Seismic–Tectonic Predictors of Landslide Susceptibility in the Western Himalayas: A Spatially Validated Machine Learning Assessment
Rock Mechanics Letters · 21.01.2027 · 4 (1) · 442-450
DOI: 10.70425/rml.202701.54 ↗Kurz gefasst
Fachartikel in der Zeitschrift Rock Mechanics Letters, online erschienen am 12.01.2027. Verfasst von Mir Mujtaba Laghari und 2 weiteren Autor:innen. Die Arbeit nennt 23 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Seismicity may contribute to regional slope instability in the Western Himalayas, but its incremental predictive value relative to a matched non-seismic baseline remains poorly quantified. This study assessed the regional contribution of seismic–tectonic predictors using 763 geocoded NASA Global Landslide Catalog reports from 2007–2017, 345 earthquakes of magnitude 5.5–7.6 from 1965–2016, mapped active faults, and hydrographic variables. Landslides were aggregated to a 0.25° grid, yielding 231 presence cells among 1,600 cells. Spatial block-permutation tests, nested five-fold spatial cross-validation, four classifiers, matched baseline comparisons, and SHapley Additive exPlanations (SHAP) were applied. Landslide-presence cells were closer to catalogued earthquakes than background cells (61.5 vs. 118.3 km; P = 0.0005) and had a higher energy-weighted seismicity index (2.323 vs. 2.003; P = 0.0015). Random forest achieved the strongest mean discrimination and calibration (ROC–AUC = 0.855 ± 0.027; PR–AUC = 0.489 ± 0.073; Brier score = 0.096 ± 0.004). Adding seismic–tectonic predictors increased random-forest ROC–AUC by 0.048 and reduced the Brier score by 0.014, although paired tests across five folds did not reach P < 0.05. The high and very high susceptibility classes occupied 16.5% of grid cells and contained 68.2% of catalogued events. Earthquake proximity, Gaussian earthquake density, and energy-weighted seismicity were the leading physical seismic predictors in SHAP analysis. These findings indicate useful regional seismic–tectonic predictive information but do not establish earthquake triggering or causation. Interpretation is limited by catalogue reporting bias, coarse mapping units, mixed magnitude scales, and unavailable terrain variables.
Zitieren
APA
Laghari, M. M., Saeed, K., & Cheng, N. (2027). Seismic–Tectonic Predictors of Landslide Susceptibility in the Western Himalayas: A Spatially Validated Machine Learning Assessment. Rock Mechanics Letters, 4(1), 442-450. https://doi.org/10.70425/rml.202701.54
Vancouver
Laghari MM, Saeed K, Cheng N. Seismic–Tectonic Predictors of Landslide Susceptibility in the Western Himalayas: A Spatially Validated Machine Learning Assessment. Rock Mechanics Letters. 2027;4(1):442-450. doi:10.70425/rml.202701.54
BibTeX
@article{Laghari2027,
author = {Laghari, Mir Mujtaba and Saeed, Khuzaima and Cheng, Nirandoal},
title = {Seismic–Tectonic Predictors of Landslide Susceptibility in the Western Himalayas: A Spatially Validated Machine Learning Assessment},
journal = {Rock Mechanics Letters},
year = {2027},
volume = {4},
number = {1},
pages = {442--450},
doi = {10.70425/rml.202701.54}
}
EUVIMED