The article describes the collapse of a large hall with a flat roof in southern France after heavy snowfall turned into rain. The analysis of such cases is very valuable on the one hand to check the design snow loads and on the other hand to identify structural weak points of a structure. The study was carefully prepared, but unfortunately no quantitative data was available on important input parameters such as the amount of snow load at the time of the damage or the condition of the structure before its collapse. Therefore, the main statements of the article are somewhat speculative. In order to finally answer the question what was the exact cause of the collapse of the building, the collapsed structure would have to be investigated in more detail (e.g. material technology tests). In several building collapses, such as the ice rink in Bad Reichenhall, material deficiencies were partly responsible.

The beginning of chapter 1 Introduction has only a small relation to the main content of the article. Dynamic effects of gravitational natural hazards on a building have completely different consequences than static effects such as snow load. I recommend making only a reference to snow loads in the introduction.

The collapse of the building would need to be described in a bit more detail:

• Time of the collapse?
• Weather at the time of the collapse: wind effects?
• Snow depth and snow distribution on the roof? Equal amount of snow on the ground as on the roof?
• Overview sketch with details of damage to the structure?
• Were there any eyewitnesses?

The building and the supporting structure would also need to be described in a more precise way:

• is it a spatial supporting structure (Fig. 7a only shows the truss construction of the supporting structure; Figs. 8 and 11 do not allow to see the necessary details of the truss.)?
• Roof pitch according to plan (line 145: nearly flat; line 151: about 1%; Fig. 9: less than 1%? What is the correct slope?)
• Drainage of the roof: location and number of outlets – along the edge of the roof? Were there emergency outlets for the water?
• Steel properties of the various structural elements - was only S235 used?
• Further, when analyzing a building collapse, comment on the loading history (e.g. max. snow loads since construction, line 329: 42 cm on 22 Jan 1992) and any adjustments made to the building (e.g. alterations to the structure).

The assumed snow depths and snow densities appear to be somewhat speculative: 

• According to Fig. 4, Montpellier is in the zone with 20 to 30 cm of snow. However, in the following, a snow depth of 30 to 35 cm is used for the analysis: a justification is missing.
• The assumed new snow density of 250 to 350 kg/m3 seems too high. I would expect a maximum of 250 kg/m3. It would be interesting if the snow depth and density could be quantified with measurements.
• The SAFRAN simulation with a SWE of 35 mm seems plausible.
• The consequences of rain on the snowpack should be discussed in much more detail. A 30 cm layer of snow is unlikely to retain 50 to 60 mm of rain. Typically, snow is saturated with about 15% water. The rain that cannot be stored runs off on the roof surface. The assumed “wet” snow density of 600 kg/m3 seems to be too high.

The calculated bearing capacity of the supporting structure seems to be rather high:

• At best, e.g. stability analyses (buckling) of the columns or the tubular truss elements could provide further insights.
• In the damage analysis of a large flat roof after a rain on snow event with questionable drainage, the assumed uniform load distribution is rather simplified. The important question is, where was the water, which could not be absorbed by the snow cover, before the collapse? The roof was sloped, I assume that a large part of the water flowed to the edge of the roof. The maximum water load would therefore be along the edge of the roof. This can be simulated with an additional load case with a trapezoidal load distribution (minimum snow/rain load roof center and maximum snow/rain load roof edge).
• The snow load can cause a sag in the middle of the roof, which may be greater than the overheight due to the roof pitch of less than 1°. This would result in the water not absorbed by the snow cover flowing to the center of the roof. This could also be studied with a load case with trapezoidal load distribution (maximum snow/rain load in the center of the roof, minimum snow/rain load at the edge of the roof).

Section 4.3 is not directly related to the damage analysis presented and is somewhat speculative. Measurements of the temporal development of snow loads in the Mediterranean region are practically non-existent. I recommend to omit this chapter. Instead, to give further hints on how such rain on snow events can be better managed on a flat roof (slope of roof? arrangement of drainage? emergency outlets?). Further, it would be helpful to indicate what information would be useful for a more complete damage analysis in future (e.g. photos immediately after damage? Snow depth measurements? Snow load measurements? ...).

Further comments:

Line 42/43: deficient building, better: insufficient material strength?

Line 53: Determining the ultimate bearing capacity of a building is similar to or more difficult than determining the possible snow load.

Line 56: What is the AROME numerical model?

Line 64: What changes are expected about the characteristic snow loads…. not clear what is the meaning.

Lines 79 – 95: Add precipitation data.

Lines 97: Why is the rain-on-snow event exceptional? Return period of event?

Fig. 4: add the location of the collapsed structure

Line 129: Explain why 30 – 35 cm were chosen (see Fig. 4: 20-30 cm). Give some reference values for new snow density: 250-350 kg/m3 seems to be too high.

