General Comments

The authors present updated landslide susceptibility maps for in the Philippines Itogon and Abuan derived from typhoon-triggered landslide inventories. Binary Logistic Regression (BLR) and a Least Absolute Shrinkage and Selection Operator (LASSO) technique for the selection of variables were used to model the susceptibility of this region. Susceptibility models derived from independent and a combination of the 2009 and 2018 events were created in Itogon. A susceptibility model was also derived in Abuan based on a 2019 typhoon event.

Their results present information that emphasizes the importance of utilizing multi-temporal inventories in developing future susceptibility maps to inform land use and hazard zonation policies.

However, there are two significant points that should be considered to improve this manuscript:

First, the results of two different study areas considering 3 events across time that aim to assess a research objective that operates on the hypothesis that the time dependence of typhoon-triggered landslides in a region would be evident in the deterioration of model accuracy.

The results of the Itogon region with events from 2009 and 2018 sufficiently address this research question and provide quantified information on the temporal behavior of landslide susceptibility across time. The analysis of these results should already merit publication.

The inclusion of the results in the 2019 landslides from Abuan deviate from the direction of evaluating time-dependent susceptibility. The comparison of the model in Abuan to Itogon veers towards investigating regional and spatial differences between the sites in which these typhoon-triggered landslide occurred. I would recommend a separate study to focuses on the spatial and not temporal aspect of typhoon-triggered susceptibility be considered for the Abuan results.

Response – we are keen to retain the Abuan work within the paper as whilst we agree with the reviewer that the spatial element is less well quantified than the spatial, we are of the opinion that it adds other important elements to the discussion. E.g. whether no susceptibility map is better than a “bad one”. It also allows us to “set the scene” and provide motivation for a separate study on the spatial and regional difference as the reviewer suggests.

Second, the authors could consider referencing an updated Landslide Hazard Atlas of susceptibility maps generated by the University of the Philippines Resilience Institute and the Nationwide Operational Assessment of Hazards (NOAH), available at https://noah.up.edu.ph, rather than the MGB susceptibility maps. The landslide hazard maps are available on a national level and are used in practice for hazard zonation and land use planning. A large section in their discussion could benefit from comparing their results to the NOAH hazard information.

Response – we agree with the reviewer that reference to and comparison with the NOAH maps will be beneficial. And will add such reference and discussion throughout the paper where applicable.

The most important contribution of this study to the community is the quantified deterioration of susceptibility model performance accuracy in Itogon that typhoon-triggered landslides display a degree of dependency across time.

Overall, I recommend that the authors update their hazard information for better context in the discussion and to highlight the improvement of susceptibility information with multi-temporal landslide inventory. I also recommend that the authors contemplate on the exclusion of the 2019 Abuan landslides in this study. The results do not support the research objective’s underlying hypothesis to consider the time-dependence of typhoon-triggered landslides.

Response – as highlighted above, we will update our hazard information to include reference and discussion of the NOAH maps/information. Following discussion with the co-authors, we prefer to keep the Abuan elements of the paper but will re-phrase and re-structure the abstract/introduction to ensure that the preliminary consideration of the spatial element is also an aim of the paper.

Specific Comments

L45-50: Consider incorporating the landslide hazard information from the NOAH Landslide Hazard atlas. This information would be beneficial to further realizing the contribution made by this study in Itogon for typhoon-triggered landslides.

M.L. Rabonza, R.P. Felix, A.M.F Lagmay, R.N. Eco, I.J. Ortiz, ang D.K. Aquino (2015). Shallow landslide susceptibility mapping using high-resolution topography for areas devastated by super typhoon Haiyan. Landslides, Volume 13, Issue 1 pp 201-210

Alejandrino, A.M.F. Lagmay and R.N. Eco (2016) Shallow Landslide Hazard Mapping for Davao Oriental, Philippines Using a Deterministic GIS ,Model. In: Communicating Climate Change and Natural Hazard Risk and Cultivating Resilience: Case Studies for a Multidisciplinary Approach Eds. Yekaterina Y. Kontar. Springer, Berlin Germany

Paul Kenneth Luzon, Kristina Montalbo, Jam Galang, Jasmine May Sabado, Carmille Marie Escape, Raquel Felix, and Alfredo Mahar Francisco Lagmay (2016) Hazard mapping related to structurally controlled landslides in Southern Leyte, Philippines. Natural Hazards and Earth System Sciences, 16, 875-883, 2016

Response – yes, we will incorporate this and agree it will be beneficial for highlighting the contributions of this study.

L61-63: The concept of spatial and temporal dependence introduced in this section could be strengthened by a connection to the path-dependence of landslides by Temme et al. (2020).

