and seems almost devoid of information gleaned from the observations. It almost reads as if the model has been derived independently of the data. Overall, the model makes sense, but what is truly new about it and how do the data support it? What (existing?) models are now excluded? The model sounds like a ‘variable plug’ model (like reference 18?)? Why do you think your gas/fluid and fluid magma description constitutes a ‘new paradigm of magma’? I think this part should be shortened considerably.

(5) The description of the different phases (5.1 and 5.2) prior to the explosive event appear ad hoc and arbitrary. I don’t see anything increasing or changing markedly at the beginning of May. The entire section 5.1 is weak; it reads okay but not clear on what the model is really based (besides sounding okay). Both sections are a mix of describing the observations in more detail with an explanation of what they could imply with their ‘model’; this is confusing.

(6) A large part of the manuscript presents the observations. There is little that connects the observations with each other and it is not clear how the observations were used to derive the model or how the observations can actually be used in an operational environment. Multidisciplinary is fine, but why this particular set of data? Because they exist or because they provide some unique insight to the events? From a seismic perspective, only a very small part (VLP events) are considered; many other seismic data products and analysis techniques exit and/or have been presented and proposed, why only look at VLP? There is no discussion of other seismic precursory signals (or their absence); why not look at and present VT or tremor rate of something like RSAM (or at least directly point to the reference showing these things). Please explain. 

(7) Studying a few months of data is insufficient to derive a robust forecasting tool. If the authors wish to do this, they would need to look at long time series of all available data (geodetic, thermal, and seismic - seismic much more in-depth than here) to identify patterns that could be used for forcasting. And to show the reader that signals seen in 2019 (geodetic, thermal, and VLP seismicity) are not common but are (are at least often) associated with paroxysms. Maybe the authors know that and assumed this is common knowledge, but it is not.

—

List of items that need clarification or changes:

- P1, L19. Suggest to remove “up to the task”

- P1, L23. Not clear what is meant “We framed our observation in a model …” — Maybe: “We developed a model that explains the observations. We call the model “push and go” where …”

- P1, L34. No need for ‘here.’ — Not clear what ‘emerging part’ means? Is it rising or the part above sea level? I’d also suggest to break up the sentence into two, it is long. The first one what you do and the second why.

- P2, L50. ‘distinguishable’ should be replaced, e.g., by ‘barely detectable signals’ or by ‘offers only weak precursory signals that are almost indistinguishable from background’ or something similar. You try to say that whatever was recorded looks not much different than ‘normal’.

- P2, L60 “The paroxysm of July 3”? Or were there multiple ones on July 3rd?

- P2, L56-70. This is basically a summary of the results with interpretation/discussion. What is this doing in the introduction? I think this part is misplaced and should be cut. Also, I don’t understand L62-66; how do the bombs, etc. emitted relate to the rest of the paragraph or anything that has been said so far. What’s the context? Are they unusual geochemically or in any other way? These 2 paragraphs don’t fit here. Why talk about the emitted material in this ‘summary’ section. What is the purpose? It becomes clear(er) later but in the introduction the reader has no idea why this is even mentions. The last paragraph is an interpretation and completely unjustified at this stage.

- P3, F1. The August part of the figure is not mentioned in the text and the caption should make it clear that the second sentence refers to the August eruption only (this is not clear). The entire thermal image needs more labels, like where is Stromboli (where the ‘hot’ colors are?), a scale, more explanation (in caption) what the image actually shows and what colors mean. Currently this part could be anything. — For people not very familiar with Stromboli (most readers?), please add a description of direction for the photo like “taken about X km SW of the island and looking towards NE” or whatever is correct.

- P4, L113-116. I do not follow the sentence. How often are pictures taken? Do you mean that the number of pictures taken depends on the level of activity? Not clear. And why? And what is Dh anyways if it is taken at different times relative to an explosion event, then the ‘dispersion’ of the thermal signal will naturally also be different; or are you interested mainly in overall increase/decrease of the signal? How can you compare images if taken at different times rel. to an explosion? This is not explained and cannot be understood from the text. 

- P4, L118. ‘related to each area,’ I don’t follow. The Figure 3  (in a) also just shows a green line, there is no ‘area’?

- P4, L119. ‘result’ rather than ‘analysis’?

