Landslide size matters: A new data-driven, spatial prototype

• DOI: 10.1016/j.enggeo.2021.106288
• Publication Date: 2021-07-24T06:28:34Z
• Authors: Luigi Lombardo; Hakan Tanyas; Raphaël Huser; Fausto Guzzetti; Daniela Castro-Ca...
• Keywords: Earthquake; 550; UT-Hybrid-D; 15. Life on land; 01 natural sciences; Integrated nested Laplace approximation (INLA); ITC-HYBRID; Landslide area prediction; ITC-ISI-JOURNAL-ARTICLE; Slope unit; Landslide hazard; Bayesian spatial modelling; 0105 earth and related environmental sciences

The standard definition of landslide hazard requires the estimation where, when (or how frequently) and large a given event may be. geoscientific community involved in statistical models has addressed component pertaining to be by introducing concept landslide-event magnitude scale. This scale, which depends on planimetric area population landslides, analogy earthquake magnitude, been expressed with single value per event. As result, geographic or spatially-distributed considered at slope disregarded statistically-based studies. Conversely, extent commonly part physically-based applications, though their implementation is often limited very small regions. In this work, we initially present review methods developed for assessment since its first conception decades ago. Subsequently, introduce time model able estimate landslides aggregated units. More specifically, implemented Bayesian version Generalized Additive Model where maximum size unit sum all sizes are predicted via Log-Gaussian model. These "max" "sum" capture spatial distribution (aggregated) sizes. We tested these global dataset expressing co-seismic due 24 earthquakes across globe. two both evaluated suite performance diagnostics that suggest our suitably predict unit. addition complex procedure involving variable selection uncertainty estimation, built over slopes triggered response seismic shaking, simulated expected failing surface did not occur past. What achieved literature provide information about failed landscape. vital studies should combined occurrence locations. could ensure governmental territorial agencies have complete probabilistic overview behave specific trigger. predictive currently valid only 25 cases tested. Statistically estimating extents still infancy stage. Many more applications successfully validated before considering such an operational way. For instance, validity verified regional catchment as much it needs different types triggers. However, envision new paradigm breakthrough and, time, even become official risk protocols.