Selected problems of at-site flood frequency analysis
Conference Paper
Overview
Identity
Additional Document Info
View All
Overview
abstract
The primary objective of flood frequency analysis (FFA) is the estimation of the upper quantiles of a probability distribution. The unknown underlying distribution function, small samples and uncertain data that are below the range of interest, make the task of achieving the objective difficult or even ill-posed. This paper discusses cognitive limitations of FFA and presents the findings of investigations on: (a) model discrimination, (b) the use of parsimonious models, and (c) the robustness of upper quantiles to largest elements of a sample. Moreover, bounded-at-zero distributions with two shape parameters for the flood variable are obtained using power transformation; then the probability density function and cumulative probability distribution for the variable are introduced. Based on the gamma distribution, two- and three-parameter heavy-tailed distributions are developed. Estimation techniques based both on the generalized probability weighted moments and the mean deviation are recommended for further investigation. Copyright 2007 IAHS Press.