- P5, F3. (i) You mean ‘(green)’ rather than ‘(blue)’ or remove (L126)? Green is the moving average of CD. Blue is labelled ‘MODIS [GW]’. (ii) The orange line has zero (0) justification; based on the line ‘potential dangerous’ occurs fairly often (in April 2019 too?). Could be interesting to see same plot covering a longer period to establish a ‘normal’? — (iii) Interesting to see that MODIS data, with two exceptions (?) do not indicate strong (precursory) increase because it is something that measures what has already happened, as such a longer period for the data would also be useful to understand what is ‘normal’. Also, are there any ‘events’ associated with the two exceptionally high values in late June 2019? — (iv) For a), the data have a 49-sample low pass (for the hourly measurements) (v) Don’t think the 3 month period is sufficient to show anything. Is there a normal with which one could compare the ‘unexpected’ explosions? Seems random so far.

- P5, L140-142. ‘Figure 3a shows a very good agreement …’ — No, I see more MODIS data points starting in June - that means more lower thermal energy ‘hotspots’? The number of the MODIS data points and their values do not change much while the green line (cumulative dispersion) goes from a low value in early June to high values at the end of the month. What ‘agreement’ is there? — Generally, explanations are lacking.

- P6, L150-151. ‘During the two explosive episodes …’ Do you mean July and August?

- P6. F4. (i) Please add/show a line or something when the explosion occurred; that way it would be easy to see that the uplift preceded the eruption by a short amount of time (can’t see the ‘ashes line) (ii) Please also mention that your time is in UTC (?) as the explosion occurred at 16:46 local time (per L72). (iii) why is the tilt signal at PLB more than twice as large as at TDF? (iv) what do positive/negative values mean with respect to uplift/subsidence of the crater? Please explain. (v) L157 “In blue, t.” Is something missing? Not clear. (vi) part f), after considerable filtering, show a ‘nice’ signal preceding the explosive event. It would be interesting to know, whether that signal actually stands out when considering a longer period than June to mid July [what is ‘normal’ is really not addressed …]. Do any of the other GNSS sites show any (similar) signal??? Need to say that. (vi) I’d make the labels for the GNSS sites bigger in d and e, they can’t be read but two are referred to in the caption.

- P7, L180-183. Don’t follow. A minimum in the strain value is not visible at 14:34:32 in Fig. 4f (a large minimum occurs after the eruption). Not clear why this is important (since it’s not visible in Fig 4f). This is the ’signal’ that has been shown previously as showing ‘something’ prior to the eruptive event? Would need to be shown differently (clearer) to make a point. Currently cannot follow based on lack of visible signal at time given.

- P7, 184-187. Not clear where this paragraph comes from. The 2019 events have a similar time history relative to each other (but there are no real precursors??)? Or do the 2019 events have a similar time history compared to a 2007 event? If so, this should be shown. But not in the middle of the data section unless you show (Figure?) similarities between the 2007 and 2019 events? So far the tilt/GNSS/etc. data of section 3.2 are focussing on the July event; so how can then the 2007 event be compared with the 2019 sequence? This is not clear. Also, how can you ‘suggest a common source mechanism’ (from the time history …) and then in the next sentence (in the same paragraph) talk about strain data (and lack of changes prior to the July event). How are those two things connected? This is starting to get tough to read.

- P7, F5. (i) L193. ’time between subsequent VLPs’ (ii) I like the longer period covered. (iii) If there is any ’signal’ then it is fairly subtle. Not sure if this would help, but (like Fig 4f) showing low-pass filtered median (or other?) values could bring out some features in parts b), c), and d)? Like in d), the large time intervals between VLPs ‘disappear’ but when does that start? Could be even as early as April? What I’m trying to say is: it’s difficult to make a case of ‘precursory signal’ from the current figure and enhancing some features could be a useful way of making something more clear

- P7, L198. I may have missed that but have you already called reference [9] before calling [10]?

- P7 L203-206. “change slightly … once … plotted together” — Explain what you mean, what is changing according to you? And “in contrast with clear variations … in strain VLP shape …,’ what do you mean? Which one of 5a) to f) shows those clear variations? Please say which part(s) you mean and point to them. You point two times to Figure 5, once to say ‘not much change’ and ’clear variations’ without telling us which parts of Figure 5 you are referring to in each case. Please clarify.

