Comparison of methodologies for flow prediction through teleconnection indices. Case study: Paute river basin

Abstract

The Paute river basin (southern Ecuador) suffers hydrological changes due to climate change and human activities. Hydrological changes cause extreme events and affect ecosystems, hydroelectric plants, and quality of life. It highlights the importance of understanding hydrological behavior to make appropriate decisions in extreme environments. This study seeks to predict discharges in the Paute river basin through global teleconnection indices. Multiple Linear Regression (MLR) was obtained using three different methodologies: multicollinearity analysis, Principal Component Analysis (PCA), and correlation with monthly delays. It was shown that the principal component analysis scenario obtained the best predictive fits, specifically by including 41 indices and 20 components. For the scenario using monthly delays, the best delay occurs within a single month for most seasons. Finally, with the multicollinearity analysis scenario, better results were obtained using 41 indices, although essentially the performance corresponds to the number and indices of each model. Teleconnection indices are not sufficient when used as the only input variable for download modeling and prediction, giving mostly unsatisfactory results. However, a clear trend links the behavior of flows and indices, and it is possible to improve the models based on more climatic variables or with other predictive methods.