Temme, A., Guzzetti, F., Samia, J., & Mirus, B. B. (2020). The future of landslides’ past—A framework for assessing consecutive landsliding systems. Landslides, 17(7), 1519–1528. https://doi.org/10.1007/s10346-020-01405-7

 Response – we agree, and will add sentences to connect and refer to the Temme paper.

L93: The use of the term time-dependence could pertain susceptibility during typhoon season, or within a sub-seasonal period. I recommend the authors to consider rephrasing this to a path-dependent perspective and connect to the concepts of Temme et al. (2020) and the results of the multi-temporal susceptibility analysis of Samia et al. (2020).

Samia, J., Temme, A., Bregt, A., Wallinga, J., Guzzetti, F., & Ardizzone, F. (2020). Dynamic path-dependent landslide susceptibility modelling. Natural Hazards and Earth System Sciences, 20(1), 271–285. https://doi.org/10.5194/nhess-20-271-2020

Response – we consider that there is a subtle difference between general time-dependence (which includes anything that makes landslides temporally dependent) and path-dependency, which predominantly considers how past landslides have a time-dependent influence on landslides. However, we agree with the reviewer that these terms could do with better defining in the context of each other, so will update the text to include reference to and better explanation of the Samia path dependency concepts.

L189: Why was the inventory slightly clipped? It also would be worth mentioning a brief qualitative comparison between this inventory 2018 Mangkhut and that of Emberson et al. (2022).

Emberson, R., Kirschbaum, D. B., Amatya, P., Tanyas, H., & Marc, O. (2022). Insights from the topographic characteristics of a large global catalog of rainfall-induced landslide event inventories. Natural Hazards and Earth System Sciences, 22(3), 1129–1149. https://doi.org/10.5194/nhess-22-1129-2022

 Response – the inventory was clipped to match the watershed boundary. We will include a brief comparison with the Emberson paper as we agree that this would be beneficial.

L371-400: These paragraphs are presented in a way that focuses on events across time, but gives the impression that the 2009, 2018 and 2019 landslides occurred on spatially similar settings or even same site. Splitting the presentation of results into two paragraphs (one for Abuan and one for Itogon) to discuss the separate geographic sites could make it clearer.

 Response – we agree that this section could be better structured, so will re-phrase / re-structure to make this clearer as suggested.

L455-473: Is there any insight on the hazard between 2009 and 2018 in Itogon that can be derived from the susceptibility models? Any insight on susceptibility or changes that could’ve caused landslides to occurred with smaller passing tropical cyclones within these 9 years? (Referring as well to insight from Figure 4)

Response – it is difficult to make any detailed inference on what caused the changes between 2009 and 2018 without any landslide data and trigger data in that 9 year period. We will add a few sentences discussing possible changes between the two dates and make reference to potential future work that could acquire more time slices of landslide data between 2009 and 2018 to properly assess what might cause changes in this period.

L474-565: Please refer to the Landslide Hazard information from the susceptibility maps of NOAH to provide an updated hazard context for this section of the discussion.

 Response – we will add reference to and discussion around NOAH as suggested.

L549-556: These are valid concerns and points of uncertainty raised for the Abuan susceptibility results. Though, the alignment of these results the objectives presented in L93-95 are not clear.

Response – we will update/ re-phrase the objectives so they better align with these results.  

L529-538: While magnitude underestimation is a limitation in the use of satellite-derived rainfall products, another factor worth discussing is the limitation to capture spatial patterns and locate the storm centers when using such products. (See Ozturk et al., 2021)

Ozturk, U., Saito, H., Matsushi, Y., Crisologo, I., & Schwanghart, W. (2021). Can global rainfall estimates (satellite and reanalysis) aid landslide hindcasting? Landslides, 18(9), 3119–3133. https://doi.org/10.1007/s10346-021-01689-3

Response – we agree that this would an interesting element to add to the discussion, so will include a few new sentences to consider this as suggested.  

L590-612: Table 1. Shows that land cover is significant for the 2009 and combined 2009+2018 model. It would worth mentioning the role of land cover change that could have an influence susceptibility over time. Itogon is estimated have had significant tree cover loss between 2010 and 2020 based on: Global Forest Watch, http://globalforestwatch.org.

1. Hansen, P. V. Potapov, R. Moore, M. Hancher, S. A. Turubanova, A. Tyukavina, D.Thau, S. V. Stehman, S. J. Goetz, T. R. Loveland, A. Kommareddy, A. Egorov, L. Chini, C. O. Justice, J. R. G. Townshend, High-resolution global maps of 21st-century forest cover change. Science 342, 850–853 (2013).

 Response – we agree that land use change could be important so will add a paragraph to the discussion considering this as suggested.

Technical Corrections

L32: ‘>30o’ to >30^o

Figure 6. Consider using ‘performance’ rather than ‘success’

Figure 7. Consider using ‘performance’ rather than ‘success’

Response – we will correct the degree symbol and change to performance.

