Resumo

CAMPOS-ARANDA, Daniel Francisco. Obtainment of empirical equations for design floods estimation in the Hydrological Region No. 10 (Sinaloa) of Mexico. Tecnol. cienc. agua [online]. 2014, vol.5, n.1, pp.125-143. ISSN 2007-2422.

Floods designing (Q_Tr) is the most common hydrological estimation, because it is basic to sizing and reviewing all types of hydraulic works. Its estimation in watersheds without annual maximum flow data is based on regional methods, which use such information from a region or geographical area, in order to obtain results that apply within it. There are two types of regional procedures: those related to the index flood and those that develop empirical equations. The first type is based on the homogeneity of the region, hence its predictability; whereas the second method allows some degree of heterogeneity, which is taken into account in the regression models generated. Four empirical equations were established, using as prediction variables: the watershed area (A), the main channel length (Lcp) and predictions of maximum daily rainfall (PMD). Only two equations were selected for each of the six return periods established. 23 records of maximum annual flow were processed to obtain values of Q_Tr based on the General Extreme Values distribution. To obtain the fit coefficients of each equation three methods were used: the least-squared residuals in the logarithmic domain, their bias correction and the numerical optimization minimizing an objective function. Six empirical equations for each of the six return periods analyzed were obtained. The regularity of the estimations was tested by numerically contrasting results from six gauging stations (not used in the derivation of the equations) and six of those processed (those with the longest period of record). Such analysis allowed to point out the predictive ability of the empirical equations developed. It is highly recommended to apply the proposed procedure in other regions to find such empirical equations for design flood estimation, because of their very simple application.

Palavras-chave : multiple linear regression; relative error; performance indices; Rosenbrock algorithm.