Modelos multivariados de predicción de caudal mensual utilizando variables macroclimáticas. Caso de estudio Río Cauca

Abstract

The linear teleconnections between the ENOS phenomeno and the Cauca ValleyColombia monthly interannual flows was studied. Two multivaritate statistical techniques was used: Empirical ortoghonal functions (EOFS) and the canonical correlation analysis (CCA). Armax models was adjusted to predict the month flows, using principal components and canonical coefficients of the macroclimatic variables, as auxiliary variables in prediction models. When includes the principal components of macroclimatics variables as predictor variables in flows models, improved in prediction was obtained, indicating that they contribute with additional information. The FPE (final prediction error) was reduced in 9.44% in average, using the first principal component of macroclimatics variables as auxiliary variable. While, the FPE was reduced in 13.35% in average using the first canonical coeficients of macroclimatic variables as auxiliary variable. The flows models presented good adjustment, for what they can be used for prediction. Likewise, The multivariate EOFS and CCA methods proved to be a valuable tools in the study of climate variability to understand the relationships between the ENOS phenomeno with the region hydrology.