Forecasting streamflows in the San Juan River Basin in Argentina

Abstract

San Juan province, located in western Argentina, presents great climate variability with arid characteristics. Mean annual rainfall averages less than 100 mm for the whole province, and snowmelt in the Andean upper basin provides the San Juan River Basin with seasonal streamflow during summer, the period of highest water demand for irrigation. Traditional streamflow forecasts for the San Juan River are based on statistical regression models that are strongly dependent on values of snowpack in winter months (July, August, and September) and streamflow values in the spring months. However, producing forecasts for San Juan River summer streamflow using the Multivariate El Niño Southern Oscillation Index (MEI) data in the preceding June of the water year as an explicative variable can improve reservoir operating system performance for irrigation. To demonstrate this, climate predictors such as the MEI were used to forecast San Juan River streamflows to provide predictability at a six-month lead time. A backpropagation neural model, based on coupled data of snowpack and a climate predictor during the winter period, proved successful in forecasting San Juan River flows during the following summer period. © 2010 Springer Science+Business Media B.V.