- P7, L206. I would start a new paragraph when taking about the two families.

- P8, 201-211. The words ‘disappears’ and ‘appears’ are used in a very different sense and that is confusing. How about L211 “The first family is a dilatational strain …’ (or ‘consists of a’, ’shows as a …’)?

- P8, L211-215. (i) what is the peak amplitude of the family 2 strain events? (it is mentioned for family 1 but not 2) (ii) Not sure what you mean by ‘balanced’ (like ’no offset’ but that depends on the filter?)? Family 1 is pretty symmetric (based on Fig. 6a left) while family 2 is asymmetric (Fig 6a right). (iii) Don’t understand sentence L213-215. What is ‘suggesting’ this to you, why?

- P8 F6. (i) is the filter the same for the strain VLPs and the seismic VLPs (I don’t think so). (ii) remove “/pinpointed” in L220. (iii) Do you imply that the strain VLPs are different than the seismic VLPs? You have 2 families for one group but 6 for the other. Why would that be? Due to strain meter being ‘much farther’ from the vent system than the seismometer STRA used for the seismic VLP analysis?

- P8, L223-226. Is this sentence directly referring to the 6 seismic VLP families shown in Figure 6 or simply a ‘general’ observation/suggestion based on VLPs seen at Stromboli? Not clear. Also, Chouet reference is [9] (and called after [10])? [All figures seem to end at 3 July, the first explosion, what about the August one?]

- P8, L227-230. What makes the 28 August event than unexpected? It occurred during an active phase. First time ‘volcanic tremor’ is mentioned. Did any precede the 3 July event (or was tremor activity different prior to it compared to ‘normal’). — A bunch of data are shown without providing context.

- P9, L236-239. (i) The waveforms in Figure 6b are not filtered in the 2.5-20 s passband, they contain significant lower frequency energy (families 2 and 5 close to 50 s, coincidentally, these are the families that become pretty active near 25 June). (ii) Why have windows 40-s long when Figure 6 shows much longer signals? Based on Figure 6b, the cross-correlation window should be about 3 minutes long; why did you use 40 s?

- P9, L248. Replace ‘pinpointed’ by ‘found’?

- P9, L252-255. Interesting, but it would help if the strain and seismic VLPs would be filtered in the same frequency band. Then similarities (or absence of them) would be more readily visible.

- P9, L255-263. Interesting, but what does this mean? The strain VLP ‘disappears’ or is it only no longer visible/identifiable at the strain meter? If it really disappeared, then why would the seismic family-1 VLP continue, even if slightly changed? A change in attenuation is certainly not a cause (L262 you mean ‘less damping’). Interesting observation but so far no attempt to explain what that could mean. Since all traces are normalized, did the VLP family 1 signals change amplitude (became smaller) [the amplitude and/or relative amplitude between and within families, or any temporal changes is not addressed at all …]

- P9, L264-270. Not clear what is meant by ‘seismic VLP component analysis’.

Section 4. Modeling the volcanic source. Overall: This part is pretty generic (Mogi-type source modeling using surface deformation signal - or lack thereof). (i) What is the point of this exercise? (ii) Modeling results have no uncertainties, comparison with previous model says ‘results differ’ without a clear explanation of why and which one’s supposedly better. Should provide range of possible solutions, discuss uncertainties, investigate parts of their modeling which are better constrained than others, should acknowledge that their depth is not allowed to be shallower than what they got and as a result, that it could be much shallower (higher up), … (iii) I don’t see the significance. (iv) Why is the volume in this manuscript so much different from the DiLieto et al. GRL paper; and which one should a reader prefer and why? 

- P9, L276-277. ‘We obtained a well-constrained pressure source’; what do you mean? How do you know it is well constrained?

- P9, L277-278 and P10, L285-300. ‘… source … under … craters … about 100 m above sea level’ — (i) Based on later text, the source could not have been shallower because of the modeling constraints (below lowest observation point). Wouldn’t that mean about 100 m above sea level is the deepest possible source location (and it could have been shallower??). (ii) Why is source deeper than in Di Lieto et al. and why is that a better result (you do not allow the source to be shallower), this makes little sense. You cannot, from you modeling, exclude a shallower source (a source above 80 m above sea level and closer to the top of the edifice). “Our solution achieves this value exactly” (‘this’ being the shallowest allowed depth) clearly means it could be shallower (or would want to be shallower if allowed …).

