*2.1. HEC-RAS Program System*

The River Analysis System (HEC-RAS), a one-dimensional model, created by Hydrologic Engineering Centre, has been designed to carry out steady flow water surface profile computations of natural rivers and networks of natural and constructed channels, unsteady flow simulations, moving boundary sediment transport computations, and water quality analysis. All these components utilize a common geometric data representation and hydraulic computation procedures. The calculations of one-dimensional moving material of the river bed causing scour or deposition over a certain modeling period establish a base for sediment transport simulations. Generally, sediment transport in rivers, channels, and streams depends on two modes: bed load and suspended load, which in turn depend on sediment particle size, the velocity of water, and river bed slope. The basic idea of evaluating sediment transport capacity by HEC-RAS is by computing sediment capacity of each cross-section as a control volume and for all particle sizes. HEC-RAS requires boundary condition data of each type for making such calculations. The boundary conditions are necessary to get the solution to the differential equations set, describing the problem over the area of interest. There are a number of boundary conditions for steady flow and sediment analysis computations in HEC-RAS. Boundary conditions can be either external, which are specified at the ends of the simulated network at the upstream/downstream, or internal, which

are to be used for connecting junctions. Background information regarding computational methods and equations used in modeling sediment transport is available in [32].

#### *2.2. Data Collection*

Owing to the noticeable global warming influence on the hydrological and river systems observed around the turn of the century, we considered to start the modeling process from 2005 onward [33–38]. For this, we collected the levels of Tarbela Reservoir and the flows of the Indus River at Besham Qila, the nearest station to the upper periphery of the reservoir located about 134 km upstream of the dam, from the project authorities for the 2005–2018 period. To hydrodynamically and morphodynamically initialize, calibrate, and validate the model, bathymetric surveys of the Tarbela Reservoir for the years 2005, 2013, and 2017 were also obtained.

To develop SRC and WA-ANN models, suspended sediment concentrations (ppm) and its gradational data at Besham Qila gauge recording station were collected for the 1969–2014 period from the Surface Water Hydrology Project (SWHP) of the Water and Power Development Authority (WAPDA), Pakistan. The raw data so collected are presented in Figure 2.

**Figure 2.** Data used in the study: (**a**) daily Tarbela Reservoir inflow and levels; (**b**) occasionally-collected suspended sediment concentration samples with observed flow.

The Tarbela Reservoir was cut into 73 cross-sections or range lines (R/Lines) to study the morphodynamics of the huge reservoir (see Figure 3).

**Figure 3.** Range lines (R/Lines) of Tarbela Reservoir used from [5].

The first comprehensive reservoir survey after the dam's construction was in 1974, and since then, each year, hydro-graphic surveys of the Tarbela Reservoir have been conducted. To cover the whole reservoir area, i.e., 161 km2, the hydro-graphic surveys were conducted using a systematic sounding method over the 73 cross-sectional range lines. Approximately 3500–4000 sounding measurements of the bed level alterations, reservoir depths, and water level elevations along these range lines are available, which were collected mostly during September–November. The distance between the cross-sections/range lines and the measured data points along these cross-sections are not identical. The average distance between each cross-section measured along River Thalweg was approximately 1000 m. However, compared to the upper periphery of the reservoir, the distances between the cross-sections nearer to the dam were smaller. The distance between measured data points along the cross-sections, i.e., lateral distance in y direction, was also variable with a mean of 39 m. The mean cross-sectional width near the dam axis was approximately 4000–5000 m, reducing to only 90–150 m near the upper periphery of the reservoir. Therefore, the major storage volume is near the dam axis, containing huge sediment deposits.

Water depths in the reservoir vary from a maximum 150 m near the dam to mostly 20 m at the reservoir inlet. To secure the stability of the dam and bank slopes along the reservoir, the maximum lowering and rising rate for the reservoir during operation is 4 m/day and 3 m/day, respectively, between reservoir levels 396 and 460 m and only 1 m/day up to the maximum conservation level of 472.5 m asl. The average slope of the river bed in 1979 was 0.0011211, which decreased in 2010 with an average slope of 0.0005988.

#### *2.3. Performance Measures for Model Evaluation*

To assess the performance of the models in terms of accuracy and consistency in simulating reservoir water depths and river bed levels, the following three statistical measures tests were made up of: (a) the coefficient of determination (R2), an indication of the level of the relationship between the observed and simulated data, ranging from 0–1; (b) the observations' standard deviation ratio (RSR), the ratio of the root mean squared error (RMSE) to the standard deviation (STDEV) of the observed data; (c) the Nash–Sutcliffe efficiency (NSE), a statistical assessment to calculate the relative magnitude of residual variance compared to the measured data variance [39]. The formulas are shown in Table 2.


**Table 2.** Statistical performance parameters used to evaluate the modeling performance.
