Overall, the paper presents a new approach to model the landslide size distribution of the so defined “large landslides”, setting a threshold at 10⁵ m². The described approach is rooted in the Bayesian framework and builds the estimation of the posterior probability distributions on previous knowledge (literature) and on several inventories that were publicly available. The approach presented builds the prior distributions using all the landslides available in the inventories, without distinguishing, by choice, among different inventories (event, multi-temporal, geomorphological), triggers, type of failures, experience of the mappers, method of mapping. The authors claim that this method proved that the power-law scaling of large landslides is not purely mechanistic, as results show no difference of statistics among a quite wide set of landslides and inventories.

The paper is written in very good English, concepts and ideas are clearly expressed, with a sound logical consequentiality. The effect is that it is easy to read and understand. Figures are good looking, self-explanatory and consistent, well captioned and properly cited in the text. There are no missing nor unnecessary figures.  

The Introduction is well-framed, and the research niche is clearly defined. Methods are accurately and clearly described, both the theoretical framework and the practical aspects, as well as the working assumptions and their legacy. Results are clearly presented and illustrated and, in general, the discussion highlights the main aspects related to the research questions posed.  

I have a few questions/concerns that need to ask the authors. The rest of my few comments can be found in the annotated pdf.

The first question is about the incompleteness of large landslides in inventories. In the Introduction, it is stated that (lines 42-47) “Still, most uncertainty remains about the large landslides that are rarely sampled. …  One reason for this knowledge gap is that large landslides are often elusive in catalogues compiled shortly after a landslide-triggering earthquake or rainstorm (Hao et al., 2020; Abancó et al., 2021; Santangelo et al., 2023). Sample sizes often involve only a handful to several dozen large landslides, and thus often remain too small for robust statistics in a given study area. Hence, inference is mostly based on the simple extrapolation of model fits beyond the observed size range.” In my experience as geomorphologist involved in several landslide mapping activities, when preparing landslide inventory maps, the source of incompleteness of inventory maps is predominantly due to missing small landslides, which are the most elusive. If an event occurs and landslides are smaller than 0.1 km², that does not necessarily mean that large landslides were under sampled, but it cannot be excluded to be a peculiar feature of that specific event, which was the case of the event in the Marche region, cited in this paper (Santangelo et al., 2023). I think in the end the problem this paper is facing does not really change for this, as the sample of large landslides is often limited in the inventories because their number is, as a matter of fact, far smaller than smaller landslides. What I have seen often, is that many geomorphological historical inventories (sensu Malamud et al., 2004; Guzzetti et al., 2012) lack very large and very dismantled landslides, where the evidence needed to recognise and map them is more complex, and often requires higher geological skills, long and wide experience and, above all, time and dedication. Also, I do not understand the reference in the title to “incomplete inventories”. I do not think I get this. I would just refer to inventories in general.

The second question is about the effect of building the priors by putting all landslides together. I am wondering what the effect would be of estimating the posteriors building on priors only based on specific types of landslide inventories. So, treating separately different types of inventories, event-based (and multi-event), geomorphological, and multi-temporal. Would the estimates be similar or different, and what would be the interpretation of that result? As the authors correctly checked the choice of the size threshold, in my opinion this one also needs to be checked and commented.

The third question is about using this kind of landslide size statistics as a tool for inferring the degree of completeness of inventories before looking at the data. This could be added as a topic in the discussion section.

Minor comments and typos can be found in the annotated pdf.

In conclusion, I think the paper is suitable for publication in NHESS pending minor revisions.

Best regards.

Dear authors,

I provide the referee comments to the manuscript entitled ‘’Size scaling of large landslides from incomplete inventories’’.

General comments:

The manuscript is clearly presented and well-structured, making it easy to read and follow. The addressed topic is relevant to the field of landslide research, particularly of landslide inventory mapping and statistical modelling of landslide inventories. The methods are clearly outlined and described with enough detail. The English is high level, contributing to the clarity of the work.

Finally, the manuscript is prepared according to the NHESS journal’s standards and can be published after minor revision.

Below, I provide a few specific comments.

Comment 1:

I suggest improving the presentation of the landslide inventories used in the study. I recommend adding this relevant information about the landslide inventories to Table 1: type of the inventory; method for its preparation; expert team, if possible; and area covered. I believe that including this information would enhance the understanding of the reliability of the input data. Moreover, since the concept of incomplete inventories plays an important role and is included in the title, it would be useful to briefly clarify the main reasons for their incompleteness.

Comment 2:

A significant number of references on the landslide inventories, cited in the Table 1, are not included in the References list. Please, thoroughly review the citations and add any missing references to the list.

Comment 3:

The statement in the Abstract (line 13) "Our model identifies several inventories with outlier scaling statistics that reflect intentional censoring during mapping" raises some important concerns. While this is a significant assertion, it would be more compelling if it were supported by concrete evidence or detailed analysis in the Results or Discussion section. Without sufficient evidence, the claim appears speculative. I recommend providing more justification or proof that clearly demonstrates the intentional censoring you refer to, ensuring that the statement is well-founded and substantiated.

Kind regards