Regionalization of annual maximum streamflow in the Mundaú River Basin using the Index-Flood Procedure

Código

I-EBHE0172

Autores

Eber José de Andrade Pinto, Mucio Valença Virães

Tema

WG 1.10: Hydrologic Design - Solutions & Communication

Resumo

The study developed focuses on the regionalization of peak flows in the Mundaú river basin, located between the Brazilian States of Pernambuco and Alagoas, with 3.600 km². It encompasses 08 streamflow gaging stations, applying the ?index-flood? procedure, utilizing L-moments, and employing the L-statistics proposed by Hosking and Wallis (1997). The steps for applying the ?index-flood? method were as follows: Organization and dimensionless transformation of the series by the mean; Definition of empirical frequency curves for each dimensionless series; Definition of homogeneous regions and regional frequency curves; Regression analysis between the physical and climatic characteristics of the basin and the dimensionless factor, the mean annual peak flows of the stations; and Estimation of quantiles of interest. To minimize subjectivity in the regionalization stages, L-statistics were applied in data consistency, with the discordancy measure Di; in the delineation of homogeneous regions, with the heterogeneity measure H; and in the selection of the regional probability distribution, with the goodness-of-fit measure Z. Initially, the consistency of the peak series was evaluated, and the discordancy measure Di was calculated. Six series ranging from 07 to 43 years in length were selected. The series were considered stationary with a significance level of 5%. Analysis of the dimensionless empirical curves and the heterogeneity measure H has showed that the Mundaú basin forms a single homogeneous region. The selection of the regional frequency distribution was carried out by estimating the parameters of the candidate distributions (GLO, GEV, LN-3P, P-III, GP) from the regional L-moments, applying the goodness-of-fit test, and evaluating the positioning of the regional L-ratios on the L-Kurtosis x L-Skewness diagram. Generalized Logistics (GLO) was chosen as the dimensionless regional distribution; the estimated position, scale and shape parameters were 0.850, 0.322 and -0.261, respectively. The regressions between the physical and climatic characteristics of the basin and the dimensionless factor were evaluated using the F-test with a significance level of 5%, a coefficient of determination, R², above 90%, and percentage deviations up to 25. The Mundaú basin presented two regression equations. One regional equation had explanatory variables of drainage area (A) and the percentage of drainage area with a slope greater than 20% (DA20abs), with R² = 0.997977 and percentage deviations less than 8.13% in absolute value, and another regional equation based on drainage area (A) and mean annual precipitation (MAP), with R² = 0.992879 and percentage deviations less than 6.95% in absolute value. The major floods recorded in the basin occurred in 1945, 1969, 2000 and 2010, with time intervals between occurrences of 24, 31 and 10 years, respectively. The selected regional dimensionless distribution, GLO, has no upper limit, as the shape parameter is negative. The parameter value results in a heavy upper tail. It is recommended to estimate quantiles for return periods of less than 200 years. The regression equations should be used with caution for areas smaller than 500 km², as the data were obtained through twice-daily readings and, for areas smaller than 500 km², the peaks of hydrographs were rarely recorded.