The manuscript is organized well, however, it needs significant revisions to be considered for publication. First of all, the title needs to be rephrased. The manuscript only deals with the hazard aspect of landslides, the risk and management parts are not even discussed.
Also, the introduction and abstract say about uncertainties, but there is no scientific quantification carried out in the whole work. This is critical as the authors claimed, and additional analysis should be conducted.
Line 32: TSS .unexplained abbreviation in abstract
Lines 59-60: this is confusing. ML models are seldom used to understand the mechanisms of initiation, rather they were used more as a black box method
Lines 131-132: are these landslides specifically associated with an event? or are they recurring? please describe the temporal variation of the selected landslide events
Also, there are 317 landslides, converted to points, such that landslides:non-landslides is 1? or how did you convert the landslide areas to points?
Line 134: 0.3 km: is it km2?
Lines 150-152: this is valid if the selected landslides are also associated with the time of LULC map considered The LULC varies with respect to time, and it is important that the considered land cover is actually associated with the triggering of the landslide. Provide more details about the LULC map used. Also, if any maps are digitized from available non-digital formats, the scale should be mentioned.
Lines 164-166: what parameter is used to represent rainfall? is it average annual rainfall? or the rainfall event just before the failure? or peak rainfall or rainfall intensity? the definition should be clear at least here
Section 2.5: the hyper-parameters used in each model should be mentioned here, and the fine-tuned values should be mentioned in results section
Lines 338-339 the definition of 'presence' is not clear. are they single points of landslides? or landslide areas?
Section 3.1: on what data is the multi-collinearity measured? with all 317 LS and non-LS points? If the data is different from the sample size, these values cannot be generalized
Lines 366-368: It is also possible that the influence of features can vary with the ML model, and also the training data that you have used. This part also needs to be explored.
Also, What happens if the less relevant features are removed? and the prediction is carried out only using these features? will it improve the performance?
I should also say that the explainability of ML models has already progressed much ahead of simple feature importance. You can refer to recent publications like https://doi.org/10.1016/j.cageo.2023.105364, and try to improve the explainability part using shapely values
Lines 400-401: is the ratio of landslides to non-landslide point always kept as 1? this is also important. Also, when you say 25 % of the inventory, is the remaining 75 % used for testing/validation?
And what is the train to test ratio for each sampling case? did you apply any cross validation techniques?
Some of these questions are answered in the figures, but they should also be mentioned in the text.
Line 403: 'In order words', or 'In other words?'
Lines 403-406: this statement is valid only with the definition of the classification method that you adopted. How did you classify the total area? Natural break method? or an equal interval or quantile-based approach?
Figure 2: replace 'enventory?' with 'inventory', also rephrase the caption
Figure 3: follow the standard reference formatting of the journal, and also provide captions for each sub figure, after the figure caption
Figure 5: if you overlap the landslide inventory on slope map, it is very evident that there are landslide point in the lowest sloping areas, close to the water body on the western part. Still, the results says slope is the most critical landslide feature. How do the authors justify this?
Also, Why is SPI values very high on the flatter portion of the water body? This should be rechecked.
Figures 6-9:please use discrete classifications from very low to very high, as mentioned in the text. And provide a suitable title for the legend. What the values presented? are they 'Landslide probabilities? What are the uncertainties associated with these values? will they vary with the sampling size or model?
Figure 11: how many landslides are present in each of these classes? are the performance metrics biased to the presence scenario or absence scenario?
Table 2: The correlation matrix should also be plotted for each case

The authors have satisfactorily addressed most of the comments. The following minor comments can also be taken care of:
The abstract is too long. It can be shortened.
Lines 179-191: Authors mention many websites here. These citations should follow the standard formatting requirements of the journal.