Multiscale Characterization and Prediction of Reservoir Inflows Using MEMD-SLR Coupled Approach: Fid-Bau Portal

Multiscale Characterization and Prediction of Reservoir Inflows Using MEMD-SLR Coupled Approach

Adarsh, S. / Janga Reddy, M.

In this study, multiscale characterization of hydroclimatic time series was performed using the Hilbert-Huang Transform (HHT) approach. First, to investigate the possible teleconnections of monthly inflows into Hirakud reservoir in India, and two potential large-scale climate oscillations, El Niño Southern Oscillation (ENSO) and Equatorial Indian Ocean Oscillation (EQUINOO), the corresponding time series were decomposed using the multivariate empirical mode decomposition (MEMD) method, and then running correlation analysis, namely time-dependent intrinsic correlation (TDIC) analysis, was applied. TDIC analysis showed that there exists a long-range negative correlation between EQUINOO and reservoir inflow in most of the time scales, whereas a positive relation prevails between ENSO and inflow at the biannual time scale in particular. TDIC analysis further proved that the association between large-scale climate oscillations and reservoir inflow is not always unique but associated with localized reversals in the nature of correlation in the time domain; also, both the nature and strength of the association vary with time scales. Stemming from this finding, this paper proposes an innovative approach that combines MEMD and stepwise linear regression (SLR) methods for prediction of reservoir inflows. In this approach, the different modes obtained through MEMD are predicted independently by SLR fitting, considering the statistically significant inputs at respective time scales, and the final summation of the predicted modes gives the monthly inflows. A statistical performance evaluation based on multiple criteria showed that the proposed MEMD-SLR approach displayed better performance, with a high $R^{2}$ of 0.905 and low values of root mean square error (RMSE) and bias for validation data over the EMD-SLR, M5 model tree, and multiple linear regression models for inflow prediction, including significant improvement in prediction of high inflows.