ANÁLISIS REGIONAL DE FRECUENCIAS DE CAUDALES DE RIOS PERTENECIENTES A CUENCAS CON NACIENTES EN LA CORDILLERA DE LOS ANDES

VICH ALBERTO I.J.; NORTE FEDERICO; LAURO CAROLINA

Meteorologica

Centro Argentino de Meteorólogos

Año: 2014 vol. 39 p. 3 - 26

1850-468X

This paper aims to determine the magnitude of flow variables for different return periods, especially in the case of flow levels much greater than those observed in the past, and to obtain a quantile for sites with scarce data using a Regional Frequency Analysis approach based on L-moment statistics (ARF-LM). The area under analysis comprises mountain basins along a broad latitudinal gradient between the Bermejo and the Santa Cruz rivers in western Argentina.Thirty-two gauging stations or sites were analyzed and five variables were determined: annual flow (Qaño) and four seasonal flows: winter QI (july, august, september), spring QP (october, november, december), summer QV (january, february, march) and autumn QO (april, may, june), from the average daily flow observed.Exploratory data analysis was performed to complete and verify some underlying assumptions in the series (independence, randomness and normalcy) and homogeneity conditions. Normalcy is rejected in 60% of the cases, summer and autumn flows being the variables with the least degree of normalcy. Lack of both independence and randomness occurs in 34% of the series analyzed, notably the winter and autumn flows with a level of significance of 5%.The Los Patos, San Juan, Cuevas, Diamante, Atuel, Grande, Valenzuela, Poti Malal, Neuquén and Santa Cruz rivers did not show significant trends in any of the flow variables under analysis. A statistically significant increasing trend was detected in some variables and locations in Argentina?s northwestern and central-western region while a decreasing trend was observed in some variables and locations in the Patagonian Andes. In general, flows Qaño, QI and QV show abrupt changes in over 65% of the cases while QP only in 44%. In general, the jump in mean values took place in the 1970s.Different clusters of sites were considered and a homogeneous region was identified for each of the five variables. Homogeneity was verified by means of discordance and heterogeneity tests and the best-fit distribution through ZDIST and PUMT measures. The distributions used were: generalized extreme values, generalized logistic, generalized normal, normal, Gumbel, generalized Pareto, exponential, and Pearson III. In all cases the homogeneous region includes only stations in the southern Central Andes, from the Tunuyán to the Colorado rivers. The Pearson III and Gumbel distributions exhibit the best fit.