|The theme addressed in the manuscript “Natural and human-induced landslides in a tropical mountainous region: The Rift flank west of Lake Kivu (DR Congo)” by Mateso et al., is an interesting and relevant study that explores how landslides occurrence since the late 50’s of the last century is impacted by anthropogenic changes related to forest cover, roads, and mining activities in a rural tropical mountainous region under high anthropogenic pressure. Nevertheless, the manuscript, presents some aspects that should be better addressed. These are detailed described in the Specific Comments and Technical Comments sections.|
1) Are some of the inventories used by authors covering landslides trigger by rainfall and by earthquakes. That should be turned clear in the landslide inventory section. In addition, the classification of landslides as mining and road landslides despite their depth and type is not the most adequate and constrain possible explanations for their occurrence.
2) What this work brings differently from the works of Depicker and co-authors for example?
3) If I understood well only exists aerial photographs from late 50´s of the last century and from 2016 land cover ESA model. How authors stablished the correlation between landslides that occurred outside the time frame of these two land cover images (part of the landslides are dated from 2005-2019 images) and the predisposing terrain conditions related with forest cover? This is, in each part of this 60-year period occurred the forest loss, forest gain? For example (L. 208-218) how can someone interpret based on figure 2a that in 1955-58 48 % was already deforested. It means that the entire area was covered with forest before? The comparison only allows to compare land use cover changes between the two-time frames. I strongly believe that these limitations should be carefully considered and the dynamic component revised by authors.
4) A composed figure with the predisposing factors should be considered by authors
5) In the methods section authors should clearly describe the susceptibility modelling strategy and with which landslide inventory partitions intend to validate the susceptibility maps. For example, validate the susceptibility maps with road and mining landslides and why.
6) In the discussion section, please address better the possible bias that could overcome from the resampling of predictor variables for the SRTM resolution
L. 54: I agree with authors regarding deep-seated and shallow landslides differentiation but a discussion regarding the type of landslide is also acknowledge in this introduction section. The same should be better addressed in L. 179-185: The classification of the landslide type in mining and rod landslides does not seem the most interesting. Even so, what type of landslides are mapped? In addition, it’s important to earlier in the manuscript to clarify which type of landslides authors are used to assess susceptibility. Only slides? Adjust terminology along the manuscript accordingly.
In section 2.1 a description of the base inventory constructed by Depicker et al (2020) is welcome. For example, what landslide inventory period is covered by the work of Depicker et al (2020)?
L. 142: Please address better what the mean of “differentiated between the processes”
L. 154-156: For authors what is consider a recent landslide? It is a landslide with less than 60 years? How distant in time is possible to relate landslides with a date of occurrence. This point regarding the definition of old and recent landslides, concerning landslide age needs to be clarified in my opinion.
L. 162-165: the visual estimation of landslide depth is discussable. Please turn clear.
L. 170-171 “by selecting representative areas with various landslide and landscape characteristics”. How the representative areas are selected and how are these areas are defined? Which percentage of study area was covered by field survey? The use of the criteria “various landslides” should also be clarified.
L. 188: Include in the manuscript the criteria that define a landslide event.
L.191-192: why a maximum of 30 landslides per cluster? The concept of minus event, containing other isolated landslides (L. 193) was not perfectly clear to me.
L. 250: Is missing from the list of predictor variables the forest dynamics between 1955-58 and 2016
L. 237-239: The difference between shallow landslides and deep-seated landslides is relevant if landslides are of the same landslide type, e.g., slides. Authors made this explanation latter on the manuscript, but that should be turned clear earlier, in section 2.1. Moreover, how is selected one point (pixel) per landslide. The landslide susceptibility predictive power provides significant differences depending on the use of the entire landslide area for training the susceptibility model considering larger usually deep-seated slides or shallow slides (usually smaller in size).
L.239: Please turn clear what is the centre of the landslide trigger area, this concept is not clear to me. Do you mean rupture zone, depletion zone, initiation area? Please, see my previous comment on this topic.
L. 247-249: How spatial predisposing predictors allow to discuss triggering conditions? I apologize, but I think the description is too general and maybe it should be preferable to direct the explanation for preparatory conditions for landsliding.
L. 270 – 272m. Authors mentioning that the few recent landslides observed along these roads confirm the assumption that the direct impact of the main roads on the occurrence of landsides is limited. This are not results? Move for the results section.
L. 287-292: I understand authors idea, but how much of these shallow landslides occurred along or near the fault zone? If authors are using the distance to faults to correlate with weathering, are not these weathering materials more prone for landslding? This should be better addressed.
