The Upper Indus River Basin (UIB), Tarbela Reservoir in Pakistan Assessment of the future changes of sediment-driven storage losses is critical for the Tarbela Reservoir operation but remains limited. Here we estimate soil erosion and sediment flux over the past using empirical model Revised Universal Soil Loss Equation (RUSLE) and Water Balance Model for Sediment (WBMsed), respectively, driven by physiographic and hydroclimatological factors. We also develop multiple AI Model-based ensemble and Sediment Rating Curve (SRC) to estimate suspended sediment concentration (SSC) and generate observation-constrained projections of sediment-driven storage losses at the Tarbela Reservoir. AI models are trained and tested using intermittent daily observations of SSC, streamflow, temperature, and precipitation over 1991–2020, which are superior to the SRC model. The AI (RUSLE) model demonstrates that the average sediment load of 168.21 (178.4) Mt/year, which aligns closely with the observed reservoir sediment deposit (170 Mt/year). The AI (SRC)-based projections show the annual sediment load of 182.67 (124.1) and 200.76 (144.5) Mt/year under stabilized and business-as-usual scenarios, respectively. These escalating sediment loads cause substantial Tarbela Reservoir infilling, potentially reducing the gross storage below 20% of the design capacity in 2040 s. Addressing sediment-induced storage loss, this multi-model framework provides an integrated methodology for hydromorphological assessment, thereby supporting sustainable water resources management in data-scarce regions. • AI-based models outperform rating curves in estimating sediment fluxes. • Future sediment loads rise to ∼183–201 Mt/yr, accelerating reservoir infilling. • Sedimentation may cut Tarbela storage below 20% of design capacity by 2040 s.
Raza et al. (Mon,) studied this question.