Robust Estimators in Hydrologic Frequency Analysis

by Fahim Ashkar, Univ of Moncton, Moncton, Canada,
Taha B. M. J. Ouarda, Univ of Moncton, Moncton, Canada,
Ren? Roy, Univ of Moncton, Moncton, Canada,
Bernard Bob?e, Univ of Moncton, Moncton, Canada,



Document Type: Proceeding Paper

Part of: Engineering Hydrology

Abstract:

This paper applies a specific robust estimation procedure known as 'trimming' in flood frequency analysis. The trimming technique consists in censoring the series of recorded floods by excluding the most extreme values. Using Monte Carlo simulation, the effect of various proportions of symmetric trimming on the estimation of moments, distribution parameters, and quantiles, is examined. The influence of the sample size and the parent distribution parameters on the estimation performance is also investigated for the log-Pearson Type 3 distribution (LP3D). The classical method of moments, the 'sundry averages method' (SAM) and the method of mixed moments are selected for the fitting of the LP3D. The utility of robust techniques in hydrologic frequency analysis is demonstrated, and recommendations are made based on the root mean square error of parameter and quantile estimates.

Subject Headings: Flood frequency | Parameters (statistics) | Hydrology | Frequency analysis | Frequency distribution | Monte Carlo method | Probability distribution

Services: Buy this book/Buy this article

 

Return to search