- P9, L278-285. Volumetric expansion of about 10,000 m^3. That’s about 20x20x20 m, which seems like a pretty small source (or not?) (and would possibly be even smaller if allowed to be shallower?). The explanation of why this volume is much smaller than in Di Lieto et al. is not clear; please describe as clearly as possible what the differences are (4-min signal considered here versus what? ‘entire signal duration’ is not clear, what is ‘entire’?). Is peak-to-peak amplitude used here too or only in Di Lieto et al.? Not clear. If different, why? What do you mean by ‘other than a different …’? — Overall this sounds like initial assumptions and modeling approach can lead to vastly different results. You would need to explain why the current assumptions and approach are better. Or is your model just one of many possible ones? If that’s the case (and I tend to think that way), then what feature is common (and thus stable) in the modeling results?

- P10, L290-294. ‘… well constrained and the data fit is good (Figure 7). (i) It would help to show (appendix) the signal that was modeled. For example, tilt from the summit seismic stations. How was it estimated and could you show it? For the GNSS sites, what was your signal, I thought they did not show signal, i.e., signal was below the noise level. That is, of course, a modeling constraint; is this how you used the GNSS data? If so please clarify. You could add a table (appendix) with all data (observations) and modeling results to say uniquely what you used and how it was fitted; currently not clear. (ii) Figure 7a shows the source location and recorded and modeled deformations. All recorded signals shown point north of your source location (i.e., clockwise for sites E of source and counterclockwise for TDF W of source). To me that means the source location should be slightly north of the location shown in Figure 7 to fit the data better? Explain why you think that is not true (like other data force the location to the one shown). (iii) The uncertainty for the strain data is much smaller than the modeling result differences (0.02 versus 0.08 micro-strain), why is that and could that be reduced with a source higher up the mountain? (iv) what type of minimization (L2 norm?) is used and are data weighted? Otherwise the modeling seems to be driven by STRA and STR1 tilt dat. Explain. (iv) What are modeling uncertainties? Or is the given source the only possible solution (even considering uncertainties)? If any modeling is kept, I suggest to give a ‘best’ result and some plausible uncertainties for each value. Okay, Fig. 7 caption has some for source depth and volume [but none for lat/lon, which I think also have uncertainties …]; depth has +-0 m uncertainty, that is not likely the case; seems modeling is rather limited.

- P10, L294-300. The summit area tilt data have limited resolution, okay. Here source depth is given with uncertainties (why not for the other model run with all data?) allowing a source basically from the surface to about 1 km depth (below the surface) which includes the 80 m above sea level result from the full dataset. It seems that the SVO strain data would possible also prefer a source higher up (than 80 m a.s.l.) but this cannot be verified with the existing modeling approach. It seems depth is thus not very well constrained. It also seems that source volume is small (when only considering the last 4 min before eruption).

- Modeling. Why only use the last 4 min before the event? Does this mean you think all preparation (volume expansion) that led to the event happened in 4 minutes? What is your justification?

- Please break up section into more than one paragraph.

- P11, L344-346. (i) What is the justification for the 70%-gas + 30%-basalt? How much does the ratio really matter for the model? (ii) Probably more crucially, the WMS-CMP sounds like a modification of the ‘weak plug’ model of reference Oppenheimer et al. [18]? Why not? (iii) The idea of two separate parts also seems unrealistic as there’s a transition (the gas escape and melt crystallization are a continuum process occurring over a wide depth range), so the model is more conceptual (why not say that)?

- P11, L365. What is ‘… the beginning of the observations’? This is not defined and plots show different time windows.

- P11, L366-368. Figure 3 shows the cumulative dispersion. There is no increase at the beginning of May. Without having established a ‘normal’ it is very much speculative when there’s an increase; simply based on Fig. 3 data (which I think is insufficient to establish ‘normal’, just to be clear!), cumulative dispersion increases in mid-to-late June. Figure 5f shows ’strain VLP’ and Figure 6c seismic VLP, neither shows an increase in May (seismic VLP data only span June-3 July anyways). Your figures contradict your sentence. So why write a sentence like that? You should also help the reader and point them back to the figures to encourage confirming (or contradicting) your text.