L. 355- Seven landslides are mining landslides? But this category is not related to road landslides? Are landslides classified in more than one category? L. 357-358: please avoid repetitions, the idea that road landslides include only the landslides located within 50 m of roads was already written above in the paragraph.
L. 367-372: how much of the 634 shallow landslides of the October 2014 rainstorm event occurred in the study area? Only 14? This was not clear to me. Where they occur specifically on the shores of Lake Kivu? The blue dots on figure 3 seems few for these 634 landslides mentioned.
L. 388-390: 25% more deep-seated landslides were identified in the field but not used in the analysis to avoid biases due to overrepresentation. Are the predisposing conditions associated to these landslides like the ones used to assess landslides susceptibility? If not, how good can be the predictive susceptibility map. Validating the produced landslide susceptibility map with those landslides could help answering this question.
L. 421: the fact that 72 % of the landslides are found in areas of forest loss really means that they occurred with that predisposing condition? These landslides occurred before or after the forest loss? The dates of the landslides not always match the date of the forest cover image.
L. 462-463: The authors analysis conclude that shallow landslides are 2.5 times more likely to occur as deforestation increases? Please see my previous comment on this topic.
L. 473: This steep was not clear from the methods section. If the susceptibility model was produced for shallow slides, why validate it with mining landslides. It seems to me from figure 5b that mining landslides are closer to the characteristics of recent deep-seated landslides than to shallow slides. This validation with landslides datasets not used to construct the susceptibility maps is not clear from the methods section.
L. 525-536: landslides in the permanent anthropogenic environment – what is the relationship between the short time for which the shallow landslides were inventoried in the study area and the fact that the study area was not altered by mechanized farming? When mechanized farming begins in the study area? Which terrain anthropogenic conditions lead to slope instability in these anthropogenic landslides. This should be better addressed.
L. 545: the model show that seismic activity play a dominant role in deep-seated landslide distribution in the study area – ok. Were all the old deep-seated landslides earthquake triggered, if not, are the terrain conditions of old deep-seated rainfall-triggered landslides like the previous ones?
L. 592-593: the regional susceptibility model shows that deforestation is the most important factor. Again, and I apologize for my repetition, this is difficult to understand since landslide dates do not have a landcover photography that allows to stablish a direct relationships with landcover. This section of drivers for shallow slides should be better addressed concerning the relationship between forest cover type and landslides.
L. 615: why authors say that colluvium deposits result in a concentration of susceptible places. This is not clear to me. Please address better.
Section 4.4. Why should the produced susceptibility models validate the mining and road landslides: Are these two categories related to shallow or deep-seated landslides?
L. 58, L. 123: Monsieurs et al., 2018 and references therein along the manuscript: please check and indicate if the references to the work of Monsieurs et al., 2018 is 2018a or 2018b.
L. 87: please check “altitude” or “elevation”, If related with relief it should be elevation.
L. 203-205: Create an appropriate reference in the reference list for the work of Smets et al., to be submitted) and cite accordingly in the manuscript.
L. 794 and L. 797 please indicate which reference is 2021a and 2021b
Items not cited in the manuscript but present in the References list: L. 760: Trumbore et al 2019….; L. 773: Fisher et al 2012….; L. 779: Geenen 2012; L. 810: Kampunzu et al 1998….; L. 928: Vanacker et al 2003….; Hungr et al., 2014. Please check references/reference list and respective citation on the manuscript.
Figures and Tables
Page 3 – Figure 1: Check graphic representation of scale bars of maps 1a and 1b, uniformize and add bottom line in scale bar of map 1b. The same should be checked in all figures. Include in legend of figure 1a the meaning of the yellow star in the upper-right part of figure. It represents the location of study area?
Page 6 – Figure 2: In figure 2a and 2b consider having the same legend for classes “no forest” in figure 2a and “permanent anthropogenic environments” on figure 2b, if representing the same class variable. The same for “Forest cover” in figure 2a and “permanent forest” in figure 2b.
Page 10 – Figure 3: Somewhere along the manuscript text clarify what criteria define a heavy rainfall event. Furthermore, the shallow slides associated to each cluster should be identified. Consider adding polygon lines (limits) grouping all the landslides for each event. Figure 3b include scale bar. The additional landslides identified only in the field could be associated with some of the rainfall events identified on figure 3a? Please turn clear along text.
Page 17 – Figure 8: What the meaning of values above the vertical bars since they do not match with both right and left side bars. Is the frequency ratio score. Please adjust caption.