Predicción de caudales en la cuenca del Machángara

Abstract

The dependence on water in all daily activities around the world has generated a desire to know how much rain and streamflow can exist in a particular place. In this study occurs a model for forecast rain and streamflow in the basin of Labrado and Chanlud. It was determined three climatic variables that most influence has with rain of the two basins through a correlation matrix.After having three climatic variables proceed to make two methodologies for predicting rain through an ARIMA model and artificial neural networks (RNAs). In the system ARIMA only was proposed one model with the rainas input variable and in the neural networks was generated 16 models for each zone.The input variables in the RNAs were the three potential predictors and the rain of each basin.The RNAs were selected for the prediction of rain in each basin through statistical parameters such as:mean square error, mean absolute errorandNash-Sutcliffecoefficient. The predicted rain was used in predicting streamflow.Two methodologies were raised; the first method is the RNAs and the other was a hybrid model that combines the ARIMA method and RNAs. We selected the best model through statistical parameters mentioned above.