Hydraulic performance of riverbank filtration: case study West Sohag, Egypt

Egypt, such as many countries worldwide, faces water-related challenges. Potable water distribution capacity is ten times smaller than the demand, resulting in a shortage of drinking water in 20% of Egyptian villages, (Abdelrady et al. 2020). Also, the existing Water Treatment Plant (WTP) could not work properly because of contamination of surface water systems that come from industrial, agricultural, and domestic activities, in addition to the impact of climate extreme events. On the other hand, water quality standards become more stringent, and consequently, the water treatment cost is increasing. Furthermore, by-products of chlorination (a conventional disinfectant method) have been related to many dangerous diseases, including a higher rate of miscarriage, as well as damage to the central nervous system. As a result, new alternative water treatment technology is needed (Abdalla & Shamrukh 2011).

Because of the difficulties with direct usage and extraction of surface and groundwater for potable water supply, another water treatment should be used. Riverbank Filtration (RBF) technology is a natural, renewable, safe, and cost-effective drinking water treatment or pretreatment technology. Water from the river flows naturally along the river banks and into the aquifer (Figure 1). Physical, biological, and chemical reactions occur through the porous media of the aquifer, so pollutant concentrations can be mitigated and can be attenuated to the standards limit (Abdel-Lah 2013). Sandhu et al. (2011) showed that the RBF sites on India's Yaman River removed about 50% of dissolved organic matter (DOM) and micropollutants in the range of 13–99%.

RBF is well-known in many countries around the world. It has been used along the Rhine and Elbe Rivers in Germany for over 140 years (Becker et al. 2015). In several European countries, RBF is commonly used, such as Poland, France, Netherlands (Przybyłek et al. 2017), also the United States (Kelly et al. 2006). In addition to applying in west Asia, it is being considered in China, where evaluation of the suitability of RBF along the Second Songhua River (Wang et al. 2016) is taking place. In India, the potential of RBF is being checked in Uttarakhand (Sharma et al. 2014).

RBF's ability to reduce and remove water pollutants such as turbidity and microorganisms before they enter the abstraction well has been assessed in Egypt along the Nile River and its branches (Shamrukh 2006; Abdelwahab & Shamrukh 2008). Due to favorable hydrogeological conditions, it was found that there is a promising potential of RBF in Egypt (Ghodeif et al. 2016). These conditions have been reported in many studies as follows: the Nile River and the surrounding aquifer have a significant connection, the aquifer thickness is greater than 10 m, hydraulic conductivity is more than 10–4 m/s, which increases the transmission capacity, there is less river bed clogging due to the nature of the Nile River's erosive flow (Salah et al. 2019; Abdelrady et al. 2020). The main snag of RBF is that it is site specific, so a comprehensive investigation has to be done before installing a new RBF well. Lessons learned from the failure of a site in Dishna, Egypt, have been introduced by Bartak et al. (2015). Assessment of the applicability of RBF for the Embaba waterworks has taken place (Ghodeif et al. 2018; Grischek et al. 2018). The efficiency and water quality of the RBF system mainly rely on three pillars: the surface water quality, the aquifer hydraulic characteristics, as well as the RBF system's configuration parameters such as the number of wells, distance between wells, and abstraction rate.

This study aims to investigate the hydraulic performance and evaluate the RBF system's applicability for waterworks in west Sohag. Monitoring of water levels and water quality is used for the evaluation, also different scenarios were introduced using MODFLOW and MODPATH 10.2.3 under the platform of GMS to evaluate the efficiency of the RBF system.

Sohag governate is located in upper Egypt and about 476 km away from Cairo in the south (Figure 2). It is located between the Assiut governate in the north and Qena governate in the south, on the eastern side is the Red sea governate and on the western side the New valley governate. The Sohag government is rapidly expanding and seeing rapid population growth. Out of an estimated 4,603,861 people residing in the governorate, 3,618,543 people live in rural areas as opposed to only 985,318 in urban areas, which leads to an increase in the drinking water demand.

In the study area, summers are humid and winters are moderate, with little precipitation. The temperature ranges from 36.5 °C in the summer to 15.5 °C in the winter, with relative humidity levels ranging from 51 to 61 percent in the winter, 33 to 41 percent in the spring and 35 to 42 percent in the summer, (Ahmed & Ali 2011) and 7 millimeters of rain a year; however, an irregular rainfall occurrence happens on occasion for a brief period, which may lead to a hazardous flash flood. Land use categories in the Sohag governate are varied along the Nile but in the study area in west Sohag the main category is urbanization (Figure 2).

