A numerical study of pumping effects on flow velocity distributions in Mosul Dam reservoir using the HEC‐RAS model

This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. © 2020 The Authors. Lakes & Reservoirs: Science, Policy and Management for Sustainable Use published by John Wiley & Sons Australia, Ltd. Civil, Environmental and Natural Resources Engineering, Luleå University of Technology, Luleå, Sweden


| INTRODUC TI ON
Sediment deposition in reservoirs is one of the most common problems for all dam storage works around the world. Sediment deposition generally reduces the water storage capacity by as much as 0.8% annually (Basson, 2009), leading to a reduced project life, even with increasing water demands. Another problem arises from sediment deposition near the gates of power generation plants, which can affect the operation efficiency of a dam.
Pumping stations are one of the main structures in some multipurpose dam and reservoir projects, especially in cases where water supply and land irrigation occur at a higher elevation than the reservoir water level. As the reservoir is the source of pumped water for such stations, the sediment concentration, sediment properties, and pumping rate affect the total quantity of water delivered, the quantity entering the intakes and the water pumping capacity. The mechanical elimination of sediment, particularly from a large reservoir, is an expensive process, compared to utilization of a suitable hydraulic method for controlling or reducing the sediment effects.
Physical models are typically used to study, analyse and formulation solutions to different sedimentation problems in reservoirs and other hydraulic structures. These types of models are scaled descriptive of water flow state and geometry, usually being used for design and optimizing the efficiency and safety of different hydraulic structures.
As an example, the effects of reservoir water levels and upstream inflow rates on sediment load transportation in the upper part of Mosul Dam reservoir were studied by Issa, Al-Ansari, Knutsson, and Khaleel (2012), using a moveable bed distortion model. This type of model, however, is considered expensive, compared to mathematical or numerical models, which are more widely used techniques. A major advantage of using computational models is that they can be employed with different physical domains more easily than physical models, the latter tending to be more site-specific (Papanicolaou, Elhakeem, Krallis, Prakash, & Edinger, 2008).
The most common models for reservoir sedimentation include: (a) the Hydrologic Engineering Centers River Analysis System models (HEC-RAS 1-D and 2-D) (U.S Army Corp of Engineers, 2016), (b) the National Center for Computational Hydro Science and Engineering models (CCHE, one-dimensional and two-dimensional) (Khan, 2003), (c) the Hydrodynamic and Sediment Transport Model, known as the Estuarine Coastal Ocean Model (ECOMSED, 3-D) (Blumberg & Mellor, 1987)

