This manuscript provides an interesting research into the use of Sentinel-1 SAR data to accurately identify the timing of earthquake-triggered landslides, their reactivations and potential precursory motions. The authors apply a combination of amplitude and coherence-based timing detection methodologies on the 2018 Lombok, Indonesia earthquake sequence which results in a multi-temporal inventory. This then allows them to interpret the different triggering conditions of new and reactivated landslides.

I believe this study is very interesting, the techniques are very relevant, and could have potential to be adapted to also identify timing of rainfall-triggered landslides. However, I think this manuscript could benefit from additional considerations on the methods and writing. My specific comments and line-by-line comments  are attached as supplement. These comments mainly come down to the following main points:

• The objectives of this paper are not that clear from the introduction. I believe that the introduction would benefit from a clear presentation of the aims and objectives of the manuscript. Currently it does not become clear that the coherence matrix approach is a novel technique in this context which you are going to explore the usage of. This has implications on the timing results and the interpretability of them.
• Given that the coherence matrix approach is a novel methodology for landslide timing detection, I believe this requires comprehensive analysis on the ability to use it for this purpose. This currently seems to be lacking. For example, once you identify timing, you seem to be 100% sure about the validity. Partly from these timings you then derive conclusions on precursory motion and reactivation. There seem to be little discussion about the actual uncertainties related to those timings and the constraints it puts on the results. What is the effect of noise? In addition, the coherence product has a relatively large spatial resolution. It is unclear how this spatial resolution affects the ability to use this product in your method. I can imagine that mixed pixels might play an important role. Given that you propose a new method, perhaps somehow a sensitivity analysis on the effect of landslide size on the ability to detect changes could be beneficial, especially when others want to use a similar technique. 
• You use a very limited amount of landslides compared to the complete inventory and derive general conclusions on the triggering conditions of new and reactivated landslides. Are these results representative? I think this should to be addressed and put into perspective.
• I believe that there are some structural changes that could improve the manuscript: (1) I think it is more relevant to first describe the SAR datasets (current section 2.3) before diving into the detection methodologies (current section 2.2). This will allow to introduce all the concepts that you will be talking about during the detection methodologies section. (2) Section 4.1 seems to consist of a mix of methodology and results that I think would better fit in the methods and results sections.
• It is not very clear which precursory motions and reactivations have been validated through optical data. This seems essential for the applicability of the methods and the interpretability of the results later.
• Figures should be improved, legends, scale and axis-labels are sometimes missing,

I find the manuscript interesting, and the topic and experiments are relevant. Overall, I believe the manuscript would benefit from some clarifications and restructuring to enhance clarity and flow. Below are some suggestions for improvement:

• The manuscript presents two key novelties: (i) the coherence matrices method and (ii) the novel insights into the specific earthquake sequences that this method, along with the amplitude-based approach, helps uncover. However, while the introduction focuses more on the method (i), the conclusions emphasize the second (ii) novelty. Aligning these sections more closely could strengthen the manuscript.
• The method could also be valuable in distinguishing landslides triggered by different rainfall peaks occurring close in time (e.g., Emilia Romagna in 2024). As you note, this introduces additional complexities, but it could be an interesting avenue for future research. The University of Bologna has open-sourced a highly accurate dataset for that event, which may be useful for further exploration.
• The choice of thresholds for discarding landslides (<2000 m² for the amplitude approach with a 22x20 pixel size and <3600 m² for the coherence approach with a 60x66 pixel size) should be better justified. Additionally, it would be helpful to clarify why the amplitude-based approach includes landslides approximately five pixels in size, while the coherence-based approach includes landslides as small as a single pixel.
• If landslides below 2000 m² and 3600 m² are discarded because they are assumed to be too small for detectable changes, how does this impact the detection of reactivations? Are reactivations generally larger than these thresholds, or do you believe smaller reactivations can still be detected?
• Regarding reactivations, do you primarily detect an increase in landslide area, or do you also observe failures within the existing scar?
• How do you account for geometric distortions? For example, do you remove shadowed areas a priori, or do you rely on the assumption that using both orbits provides a high probability of capturing meaningful data?
• Figure 2 is crucial but could be made clearer. A suggested improvement would be to plot coherence on the y-axis and real dates on the x-axis, with coherence values represented as horizontal lines extending from the date of the first image to the last. Not sure this would help, but it is worth a try.
• The validation of this approach is thorough and well-executed. However, I have some questions regarding the terminology. In Section 2.4, you state that optical and SAR data can agree, disagree, or partially agree. Could you expand on what "partially agree" means? Additionally, framing it this way implies there is no ground truth. However, in cases where high-resolution optical imagery confirms a reactivation linked to a specific shaking event, wouldn’t that be considered ground truth? Even a few well-validated cases could be sufficient to support the analysis and conclusions.
• It would be helpful to explicitly state that the conclusions are valid for landslides above a certain size threshold, as smaller failures may behave differently.
• In the abstract, I recommend opening with the importance of attributing landslides to specific triggers to highlight the study’s relevance.
• Line 11: "sequences of triggers" → "sequences of earthquakes."
• Section 2.2 could be removed, as it does not seem directly relevant to the study.
• Line 103: "somewhat simpler" → Consider rewording to "We modify the approach as we can assume landslides occur concurrently..." for clarity.
• Line 465: Could you clarify what kind of precursory activity you are referring to?