- P11, L369-371. (i) Since a ‘normal’ has not been established for the MODIS heat flux, it is difficult to argue that the higher values starting on 6 June represent ‘anomalies’ (relative to other events on Stromboli). (ii) Figure 3a shows the MODIS data (blue circles). The maximum value is definitely not 0.45 GW and definitely not on 2 July. Do you mean preceding the eruption? (iii) Okay, Figure 8 combines the data that are discussed, thanks. Still, don’t see an increase in cumulative dispersion in May 2019 [for times without a green line, there’s no signal or no data? You should say that, I assume no signal]. I see MODIS data but they are slightly different from the data shown in Figure 3 [like around 22 May], why?

- P13, F8. (i) Please add to caption that red and blue dashes mark the two explosive events. (ii) Why stop at the second event and not show data beyond that? Are things quieting down then? Not clear. (iii) (strain) VLP daily occurrence rate should be the same as in Figure 5f? Why is it not the same? Figure 5f does not show any significant rate increase but this figure does, so what is different? And if you make a story out of the different strain VLPs, then family 2 starts in early April (when also cumulative dispersion in not zero any more). (iii) What do the yellow, blue, and red fill colors represent? Two stages of precursory signal followed by the explosive phase? Please explain. (iv) caption ‘all … data show convergence toward an unrest phase …’ How can you say that? You say GNSS data do not show unusual signal, you say tilt data show nothing coming from depth attributing the TDF signal as ‘local’, and you do not actually mention or describe the areal dilatation. (later, L445-447, you suggest conduit pressure source was responsible for tilt, but why no signal at second tilt site?) (v) You mean “SPLN-STDF-SPLB triangle” (as the Excel label says)? That leaves daily strain VLP and heat. If the prevent data are representative (repeat: this has not been established), then the heat and VLP data start to be ‘different’ in early April, and again in early June (family 1 VLP) and late June (stop of family 1 VLP).

- P11-12, L365-394. (i) Please clearly state when the precursory phase started (I think you suggest early May 2019?) (ii) The signals (Figure 8) are interpreted as suggesting a shallow origin and increase in cumulative dispersion. The discussion then is entirely based on the latter. However, the cumulative dispersion does not start to increase in early May. Do you mean early June? That could make some sense. Otherwise the entire paragraph does not make sense. (iii) L390, please add year (’28 December 2014’) or is there only one 28 December Mt. Etna eruption and all readers (should) know about it?

- P13, L401. What is a ‘regime state’?

- P14, L411-416. If there’s a start to a change then I would place it closer to 21 June when “all of a sudden” several families become active. A second change is evident on 25 June, when families 2 and 5 become prominent (in addition to family 1) (family 6 looks like a stretched version of 5? And 5 like a stretched version of 2? Could that be interesting and what would it possibly mean?)

- P13-14, L405-416. Mainly a repeat of the seismic VLP section. Repeat only relevant parts here?

- P14, L417-428. Just to clarify: all seismic VLPs (Figure 6) show little variation in incidence angle and azimuth and starting June 25, become very localized (per Giudicepietro et al.)? Please make sure that this main point of the paragraph comes out clearly.

- P14, L429. Please specify when the second phase started. Not clear from the text (L400-401 indicates start in early June).

- P14, L445 and P15, L468. (i) What do you mean by ‘isotropic pressure’. Pressure is not a vector. (ii) What is the pressure at the source and why would a cylindrical column of equal volume ‘balance the pressure of the modelled source’?

- P14, L449-455. (i) Where do you see a ‘notably’ increased radial tilt? (ii) If SPLN shows uplift, would that go unnoticed at the tilt sites? One should be able to model that?

- P14, L459. Where did the VT occur? 0.8 x 10^4 m^3 has units for volume not depth. Do you mean 800 m deep? And then relative to surface that includes the topography or relative to what?

