Prediction of the spatial variability of rainfall- and human-induced landslides in the Bhagirathi Valley of the Indian NW Himalayas using machine learning techniques

Landslides are recurring and catastrophic natural hazards that cause tremendous loss in mountainous regions. Landslides in the Indian Himalayas are mainly due to rainfall, earthquakes, and anthropogenic activities such as deforestation, road widening, tunneling, and urbanization. Landslide susceptib...

Celý popis

Uložené v:

Podrobná bibliografia
Vydané v:	Environmental earth sciences Ročník 84; číslo 21; s. 598
Hlavní autori:	Gupta, Neha, Das, Josodhir, Kanungo, D. P.
Médium:	Journal Article
Jazyk:	English
Vydavateľské údaje:	Berlin/Heidelberg Springer Berlin Heidelberg 01.11.2025 Springer Nature B.V
Predmet:	Algorithms Anthropogenic factors Artificial neural networks Back propagation Biogeosciences Decision trees Deforestation Earth and Environmental Science Earth Sciences Earthquakes Environmental Science and Engineering Geochemistry Geological hazards Geology Hazards Human influences Hydrology/Water Resources Landslides Landslides & mudslides Learning algorithms Machine learning Mountain regions Mountainous areas Neural networks Original Article Precipitation Predictions Rainfall Regression analysis Risk assessment Roads & highways Seismic activity Seismic response Spatial variability Spatial variations Support vector machines Terrestrial Pollution Urbanization Valleys Widening Brazil Ecuador Himalaya Mountains Uttarakhand India China India Ensemble models Machine learning algorithms Statistical models Landslide susceptibility zonation mapping Bhagirathi valley
ISSN:	1866-6280, 1866-6299
On-line prístup:	Získať plný text
Tagy:	Pridať tag Žiadne tagy, Buďte prvý, kto otaguje tento záznam!

Popis
Shrnutí:	Landslides are recurring and catastrophic natural hazards that cause tremendous loss in mountainous regions. Landslides in the Indian Himalayas are mainly due to rainfall, earthquakes, and anthropogenic activities such as deforestation, road widening, tunneling, and urbanization. Landslide susceptibility mapping is a prerequisite for vulnerability and risk assessment. The present study aimed to evaluate the prediction of the landslide spatial variability for the Bhagirathi Valley in the Uttarakhand Himalayas, India, using machine learning and statistical algorithms, namely, Random Forest (RF), Extreme Gradient Boosting (XGBoost), K-Nearest Neighbour (KNN), Artificial Neural Network (ANN), Frequency Ratio (FR) and Certainty Factor (CF). A geospatial database with 3098 landslide polygons that are either rainfall-triggered or due to road widening, as well as thematic layers pertaining to 12 landslide causative factors, were prepared for this study. The earthquake-induced landslides, referring to the 20th October 1991 Uttarkashi earthquake with a moment magnitude of 6.8 and a maximum Mercalli intensity of IX, were excluded from the database. Hence, the present spatial variability prediction has been attempted using the rainfall- and anthropogenic-induced landslides database. In terms of model robustness, the area under the curve (AUC) reached its highest value of 0.91 for XGBoost. The AUC values for the other models were as follows: RF at 0.90, KNN at 0.89, ANN at 0.89, CF at 0.89, and FR at 0.88. The study demonstrated that ensemble tree-based models are more effective in predicting landslide spatial variability than other techniques. The results emphasized that proper model selection and balanced complexity can significantly enhance performance. The landslide susceptibility maps thus produced will be helpful for decision-makers and planners in assessing the vulnerability and risk of the Bhagirathi Valley. The Bhagirathi Valley is considered as one of the northwestern Himalayan valleys most affected by landslide hazards.
Bibliografia:	ObjectType-Article-1 SourceType-Scholarly Journals-1 ObjectType-Feature-2 content type line 14
ISSN:	1866-6280 1866-6299
DOI:	10.1007/s12665-025-12568-8