and (d) the Simulation of Sediment Movements in
Water Intakes with Multi Blocks Options model (SSIIM, 3-D) (Olsen, 2018).
Some previous studies applied 1-D, 2-D or 3-D models to simulate flow, sediment transport and reservoir bed formation. (Jungkyu & Song, 2018) applied a 1-D numerical sediment transport model to study the effects of the number stream tube on sediment on the model performance by comparing the results with 14 years of actual sediment data. Reasonable results were obtained with three stream tubes, with the numerical results varying depending on the densities of deposited sediments. Hu, Cao, Pender, and Tan (2012) presented a coupled layer-averaged 2-D model that absorbed both the current sediment transport and morphological formation. It was applied to simulate two measured turbidity current events in the Xiaolangdi Reservoir on the Yellow River, with the results indicating the model was a suitable tool to define both the current front location and the suitable time of opening the bottom outlets to implement reservoir sediment management. Isaac & Eldho (2019) presented a 3-D model and experimental work focusing on hydraulic flushing of the reservoir to identify the suitable orientation intake and to optimize hydraulic flushing. The variations between the 3-D model results and experiments were between 4% and 6% for the different discharges and durations considered in their analysis. Their results also indicated the 1-D model for sedimentation and the 3-D model for sediment flushing can be efficiently applied for planning and design purposes.
Previous studies have applied HEC-RAS to evaluate the sedimentation process in the Tuttle Creek Reservoir (Shelley et al., 2015) and the Olmsted Locks in the United States (Ghimire & DeVantier, 2016), with the results indicating the model is an efficient tool for sediment studies. The HEC-RAS model as a one-dimensional (1-D) model was applied to simulate the flows and evaluate the total sediment deposited in Mosul Dam Reservoir within 25 years of its operation from 1986 to 2011 (Mohammad et al., 2016). The results indicated the model performance was good for estimating the total load deposited in the reservoir, compared to the bathymetry survey , 2013a.
For some storage reservoirs with a pumping station as one of the main attached structures, the pump operation changes the flow regime and sediment transport in the area. Sediment movement today and inside the pump results in a reduced pumping efficiency and capacity. A number of previous studies looked at sedimentation problems near intakes and pumping stations. As one example, the CCHE two-dimensional model was applied to analyse the sedimentation problem at the Rowd El-Farag pump station intake (Moussa, 2011). The SSIIM2 model was applied to simulate the flow velocity distribution and sedimentation in the intake in a rectangular channel (Ashari & Merufinia, 2015), with the 3-D model exhibiting a good agreement with measured values from previous studies.
The effects on the flow pattern resulting from pumping from a reservoir was studied, and a numerical simulation model was compared with field measurements to better understand the flow pattern near a reservoir pumping station intake (Michael, Cesare, & Schleiss, 2018). The results indicated the pumping effect on the flow pattern in front of the intake was not significant, while injected water by a turbine model had an important effect in the nearby area.
The problem of sediment deposition near the Mosul Dam Reservoir pumping station was studied by Mohammed (2001), including field measurements of the flow velocity at selected points, sediment depth, grain size distribution of the deposited sediment material and grain size distribution of the suspended load.
The general objective of the present study was to analyse the flow velocity and direction near the pumping stations in reservoirs using a two-dimensional model, which help identify extent of pumping effects on the reservoir's flow regime. This is in addition to better understanding the effects of the stream power as an indication of the sediment transport capacity of the flow towards the pump intakes. The flow velocity and direction are the most important factors driving the sediment movement, which leads to deposition or erosion. Further, identification of the sediment concentration along the reservoir helps to evaluate the effects of the main river flow and different additional run-off on the reservoir sediment concentration. The Tigris River, as well as the sediment carried in run-off flows from the ten valleys on both sides of the reservoir, comprises the main sources of flows and sediments delivered to the reservoir.

| S TUDY ARE
These valleys are seasonal, carrying significant rainfall run-off flows and sediment loads in the winter, and being mostly dry in the summer. The reservoir was divided into two zones for the present study.
Zone 1 is the upper part, extending for 24 km from the reservoir inlet ( Figure 1). The flow velocity and direction in this zone were analysed as a two-dimensional flow. In contrast, the lower part (zone 2) was considered as a one-dimensional flow to reduce the simulation computation times.

| Pumping station
The pumping station within Mosul Dam was constructed in 1985 at the time of the dam's construction. It is located on the right bank of the reservoir, about 20 km from the simulated reservoir along the thalweg path ( Figure 1). The station started to supply supplementary irrigation water to the North Jazeera Irrigation Project in 1991. The project is located in the northwest of Ninawa Governorate's Rabeaa Sector, covering an area of about 600 km 2 . The main station components are the stilling basin, the intake structure and the central pumping station. The pumping unit consists of twelve pumps with a maximum pumping capacity of 4 m 3 /s for each unit. One of the station's main problems is sediment accumulation in the approach to the intake structure and in the suction pipes, affecting the pumping efficiency and capacity, and threatening the continuation of the station's operations (Mohammed, 2001).

| ME THODOLOGY
The present study was undertaken using the River Analysis System developed by the Hydrologic Engineering Canter (HEC-RAS model, 5.06). The model simulates the flow in rivers, channels and reservoirs with different hydraulic structures, including embankments, bridges, culverts and weirs. The model can also simulate the sediment transportation as a one-dimensional flow, including deposition and erosion for long periods. In the case of reservoir flow, the plane flow distance is much greater than the perpendicular distance; thus, the vertical velocity was considered to be low, meaning the vertical velocity and its derivative in the mass and momentum equations can be neglected in the analysis.  (Molinas and Yang, 1986).
The deposition or erosion (i.e. change in bed section) that will occur for each control volume can be estimated on the basis of the difference between the load entering and released by the control volume. The released load depends on the flow transport capacity, which can be estimated at more than one section for non-cohesive sediment, depending on the hydraulic and sediment particle properties.
For all approaches, the instigation of bed particle movement generally depends on such flow characteristics as the flow velocity, depth and sediment properties. The shear stress can be used to identify the creation of particle motion and the flow transport capacity, expressed using the stream power. The stream power is the product of the average flow velocity and the average shear stress.
Assessment of the shear stress helps confirm whether or not the flow can detach and transport sediments on the basis of a critical shear stress value.

