The article addresses a very current topic, i.e. the analysis of processes and mechanisms responsible for the instability of rock masses in permafrost, with particular emphasis on climatic drivers, and therefore on the potential impacts of climate change, and on monitoring/modelling procedures potentially useful for warning purposes. While many studies have shown a significant increase in rock slope instability under ongoing climate change, the specific mechanisms that cause individual instabilities are still poorly understood and little investigated, especially with regard to small-scale (but frequent) events. As a result, warning procedures still remain a challenge. The authors propose an approach based on the integration of field data, laboratory tests and numerical modelling, applied to the case study of the collapse of a rock pillar in June 2023 on the Matterhorn. The paper highlights the role of snow-melt water percolation within frozen rock masses, pointing out the occurrence of precursory signals a few days before collapse. The paper is structured in a clear and rigorous way; objectives, data and methods are well illustrated, and the conclusions are adequately supported by the results. Below are some considerations aimed at clarifying some points and further improving the excellent overall quality of the manuscript.

L140: “A Vaisala WXT520 weather station is available both at the rockfall zone at 3500m asl as well as at the Solvay Hut at 4003m asl”: What kind of data are recorded by these stations? Is precipitation included? Precipitation data are not mentioned in the text, although rainfall has been indicated as one of the potential drivers of thawing (see L383): the paper doesn’t specify whether rainfall occurred in the days preceding the collapse, and only snowmelt is considered responsible for the percolation of water into the frozen rock mass.

L255: “While both the rock and air temperatures in spring 2023 and the month of failure (June 2023) are within the range of the longterm average (see Fig. 3 and supplementary Fig. S3)…”: actually, according to figure S3, the rock temperature shows an interesting trend in 2023, always remaining around -0.5 °C in the months 1-6 at 2m depth, which corresponds to the ALT in 2022, while in previous years it reached a minimum between -1.5 °C to -2.5 °C in April. Furthermore, Fig. S3e shows that at 0.6 m rock temperature increased significantly in the 2 weeks preceding the event. I think these data should be shown in Figure 3 and mentioned and discussed in the paper.

L256: “the humidity in spring and June 2023 is remarkably high”: this observation has not been discussed nor taken into account in the rest of the paper, you may consider to remove this information or you should include in the discussion.

L366: “rapid decrease in snow depth at the nearby IMIS weather station Stafelalp/ZER4, see supplementary Fig. S7”: to make this observation more evident, it would be useful to add a bar in Fig. S7 corresponding to the day of the rock pillar collapse, as it has been done in most of the figures. Furthermore, the dynamics of the snowpack does not appear different compared to previous years: it would be appropriate to mention this point and take it into account in the discussion.

L414: “The clear diurnal pattern of short (< 9s) seismic pulses implies freeze and thermal contraction as the main drivers of those events”. Isn't the observation (Fig. 10b) that the short seismic pulses increase in the pre-collapse period, which in the diurnal cycle are linked to the nocturnal freezing, in contradiction with the thesis supported by the paper that it is the melting triggered by the percolation of snow melt water in the frozen mass that causes the triggering of rock pillar instability?

For data and method comparison purposes, did you consider the work of the group of Magnin et al. on similar topics? (e.g. Magnin, F., & Josnin, J. Y. (2021). Water flows in rock wall permafrost: A numerical approach coupling hydrological and thermal processes. Journal of Geophysical Research: Earth Surface, 126(11), e2021JF006394; Ben-Asher, M., Magnin, F., Westermann, S., Malet, E., Berthet, J., Bock, J., ... & Deline, P. (2022). Estimating surface water availability in high mountain rock slopes using a numerical energy balance model. Earth Surface Dynamics Discussions, 2022, 1-25).

Technical corrections:

L44: “Scandroglio et al., 2025”: not in the reference list.

L77-78: “Together with similar instrumentation and monitoring techniques on the Italian side”: if possible, please add at least one reference.

L91: “1981–1990”: did you mean 1981-2010? Usually climatic reference periods are 30-yrs long.

L277: “Both local terrestrial”: change to “both local terrestrial”.

Figure 1: “Hörnli hutte” change to “Hörnli hut” to be consistent with the rest of the text.

Figure 3, caption: “since 23 June 2023”: you probably mean “until 23 June 2023”.

Figure S8, caption: please complete the caption explaining each subfigure (a, b, etc.).

This study provides a comprehensive, integrated multi-method analysis including field observations, laboratory findings, and mechanical modeling investigating the early destabilization to the final failure of a single free-standing rock pillar. 

There are very few rockfalls with precursory observations, therefore this study is particularly interesting as it provides multi-disciplinary in-situ observations (GNSS, inclinometers, seismometers, weather data). These observations are essential to develop methods for forecasting such events and to understand the physical mechanisms that promote failure.

In my field, seismology, I found the analysis and the choice of parameters were appropriate and correctly justified. 

Overall, I found the manuscript very clear. I only have minor suggestions to improve this study. 

While the displacement data show a clear acceleration before failure, which can be used to estimate the failure time, the results of the seismic analyses are not so convincing.

The data is only analyzed for 13 days before failure. Why not including all data since installation in 2019?

I understand that the manual classification of seismic events is time consuming, but only a very small fraction of events is removed. 

Would the results change significantly if the authors used only the results of STA/LTA detection without manual validation?

The results of the dv/v and energy rate are fully automatic and rather fast to compute. I strongly encourage the authors to look at all available data (at least for dv/v and energy rate) in order to analyze seasonal fluctuations, to quantify the normal variability of each variable and to analyze wether the precursory fluctuations observed in the last days before failure are really unusual.

On l308-309, the authors claim that "The short events reach a short-term maximum rate around 4 June, followed by a second rise towards the failure of the rock pillar."

I don't agree with this statement. In Fig 10b, the strongest peak occurs about 2 days before failure, then the rate of events decreases until failure.

How do you explain this pattern? Possibly seismic events are associated with fracture propagation; and during the latest days before failure, 

displacement may occur as aseismic shear along these fractures? See also l415-417.

Evolution of seismic energy rate (Figure 8). Did you filter the data before estimating the energy? Could you test different frequency ranges?

Could you add a figure showing typical waveforms at all sensors for both short and long events?

Is the amplitude generally stronger at the station closest to the rock pillar?

Can you identify P and S waves?

Is the temporal evolution of short events different if you select only events with a strongest amplitude at the closest station?

Experiment: How do you explain the large variability between friction angles (50-80°) for air-dried no-cohesion frozen samples? 

Numerical model. How to explain the difference in amplitude displacement by a factor of 1000 between the model and observations?

Which parameter of the model could you modify to better explain the observations?

Figure 10. Why is there a peak at 0 and 24 hrs for both short and long events? Is it real or a problem of side effects?

Figure 7. "Boxplots give information on all forecasts since the OOA per velocity window." I don't understand precisely what is shown in these plots?