- P15, F9. (i) I would plot seismic and strain VLP beneath each other (move SPLN uplift either up or down). SPLN seems to suggest 2 cm uplift starting near 25 June, why do other sites not show any (similar) signal? (ii) It looks strain and seismic VLP see almost the same number of VLP events. Are they seeing the same events (like is their timing the same)? Or do the two types of data see something (slightly) different? If it’s the same, you may want to look at the strain VLPs a bit more closely to verify that they are truly belong to 2 families (and not 6 as the seismic VLPs). Along the same line, strain VLP is dominated by family 2 after 25 June, but seismic VLPs are mainly made up of the 2 groups family 1 and families 2+5, which would indicate the two types of data see something slightly different? (iii) What is the relevance of separating cumulative dispersion for the 2 craters? This is not really addressed in the text. The dispersion shows no change on 25 June. Also, the early data seem to suggest that an observation attributed to one crater has a corresponding observation from the other. Does this mean you can’t really distinguish uniquely from which crater the signal came or (less likely) that each event at one crater actually had a corresponding event at the other? (iv) Tiny thing, but why do you change color of strain VLP from Figure 8 to 9 (red blue to blue red)?

- P16, L487. Do you mean Figure 5 or Figure A6? The scale of Figure 5 would not show anything at the minute-level.

- P16, L491. ‘This last occurrence is the evidence …’ Why ‘evidence’ [that is too strong, ’suggests’ or ’supports’]? What do you mean by ‘last occurrence’?

- P16, F10. (i) text (“Density”), values, and arrows are too small. (ii) Please add ‘sketch, not to scale’ or something similar. It is interesting that your CMP-WMS boundary is near 250 m a.s.l. if one assumes vertical axis is to scale; this is significantly (?) above the about 80 m a.s.l. from the modeling results (Z=80 +- 0m).

- P16, L522. ‘slightly similar’ is not clear. You mean the two curves are ‘quite similar’? ‘Slightly similar’ means they are not similar.

- P16/17, L520-524. This is one of the very few instances where the second explosive event is mentioned (past the introduction). Is there nothing else to be said about it? Apparently it was “unexpected” too. Were there any other similarities to the 3 July event or is that impossible to say because of the heightened unrest flowing the 3 July event. Also, did the system ‘die down’ after the 28 August event or did unrest continue, diminish, and then end (when)?

- P17, Conclusions, L526-530. ‘without any meaningful variation of the main geophysical parameters used for the evaluation of the activity …’ — First, do the data shown represent the ‘main geophysical parameters’? If so, say it. If not, say what they are and either show/include them in this manuscript or point to publications/web pages etc. where plots of these parameters can be found. I assume additional seismic observations are considered ‘main geophysical parameters’ exist that are not even mentioned here. The manuscript title includes ‘multidisciplinary approach’ (which the manuscript does follow) but it is equally important to include all of the relevant data (or point to them). For example, volcanic tremor is mentioned but nowhere shown, similarly for earthquakes (I assume VT).

- P17, L530-533. ‘careful revision of “hidden” characteristics of these signals …’ — One of the fundamental shortcomings of the manuscript is that the introduction fails to explain what the normal signals are (that are being recorded and analyzed in normal observatory operations). It seems none of the datasets shown here are used for regular operations? What is used? And, giving some example, what did datasets used for regular operations show or not show? — It is clearly a problem that explosive events happen at Stromboli for which no warnings can be issued and the manuscript tries to address this.

- P17, L533-546. (i) Sounds great. But in practical terms: What data and data products would you now produce for observatory operations to get an idea whether a paroxysmal explosive event is about to happen? This is a single case study. Ideally you would identify certain patterns in the data and then look at previous similar explosive events to find out whether they were also preceded by similar signals (signal anomalies). What, specifically, are the signal anomalies that you found preceding the event? If the signals change from one to the next, then they have little/no predictive power. The list presented later (L551-566) simply repeats the data analysis parts presented; so all of them are of equal importance and significance?

(ii) A forecasting tool for operations needs to be thoroughly vetted establishing that a combination of observations (at least) tends to results in a paroxysm. I don’t see “an interesting proposal to improve our capability to forecast dramatic events” in Figs. 8 and 9, I see some data that (possibly) show some variations prior to and coinciding with the timing of the 3 July 2019 event (singular event, singular observations?). A proposal would be “A high rate of (N>100/day) and predominance of family 2 strain VLPs over family 1 strain events indicates a high likelihood for an explosive event within the next week to 10 days” (or a similar sentence - this is simply an example sentence, I’m not saying that this sentence is true or that you should use it …).

