Decomposing five decades of river flow: multiscale discharge variability, wavelet decomposition, and change-point detection in the upper indus basin (1973–2020)

The Upper Indus Basin (UIB) serves as a critical freshwater source for millions in South Asia, yet exhibits complex hydrological responses to climate change, cryospheric dynamics, and anthropogenic influences. However, wavelet-based and change-point detection methods in glacio-hydrological basins remain constrained by the lack of consensus on optimal wavelet families for snow-glacier melt signatures and insufficient evaluation against seasonal variability and stochastic noise. This study applies Pettitt change-point detection and Discrete Wavelet Transform (DWT) decomposition using Daubechies-4 (DB4) and Symlet-4 (SYM4) wavelets to monthly discharge records (1973–2020) from four UIB stations (Partab Bridge, Kachura, Daniyor, Yugo), preceded by a controlled synthetic-discharge reconstruction check. Synthetic series generated from observed harmonic seasonal components and residual distributions reproduced the observed records with Nash–Sutcliffe Efficiency values of 0.84, 0.80, 0.69, and 0.65 at Partab, Kachura, Daniyor, and Yugo, respectively; these values represent reconstruction similarity rather than independent validation. The discharge records exhibit substantial serial dependence (lag-1 autocorrelation: 0.642–0.740), necessitating block-bootstrap correction. Kachura (p = 0.021) and Yugo (p = 0.046) showed evidence of a change point in 1987, whereas Daniyor (p = 0.362) and Partab (p = 0.088) did not meet the 5% threshold, based on dependence-calibrated (exploratory) block-bootstrap Pettitt tests. At Daniyor and Partab, the Pettitt years were retained only as candidate split points for an exploratory, descriptive comparison. Exploratory Brown–Forsythe comparisons with Benjamini–Hochberg control showed scale- and wavelet-dependent changes. Semi-annual (∼6-month) detail-coefficient SD increased at Kachura (+132.8% DB4, +148.4% SYM4) and Yugo (+49.6% DB4, +66.6% SYM4). Sub-seasonal (∼3-month) changes were opposite in sign at Kachura (−42.8% DB4, +34.2% SYM4) and Yugo (−53.5% DB4, +31.5% SYM4). Multi-year (∼8-year) detail-coefficient SD decreased under SYM4 at Kachura (−34.1%), Yugo (−46.5%), and Partab (−59.2%). Daniyor showed only semi-annual change (+30.8%, SYM4). These findings provide a reproducible descriptive framework for scale-dependent hydrological variability while emphasizing the need for independent hydroclimatic observations to establish physical attribution.