A worldwide comparison of regional flood estimation methods and climate

Abstract

A summary of the results of regional flood frequency analyses that have been carried out in many tropical and sub-tropical countries is presented. The results provide a set of flood estimation methods which should be of immediate usefulness in practical applications, bearing in mind the errors of generalization. By defining the representative values of the mean annual flood and of the flood frequency curves for each region, an attempt has been made to develop relationships between the methods and climatic characteristics. These show that adequate relationships can be obtained based solely on the median annual average rainfall and the median catchment area for the regions. In particular, the steepness of the regional flood frequency curve can be estimated reasonably well from these two parameters. The relationships confirm that, broadly speaking, for humid regions the "average" flood -the flood that occurs regularly every one or two years -is relatively large, but the flood frequency curve is not very steep; rare floods (occurring once in every 100 to 1000 years) are not very much larger than the average flood. Conversely, in arid regions, the average flood is small, but rare floods can be extremely large multiples of the average. These relationships provide a useful initial indication of the type of regional flood frequency estimation method likely to be found in tropical and sub-tropical regions throughout the world.Comparaison à l'échelle mondiale entre les méthodes régionales d'estimation de crue et le climat Résumé On présente ici un résumé des résultats d'analyses régionales de fréquence des crues appliquées à de nombreux pays tropicaux et sub-tropicaux. Les résultats fournissent un ensemble de méthodes d'estimation des crues qui devraient être d'un intérêt immédiat pour des applications pratiques, tout en gardant à l'esprit les dangers d'une généralisation abusive. En définissant les valeurs représentatives de la crue moyenne annuelle et des distributions de fréquence des crues pour chaque région, on a essayé de mettre en évidence une relation entre les méthodes utilisées et les caractéristiques climatiques. Ceci montre que l'on peut obtenir des relations régionales satisfaisantes basées uniquement sur la médiane de la pluie moyenne annuelle et sur la médiane de la superficie des bassins. En particulier, la pente de la distribution de fréquence régionale peut être assez bien estimée à l'aide de ces deux paramètres. Ces relations confirment qu'en général, pour des régions humides, la crue "moyenne" -la crue qui se produit tous les un ou deux ans -est relativement importante, mais que la courbe de distribution de fréquence n'est pas très pentue; les crues très rares (qui se produisent tous les 100 ou 1000 ans) ne sont pas beaucoup plus importantes que la crue moyenne. Par contre, dans des régions arides, la crue moyenne est très faible, mais les crues rares peuvent être plusieurs fois plus fortes que la crue moyenne. Ces relations fournissent une première indication sur le type de distribution régio...

Flood frequency analysis at ungauged sites in the KwaZulu-Natal Province, South Africa

“…To compare the results from this study with similar OLS models reported in the literature by van Bladeren (1993), Mkhandi and Kachroo (1997) and Meigh et al (1997),the R 2 statistic was calculated for the OLS models.…”

“…Based on the previous studies by Meigh et al (1997) and Mkhandi and Kachroo (1997) and similar studies from other parts of the world, e.g. NERC (1975), AREA is the single most important variable and should always be included in the models.…”

“…(17) and (20), it would be preferable to use all catchments when estimating MAF in Region 1, but only data from Region 2 should be used when estimating MAF in Region 2. A separation according to AREA as suggested by Meigh et al (1997) did not improve the average prediction variance and, hence, was not used.…”

“…Previous attempts to develop regression models in South Africa have been reported by van Bladeren (1993), Mkhandi and Kachroo (1997) and Meigh et al (1997). All three studies used explained variance (R 2 ) to quantify model performance.…”

“…The model for region S, T and U obtained an R 2 value of 0.727, and for drainage region V an R 2 value of 0.707. Meigh et al (1997) developed a general regression model for South Africa and Botswana relating MAF to AREA and mean annual precipitation (MAP). They divided the considered catchments into two groups based on AREA as AREA > 1 250 km 2 and AREA < 1 250 km 2 and reported R 2 values of 0.542 and 0.593 respectively.…”