 |
| Titel |
Derivation of landslide-triggering thresholds by Monte Carlo simulation and ROC analysis |
| VerfasserIn |
David Johnny Peres, Antonino Cancelliere |
| Konferenz |
EGU General Assembly 2015
|
| Medientyp |
Artikel
|
| Sprache |
Englisch
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 17 (2015) |
| Datensatznummer |
250110778
|
| Publikation (Nr.) |
 EGU/EGU2015-10815.pdf |
|
|
|
|
|
| Zusammenfassung |
| Rainfall thresholds of landslide-triggering are useful in early warning systems to be
implemented in prone areas. Direct statistical analysis of historical records of rainfall and
landslide data presents different shortcomings typically due to incompleteness of landslide
historical archives, imprecise knowledge of the triggering instants, unavailability of a rain
gauge located near the landslides, etc.
In this work, a Monte Carlo approach to derive and evaluate landslide triggering
thresholds is presented. Such an approach contributes to overcome some of the above
mentioned shortcomings of direct empirical analysis of observed data. The proposed Monte
Carlo framework consists in the combination of a rainfall stochastic model with
hydrological and slope-stability model. Specifically, 1000-years long hourly synthetic
rainfall and related slope stability factor of safety data are generated by coupling the
Neyman-Scott rectangular pulses model with the TRIGRS unsaturated model (Baum
et al., 2008) and a linear-reservoir water table recession model. Triggering and
non-triggering rainfall events are then distinguished and analyzed to derive stochastic-input
physically based thresholds that optimize the trade-off between correct and wrong
predictions. For this purpose, receiver operating characteristic (ROC) indices are used.
An application of the method to the highly landslide-prone area of the Peloritani
mountains in north-eastern Sicily (Italy) is carried out. A threshold for the area is
derived and successfully validated by comparison with thresholds proposed by other
researchers.
Moreover, the uncertainty in threshold derivation due to variability of rainfall intensity
within events and to antecedent rainfall is investigated. Results indicate that variability of
intensity during rainfall events influences significantly rainfall intensity and duration
associated with landslide triggering. A representation of rainfall as constant–intensity
hyetographs globally leads to non-conservative thresholds. When a time-variable rainfall-rate
event is considered, the simulated triggering points may be separated with a very good
approximation from the non-triggering ones by an I–D power-law equation. Such
an equation thus turns out to be adequate to represent the triggering part due to
transient infiltration produced by rainfall events of variable intensity. This provides a
physically based justification for the I-D power-law threshold form, which is valid when
landslide occurrence is mostly due to that part, i.e. in hillslopes of low specific upslope
contributing area, with soils of relatively high hydraulic conductivity and high critical
wetness ratio. Otherwise, rainfall time history occurring before single rainfall events
influences landslide triggering, determining the need to take into account antecedent
rainfall. |
|
|
|
|
|
|