Line 134: The rain on snow event should be discussed in more detail. How much water can the snowpack absorb? What can be the density of a wet snowpack?

Fig. 5: How was snowfall measured? Where is the meteo station Lavalette?

Line 145: (nearly) flat – the slope of the roof is in a rain-on-snow event very decisive. What means nearly? 1°? 3°?

Line 154: The drainage system should be explained in detail.

Line 157: central part of the structure: where is that? Indicate the location on a Fig. e.g. 11; The western and eastern facades were “heavily” (not hardly) damaged.

Line 161/162: was there an element that was clearly the weakest?

Tab. 1: Steel quality of the different structural members?

Line 178: “dead” weight

Line 204: better snow load not snow pressure

Tab. 3: Better snow load instead of pressure value

Line 208-209: …by construction the applied pressures. Difficult to understand.

Line 232: snow load on the ground was estimably 30-35 cm – how was the snow load on the roof? Was a shape factor of 0.8 applied? Was there wind during the snow fall event which reduced the snow height on the roof?

Line 255: difficult to understand: the highest height scenario…

Line 278: which were the detected structural weaknesses?

Line 285: “a maximum range of span to be on the safe side”…difficult to understand: If the planned geometry of a building is taken into account in the design of the load-bearing structure, a structural failure should not occur. In connection with the drainage, the roof pitch and the sags in the service state would need to be checked.

Line 334: Is there some evidence that the drainage openings were blocked by ice? With temperatures around 6° C might be hardly the case?

Comments attached in the pdf file.

The paper investigates the collapse of a steel building, built in 1982 in Montpellier in the south of France, under snow and rain loads occurred in 2018, providing detailed information of the meteorological event, its features, influence on snow accumulation on the building as well as on the subsequent rain event, heavily affecting the snow density and the resulting load acting on the roof.

The paper continues with the FE modelling of the structure, to simulate the collapse condition trying to estimate, by means of back analysis, the actual load intensity which led to the collapse.

On the following parts clarifications are needed.

1. In Annex C it is stated that the structure is not known in detail and some simplifications and assumptions on the real geometry are introduced in the FE model, the influence of which in the results is also checked by means of a “virtual” This kind of assumptions may significantly affect the validity of the FE results and more explanations are needed. In particular, a detailed list of missing information should be added, commenting on the potential impact of the induced uncertainty in the FE model. Some drawings showing the structure and its elements (cross sections, dimensions, etc.), possibly from the time of the construction, could help in better understanding the structural behaviour.
2. Steel properties are reported in Table 1, clearly referring to nominal values for S235 steel. In a static non-linear analysis, the actual mechanical properties of steel play a fundamental role in the determination of ultimate loads leading to the structural collapse. A clarification on this aspect should be introduced, possibly referring to test results on specimens extracted from the steel members after the collapse or, at least, by making reference to mean values of resistances instead of characteristic values, as it is the case in Table 1.
3. The mesh sensitivity study, mentioned in line 170-175 and illustrated in Annex A, does not seem appropriate for a truss system, with hinged beams
4. Collapse criteria illustrated in 3.3 (lines 195-203) are not clear, as it could be interpreted that the collapse is reached as soon as one steel beam yields or reaches the ultimate strength (which is then not expected to occur as this happens only after yielding). In pushover analyses the final collapse mechanism is identified under non-linear static analysis under increasing loads, which is not evident in methodology illustrated in the paper. A clarification is needed.
5. Among the collapse criteria no mention is made on buckling of compressed members, which as expected and as confirmed by the photos of the collapsed structure, has occurred. Buckling anticipates the failure of members with respect to the uniform compression till yielding and this aspect could lead to a significant reduction of the ultimate load in the FE analysis. A clarification is needed.
6. Considering the flexibility of the structural system of the roof, the assumption of the uniform distribution of the snow load, and moreover of the rain load, all over the roof surface is a strong assumption, which could lead to wrong unconservative results. This aspect is only mentioned as a limitation of the analysis, but should be better illustrated as ponding effects could have caused a significant redistribution of the load, with concentration in the centre of the roof area, i.e. where its effects are more onerous for the system.
7. Based on the above comments the discussion of the FE results in 4.4 may need to be reconsidered.
8. Paragraph 4.2. It is claimed that the structure respected the structural design codes at the time of the construction as well as at the time of the collapse (2018). Later on in the Appendix it is stated that the SLS limit states were not verified. The particular structural scheme, a steel 3D truss plate with no intermediate supports, is particularly prone to deformation effects, which generally end up in governing the design. Some more details on these aspects are needed, to better understand the validity of the drawn conclusions about the compliance with the design standards.
9. Line 315: recent climate models provide also snow variables, such as snow depth or SWE.
10. In the conclusions it should be better highlighted which are the main outcomes of the study, i.e. which sort of recommendations are proposed by the Authors also in view of the revision of structural design standard or for the analysis of existing buildings.