Wavelet transform for medium-range streamflows projections in national interconnected system

This study aimed to analyze the variability of average annual streamflow time series of the National Interconnected System (NIS) (Brazil) and create a projection model of future streamflow scenarios from 3 to 10 years using wavelet transform (WT). The streamflow time series were used and divided into two periods, namely, 1931–2005 and 2006–2017, for calibration and verification, respectively. The annual series was standardized, and by the WT, it was decomposed into two bands plus the residue for each base posts (BP) for later reconstruction. Then, an autoregressive (AR) model per band and residue was made. The projection was obtained by adding the AR models. For performance evaluation, a qualitative analysis of the cumulative probability distribution of the projected years and an analysis of the likelihood were performed. The model identified the probability distribution function of the projected years and obtained a likelihood ratio of > 1 in most SIN regions, indicating that this methodology can capture the mediumrange variability.