Predicting the thickness of shallow landslides in Switzerland using machine learning
<p>Landslide thickness is a key variable in various types of landslide susceptibility models. In this study, we developed a model providing improved predictions of potential shallow-landslide thickness for Switzerland. We tested three machine learning (ML) models based on random forest (RF) mo...
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Copernicus Publications
2025-02-01
|
| Series: | Natural Hazards and Earth System Sciences |
| Online Access: | https://nhess.copernicus.org/articles/25/467/2025/nhess-25-467-2025.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1832098353768300544 |
|---|---|
| author | C. Schaller C. Schaller L. Dorren M. Schwarz C. Moos A. C. Seijmonsbergen E. E. van Loon |
| author_facet | C. Schaller C. Schaller L. Dorren M. Schwarz C. Moos A. C. Seijmonsbergen E. E. van Loon |
| author_sort | C. Schaller |
| collection | DOAJ |
| description | <p>Landslide thickness is a key variable in various types of landslide susceptibility models. In this study, we developed a model providing improved predictions of potential shallow-landslide thickness for Switzerland. We tested three machine learning (ML) models based on random forest (RF) models, generalised additive models (GAMs), and linear regression models (LMs). Next, we compared the results to three simple models that link soil thickness to slope gradient (Simple-S/linear interpolation and SFM/log-normal distribution) and elevation (Simple-Z/linear interpolation). The models were calibrated using data from two field inventories in Switzerland (HMDB with 709 records and KtBE with 515 records). We explored 39 different covariates, including metrics on terrain, geomorphology, vegetation, and lithology, at three different cell sizes. To train the ML models, 21 variables were chosen based on the variable importance derived from RF models and expert judgement. Our results show that the ML models consistently outperformed the simple models by reducing the mean absolute error by at least 20 %. The RF models produced a mean absolute error of 0.25 m for the HMDB and 0.20 m for the KtBE data. Models based on ML substantially improve the prediction of landslide thickness, offering refined input for enhancing the performance of slope stability simulations.</p> |
| format | Article |
| id | doaj-art-c064e68bce3447dea78e2490439e6f77 |
| institution | Kabale University |
| issn | 1561-8633 1684-9981 |
| language | English |
| publishDate | 2025-02-01 |
| publisher | Copernicus Publications |
| record_format | Article |
| series | Natural Hazards and Earth System Sciences |
| spelling | doaj-art-c064e68bce3447dea78e2490439e6f772025-02-05T11:23:26ZengCopernicus PublicationsNatural Hazards and Earth System Sciences1561-86331684-99812025-02-012546749110.5194/nhess-25-467-2025Predicting the thickness of shallow landslides in Switzerland using machine learningC. Schaller0C. Schaller1L. Dorren2M. Schwarz3C. Moos4A. C. Seijmonsbergen5E. E. van Loon6Bern University of Applied Sciences – HAFL, Länggasse 85, 3052 Zollikofen, SwitzerlandUniversity of Amsterdam UVA – IBED, Science Park 904, 1098 XH Amsterdam, the NetherlandsBern University of Applied Sciences – HAFL, Länggasse 85, 3052 Zollikofen, SwitzerlandBern University of Applied Sciences – HAFL, Länggasse 85, 3052 Zollikofen, SwitzerlandBern University of Applied Sciences – HAFL, Länggasse 85, 3052 Zollikofen, SwitzerlandUniversity of Amsterdam UVA – IBED, Science Park 904, 1098 XH Amsterdam, the NetherlandsUniversity of Amsterdam UVA – IBED, Science Park 904, 1098 XH Amsterdam, the Netherlands<p>Landslide thickness is a key variable in various types of landslide susceptibility models. In this study, we developed a model providing improved predictions of potential shallow-landslide thickness for Switzerland. We tested three machine learning (ML) models based on random forest (RF) models, generalised additive models (GAMs), and linear regression models (LMs). Next, we compared the results to three simple models that link soil thickness to slope gradient (Simple-S/linear interpolation and SFM/log-normal distribution) and elevation (Simple-Z/linear interpolation). The models were calibrated using data from two field inventories in Switzerland (HMDB with 709 records and KtBE with 515 records). We explored 39 different covariates, including metrics on terrain, geomorphology, vegetation, and lithology, at three different cell sizes. To train the ML models, 21 variables were chosen based on the variable importance derived from RF models and expert judgement. Our results show that the ML models consistently outperformed the simple models by reducing the mean absolute error by at least 20 %. The RF models produced a mean absolute error of 0.25 m for the HMDB and 0.20 m for the KtBE data. Models based on ML substantially improve the prediction of landslide thickness, offering refined input for enhancing the performance of slope stability simulations.</p>https://nhess.copernicus.org/articles/25/467/2025/nhess-25-467-2025.pdf |
| spellingShingle | C. Schaller C. Schaller L. Dorren M. Schwarz C. Moos A. C. Seijmonsbergen E. E. van Loon Predicting the thickness of shallow landslides in Switzerland using machine learning Natural Hazards and Earth System Sciences |
| title | Predicting the thickness of shallow landslides in Switzerland using machine learning |
| title_full | Predicting the thickness of shallow landslides in Switzerland using machine learning |
| title_fullStr | Predicting the thickness of shallow landslides in Switzerland using machine learning |
| title_full_unstemmed | Predicting the thickness of shallow landslides in Switzerland using machine learning |
| title_short | Predicting the thickness of shallow landslides in Switzerland using machine learning |
| title_sort | predicting the thickness of shallow landslides in switzerland using machine learning |
| url | https://nhess.copernicus.org/articles/25/467/2025/nhess-25-467-2025.pdf |
| work_keys_str_mv | AT cschaller predictingthethicknessofshallowlandslidesinswitzerlandusingmachinelearning AT cschaller predictingthethicknessofshallowlandslidesinswitzerlandusingmachinelearning AT ldorren predictingthethicknessofshallowlandslidesinswitzerlandusingmachinelearning AT mschwarz predictingthethicknessofshallowlandslidesinswitzerlandusingmachinelearning AT cmoos predictingthethicknessofshallowlandslidesinswitzerlandusingmachinelearning AT acseijmonsbergen predictingthethicknessofshallowlandslidesinswitzerlandusingmachinelearning AT eevanloon predictingthethicknessofshallowlandslidesinswitzerlandusingmachinelearning |