PROBABILITY DISTRIBUTION OF MONTHLY STREAMFLOWS OVER CUYO AND PATAGONIA REGIONS (ARGENTINA). HYDROLOGICAL DROUGHT MONITORING APPLICATION
Juan R. Rivera y Olga C. Penalba
Instituto Argentino de Nivología, Glaciología y Ciencias Ambientales, CCT-Mendoza, CONICET
Universidad Juan Agustín Maza
Departamento de Ciencias de la Atmósfera y los Océanos, FCEyN, UBA
Manuscript received on 16th March 2017, in final form on 14th July 2017
The standardized streamflow index (SSI) was developed following the theoretical concept of the standardized precipitation index (SPI) for monitoring and comparison of spatial and temporal patterns of the hydrological conditions, considering monthly streamflows as input variable. A fundamental step for its calculation is to select the probability distribution that best fits the frequencies of observed streamflows. This work compares 6 widely used probability distributions (gamma, lognormal II, Pearson III, Weibull, logistic and loglogistic) for the representation of hydrological variables in order to determine the most adequate over the watersheds of Cuyo and Patagonia (Argentina). The Anderson-Darling (AD) test was considered to evaluate the goodness of fit, which guarantees rigorous testing for extreme events analysis. Recommendations are given considering the fits based on the AD statistics and the analysis of the normality of the SSI time series, considering that none of the analyzed distributions were adequate in all the basins and months of the year. These results highlight the use of the lognormal II distribution for the representation of monthly streamflows in the Cuyo region and the assessment of hydrological droughts based on this distribution. In the case of Patagonia region, the loglogistic distribution is the most adequate for the representation of the monthly streamflows. However, it was observed that the SSI time series based on this distribution have biases in their mean values, standard deviation and percentage of months under drought conditions. These factors may affect their quality for the monitoring of hydrological droughts over Patagonia. Therefore, it is recommended the use of the lognormal II distribution for the representation of monthly streamflows over the study area and its application to the monitoring of hydrological droughts.