 |
| Titel |
How can we assess the processes that control recurrence intervals of landslide-generated turbidites? |
| VerfasserIn |
Michael Clare, Peter Talling, Peter Challenor |
| Konferenz |
EGU General Assembly 2013
|
| Medientyp |
Artikel
|
| Sprache |
Englisch
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 15 (2013) |
| Datensatznummer |
250073697
|
|
|
|
|
|
| Zusammenfassung |
| We analyse sedimentary deposits from three distal basin plain settings that are disparate in
terms of their geography, physiography and age. These include an outcrop study (N=696
turbidites) from the Miocene Marnoso Arenacea Formation in the Italian Apennines,
cores from the Balearic Abyssal Plain (N =151) and boreholes from the Madeira
Abyssal Plain (N=108). In each case, stacked sequences of turbidity current deposits
intercalated with subordinate hemipelagite beds are identified. On the basis of bed
geometries and inferred deposit volumes (0.7 to 500 km3), the turbidity currents are
interpreted to have been triggered by slope failures on the basin margins, rather
than in relation to hyperpycnal flood discharge. Inter-event times (i.e. recurrence
intervals) are determined from hemipelagic bed thicknesses and from a calibrated
age-stratigraphic framework. Erosion of hemipelagite beds beneath turbidites is non-existent
or negligible; hence these distal depositional sequences can be regarded as long
term catalogues of landslide activity that may also be used to understand tsunami
risk.
The distribution of inter-event timings is indicative of an exponential relationship (i.e.
Poisson process). This satisfies two conditions: 1) a lack of memory, and 2) a constant
probability of event occurrence through time. An exponential distribution can be
characterised by just one parameter, λ, which is defined as the rate parameter, or mean
inter-event time. To test this inferred distribution, each data set is normalized to λ and
exceedance plots are generated in relation to RT (a dimensionless expression of inter-event
time). Remarkably, the data align extremely well, despite their disparity in age, location and
setting. A fit with an exponential growth curve is good (R2=0.98); however, the longest
inter-event times (RT >3) are slightly underrepresented. A similar observation is made by
comparing suites of randomly generated synthetic data against the actual data. A Generalised
Linear Model is fitted to the data, using a Gamma distribution (of which the exponential is a
special case). This indicates a dispersion parameter, α, of between 1.2 and 1.8 is appropriate
for the data sets. This can be regarded as an exponential distribution and the small
deviation in shape parameter accounts for the slight overpopulation in the tail of the
data.
Statistical tests, including survival analysis, Cox’s proportional hazards model
and Hurst statistic, indicate that event timings do not show significant clustering
and, surprisingly, occur independent of variations in sea level. Despite this, event
magnitude does appear to be related to eustatic variations. It is suggested, that a
governing discrete stochastic process controls the timing of landslide-triggered
turbidity currents in distal basin plain settings, regardless of their locale or age.
This may be explained by exponential distributions that have been identified for
recurrence times of large magnitude earthquakes globally. A determination of event
frequency distribution and related dispersion parameter, can inform predictions of
future events or inference of larger data populations from small samples. This will
support statistically meaningful analyses for forward-looking geohazard assessment. |
|
|
|
|
|
|