I read with great interest the paper of Sandri et al. about probabilistic hazard assessment due to lahars on Vesuvius volcano. As mentioned by the authors, lahars are one of the most dangerous volcanic phenomena and this is a welcomed contribution to advance our knowledge on this topic. In general, I think this is a very clear, concise, and useful contribution and I strongly recommend publication. I have some remarks that I think will improve the final product.

1. I think it is worth to mention in the introduction some of the codes and publications that have been used in the past to simulate lahars and build hazard maps such as LAHARZ, Laharflow, etc. This can be included in lines 40-45.
2. In the discussion on how to estimate the volume of material removed from slopes, the publication of Pierson et al. 2013 (bull volcanol 75:723) about lahars due to rainfall in Chaiten volcano can be useful to strengthen the criteria used by the authors.
3. I think that in some place, some numbers regarding the range of discharge rate (Q, m3/s) and initial or total volume (V) of the simulated lahars should be given. As most people working on lahars are familiar with and use these measurements, it would be useful to translate the degree of initial erosion, rainfall, etc into Q and V to compare with other volcanoes or lahar cases around the world.
4. As the authors claim that this is a novel methodology for probabilistic lahar hazard assessment, in my opinion there is a lack of discussion and/or comparison with previous works on the subject on Vesuvius or elsewhere. For example, it would be interesting to compare with a) traditional (field studies) methodologies for lahar hazards, how the quantitative probabilities given in this work can translate (or not) to the typical high, medium, low quantitative probabilities of hazard maps b) similar previous studies, in particular with Tierz et al, 2017. How it compares in terms of the physical model used, how probabilities are obtained and analysis of final results.

I think these additions can be incorporated easily and without substantially altering the original structure of the draft, but I think they will improve the impact among other people that study lahars.

Dear Authors

I have reviewed the paper, “Lahar events in the last 2000 years from Vesuvius eruptions. Part 3: Hazard assessment over the Campanian Plain.”

The paper was well written and logically ordered.  I enjoyed reading the probabilistic approach to lahar hazards and how the outputs of the analysis have been presented.  This paper was difficult to review.  The probabilistic model and outputs presented are good and would be of interest to the wider community as well as presenting a very efficient and effective method of presenting lahar hazard analysis.  The manuscript though was very short and due to that tended to “gloss over” key aspects of lahar hazard as well as determination of key inputs of the probabilistic hazard analysis.  In places this resulted in the paper seeming to be incomplete in understanding lahar hazard.  Of note is that the abstract itself seems to read more like a “highlights section” rather than a proper abstract that succinctly reports on the methods and findings of the study.  While I also understand that this manuscript is part of a series of publications focussed on the Campanian Plain (Southern Italy) this particular manuscript is focussed on such a localised issue that the manuscript does not provide any indication on how this analysis may be of relevance internationally.   Of note is that referencing throughout the manuscript seems very limited.  the biggest annoyance in this aspect is the amount of self-citation to unpublished work (“this issue”) which makes the paper very hard to review particularly when justifying input data for the probabilistic analysis. 

In general, the manuscript, the lack of referencing, seems to disregard all previous work on lahar sedimentology, lahar trigger mechanisms, understanding of lahar rheology, lahar simulations and hazard analysis in favour of papers included in the same issue from the same authors.  Not having that robust background information makes it hard to assess the manuscript and hazard models produced.  Essentially most of the input information for the probabilistic analysis is derived from data that is difficult to assess for its applicability to the model presented.

More specific comments are as follows.

• The abstract mentions some representation of how uncertainty on the initial flow conditions is treated or accounted for. I am assuming that is related to the initial flow volumes, materials and rainfall required to trigger a lahar event, but it seems to be hard to find how this was accounted or in the body of the manuscript over than line 194-196 which does not address uncertainty.
• Lines 30-32 identify bulking and debulking or other concepts of variability in the flow, but I do not see how the modelling approach; ImexSflow 2D (not referenced; having to search Github) accounts for some of these. In many respects the simulation strategy does not differ from other Voellmy-Salm based simulations (e.g., Titan 2D, Volcflow, Avaflow) so why the need to recreate these tools with no real improvement to the initial starting and flow conditions or its specific applicability to rain-triggered lahars remobilising material.
• A minor point in lines 33-34 is the listing of a sedimentological characteristics of lahars which may not be factors required for hazard analysis but more relevant for numerical simulation. Hazard analysis inputs should be focussed on occurrence and magnitude with related hazard impacts characteristics being inundation depth, velocity, dynamic pressure etc.
• In lines 38-39, 56 and again in the discussion there seems to be no mention of recent statistical surrogate and emulation approaches (Mead et al., 2023; Tierz et al., 2017; Bayarri et al., 2009, etc.) to account for variability in trigger mechanism and source conditions. Aligned to this is the complete lack of recognition of Mead & Magill (2017) and George et al. (2022) (D-Claw simulations) to provide better constraints over the trigger mechanism of rain triggered lahars (e.g., line 70-) that consider source pore porosity and liquefaction processes.
• Several times throughout the paper terms like “a large number of lahar simulations”, “a massive set” were used when these should be quantified.
• An issue that needs addressing is the use of past deposit data being analysed to inform model simulation inputs or initial starting conditions (line 90, section 2.3, etc). I struggle to see how the porosity of deposited volcani-clastics which is a mixture of material consisting of source material plus entrained clasts that has been saturated, sorted and probably lithified is representative of the original source material.  Defining the porosity of the source material needs to be better defined.
• In terms of the modelling a 50x50 m computational grid seems quite course in relation to the assumptions made on slope and discrimination of catchments and hazard zones.
• An aspect of the study which could with further explanation which i struggled with was the definition of the hazard over time. As I understand it the hazard analysis is simply related to flow depth and dynamic pressure from two subplinian eruptions rather than a more complete record of events over the history of the volcanic centre. Should the probability maps and outputs also be prefaced or conditional on the probability of those eruptions occurring again producing a specific volume of available material that could be remobilised.  

I enjoyed reading the manuscript and I believe the manuscript should be published but it needs to be placed in context with the considerable volume of international literature available on the topic of lahar hazard analysis.