 |
| Titel |
Probabilistic classification of local rainfall-thresholds for landslide triggering |
| VerfasserIn |
Zenon Medina-Cetina, Jose Cepeda |
| Konferenz |
EGU General Assembly 2010
|
| Medientyp |
Artikel
|
| Sprache |
Englisch
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 12 (2010) |
| Datensatznummer |
250044237
|
|
|
|
|
|
| Zusammenfassung |
| A procedure is formulated for the probabilistic classification of an empirical threshold model
for rainfall-triggered landslides. The threshold model t(Î) with a set of parameters Î,
integrates rainfall intensity I, duration D and n-day antecedent precipitation An. The
probabilistic classification based on this model is the solution of an inverse problem under a
Bayesian theoretical framework. This approach allows for identifying, characterizing and
propagating the uncertainty encapsulated on the evidence use for the calibration of the
classification model. The use of the Bayesian paradigm results in the integration of the
joint probability distribution of the model parameters. From the parameters’ joint
probability distribution, it is then possible to assess the correlation between the
model parameters and to populate likely realizations of threshold curves, so that a
smooth probability threshold definition can be obtained, as opposed to deterministic
optimal threshold curves. A case study is discussed where observations are provided
for a dataset of I - D - An records considering both landslide-triggering and
non-triggering rainfall events. A fundamental application of this exercise, is the
potential probabilistic definition of alert levels in the context of early-warning systems. |
|
|
|
|
|
|