Regional flood frequency analysis in arid and semi-arid areas

Abstract

Dimensionless regional flood frequency curves use a number of records from a reasonably homogeneous area to overcome the variability of site records. The problems posed by short individual records and the variability of estimates are particularly acute in arid areas. Regional flood frequency curves in a number of semi-arid and arid areas have been assembled to illustrate their extreme slope and skewness, and the mutual similarity of these curves. Annual maxima from 162 stations with annual basin rainfall below 600 mm are assembled from NW Africa, Iran, Jordan, Saudi Arabia, Botswana and South Africa, and compared with records from Australia, southwest USA and Russia. The general extreme value (GEV) curves from each region, fitted by probability weighted moments (PWM), are compared with the combined curve. Since floods in arid areas, if nowhere else, result from storm intensity exceeding soil infiltration, the similarity of regional flood curves may indicate similar storm characteristics in different regions. The mean annual flood (MAF) for each catchment has been related to both basin area and mean annual rainfall by regression in an attempt to provide a preliminary means of estimating the scaling factor, MAF, for catchments with no flow data. Whilst in many of the regions studied, insufficient data are currently available to permit reliable estimation of the MAF by such an approach, it is hoped that the regression equations presented may permit estimation of the MAF for Iran, Saudi Arabia and Yemen, Queensland and more tentatively South Africa and Botswana. In addition, a combined regression equation for all semi-arid and arid regions is presented. This combination of estimated MAF and regional flood frequency curves permits flood magnitude with a known risk of exceedance to be derived within arid zones of the world even where local flow data are unavailable. Specific problems associated with the analysis of flood records from arid areas include gauging problems and measurements, low or zero annual maxima in a number of years, and the suitability, or otherwise, of the mean annual flood as the appropriate scaling factor for the dimensionless curves. In some cases, the 5-year flood may be a more appropriate scaling factor.