Abstract

ROJO-HERNANDEZ, Julián David  and  CARVAJAL-SERNA, Luis Fernando. Nonlinear river flow prediction using macroclimatic variables and singular spectral analysis. Tecnol. cienc. agua [online]. 2010, vol.1, n.4, pp.59-73. ISSN 2007-2422.

The close nonlinear relationship between Colombian hydrology and global climatic processes justifies the study of nonlinear river flow models. This paper shows the nonlinear river flow dynamics in Colombia using a periodic prediction model based on the Singular Spectrum Analysis (SSA). The prediction window length is 3 and 6 months. The prediction includes both univariate and multivariate cases. The multivariate case takes into account macro-climatic variables. The proposed model is studied through the Principal Component Analysis, the correlation structure between macro-climatic variables and river flow time series and the reconstruction of series. This model is applied to the San Carlos, Riogrande II, Guatapé, Magdalena, Guavio, and Batá rivers. The results show an improvement in river flow prediction and indicate that river flow multivariate prediction using macro-climatic variables is better.

Keywords : Singular Spectral Analysis (SSA); time series modeling; nonlinear river flow prediction; prediction with exogenous variables.