Numerical and field investigation to improve an existing distributed-sediment-excluder sand trap performance

Sand traps play a crucial role in systems requiring an uninterrupted water supply, such as run-of-river hydropower plant intakes or high-demand irrigation schemes without a pumping system. Positioned downstream of irrigation intakes, sand traps serve to deposit and flush sediment before it reaches the primary irrigation channel. The mean flow velocity within these traps is a critical parameter, influencing the size range of sediment that can effectively deposit within the sand trap (Bouvard 1992). Sand traps generally consist of an upper rectangular area and a lower trapezoidal sediment deposition space. Sediment removal occurs through a sediment excluder positioned at the base of the deposition area, while deposited sediment can be flushed away via a dedicated flushing canal (Schleiss 2012). The sediment within the sand trap is flushed by a continuous system that is located either at the downstream end of the trap or via distributed sediment scour holes in the sediment excluder. Efficient operation of sand traps relies on the maximum flow velocity and turbulence.

This study investigates the velocity distribution and suspended sediment concentration distribution of the Dufour Type II (Dufour 1954) sand trap at the Lower Usuthu Smallholder Irrigation Project (LUSIP) in Eswatini. Dufour sand traps, prevalent in Europe (Swiss, French, and Italian Alps) and other regions, utilize a sediment excluder flushing conduits system along the base of the trap to manage sediment accumulation (Bouvard 1992). The Dufour sand trap needs a significant water volume for sediment flushing. These sand traps feature internal sediment excluders comprising multiple parallel canals with sloped bottoms and distributed scour outlet ports. Sediment scours through these scour outlet ports into a flushing conduit situated along the bottom of the sand trap.

Previous physical studies aimed at enhancing the operation and flushing efficiency of Dufour sand traps include research on various systems and innovations. Notable examples are the HSR system (Truffer et al. 2009), the Serpent Sediment Sluicing System (4S) (Lysne et al. 1995), the Slotted Pipe Sediment Sluicer, and the Saxophone Sediment Sluicer (Jacobsen 1999). Additionally, Daneshvari et al. (2012) investigated the hydraulic efficiency of the Dufour sand trap at the Mörel HPP in Fiesch, Switzerland, and suggested an improved sand trap flushing system that utilized mobile cylinders and probes. The study focused on determining the required flushing velocity within the flushing conduit for fully suspended sediment transport to occur using flow-3D and ANSYS CFX hydrodynamic models. It emphasized fluid flow characteristics, employing a purely hydrodynamic modeling approach.

Sawadogo (2015) developed a coupled, fully 3D numerical model designed to simulate turbulent suspended sediment transport processes and bottom outlet flushing in reservoirs. The numerical model was tested and validated using experimental data from various laboratory studies, including the work of Jobson & Sayre (1970), Van Rijn (1981), and Ashida & Okabe (1982). This study utilizes the numerical model developed by Sawadogo (2015), to investigate the flow velocity and suspended sediment concentration distribution in the LUSIP sand trap.

The primary objective of this study is to evaluate the trapping and flushing efficiency of the sand trap through comprehensive field investigations and to analyze the flow velocity and suspended sediment concentration distribution within the sand trap using advanced numerical modeling. The modeling efforts encompass both maximum design discharge and low-flow conditions. Based on these analyses, the study aims to propose technical design recommendations to enhance the performance and operational efficiency of the sand trap. A notable innovation of this research lies in its capacity to provide comprehensive insights into three-dimensional flow velocity patterns and suspended sediment concentration dynamics within a distributed-sediment-excluder sand trap. These findings aim to inform design improvements that optimize sediment management practices in irrigation infrastructure.

LUSIP addresses eSwatini's socio-economic and environmental challenges, focusing on poverty reduction, sustainable water resource management, and agricultural development (Intecsa-Inarsa 2006). By providing rural communities with access to irrigation water, LUSIP fosters economic growth, enhances food security, and promotes environmental sustainability. Sediment-laden water from the Usuthu River is diverted at the Bulungapoort River diversion works and flows through a feeder tunnel into the sand trap. Beyond the sand trap, water is conveyed through a 21 km concrete-lined canal to three downstream dams for agricultural use.

