The paper is dedicated to the description of the construction and structure of a global database of lightning-ignited wildfire (LIW) holdover times collected from a large number of studies during the last century in different regions. The investigation of LIWs characteristics, especially on the background of global climate change is a topic of highly relevance and interest. Moreover, this database is the first freely available and usable in its current format. Since this data set is the first one and unique, it is probably incomplete, but it can be assumed that it will be constantly updated in the future, of course. Even now, the database on holdover has diverse potential applications. This database undoubtedly may become a significant data source for researchers interested in studying LIWs.

However, I have some points that are unclear for me.

1. May be, it is worth to exclude Table 3 and Table 4 from the text? They are already presented in the database. Or if the authors want to provide an example of data, probably it will be better to make the Tables shorter? However, this is just a suggestion, it is on the authors’ decision.
2. What concerns Table 4, as I understood non-censored data also could be derived from the censored data? For what purposes non-censored data are provided (in Table 4 and in the database, in general)? Moreover, non-censored data are not discussed and analyzed in the text. How one can use them?

In general, the article is well structured and clear, the language consistent and precise. The article is appropriate to support the publication of a data set.

Best regards 

This study developed a database on holdover time of lightning-ignited wildfires (LIWs) based on 29 studies through a literature review. The manuscript provided a summary of the proposed database. Holdover time is a valuable parameter to understand the behavior of LIWs. However, there are a few questions in this manuscript to be solved.

1. The holdover time and the location of each lightning-ignited wildfire are more valuable than the frequency distributions. Actually, readers can use holdover time and locations to calculate frequency by themselves and do more analysis. Can you provide the accurate holdover time and location of each LIW?
2. Some wildfires didn’t go to the arrival stage, and their holdover time was missing. Can you remove the biases in this database?
3. Figure 1a and Figure 1b showed relative frequencies from different study areas, i.e., Australia, Arizona and New Mexico. It is very difficult to compare the differences between daily and hourly frequency distributions because the study areas were also different. It is necessary to use the same study area.
4. In this database, some holdover time was shown by day while others were by hour, which was not consistent. Can you convert all data to the hourly scale?
5. Figure 2, the biome ‘Mediterranean forests’ is not accurate. In Olson et al. (2001), it was ‘Mediterranean forests, woodlands and scrub’.
6. Table 3 showed very ambiguous study areas, e.g., California, which made it not user-friendly because there were different climate zones and vegetation distributions in California. Readers need more detailed locations. There were also study areas like Finland and Switzerland in Table 3, which were too coarse to use. Readers can't use this data if they want to establish the relationships between holdover time and climate, vegetation, terrain, etc. 
7. Please add citations and more explanations to the maximum proximity index and the minimum holdover time on line 205.
8. The diversity of the database is small. There were only 5 biomes in this study, where temperate broadleaf and mixed forests and flooded grasslands and savannas covered only two samples and one sample, respectively. However, there were 14 biomes globally. It is necessary to add more data. 

Best regards