The authors present an interesting and well-written manuscript presenting an analysis of typhoon-triggered landslides as input to susceptibility maps for two regions in the Philippines.  The paper focuses on three separate typhoon events, two in one region and one in another, and discusses spatial and temporal effects on landslide susceptibility.  Overall the reviewer thinks the manuscript is well written and explains the problem, method and conclusions in a clear and understandable manner.  

Two general comments:

1. The discussion about the accuracy of the Abuan model includes discussion on whether "something is better than nothing" with regard to susceptibility mapping.  The reviewer thinks this is a very relevant comment, and should be highlighted in the conclusions of the manuscript.  If the authors believe the input data into the model is not sufficient enough to produce a reliable susceptibility map, it should not be concluded that a new susceptibility map is produced, particularly when the region did not already have a map available. 

Response – we will highlight this better in the conclusion and agree that it should be concluded that a map should not be considered produced and useable if the input data is insufficient. So we will make sure this point is clear in the discussion / conclusion.

1. The aim of the paper was to use data from multiple typhoon events to assess typhoon-triggered landslide susceptibility in the Philippines.  The reviewer thinks the topic of time-dependance is discussed sufficiently within the Itogon region, with analysis completed on two individual typhoon events and then a combination of the two events.  These three models are then tested on the 2019 data from the Abuan region, with poor results (AUROC between 0,54 and 0,59 according to Figure 6 and 7).  To the reviewer this seems that some discussion is warranted on the spatial dependancy, although it is mentioned in Line 574 that this is not the focus of the paper (but not mentioned or excluded from the paper in the introduction or abstract).  If the focus of the paper is really only discussing time depenance, it may not be relevant to include the Abuan region, which is only analysed using one typhoon event.

Response – we do wish to retain the Abuan work as it is fundamental to some of the important discussion points about whether “bad maps are better than no maps”. However, we agree that the introduction / abstract perhaps don’t fully capture why Abuan is included, so we will update the introduction/abstract and other relevant places to ensure that the reasoning for including Abuan is clear – i.e., that whilst it is only based on one event, it gives some preliminary insight into spatial dependency that could be further investigated in future work, and allows important discission on the importance of model input data and how this impacts “bad vs no models”.

Specific comments:

- Section 2  general comment - There is a mixing of unit systems here.  Amount of rain is listed in millimeters, while wind speeds are noted in mils per hour.  

Response – will change all to SI units.

- The accuracy of the model from the 2009 typhoon was classified as "good/excellent", and the combined 2009/2018 model as "good", are there any complications with building a susceptibility model from a typhoon event which was described in Section 2.1 as influenced by the Fijiwhara effect, where the typhoon was impacted/worsened by a nearby typhoon?  

Response – yes, there are likely some complications as it is hard to assess the specific impacts of the Fijwhara effect. We will add a few sentences to properly discuss how the Fijwhara effect might complicate the process.

- Line 227 - the reviewer thinks the toolbox in ArcGIS may be called "Spatial analyst", not Spatial Analysis. 

Response – correct, this was a typo on our part and we will correct it.

- Line 286-289 - In point 4 it would be nice to mention what the other predisposing factors are.  Here it is listed that three factors are categorical, but one must look to Table 1 to find the other factors.

Response – agreed, we will list all the other factors here to make it easier for the reader to know what they are from the offset.  

- Line 443-447 - The figure caption for Figure 7 was challenging to read, with similar years being discussed.  Perhaps make it more clear on the figures that a and b are using a different model year than c and d. 

Response – we will update the figure captions / labels to make this clearer as suggested.

- Line 449 - the word "models" is missing after 2009.

Response – we will correct.

- Figure 8 - In the plot for e) Aspect, the reviewer does not understand why the distributions for the Itogon catchments have a peak at E/SE aspects, when the bar charts are approximately equal to the Abuan catchment data. 

Response – The landscape distributions of aspect are similar in the Itogon and Abuan regions. However, in the Itogon case, landslides were found to predominantly occur at E/SE aspects. I.e. they were occurring in these aspects proportionally more than would be expected if their distributions were random. This is not surprising, as many factors such as thermal weathering, rain and wind direction etc, have been shown to cause landslides to preferentially occur at different aspects. Conversely, in the Abuan case, landslides were found to occur across all aspects, with seemingly no predisposition for landslides to occur at any particular aspect of slope. We will add a sentence or two to better explain this in the paper.

- Line 520-527 - The sentences discussing the three main zones (core zone, middle zone and peripheral zone) are not really sentences and are slightly challenging to read.  Consider restructuring. 

Response – we agree that these sentences aren’t the easiest to read and will re-structure them.