On the Choice Between Annual Flood Series and Peaks Over Threshold Series in Flood Frequency Analysis

by Fahim Ashkar, Inst Natl de la Recherche Scientifique, Ste. -Foy, Quebec, Canada,
Nassir El-Jabi, Inst Natl de la Recherche Scientifique, Ste. -Foy, Quebec, Canada,
Bernard Bobee, Inst Natl de la Recherche Scientifique, Ste. -Foy, Quebec, Canada,



Document Type: Proceeding Paper

Part of: Engineering Hydrology

Abstract: Statistical techniques are often used to estimate extreme flood events for the design of a flood-control structure or for any other flood-related purpose. Two kinds of models are commonly used in this estimation: Annual maximum (AM) series and peaks over threshold (POT) series. It is not yet clearly known which of these two approaches is more efficient from the statistical point of view. A major drawback of past attempts to compare the statistical efficiency of these two approaches was that these were totally or partially based on large-sample theory. Previous studies have compared variances, and have concluded that POT is more efficient than AM only if the average number of exceedances per year in models of the latter is higher than about 1. 65. This paper shows that this conclusion is not valid for samples of small size, such as those frequently found in hydrology. Results based on small samples are shown to depart considerably from large-sample results.

Subject Headings: Flood frequency | Statistics | Frequency analysis | Model analysis | Hydraulic design | Comparative studies | Hydrology

Services: Buy this book/Buy this article

 

Return to search