This paper proposes a method for calculating rainfall thresholds for rainfall-induced landslides using the Multilayer Perceptron regression method and the feasibility of this method has been verified. In addition, the authors use a large amount of data and various data-driven modeling techniques. The research results have practical significance for the early warning and prevention of rainfall-induced landslides. It is recommended that the paper be published after revisions, addressing the following comments:

1. The originality of the study is not prominently highlighted in the abstract. It is recommended to enhance this aspect.

2. The use of the Multilayer Perceptron (MLP) for analyzing rainfall thresholds is a commendable innovation. However, since this method has been widely used in other fields, it would suffice to mention it with appropriate references. The MLP framework in Figure 1 is relatively simple and takes up significant space; consider removing it.

3. Given the length of the article and the complexity of the methods and procedures involved, it is suggested that the authors create a flowchart to further elucidate the methodological steps.

4. The paper mentions dividing the study area based on topography and climate, followed by partial merging based on the number of historical disasters. It is suggested to include the final regional division results in Figure 4 to avoid any ambiguity.

5. In Table 6, the categories of slope structures are represented by A-H, which is unclear and it is recommended to change them to professional terms.

6. In Table 7, the units of some landslide susceptibility factors are given, but the units for factors such as road density are missing.

7. There are many images in the article. Consider combining some of them for display.

8. Some descriptions of the figures, such as the explanation of different colors in Figure 9 in lines 223-224, should be moved from the main text to the figure captions.

9. The clarity of Figure 14 is insufficient. It is recommended to change the layout from three columns to two columns.

In the MS titled "Optimizing Rainfall-Triggered Landslide Thresholds to Warning Daily Landslide Hazard in Three Gorges Reservoir Area", the authors tried to propose a rainfall threshold for predicting landslides on a daily scale. The topic fits the journal's scope while the entire MS was poorly structured, and the methods were not clearly explained. Hence a major revision is suggested.

Major comments:

1. There were too many abbreviations, making the MS hard to follow. Please reduce them to a reasonable number (less than 10).

2. The historical landslides were divided into 2 groups. For those that occurred during the dry season (41), "only rainfall thresholds for dry season landslides were calculated for the entire study area". The authors should explain what are the contributing factors for those dry season landslides? What kind of role of these factors are playing during the rainy season?

3. Following the above question, the water level fluctuation, and the underground water level might be important factors. But these factors had not been considered in section 4.2.1 or Table 5 "Source of data on landslide inducing factors".

4. In Table 5, the human engineering activities were indicated using "road density", which seems not reasonable, unless it can be clearly figured out from the landslide inventory.

5. The methodology and the framework should be elaborated using a figure. It reads confusing as it includes too many results (Figs. 7-20, Tables 1-8) using several methods from machine learning to threshold curve yielding in different zones. 

6. Some of the figures are useless, such as Figures 1-2. Some of the figures should be combined.

7. Try to find a fault map and include it in Figure 3, it's important and should not be ignored in this mountainous area.

8. The details of the rainfall data should be introduced, including the covering period, the temporal resolution, etc.

9. The Thiessen polygon method was adopted to delineate the study area and the rainfall station (Figure 5), but it is not convincing. The zonation of the rainfall was also conducted (Figure 4). Why two methods were applied for one factor? 

10. In Figure 8, as the landslide data is not sufficient, the rainfall threshold results are derived using scattered E-D points. So, why do the authors have to conduct the zonation of rainfall stations?

11. Figure 9 shows the E-D-R threshold model for a specific zone, that is, Z13. Why only Z13?

12. "Warning Daily Landslide Hazard" reads confusing. I guess the authors want to emphasize the warning was daily, but why?

Specific comments:

The writing should be significantly improved. There were too many grammars and typos. The terms should be defined accurately, for instance, "the third dimension indicator 'rainfall for the day' (R)".