![Hier klicken, um den Treffer aus der Auswahl zu entfernen](images/unchecked.gif) |
Titel |
Uni- and multi-variable modelling of flood losses: experiences gained from
the Secchia river inundation event. |
VerfasserIn |
Francesca Carisi, Alessio Domeneghetti, Heidi Kreibich, Kai Schröter, Attilio Castellarin |
Konferenz |
EGU General Assembly 2017
|
Medientyp |
Artikel
|
Sprache |
en
|
Digitales Dokument |
PDF |
Erschienen |
In: GRA - Volume 19 (2017) |
Datensatznummer |
250150425
|
Publikation (Nr.) |
EGU/EGU2017-14887.pdf |
|
|
|
Zusammenfassung |
Flood risk is function of flood hazard and vulnerability, therefore its accurate assessment
depends on a reliable quantification of both factors. The scientific literature proposes a
number of objective and reliable methods for assessing flood hazard, yet it highlights a
limited understanding of the fundamental damage processes. Loss modelling is associated
with large uncertainty which is, among other factors, due to a lack of standard procedures; for
instance, flood losses are often estimated based on damage models derived in completely
different contexts (i.e. different countries or geographical regions) without checking its
applicability, or by considering only one explanatory variable (i.e. typically water depth). We
consider the Secchia river flood event of January 2014, when a sudden levee-breach caused
the inundation of nearly 200 km2 in Northern Italy. In the aftermath of this event, local
authorities collected flood loss data, together with additional information on affected private
households and industrial activities (e.g. buildings surface and economic value, number of
company’s employees and others). Based on these data we implemented and compared a
quadratic-regression damage function, with water depth as the only explanatory
variable, and a multi-variable model that combines multiple regression trees and
considers several explanatory variables (i.e. bagging decision trees). Our results show
the importance of data collection revealing that (1) a simple quadratic regression
damage function based on empirical data from the study area can be significantly
more accurate than literature damage-models derived for a different context and
(2) multi-variable modelling may outperform the uni-variable approach, yet it is
more difficult to develop and apply due to a much higher demand of detailed data. |
|
|
|
|
|