Machine learning-based debris flow susceptibility and risk index mapping in Quang Nam Province, Vietnam
Acta Geodynamica et Geomaterialia · 29.06.2026 · 23 (2) · 301
DOI: 10.13168/agg.2026.0020 ↗Kurz gefasst
Fachartikel in der Zeitschrift Acta Geodynamica et Geomaterialia, online erschienen am 29.06.2026. Verfasst von Nguyen-Vu Luat und 3 weiteren Autor:innen. Die Arbeit nennt 49 Literaturangaben.
Abstract (Originalsprache)
This study develops an integrated machine learning (ML) framework for debris flow susceptibility and risk index mapping across Quang Nam Province, a region prone to high-intensity monsoonal rainfall and strongly weathered metamorphic bedrock. Four ML algorithms including Gradient Boosting (GB), Random Forest (RF), Support Vector Machine (SVM), and Artificial Neural Network (ANN), were trained on 228 stream-channel samples described by eight causative factors, under two class-imbalance correction strategies. Model performance was assessed through stratified 10-fold cross-validation, and feature importance was quantified using Shapley Additive Explanations (SHAP). Gradient Boosting with cost-sensitive learning (GB-CS) achieved the best overall performance (AUC = 0.9902, F1 = 0.9286, only 4 misclassifications in 228 instances), outperforming the Modified Analytic Hierarchy Process (M-AHP) expert baseline by 0.249 AUC units (33.5 % relative improvement). SHAP analysis identified channel gradient as the dominant predictor (mean normalized importance = 0.442) across GB, RF, and SVM, followed by elevation (0.142) and geology (0.108). The GB-CS susceptibility map correctly classified 92.9 % of inventory sites within the High or Very High susceptibility classes, covering 20 % of the stream network. Integration of susceptibility with a land-use and elevation-weighted vulnerability index revealed that the highest-risk reaches occur not in the most hazardous western highlands but in transitional mid-elevation agricultural corridors - a finding with direct consequences for early-warning system prioritization.
Zitieren
APA
Luat, N., Nguyen, L. C., Do, T., & Kien, N. T. (2026). Machine learning-based debris flow susceptibility and risk index mapping in Quang Nam Province, Vietnam. Acta Geodynamica et Geomaterialia, 23(2), 301. https://doi.org/10.13168/agg.2026.0020
Vancouver
Luat N, Nguyen LC, Do T, Kien NT. Machine learning-based debris flow susceptibility and risk index mapping in Quang Nam Province, Vietnam. Acta Geodynamica et Geomaterialia. 2026;23(2):301. doi:10.13168/agg.2026.0020
BibTeX
@article{Luat2026,
author = {Luat, Nguyen-Vu and Nguyen, Lan Chau and Do, Tuan-Nghia and Kien, Nguyen Trung},
title = {Machine learning-based debris flow susceptibility and risk index mapping in Quang Nam Province, Vietnam},
journal = {Acta Geodynamica et Geomaterialia},
year = {2026},
volume = {23},
number = {2},
pages = {301},
doi = {10.13168/agg.2026.0020}
}
EUVIMED