Dynamic path-dependent landslide susceptibility modelling

Jalal Samia,Arnold Bregt,Fausto Guzzetti,Francesca Ardizzone,Arnaud Temme,Jakob Wallinga

doi:10.5194/nhess-20-271-2020

Abstract

Abstract. This contribution tests the added value of including landslide path dependency in statistically based landslide susceptibility modelling. A conventional pixel-based landslide susceptibility model was compared with a model that includes landslide path dependency and with a purely path-dependent landslide susceptibility model. To quantify path dependency among landslides, we used a space–time clustering (STC) measure derived from Ripley's space–time K function implemented on a point-based multi-temporal landslide inventory from the Collazzone study area in central Italy. We found that the values of STC obey an exponential-decay curve with a characteristic timescale of 17 years and characteristic spatial scale of 60 m. This exponential space–time decay of the effect of a previous landslide on landslide susceptibility was used as the landslide path-dependency component of susceptibility models. We found that the performance of the conventional landslide susceptibility model improved considerably when adding the effect of landslide path dependency. In fact, even the purely path-dependent landslide susceptibility model turned out to perform better than the conventional landslide susceptibility model. The conventional plus path-dependent and path-dependent landslide susceptibility model and their resulting maps are dynamic and change over time, unlike conventional landslide susceptibility maps.

Highlights

Landslide susceptibility modelling calculates the likelihood of landslide occurrence in a certain location (Brabb, 1985)
We found that the values of space–time clustering (STC) obey an exponentialdecay curve with a characteristic timescale of 17 years and characteristic spatial scale of 60 m
We compared the performance of these models using area-under-thecurve (AUC) values from the receiver operating characteristic (ROC; Mason and Graham, 2002) and selected the optimal model using the Akaike information criterion (AIC; Akaike, 1998), which penalizes the use of additional variables in a model

Summary

Introduction

Landslide susceptibility modelling calculates the likelihood of landslide occurrence in a certain location (Brabb, 1985). The resulting landslide susceptibility maps from landslide susceptibility models indicate where landslides are likely to occur (Guzzetti et al, 2005). These maps are useful in land use planning and insurance, among other functions. In this context, different methods and techniques have been used for landslide susceptibility modelling. The spatial distribution of historic landslides, documented in landslides inventories, is a crucial input for statistically based landslide susceptibility modelling (Guzzetti et al, 2012; Van Westen et al, 2008).

Objectives

Methods

Results

Discussion

Conclusion