Surface water in the west Sohag WTP is treated by conventional processes. The holding company for water and wastewater (HCWW) and its branch, the Sohag Company for Water and Wastewater (SCWW), manage some waterworks that supply drinking water from 11 conventional WTPs, 52 compact units, and 198 groundwater treatment plants (HCWW 2018). West Sohag WTP is located on the Nile River's western side, Sohag (Figure 3). Riverbank filtration (RBF) was found to be the simplest technique to increase the production capacity, with a low cost and better quality than conventional water treatment plants. The existing RBF system consists of two vertical abstraction wells 14 inches in diameter and 36 m in depth, with an observation well located in the middle between the two abstraction wells, a submersible pump, and a disinfection unit. The abstraction wells are located on the Nile River's western side (Figure 3). The abstraction rate of wells is about 35 L/s (3,024 m³/d).

SCWW made a comparison of all water treatment technology used in the Sohag governate. Table 1 shows the capital and operational cost for each treatment. The total cost of the suggested RBF system in Sohag was lower than the other methods. The cost of the RBF system depended on the component where the RBF units commonly consist of abstraction wells and a disinfection unit.

Many researchers have made geological investigations for the study area, which is located within the Nile (El-Haddad & Youssef 2013; El-Sayed 2015). The study area's sedimentological sequence contains Lower Eocene limestone, Plio–Pleistocene sands, gravels, and clays, as well as recent sediments that were found to represent the sedimentological layers' order (Ahmed 2009) (Figure 4).

The Nile valley hydrogeology and groundwater flow system at different scales have been investigated by many researchers (Shamrukh 2006; Ahmed & Ali 2011; Abdalla & Shamrukh 2016). The Nile River, irrigation canals (West and East Nag-Hammadi canals), and drains are all part of the Sohag area's surface water hydrology. The water surface elevation of the west Nag-Hammadi irrigation canal at the study area is 59 m and the bottom elevation is 56 m above sea level. The canal has a cross-section dimension of depth 3 m, width 33 m, with a flow rate of 125 m³/s at the study area. A quaternary system is the main aquifer system in the research area, which stretches west and east of the Nile. The thickness and width of this aquifer range from one location to the next. The aquifer is a semi-confined or leaky aquifer formed by the Nile's alluvial deposits, and it is categorized into two main layers (Pliocene, Pleistocene). It largely consists of sands and gravels at the bottom, with clay–silt deposits at the top, each with its own hydraulic properties. The Nile valley's Quaternary aquifer is thickest (about 400 m) at Sohag. The smallest thickness (about 30 m) is located in Maghagha in north El-Minya (Ghodeif et al. 2016).

Ghodeif et al. (2016) showed that since the river bed penetrates the clay cap, the Nile River and the surrounding aquifer have a significant connection in upper Egypt (Figure 5).

Both observation and abstraction wells were drilled, and sediment samples were taken approximately every 1 m, with sampling ending at 36 m. Table 2 shows the type of sediment layers.

The Nile River and groundwater levels at Sohag were continuously monitored by the irrigation ministry. SCWW measured the water level in the abstraction and observation wells. The water table was referenced to the World Geodetic System (WGS 84). Additional manual water level measurements were recorded by water level meter, which is used to verify the measured level and levels from the model at the surface water intake, abstraction wells, and observation well.

Water samples from the Nile River, abstraction wells, and groundwater were collected regularly following the Egyptian Higher Committee for Water (EHCW 2007), significant parameters including nitrate, ammonia, iron, and manganese were analyzed. Water samples from the Nile River and abstraction wells were analyzed at the local waterworks laboratory in the west Sohag WTP by SCWW. Groundwater was analyzed from different wells in Sohag by the ministry of irrigation.

Based on detailed hydrogeological, site visit, sediment layer, and aquifer data, a 3-D conceptual model of the riverbank filtration site was established. The model consists of four layers: an upper unconfined layer, three leaky confining layers. Unconsolidated coarse sand and gravel dominate the aquifer, with clay lenses occurring rarely. Therefore, the aquifer is subdivided in the vertical direction into four layers assuming that the depth of each layer is constant; the top layer is a 7-m-thick clay cap, the second is fine sand with a thickness of 8 m, the third a moderate sand layer with a thickness of 5 m and the fourth layer is coarse sand and gravel with a thickness of 30 m, which was assumed to be deeply extended to the impermeable layer. The model assumptions are that this aquifer is homogeneous.

