![Hier klicken, um den Treffer aus der Auswahl zu entfernen](images/unchecked.gif) |
Titel |
A new function for estimating local rainfall thresholds for landslide triggering |
VerfasserIn |
J. Cepeda, F. Nadim, K. Høeg, A. Elverhøi |
Konferenz |
EGU General Assembly 2009
|
Medientyp |
Artikel
|
Sprache |
Englisch
|
Digitales Dokument |
PDF |
Erschienen |
In: GRA - Volume 11 (2009) |
Datensatznummer |
250030091
|
|
|
|
Zusammenfassung |
The widely used power law for establishing rainfall thresholds for triggering of
landslides was first proposed by N. Caine in 1980. The most updated global thresholds
presented by F. Guzzetti and co-workers in 2008 were derived using Caine’s power
law and a rigorous and comprehensive collection of global data. Caine’s function
is defined as I = α-
Dβ, where I and D are the mean intensity and total duration
of rainfall, and α and β are parameters estimated for a lower boundary curve to
most or all the positive observations (i.e., landslide triggering rainfall events). This
function does not account for the effect of antecedent precipitation as a conditioning
factor for slope instability, an approach that may be adequate for global or regional
thresholds that include landslides in surface geologies with a wide range of subsurface
drainage conditions and pore-pressure responses to sustained rainfall. However, in a
local scale and in geological settings dominated by a narrow range of drainage
conditions and behaviours of pore-pressure response, the inclusion of antecedent
precipitation in the definition of thresholds becomes necessary in order to ensure their
optimum performance, especially when used as part of early warning systems (i.e.,
false alarms and missed events must be kept to a minimum). Some authors have
incorporated the effect of antecedent rainfall in a discrete manner by first comparing the
accumulated precipitation during a specified number of days against a reference
value and then using a Caine’s function threshold only when that reference value is
exceeded. The approach in other authors has been to calculate threshold values as linear
combinations of several triggering and antecedent parameters. The present study is aimed to
proposing a new threshold function based on a generalisation of Caine’s power
law.
The proposed function has the form I = (α1-
Anα2)-
Dβ, where I and D are defined as
previously. The expression in parentheses is equivalent to Caine’s α parameter. α1, α2 and β
are parameters estimated for the threshold. An is the n-days cumulative rainfall. The
suggested procedure to estimate the threshold is as follows:
(1) Given N storms, assign one of the following flags to each storm: nL (non-triggering
storms), yL (triggering storms), uL (uncertain-triggering storms). Successful predictions
correspond to nL and yL storms occurring below and above the threshold, respectively.
Storms flagged as uL are actually assigned either an nL or yL flag using a randomization
procedure.
(2) Establish a set of values of ni (e.g. 1, 4, 7, 10, 15 days, etc.) to test for accumulated
precipitation.
(3) For each storm and each ni value, obtain the antecedent accumulated precipitation in
ni days Ani.
(4) Generate a 3D grid of values of α1, α2 and β.
(5) For a certain value of ni, generate confusion matrices for the N storms at each grid
point and estimate an evaluation metrics parameter EMP (e.g., accuracy, specificity,
etc.).
(6) Repeat the previous step for all the set of ni values.
(7) From the 3D grid corresponding to each ni value, search for the optimum grid point
EMPopti(global minimum or maximum parameter).
(8) Search for the optimum value of ni in the space ni vs EMPopti .
(9) The threshold is defined by the value of ni obtained in the previous step and the
corresponding values of α1, α2 and β.
The procedure is illustrated using rainfall data and landslide observations from the San
Salvador volcano, where a rainfall-triggered debris flow destroyed a neighbourhood in the
capital city of El Salvador in 19 September, 1982, killing not less than 300 people. |
|
|
|
|
|