 |
| Titel |
Statistical Modelling of Compound Floods |
| VerfasserIn |
Emanuele Bevacqua, Douglas Maraun, Mathieu Vrac, Martin Widmann, Colin Manning |
| Konferenz |
EGU General Assembly 2016
|
| Medientyp |
Artikel
|
| Sprache |
en
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 18 (2016) |
| Datensatznummer |
250132945
|
| Publikation (Nr.) |
 EGU/EGU2016-13500.pdf |
|
|
|
|
|
| Zusammenfassung |
| In the recent special report of the Intergovernmental Panel on Climate Change (IPCC) on
extreme events it has been highlighted that an important class of extreme events has received
little attention so far: so-called compound events (CEs) (Seneviratne et al., 2012).
Compound events (CEs) are multivariate extreme events in which the individual
contributing events might not be extreme themselves, but their joint occurrence
causes an extreme impact. Following Leonard et al., 2013, we define events as CEs
only when the contributing events are statistically dependent. For many events
analysed so far, the contributing events have not been statistically dependent (e.g. the
floods in Rotterdam, Van den Brink et al., 2005). Two typical examples of CEs are
severe drought in conjunction with a heatwave, and storm surges coinciding with
heavy rain that cause the so-called Compound Floods in the lower section of a river.
We develop a multivariate statistical model to represent and analyse the physical
mechanisms driving CEs, and to quantify the risk associated with these events. The model
is based on pair-copula construction theory, which has the advantage of building
joint probability distributions modeling the marginal distributions separately from
the dependence structure among variables. This allows to analyse the individual
contributing variables underlying the CE separately to their dependence structure. Here is
presented an application of the statistical model for Compound Floods, based on a
conceptual case study. For these particular events it is not trivial to find satisfying
data. Usually, water level stations are not present in the area of the river where
both the influence of the sea and river are seen. The main reason being that this
critical area is small and stakeholders have little interest in measuring both effect
from the sea and from the river. For these reasons we have developed a conceptual
case study which allows us to vary the system’s physical parameters of interest.
This is based on real data for River discharge (Y RIV ER′) and Sea level (Y SEA),
from the River Têt in south of France. The impact of the compound flood is the
water level in the area between the River and Sea station, which we define here as
h = αY RIV ER + (1 − α)Y SEA. Here we show the sensitivity of the system to
a changes in the two physical parameters. Through variations in α we can study
the system in one or two dimensions which allows for the assessment of the risk
associated with either of the two variables alone or with a combination of them. Varying
instead the second parameter, i.e. the dependence among the variables Y RIV ER
and Y SEA, we show how an apparently weak dependence can increase the risk
of flooding significantly with respect to the independent case. The model can be
applied to future climate inserting predictors into the statistical model as additional
conditioning variables. Through conditioning the simulation of the statistical model
on the predictors obtained for future projections from Climate Models, both the
change of the risk and characteristics of compound floods for the future can be
analysed. |
|
|
|
|
|
|