The modeled area is about 1.913 km². The model's boundary conditions and extensions are constructed depending on Figure 2. Vertically, the model is divided into 4 layers. Horizontal grid spacing ranges from 0.5 m near the well to 10 m around the model's sides. Negative discharges of −35 L/s (−3,024 m³/d) are used to represent the abstraction well using screen depths from 45 to 27 m. The distance between the two vertical abstraction wells is about 10 m and around 8 m is the distance between the Nile bank and vertical abstraction wells; this short distance minimizes the impact of groundwater ratio on the infiltrated water. The abstraction wells are expected to lower the groundwater level, allowing a significant portion of Nile water to flow into the wells. At the western boundary, the general head was defined along the west Nag-Hammadi irrigation canal with a level of 59 m and conductance of 2.10 m²/d. No flow boundaries are applied along the northern and southern sides. The study area is urban so no recharge was assigned in the model. The groundwater system in the Nile Delta and valley aquifer is in a steady-state equilibrium with negligible change in storage, particularly after the Aswan High Dam construction. Therefore, the steady-state case was adopted to simulate the groundwater. There are two scenarios based on the eastern boundary for the Nile River.

• A.

  The Nile River is considered as a constant head boundary with a water level of 57 m; in this scenario, the Nile doesn't have a clogging layer, and the Nile River and the surrounding aquifer have a significant connection (Figure 6(a)).

• B.

  The Nile River is considered as a head-dependent boundary/river package (Figure 6(b)). The Nile has a clogging layer with an assumed thickness of 0.2 m, k_clogging 0.085 m/d, and conductance of 7 m²/d (Bartak et al. 2015; Grischek et al. 2018).

Different hydraulic properties' values for the aquifer in west Sohag were collected from the literature; Table 3 shows the properties of each layer. MODPATH expects the legitimacy of Darcy's law and the conservation of masses. An assumption of an effective porosity of 0.30 was used for all layers.

Water levels for the Nile River, abstraction wells, and observation well were monitored from Jun 2019 to Feb 2020 (Figure 7). The levels depend on the elevation of the ground surface (WGS84), as mentioned before. The Nile water levels range from 57 to 54 m, the static groundwater level is about 59 m next to west Nag-Hammadi irrigation canal and decreases until arriving at the Nile; abstraction well water levels range from 56.2 m to 53.5 m. The water level of the West Sohag irrigation canal in the study area is 59 m, which is controlled and unaffected by the Nile River stage. The aquifer and the Nile have similar water level fluctuations over time, suggesting a significant hydraulic connection.

First, the simulation of the groundwater system in west Sohag was built without the abstraction wells to obtain the initial head to use in the two scenarios. The results of the model showed that the groundwater flow direction in west Sohag is from west to east as a result of the groundwater level and the West Sohag canal water level is higher than the Nile water level, where the Nile River works as a natural drain (Figure 8). In general, rivers in arid areas recharge the aquifer. But, in the Sohag area, irrigation water is the primary cause of aquifer recharge in the Nile Valley. So, the Nile River's water levels are usually lower than groundwater levels. There are some exceptions in places where hydraulic structures are present (Ahmed 2009).

Second, abstraction wells were implemented in the model for the two eastern boundary condition scenarios. For scenario A, groundwater levels have not changed much and the influence zone is slight, which makes the infiltrated water from the riverbank higher than from groundwater, therefore this simulation is considered ideal for the RBF conditions (Figure 9). For scenario B, groundwater levels have not changed much but the influence zone is large (Figure 10), which means the infiltrated water from the riverbank may be lower than the groundwater; therefore, this simulation and conditions are considered not ideal but more real for the RBF conditions.

Table 4 shows the measured and simulated groundwater levels. The results of Scenario B and the on-site measurement are almost similar, which makes scenario B more valid than scenario A.

Flow budget in GMS is used to obtain the riverbank filtration ratio in the abstraction wells, where the connection between the aquifer and the river is explained through this ratio. The groundwater flow budget for the study area was performed to get the inflow and outflow from the model in addition to calculating the percentage of riverbank filtration (Figure 11). The abstraction rate from wells is 6,048 m³/d. For Scenario A, the inflow from the constant head is 5,129 m³/d and the inflow from GW is 919 m³/d. For scenario B, the inflow from the general head (river package) is 2,416 m³/d and the inflow from GW is 3,632 m³/d.

The particle tracking technique was used to establish pathlines of river water in the aquifer to show and verify infiltration into the aquifer, and flow toward the abstraction wells, as well as to estimate the travel time. To decide which particles are collected by the abstraction wells from the river, a forward tracking technique was used. Particles were placed along the border of the Nile. After making the simulation, some particles were captured by abstraction wells making the capture zone of the well from the river (Figure 12). For scenario A, the Nile River and the surrounding aquifer have a significant connection. Therefore, the capture zone and pathlines are small compared to scenario B, where the connection between the Nile and aquifer isn't good making, the capture zone and pathline extended. A new simulation was performed to show the mixing water between the Nile and groundwater by using backward flow paths of particles (Figure 13).

