Dear Editor, dear Authors

This manuscript by Peruccacci et al. presents a very valuable, high quality dataset on rainfall induced landslides in Italy. These data are crucial for the development of warning systems based on empirical rainfall thresholds with high spatial resolution. In addition, inventories of landslide events are an important cornerstone for improving scientific understanding of the processes involved and for supporting hazard assessment. This article shows well how time-consuming it is to compile such a database, a fact that tends to be forgotten by the general public (and sometimes, unfortunately, by users). The contribution represents a milestone for the development of future landslide early warning systems in Italy and it is to be expected that it will be widely cited in the future (together with the actual dataset by Brunetti et al., 2023).

The article is (with a few exceptions) well-structured and quite well written, the length of the text is adequate and the explanations and statements are well illustrated. Below I list three comments of a more general nature and ask the authors to comment briefly. In the second part of my review (see supplement), I go through the individual sections/chapters of the article. The specific comments that arise are not of a fundamental nature and should be relatively easy for the authors to answer. They reflect questions I asked myself while reading or text passages I stumbled over. Finally, the third part (see supplement) contains a fairly extensive list of technical details that I encourage the authors to consider carefully.

• As can be seen in Figure 4 and made clear in lines 297-298, only a few events could be recorded for the first six years of the study period (65). Of course, this can have various reasons. It may be that only a few landslides really occurred in Italy in these years (which is probably rather unlikely), or the data basis for these early years was significantly worse than for the following years. Especially if the latter is the case, the question arises as to why 1996 was chosen as the start of the study. In my opinion, this is not explained in the text and would be interesting. Please also note my comment on Figure 4 in section (2) of this review.
• It is implicitly clear from the text in lines 185-188 as well as from Table 1 that triggering rainfall (as well as its reconstruction) is not part of ITALICA (or is not part of the data collection to feed ITALICA). This is probably due to the fact that a central elaboration of the triggering rainfall amounts for all events is very time-consuming and this step is probably done at a later stage by the landslide risk managers (for a selection of events that are of interest e.g. for a province or a region). For this article this information is not crucial, nevertheless I would briefly state/describe this fact in section 4 (just before or after table 1) in one or two sentences.
• Both the Abstract and the Concluding Remarks point out the importance of spatially and temporally accurate information on landslide events for their use in early warnings in Italy. Are there already applications that demonstrate or support these statements and could be briefly mentioned in the Concluding Remarks? If not, do the authors know if such applications are planned that could be mentioned? If there is something more detailed to report here, a short additional section on "Research Opportunities" or "Research Applications" would be quite conceivable and would further strengthen the contribution.

In summary, this is an important contribution that will be of considerable interest to the landslide research community as well as to local and regional decision makers in landslide risk management. In my opinion, it should therefore be published in ESSD. I propose that the paper be accepted subject to minor revisions.

Dear Editor and Authors.

Thank you for your valuable and detailed presentation and examination of the ITALICA dataset. I feel this is a good paper that clearly presents the difficulties, resources, and data variability available in compiling such an inventory. The focus of the work is clearly defined, and the steps/ methodologies adhered to in order to compile such a dataset are clearly established.

I found the paper to be engaging and comprehensive, with questions arising mostly being answered as the paper progressed. The Figures and Tables all support the text and are positioned well to support the narrative.

I have a few very minor comments that I would like to see addressed to support the paper, most of which appear to fit with the Background section. Please see comments supplied.

Dear Editor, dear Authors

The manuscript by Peruccacci et al. highlights how essential reliable landslide datasets are in the elaboration of landslide thresholds, to be used in the development of landslide early warning systems. The authors also emphasize how landslide inventories are important tools in the hazard characterization and assessment and, they provide a very useful and quite up-to-date overview of existing landslide inventories around the world.

Despite some few examples, most of the landslide inventories have been prepared after the year 2000 and were created with the purpose to show where historical landslide events have occurred and the damages they have caused. In the compilation of these inventories, the main concern was mainly the quantity of data rather than the quality, or information about triggering mechanisms. Most of these inventories were prepared as part of specific projects and created during the duration of the project, not always updated, with new entries or improved in their quality, in later years.

The authors highlight quite well the difficulties in the compilation of landslide inventories and the fact that this is a time-consuming task. And this is, in my opinion, the reason why the compilation of landslide inventories is often a forgotten task in the national framework for landslide risk assessment, with very limited resources and personnel assigned, well indicated by Reviewer 1. 

We must also remember that in the past many previous authors have prepared landslide thresholds by using separated landslide dataset. It is with the most recent need of operational landslide early warning systems that we demand reliable landslide data to use in thresholds analyses.

The authors described well the existence of other databases in Italy, their limitations and the need for a “separate” dataset for rainfall-induced landslides. This may seem a good solution at the beginning, but in a long run it will create problems for those who manage landslide prevention at national level. Therefore I have some general comments for the authors that I would like to be adressed in the document. 

• It is not clear if and how ITALICA communicate with the other existing inventories. If a new rainfall-induced landslide should be registered in ITALICA, will this be send automatically in the other databases and viceversa? How to deal with future updates? Who is responsible for these updates?
• Have you considered the possibility to improve one of the existing databases instead of making a new one? Adding for example a “quality level code” or create a separate module with specific parameters for those landslides  that should be used for thresholds?
• Some operational early warning systems send alert messages also for landslide triggered by snow melt or only for high soil water saturation, what do you suggest based on your experience? do we need to create a separate database for each triggering?
• Many countries have one database, managed by one national institution, with limited resources and personal, and efforts are being made to improve the quality of the registered events, by assigning a quality level to each entry and to provide information about triggering. Based on your experience and take into account the problems with future updates, what do you recommend? To create separate databases or improve the existing inventories?
• At the end you have created a database with both low and high quality data. Could you explain better if you could use the low quality data in your threshold analyses (maybe for threshold at regional scale requiring low accuracy)? or do you need always high quality data? 

I also agree that the article is quite well structured and written, with adequate length and, with explanations and statements well illustrated.

It can be an important contribution and of interest to the landslide research community working in particular with landslide early warning systems. I would recommend this paper to be accepted for publication, with very minor edits (see supplied comments).