| Geometry data
The geometry of Mosul Dam was based on available contour maps for the area prior to the construction of the dam, as well as available digital elevation model data (USGS). The reservoir length along the river basin at a maximum operation level is about 80 km (335 m.a.s.l.) and was divided into 80 sections with an average length of 1-km interval between each section. The projected section was then converted to a terrain map to represent the reservoir geometry for consideration in the HEC-RAS model. The elevation-storage curve of the reservoir geometry (considered in the model as a terrain file) was compared with F I G U R E 1 Mosul Dam Reservoir zones 1 and 2, valleys and pumping station location F I G U R E 2 Storage-elevation curves for Mosul Dam at dam construction in 1986 the rating curve of the reservoir before the dam operation began in 2,983  with the two curves exhibiting good agreement with a determination coefficient (r 2 ) of 0.98 ( Figure 2).

| Flow and sediment data
The recorded daily flow rate of the Tigris River upstream of the reservoir and the sediment rating curve was created to represent the sediment load carried by the river upstream of the reservoir. In addition to the main river flow, seasonal valleys also add to the reservoir. The estimated daily flow rate and sediment load for the valleys located on the right and left banks of the reservoir were previously estimated and considered as model input (Mohammad et al., 2012; Mohammad, Al-Ansari, & Knutsson, 2013).

| Model calibration
The model was calibrated for both flow and sediment load. The measured inflow and outflow data were considered, with the reservoir water level determined from the model simulation. The simulated reservoir levels were compared with measured values through the study period (1986 to 2011) to evaluate the model performance. The determination coefficient, Nash-Sutcliffe model efficiency, and paired t test values were used to evaluate the model performance. The obtained determination coefficient was 0.87, the Nash-Sutcliffe model efficiency was 0.88, and the paired t test value was 0.37 (tabulated value at a significance level of 0.05 is 1.96), indicating there was no significant difference between the observed and simulated values.
For sediment routing, the model can simulate the continuity, momentum and sediment transportation equations only as a one-dimensional flow. Nine scenarios were considered to simulate the sediment trapping in the reservoir and the sediment distribution as deposition or erosion. The nine scenarios resulted from combining three sediment transport capacity approaches with three fall velocity approaches to assess how rapidly the particles can drop out of the water control volume and be deposited.
The three sediment transport approaches considered were the Ackers-White, Laursen and Toffaleti models. The three equations considered for the fall velocity were the Rubey, Toffaleti and Van Rijin approaches.
The selected approaches for both the sediment transport capacity and fall velocity are suitable for the size of the sediment load entering Mosul Dam. The considered material size of the sediment load for the Tigris River ranged from 0.0008 to 0.6 mm, while the sediment load from the valleys on both sides of the reservoir ranged from 0.001 to 0.4 mm. The soil classifications for both the right and left valleys generally were clay, clay loam, silty clay, silty clay loam and silty loam.
The evaluation of each simulation scenario was based on comparison of the model results with bathymetry survey data for 54 sections (Issa, 2015). The average bed level of the sections extending from the dam axis to 60 km upstream was compared with the measured values, with the final simulation values being based on statistical criteria, the determination coefficient, Nash-Sutcliffe model efficiency coefficient and t test values. The average change in bed level of the sections was considered in a calibration process to evaluate each scenario.
The results in Table 1

| S IMUL ATI ON OF FLOW AND S ED IMENT
The applied model has the ability to simulate the flow hydrodynamics as an unsteady flow for one-dimensional and two-dimensional cases, as well as a hybrid between these two cases. Scenario 5, which is combined between Laursen as a sediment transport capacity approach and Toffaleti as fall velocity approach, gave the best statistical values of the considered criteria, therefore, justifying its use for sediment analysis as a one-dimensional model.

| Flow analysis
The  Thus, the hydraulic performance of the pumping station at any other location will remain similar to the current designated location for same operation plan.

| Sediment transport analysis
The The sediment concentration along the reservoir and the average velocity from a simulation on a dry, non-rainy day in April 1988 is illustrated in Figure 5b. This time period was selected because it has a similar sediment concentration at the inlet illustrated in To compare the inverted levels along the reservoir bed, Figure 6a illustrates the original levels before dam construction (1986), the bathymetry survey results (Issa,2,905)

| CON CLUS IONS
The HEC-RAS 5.06 model was applied to simulate the depth-aver-