Back to overview

Origin of the power-law exponent in the landslide frequency-size distribution

Type of publication Peer-reviewed
Publikationsform Original article (peer-reviewed)
Author Jafarimanesh Ahoura, Mignan Arnaud, Danciu Laurentiu,
Project Risk Governance of Deep Geothermal and Hydro Energy
Show all

Original article (peer-reviewed)

Journal Natural Hazards and Earth System Sciences Discussions
Page(s) 1 - 28
Title of proceedings Natural Hazards and Earth System Sciences Discussions
DOI 10.5194/nhess-2018-167

Open Access

Type of Open Access Publisher (Gold Open Access)


Abstract. Landslide statistics is characterized by a power-law frequency-size distribution (FSD) with power exponent α centered on 2.2–2.4, independently of the landslide trigger. So far, the origin of the α-value, critical to probabilistic hazard assessment, remains hypothetical. We present a generic landslide cellular automaton (LSgCA) based on the rules of Self Organized Criticality and on the Factor of Safety (FS) concept. We show that it reproduces the power-law FSD for realistic parameter ranges (i.e. cohesion, soil unit weight, soil thickness, angle of friction, slope angle, pore water pressure) with LSgCA simulations yielding α = 2.17±0.49, which is in agreement with α = 2.21±0.53 obtained from an updated meta-analysis of the landslide literature. The parameter α remains stable despite changes in the landslide triggering process, with the trigger only influencing the spatial extent of the landslide initiation phase defined from an FS contour. Furthermore, different FS formulations do not significantly alter the results. We find that α is constrained during the initiation phase of the landslide by the fractal properties of the topography, as we observed a positive correlation between fractal dimension and α while α did not change during the propagation phase of the LSgCA. Our results thus suggest that α can be estimated directly from the FS map for probabilistic landslide hazard assessment. However full modeling (including the propagation phase) would be required to combine the spatial distributions of landslide and exposure in probabilistic risk analysis.