Statistical seasonal rainfall forecast in Neuquen river basin (Comahue Region, Argentina)

Abstract

A detailed statistical analysis was performed at the Neuquén river basin using precipitation data for 1980-2007. The hydrological year begins in March with a maximum in June, associated with rainfall and another relative maximum in October derived from snow-break. General features of rainy season and the excess or deficits thereof are analyzed using standardized precipitation index (SPI) for a six-month period in the basin. SPI has a significant cycle of 14.3 years; the most severe excess (SPI greater than 2) has a return period of 25 years while the most severe droughts (SPI less than -2) has return a period of 10 years. The SPI corresponding to the rainy season (April-September) (SPI9) has no significant trend and it is used to classify wet/dry years. In order to establish the previous circulation patterns associated with interannual SPI9 variability, there are compared the composite fields of wet and dry years. There is a tendency for wet (dry) periods to take place during El Niño (La Niña) years, and/or when there are positive anomalies of precipitable water over the basin and/or when the zonal flow over the Pacific Ocean is weakened (intensified) and/or when there are negative pressure anomalies in the southern part of the country and Antarctic sea. Some prediction schemes, using multiple linear regressions were performed. One of the models derived using forward stepwise method explained the 42% of the SPI9 variance and retained two predictors related to circulation over the Pacific Ocean: one of them shows the relevance of the intensity of zonal flow in mid-latitudes and the other is because of the influence of low pressure near the Neuquén River basin. The cross-validation used to prove model efficiency showed a correlation of 0.41 between observed and estimated SPI9; a probability of detection of wet (dry) years of 80% (65%) and a false alarm relation of 25% in both cases.