Author(s): P. H. A. J. M. Van Gelder

Linked Author(s):

Keywords: L-moments; Inhomogeneity; Sea dikes; Extreme events; Quantiles

Abstract: Hosking and Wallis (1997) have introduced a regional statistical method (called Lmoments) for probability distribution functions in a multivariate setting (when data is given about a particular quantity at more than one site). Generally, L-moments are linear combinations of ordered observations, which are unbiased regardless of the parent population. In this paper the L-moments technique will be analyzed for univariate datasets. With help of Monte Carlo simulations, results will be presented that the L-moments technique perform better than the ML-and LS-techniques, even if the data is contaminated or the model assumptions (such as homogeneity) are slightly violated. A Case Study is presented on a database of extreme water levels at various locations along the Dutch coast. The Netherlands is a low lying country which has to protect itself by sea-and river dikes (Van Gelder et. al., 1995). An accurate determination of the flood quantiles is extremely important in the design of the dikes. In this paper, the L-moment technique will be applied to the database in order to determine the quantiles. If the model assumption of homogeneity of the database is violated the L-moment technique yields better results than the traditional estimation techniques.

DOI:

Year: 1999