- P17, L563-570. Could you describe what would constitute a warning? What sequence of events would lead to a warning?

- P17, L549-566. Which parts of the ‘proposed’ parameters are already available? Based on appendix A1, for example, it seems the FLIR camera evaluation is already in place? Same for the satellite radiant heat data (appendix A2)

———

Appendix.

L685. Second occurrence of “A2.” This should be “A3. Ground …” [and all subsequent headers should be updated too like L686 “A3.1 …”]

L699-703, Fig A4. Thanks for the caption, so both PLB radial and TDF N275E show (some) upward tilt starting near early June 2019 (PLB only small and maybe not a real signal related to the later 3 July event). The main manuscript should include (caption for tilt figure for example) the definitions of up/down for the components that are approx. radial relative to the craters.

L723-727, Fig A5. Interesting, family 1 only (?) occurred prior to the 3 July 2019 event.

L739. ’stations operated …’

L760-762, Fig A7. Do the daily SPLN records show the uplift depicted in Figure 4f? Daily solutions are a low pass filter (?) and should thus show similar trends as the low-pass filtered 30s data. Could be interesting to apply a median (or similar) filter with 3 or 5 days filter length (similar to Figure 4f) to all vertical components? Well, in some way this is done in the appendix (Fig A8) except for a relatively short time period and again only for site SPLN.

L795. “COA”? You mean “SVO”?

Author Response

Dear Reviewer,

we appreciate your comments. Thank you for your constructive suggestions and comments.

Please see the attachment for our response to the comments.

Author Response File: Author Response.pdf

Reviewer 2 Report

Review on manuscript « The 2019 eruptive activity at Stromboli volcano: a multidisciplinary approach to reveal hidden features of the “unexpected” July, 3 paroxysm”.

            The study aims at investigating geophysical signals acquired before the two paroxysms that occurred in July and August 2019 at Stromboli in order to see whether there were detectable signs of unrest or not. In particular, this work highlights interesting changes of geophysical signal (seismic, tilt, thermal) prior to the first paroxysm that are not related to deep changes in the volcano plumbing system (i.e., input of a new magma), but mostly to physical modifications of the magma beneath the source vent, e.g., through an over-pressurization of the top layer of the magma column.

            The entire manuscript is extremely well written, innovative and I think it can be accepted after some minor corrections. I would just regret that not more details for the August paroxysm are provided. But I guess this is a direct consequence of the monitoring material having been destroyed by the first paroxysm. Hence, I have no particular major comments. This is a great work. All my minor comments are displayed in the pdf file of the manuscript.

Best regards,

Comments for author File: Comments.pdf

Author Response

Dear Reviewer,

we appreciate your comments, thank you very much.

• We made the recommended corrections to the text;
• We modified the figures 4 and A6;
• The choice of 0.8 for Cross-Correlation threshold is due to the family detection. 
  If the value is higher, the seismic VLPs families number increases too much. instead, if it is lower, the seismic VLPs are grouped in one single family.

Reviewer 3 Report

This well written and carefully argued paper makes a compelling case for understanding the processes involved in the development of paroxysmal eruptive episodes at Stromboli volcano; in this case, with a focus on the major paroxysmal couple in July and August 2019. The authors present the evidence for medium to long-term changes in measured geophysical parameters at Stromboli, and propose a model (involving a strong/degassed magma cap, overlying a pressurising, gas-rich magma) to provide a framework for understanding paroxysmal explosions, and the conditions under which they develop. This case is well made, and certainly plausible. The paper needs little modification before it can be published.

Comments: 

The only comments I have are that there a small number of recently-published papers on the 2019 paroxysms that could be usefully cited and incorporated into this paper. Notably, Andronico et al (Nature Communications 12, Article number: 4213 (2021)) make a similar point to the authors of the current manuscript about the lack of 'immediate' precursors, and point to some longer-term changes (but without offering a new interpretation); Caracciolo et al (EPSL, 2021) 559, art 116761, 10.1016/j.epsl.2021.116761 describe petrological evidence for a 'soft plug' model, based on observations of earlier explosions; and Inguaggiato et al (2021; GEOSCIENCES, 11, 169, 10.3390/geosciences11040169) provide an overview of the CO2 flux trajectories at Stromboli over the past couple of decades - again, with conclusions that strengthen the authors' suggestions. 

