the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 13 Aug 2024
Shaping shallow landslide susceptibility as a function of rainfall events
Abstract. This paper tests a multivariate statistical model to simulate rainfall dependent susceptibility scenarios of shallow landslides. To this end, extreme rainfall events spanning from 1977 to 2021 in the Orba basin (a study area of 505 km2 located in Piedmont, northern Italy), have been considered. First of all, the role of conditioning and triggering factors on the spatial pattern of shallow landslides in areas with complex geological conditions is analysed by comparing their spatial distribution and their influence within logistic regression models, with results showing that rainfall and specific lithological and geomorphological conditions exert the strongest control on the spatial pattern of landslide.
Different rainfall-based scenarios were then modelled using logistic regression models trained on different combinations of past events and evaluated using an ensemble of performance metrics. Models calibrated on multi-events outperform the ones based on a single event, since they are capable of compensating for local misleading effects that can arise from the use of a single rainfall event. The best performing developed model considers all the landslide triggering rainfall scenarios and two non-triggering intense rainfall events, with a score of 0.90 out of 1 on the multi-criteria TOPSIS-based performance index.
Finally, a new approach based on misclassification costs is proposed to account for false negatives and false positives in the predicted susceptibility maps.
Overall, this approach based on a multi-event calibration and on a misclassification costs analysis shows promise in producing rainfall dependent shallow landslide susceptibility scenarios that could be used for hazard analyses, early warning systems and to assist decision-makers in developing risk mitigation strategies.
- Preprint
(2976 KB) - Metadata XML
-
Supplement
(1173 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on nhess-2024-140', Jürgen Mey, 07 Sep 2024
Review for: “Shaping shallow landslide susceptibility as a function of rainfall events” by Fumagalli et al. submitted to NHESS 2024.
This manuscript deals with the high frequency–low magnitude phenomena of shallow landslides that occur after heavy rainstorms at the northern edge of the Ligurian Apennines in Italy. A set of 5 events (3 with landslides, 2 without landslides) are used to identify the conditioning and triggering factors for shallow landslides with emphasis on rainfall. The authors first use multiple logistic regressions to develop five different models, whose performance they evaluate based on a method called TOPSIS that accounts for a set of classification quality criteria. The final definition of susceptibility thresholds for their best model, incorporates the economic costs of misclassification for three different scenarios. I.e. they come up with three different susceptibility maps for each of the 5 rainfall events.
I agree that their approach would be helpful in early warning and mitigation of such small but still dangerous landslides under future rainfall scenarios.
The main points I took away from this study was that (i) the best model was trained using the landslide and rainfall data from all 5 events, (ii) including heavy rainfall events that did not trigger landslides helps to account for the rainfall threshold below which landslides are less likely, (iii) the pre-event accumulated rainfall as a proxy for the soil saturation is an important conditioning factor and (iv) that the susceptibility maps are sensitive to the assumed misclassification cost scenario.
The manuscript is in fairly good shape but needs some improvement concerning the description of the data and the methodology and a harmonization between text and figures/tables. For example, I think it would have helped to include a map of the slope units used for the analysis at least in the supplements. At several places the authors refer to information in a figure/table that is actually missing. After addressing these and the following points, I think the manuscript would be ready for publication.
Jürgen Mey
More specific comments:
L34: “a” missing
L78: the difference between “Precision” and TPR is not clear
L94: You should define here what you mean by “costs”.
L119: peridotite is not a metamorphic rock
L125: I doubt that “dipping” is the right terminus for describing the orientation of relief. In fact, relief has no orientation at all. You write about the strata (also in the following sentence) but you do not mention the actual orientation in terms of strike and dip. It would be interesting whether (or where) the strata (or other discontinuities) are oriented parallel to the hillslopes.
L129-130: give reference
L136: You write that you have analyzed GE images, orthophotos, event maps and field observations but your manuscript lacks any description of how you analyzed GE images, orthophotos etc.. What do you mean by field reconnaissance? Have you done field work yourselves or do you refer to the work of others?
L141-142: Disregarding the totally unclear (for me) use of “slope units” as “mapping units”, you obviously jump to the conclusion that the difference is negligible before justification is given.
L167: “(Frattini and Crosta, 2013)” not in parentheses
L181: What kind of DEM is this? Please give a reference. Which software did you use to extract the morphometric parameters from the DEM?
L184-187: It would be good to have a table that shows exactly, which units from the original map have been aggregated.
L188-190: Given that this is a global database, is there any validation data (ground truth) available for the SoilGrids map in your study area?
L197: How did you interpolate the rainfall data and to what resolution?
