The efect of outliers and length of records on design flood magnitude

Abstract

Flood frequency analysis is of paramount importance for the design of hydraulic structures, bridges, flood mapping, and floodplain management. At the same time, it is a long and complicated procedure evaluating uncertainties originating from different sources, because of a variety of data employed. They are collected using different techniques and methods. Here we need digital terrain models (river bed bathymetry and floodplain elevation), must have a detailed hydrological analysis of existing data and finally, decide on methods to be used for flood mapping (hydrological and hydraulic models, spatial data analysis and interpretation). Design of hydraulic structures or production of flood risk maps are based on flood magnitude-frequency relationship at the site using observed data, estimating the magnitude of flood for recurrence period (e.g. 10, 100 or other) and converting it to flood extent. The length of time series and outliers can affect the results significantly and must be considered with the highest attention when performing flood frequency analysis. The term "outlier" is generally used to refer to single data points that appear to depart significantly from other data and may originate as from incorrect observations or appear as rare hydrologic evens. Including or excluding outliers may have a significant effect on design flood. The 60-year length of annual peak flow time series of 20 observation points was involved for data frequency analysis. Four scenarios of analysis performed: two different length of data series and series with and without outliers. The estimated effect of both factors is discussed