Flood frequency analysis based on a mixed GEV distribution with upper limit applied to the Hydrological Region No. 10 (Sinaloa), Mexico

Abstract

Flood Frequency Analysis (FFA) processes the available record of annual maximum flows of a river, to estimate predictions associated with low probabilities of exceedance, its reciprocal is the return period (Tr) in years. These predictions are the design floods, with which all hydraulic works are planned, designed, and reviewed hydrologically, such as reservoirs, protective embankments, channel rectifications, bridges and urban drainage works. In this work, a novel method of the FFA is described and applied, which incorporates additional hydrometric information in a mixed General Extreme Values (GEV) distribution; this procedure is aimed to accurately define the flood of Tr = 1 000 years. Initially, the information on average annual flow and annual maximum flow of all the hydrometric stations that make up the homogeneous region under study are processed. The above, with the approach of the enveloping curves and with the objective of defining an enveloping curve with zero probability of exceedance; whereby, defines the extreme maximum flow that is approached as asymptote by the upper part of the mixed GEV, which avoids an unreal increase in predictions. Lastly, based on each record of floods, synthetic sequences of 1 500 values are generated and one with similarity to the available data and more than ten floods of Tr > 150 years is chosen. To such a synthetic sequence a GEV distribution is adjusted, with the methods of moments L and LH, to select the one with the lowest standard error of fit. This distribution forms the lower part of the mixed GEV, until the point of inflection of Tr = 500 years. The described method was applied to the seven largest flood records of the Hydrological Region No. 10 (Sinaloa), Mexico and based on its results, the Conclusions were formulated, which highlight its advantages and suggest its application to estimate predictions of high return periods (100 ≤ Tr ≤ 1 000 years) more accurately, by incorporating regional hydrometric information.