L199: Table 1 is not showing the “total rainfall of the events”
L201: Maximum daily rainfall intensities for each event?
L202: Which values did you use for the daily rainfall with a return period of 10 years and the mean annual precipitation?
L203: “areas”: Do you really use different study areas for the events?
L215-223: I am struggling with this paragraph. Can you give a definition of slope unit? Is this a subset of hillslopes that have a certain range of orientation in terms of azimuth and inclination? Since slope unit is a terrain unit I suggest rephrasing (i). What is meant by the “percentage within a unit”? Percentage of grain sizes or lithological units etc.?
L232-235: Isn’t this already part of the results?
L259: Which indices? The latter four?
L261: What is the difference between these 50 analyses? Do you change the training and validation subsets?
L295-296: Fig. 5 does not show the cumulative rainfall during the event.
L297-301: This paragraph is hard to comprehend given Fig. 5. What do the individual points in Fig.5 represent? Why are there only 3 of them for the 1977 event in panel (a) but 10 for the 2019 event.
L298-299: “..for the same maximum rainfall intensity (Fig. 5a), the landslide density is offset for the three inventories..” This cannot be judged from the figure because events 1977 and 2014 have not a single max rainfall intensity data point in common.
L300-301: “the higher the antecedent cumulative rainfall, the higher the sensitivity” à In Fig. 1 the cumulative rainfall curves show a very similar 90 days antecedent cumulative rainfall for 2014 and 2019. Why do these events come up so different in panel (b).
L309-310: But there is one point in Fig. 5a below 100 mm/day for the 2019 event.
What does ID stand for?
L314: This is not shown in Fig. 5.
L358: What is (Ci)? You do not use it at any other place, so I suggest to delete it. How are the ranks of the evaluated models reported? I don’t see it.
L429: First use of “test dataset”. Can you define this earlier?
L442: “Interesting…” I do not understand this sentence. Suggest rephrasing.
L448: I guess you mean that the FNR and FPR costs are assumed to be equal.
L459: This sentence about costs should come much earlier.
Figures:
Fig. 1: Coordinate grid is missing
Fig. 2. The last two classes in the cum. rainfall scale are incorrectly labeled.
Fig 4.: What are the numbers in parentheses? For lithology, there is the yellow bar for the 2014-2019 comparison missing.
Fig.5: see comment above. Why are there four correlation coefficients given but there are only 3 different data sets, i.e. the last (black) one is confusing.
Fig. 8: The scale bar given here seems to have the same length wrt the maps as the one shown in Fig. 2, yet here it’s 24 km whereas in Fig. 2 it is 32 km long.
Citation: https://doi.org/10.5194/nhess-2024-140-RC1 - AC1: 'Reply on RC1', Micol Fumagalli, 20 Nov 2024
-
RC2: 'Comment on nhess-2024-140', Anonymous Referee #2, 07 Sep 2024
Review of the manuscript #NHESS-2024-140 entitled "Shaping shallow landslide susceptibility as a function of rainfall events" by Micol Fumagalli and colleagues.
The manuscript entitled "Shaping shallow landslide susceptibility as a function of rainfall events" by Micol Fumagalli et al. presents an interesting rainfall-based shallow landslide susceptibility analysis, in Orba basin at Piedmont Region, Italy, using logistic regression. Five extreme rainfall events (3 with landslides, and 2 without) have been considered for both training and validation purposes. Three susceptibility scenarios have been performed for each of the five rainfall events.
This work consists of a good exercise for slope instability analysis, considering the modeling and validation processes, being also useful for civil protection warning and mitigation purposes.This is an original and very good work. Nevertheless, there are few minor issues that should be considered.
Thus, I suggest accept with minor revisions.
Next I put some issues:
L. 158-159: Please, revise and correct the legend of cumulative rainfall, in Fig. 2.
L. 167: Frattini and Crosta (2013), instead of "(Frattini and Crosta, 2013).
L. 196-201: Which was the interpolation method and which is the resolution of the interpolated raster retrieved from rainfall data?
L. 203: when you refer to study areas, do you mean the areas where the meteorological stations are, or do you refer to the different lithological subdivisions?
L. 358: Please explain the meaning of "Ci".
L. 388: Could you write in parentheses the meaning of VL (very low), L (low), M (...), H, VH in Table 2, like you did for HST?
Citation: https://doi.org/10.5194/nhess-2024-140-RC2 - AC2: 'Reply on RC2', Micol Fumagalli, 20 Nov 2024
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 229 | 66 | 63 | 358 | 24 | 6 | 6 |
- HTML: 229
- PDF: 66
- XML: 63
- Total: 358
- Supplement: 24
- BibTeX: 6
- EndNote: 6
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1