A data-based comparison of flood frequency analysis methods used in France

• DOI: 10.5194/nhess-14-295-2014
• Publication Date: 2014-02-20T13:06:18Z
• Authors: K. Kochanek; B. Renard; P. Arnaud; Y. Aubert; M. Lang; T. Cipriani; E. Sauquet
• Keywords: Gumbel distribution; Quantile; Implementation; Continuous simulation

Abstract. Flood frequency analysis (FFA) aims at estimating quantiles with large return periods for an extreme discharge variable. Many FFA implementations are used in operational practice France. These range from the estimation of a pre-specified distribution to continuous simulation approaches using rainfall simulator coupled rainfall–runoff model. This diversity raises questions regarding limits each implementation and calls nation-wide comparison their predictive performances. paper presents results national main More accurately, eight considered, corresponding local, regional local-regional Gumbel Generalized Extreme Value (GEV) distributions, as well local versions approach. A data-based framework is applied these competitors evaluate performances terms reliability stability, daily flow data more than 1000 gauging stations Results this comparative exercise suggest that two dominate performances, namely version approach GEV distribution. specific conclusions include following: (i) not suitable Mediterranean catchments, since demonstrably leads underestimation flood quantiles; (ii) recommended, because difficulty shape parameter frequent failures; (iii) all purely evaluated study displayed quite poor reliability, suggesting prediction completely ungauged catchments remains challenge.