 

 

Author Response

Dear Reviewer,

we appreciate your comments, thank you very much.

We inserted the proposed references in the Introduction (Andronico et al.) and Discussion (Caracciolo et al. and Inguaggiato et al.) sections.

 

Round 2

Reviewer 1 Report

Review of revised manuscript “The 2019 eruptive activity at Stromboli volcano: a multidisciplinary approach to reveal hidden features of the “unexpected” July 3, paroxyxm” by M. Mattia et al. submitted for publication in Remote Sensing.

I would like to thank the authors for their responses to my substantiative critique. The authors replied to all of my initial comments and changed the manuscript considerably. The changes, particularly the added explanations and also the removal of overly assertive statements, did improve the manuscript. The applied modifications are still modest in scope and some of my previous objections remain such as providing more context about the focus of the manuscript and how it relates to other previous - and possibly planned future - studies of this paroxysm and paroxysms at Stromboli in general. In fact a clear exposition about the aim of the manuscript would possibly take care (remove) my concerns. Ideally, the authors would provide a clear roadmap at the end of the introduction about the content and aim of the manuscript: Present various data collected prior to the paroxysm, conduct some modeling of a Mogi source - I still think that part is weak - and culminate in presenting a model how Stromboli paroxysms may work and what part of existing datastreams could easily be incorporated in monitoring efforts to see if their potential usefulness for medium-term forecasting of such events holds.

 

Below follow a few style related suggestions (and more could be added):

L42-44. (1) Do you mean ‘very different’ rather than ‘very diverse’? I think you want to emphasize the difference between the two rather than saying that each facies has very diverse texture. (2) “reflect a peculiar feature often associated with …”, I would remove the “being” and “found” in that part of the sentence.

L47. You may want to change ‘above sea level’ to ‘subaerial’?

L48. Do you mean ‘because they were seemingly not preceded by any long- or medium-term precursors’? Like were the people apprehensive because there were no precursors (my suggested sentence) or were they generally apprehensive because the paroxysm happened and then “without the …” is simply some additional information (independent of apprehension).

L53 ’that started just …’ please add ‘that’

L55 ‘… without offering a new interpretation.’ Of what? ‘for their cause’? Seems sentence stops in middle and something’s missing.

L56 ‘medium-term’ rather than ‘mid-term’

L80 ‘(possible due to a fortuitously taken photo, Figure 1)’?

L89 ‘… from the sea looking towards NE with the Ginostra village in the foreground’

L92 ‘… Fuoco, lower left, and the eruptive column are visible.’

L99-103 You could add a sentence why looking at the different datasets and combining them is of advantage. Something along the line of (i) using all/most available data on Stromboli and (ii) advantage of a multidisciplinary approach.

L187 ‘probably due to its …’

L194 ‘before both events’ - remove ‘of’

L195 As in L193, add ‘that’ before ‘occurred’

L231-234. Sentence is very difficult to follow and too long. Please split into two or more sentences.

L238 ‘evidencing’ — sounds awkward. Simply ‘showing’ would suffice here. In general, I would not use ’to evidence’ but rather ‘to show’ or similar throughout the manuscript (there are several occurrences).

L242-243 ‘… who found a correlation with activity at different vents’?

L244 Don’t understand ‘In respect with previous studies’

L272 ‘between’ rather than ‘among’? Would remove ‘method’, do you mean ‘splitting’? Sentence is not clear to me. Is the sentence about 6 seismic vs. 2 strain VLP families and why you think that is?

L275 ‘and computed stacks.’?

L278 ‘decreased’ instead of ‘lesser’ (sorry, it seems I’m starting to correct my own suggestions …)? Do you mean ‘frequency content’ or (simply) ‘frequency’?

L384 Do you mean ‘punctual’ (=on time)? You basically look at all the different data jointly? Not sure how you would want to phrase this.

L389 either ‘that occurred’ or remove ‘occurred’ (probably better).

Author Response

Dear Reviewer,

we appreciate your comments, thank you very much.

Please see the attachment for our response to the comments.

Author Response File: Author Response.pdf