Capture zone simulation was performed with specified travel times of 1, 5, 10, 20 and 50 days (Figure 14). This simulation is very important to make a buffer zone around the RBF system depending on the pollutant lifetime and mitigating the effect of anthropogenic activities in the vicinities. It is evident that the abstraction well capture zone occurs less than one day after pumping because the pathlines cross the region occupied by the Nile River in west Sohag, indicating that a correlation between the Nile River and the abstraction wells occurred within one day. The influenced area for scenario A is less than scenario B for all travel times, and therefore will affect the buffer zone.

Six hydraulic simulations were carried out to evaluate the performance of the RBF system in the study area using six abstraction rates (10, 20, 30, 40, 50, and 70 L/s) for the two mentioned scenarios. The number and properties of wells in each simulation and all hydrogeological properties are constant. Figure 15 shows the relation between the abstraction rate and riverbank filtration ratio RBF% where abstraction rate increased the RBF%, which mean that the abstraction rate noticeably influenced the filtration efficiency. For scenario A, increasing the abstraction rate increases the RBF% but not as much as the rate of abstraction so, in case of a strong connection between the aquifer and the river, it's not favorable to increase the abstraction rate. For scenario B, increasing the abstraction rate increase the RBF% and there is a proportional relationship between them so, in case the connection between the river and aquifer isn't good, it's favorable to increase the abstraction rate.

Figure 16 shows the relation between the abstraction rate and minimum travel time where the abstraction rate increases, the travel time decreases, and this decrease in travel time will directly affect filtrated water quality. Increasing the abstraction rate above 35 L/s doesn't significantly affect travel time. Figure 17 shows the relation between the abstraction rate and the farthest distance between the abstraction wells and the riverbank particles where the abstraction rate increases, the distance increases, and the influence zone of the abstraction well will be extended.

The RBF system's water quality was tracked for ten months (7/2019–5/2020), as well as the Nile River and groundwater, over three years of sampling (2016–2019). Table 5 shows the results of water quality assessment depending on the physio-chemical and bacteriological characteristics of water.

Except for turbidity, HPC, Fecal Coliform, and total coliform, the Nile River's water quality parameters are within limits and follow Egyptian standards (EHCW 2007). The high load of turbidity in the Nile water is due to suspended organic and non-organic material from bed scour, bank erosion, or during flash flood, where the Nile act as a drain. The high concentration of bacteria may be due to agricultural waste or household raw sewage.

Except for ammonia, all groundwater quality parameters are within the limits and matched with the standards of Egypt (EHCW 2007). The high ammonia content may be related to the sewerage system (latrines and septic tanks) surrounding the abstraction wells, which are about 60 meters away from the septic tanks or seepage of nitrogen fertilizer-containing irrigation water.

Conservative chemical concentration could be used to estimate the bank filtration ratio such as temperature, EC, and chloride. In this study, EC was used to estimate the RBF ratio and to validate the model results of RBF % for both scenarios A and B using Equation (2) where C_aw is the concentration in the abstraction wells, C_Gw is the concentration in the groundwater, C_R is the concentration in the river and RBF_ch % is the bank filtration ratio using conservative chemical concentration. The estimated RBF_ch % is about 56.5%.

The hydraulic connection between the Nile and the aquifer was investigated in this study to evaluate the hydraulic efficiency of the RBF system in west Sohag. The results show that the Nile and aquifer have a strong hydraulic connection, and the riverbank filtration ratio is high (40–85 percent). Water quality changed between the Nile raw water and the bank filtrated water in the abstraction wells where microbial and chemical results analysis demonstrated the RBF system's efficiency. Hardness, nitrate, sulfate, TDS, and turbidity were all reduced to within acceptable limits. More study is needed to determine the efficiency of RBF for iron and manganese removal. Higher manganese concentrations in the abstraction wells can require additional treatment. The distance of 8 m between the Nile and the abstraction wells was found to be suitable for receiving a large share of the Nile water while reducing or removing contaminants and suspended matter. The capital and operation cost of this RBF system comparing to conventional treatment and municipal wells is lower. Finally, the RBF system in Upper Egypt has proven that it is a cheap, successful, and safe way for water supply without the need for any additional treatment, only disinfection, or as primary treatment in a specific condition.

All relevant data are included in the paper or its Supplementary Information.