*3.4. Models' Application for Sediment Delta Prediction*

More than 200 million people of Pakistan directly or indirectly depend on the irrigation supply and power generation from the Tarbela Dam. Therefore, it is very important to assess the future delta movement and sedimentation scenarios in the reservoir. It is pertinent to mention here that SRC-generated sediment load boundary conditions are being used for all types of sedimentation modeling in the upper Indus River projects [21,42]. Therefore, to check and ascertain the long-term application of the SRC and WA-ANN techniques, the HEC-RAS model was run up to the year 2030 using future discharges calculated by [35] employing the University of British Columbia (UBC) watershed model. UBC is a less data-extensive semi-distributed watershed model developed by the University of British Columbia. As discharge alone with one level of decomposition represents more accurately the transport processes at Besham Qila, calculated future discharges by [35] were used in the trained WA-ANN model for obtaining future sediment loads. Reservoir water levels from 2005–2018 were repeated for 2019–2030. The simulated/forecasted levels of the Tarbela Reservoir for 2022 and 2030 along with observed levels of 2013 and 2017 showed a huge volume lost due to sedimentation (see Figure 11). As SRC showed overestimation (190 Mt of suspended sediment load (SSL) compared to 160 Mt SSL using WA-ANN) for the Indus River (Table 1), therefore, using SRC as the boundary condition in the modeling process also overestimated the bed level variations in the major ponding area of the reservoir near the dam. As SRC has been used for sedimentation modeling of all studies of the Upper Indus River, and it has been predicting similar results. For example, the 4320-MW Dasu Hydropower Project, which is under construction upstream of Tarbela Dam, will be silted up just 20–25 years after its commissioning without conducting yearly flushing operations [21,41]. The predicted short life of the Dasu project could very likely be a result of the overestimation of sediment load using SRCs. Initially, the work in [13] in 1970 also estimated 400 Mt of sediment load using SRC for the Tarbela Reservoir, which showed a shorter life of the reservoir. However, later studies estimated 50% lower sediment load for the Indus River at Tarbela Dam (see Table 1). Due to less sediments entering the reservoir, it is still operational and not silted-up. It might be possible that in 1970, very limited sediment concentration data were available, which might have consisted of high-flow hydrological years. However, the availability of long data series of sediment sampling cannot help SRC to model the hysteresis phenomenon and hydrological variations related to shifting in high flows from summer to post- and pre-summer months at the Upper Indus Basin [41]. Therefore, the WA-ANN-generated sediment load boundary condition, using future projected discharges, can more precisely represent the sedimentation modeling processes.

**Figure 11.** Comparison of simulated bed levels for 2022 and 2030 with the SRC and WA-ANN techniques, along with observed levels of 2013 and 2017. The longitudinal profile is only showing the sediment delta region of Tarbela Dam.