The Usuthu River, a vital watercourse, experiences varied climatic conditions, with Eswatini's subtropical climate bringing wet summers (October to March) and dry winters (April to September). Rainfall ranges from 500 mm in the Lowveld to 1,200 mm in the Highveld, with peak river discharge during the rainy season often accompanied by sediment and debris-laden floods. The Usuthu River sub-catchment, currently experiences significant sediment yield, with a long-term estimated effective sediment yield of 304 t/km²/a. Projections for the year 2070 indicate that future land use changes, particularly deforestation and the expansion of arable land, could increase sediment yield to 412 t/km²/a, representing a 36% increase (Basson & Sawadogo 2018). This anticipated rise in sediment load underscores the importance of effective sediment management strategies, such as optimizing the performance of sand traps, to mitigate the impacts of increased sedimentation on irrigation infrastructure and freshwater ecosystems in the region.

According to the eSwatini Water and Agricultural Development Enterprise, the sand trap achieves 40% effectiveness in trapping and flushing sediment, with 2 mm sand particles being detected in the downstream feeder canal. As a result, manual cleaning is required monthly for approximately 3 days during the dry season and twice a month during the rainy season. A properly functioning sand trap is essential for effective sediment management, safeguarding irrigation infrastructure, and boosting agricultural productivity. By providing consistent water quality and quantity, a functional sand trap supports the economic empowerment of smallholder farmers, enabling them to transition from subsistence to commercial farming. This creates employment opportunities along agricultural value chains, from farming to processing and export. A functional sand trap demonstrates the importance of sediment management in large-scale irrigation projects, encouraging policymakers to prioritize sediment control in future projects. This can lead to the development of policies promoting sustainable water and land management practices. The sand trap's role is vital for supporting socio-economic development, guiding policymaking, and contributing to the achievement of Sustainable Development Goals (Table 1), promoting rural development, climate resilience, and a sustainable future for Eswatini.

The Avio gate is a specialized hydro-mechanical gate designed for automatic downstream control of irrigation canals. Its primary function is to maintain a constant downstream water level by automatically adjusting its position based on changes in the flow conditions (Hydrostec 2019). Although the gates require regular maintenance, including lubrication, they are highly suitable for challenging environments, such as those found in developing countries or remote locations. The gate's operation is highly sensitive to downstream water levels, therefore, if the downstream water level is not properly regulated or if significant flow changes occur, the gate's response might not be as precise as required, leading to suboptimal flow management.

The Avio gate ensures minimal water level variation within the sand trap despite fluctuations up to 8 m in the Lower Usuthu River, thereby maintaining a constant head on the three manually controlled flow regulating gates. These gates, located immediately downstream of the sand trap, are used to set the required flows into the Feeder Canal (Intecsa-Inarsa 2006). The downstream vertical gates and upstream Avio gate work in tandem to control the flow within the sand trap.

Observations conducted during the dry flow season indicate that turbulence within the system is primarily generated by the operation of the Avio gate. The gate's mechanism causes fluctuations in water velocity and direction, leading to localized turbulence that affects flow dynamics. However, it is important to note that these observations were limited to the dry season. Seasonal variations in flow characteristics and sediment load transport into the sand trap are expected, and these differences could significantly influence turbulence patterns and sediment dynamics across different times of the year. Consequently, further observations during other seasons would provide a more comprehensive understanding of the system's behavior.

A United States Geological Survey (USGS) depth-integrated sampler was used for the fieldwork sampling, featuring a 500 mL bottle inserted in a fish-shaped frame mounted on a rod. During operation, the sampler is systematically lowered from the water surface to the bed and back up, maintaining a continuously uniform sampling rate against the flow direction. The intake nozzle ensures that the intake velocity is nearly equal to the local stream velocity. Once filled, the sampler is removed from the water flow, and its contents are carefully transferred into another sampling bottle for further analysis.

The total suspended sediment (TSS) samples were collected at multiple locations throughout the trap (refer to Figure 1) to assess suspended sediment concentration gradients from upstream to downstream. The TSS sample before the trap outlet should exhibit a very low concentration, indicating that clear, sediment-free water passes through the trap to the feeder canal.

A Van Veen bed grab (BG) sampler was used to collect samples of the upper sediment layer on the streambed. These BG samples, weighing approximately 800 g in dry weight, underwent drying, weighing, and subsequent sediment size analysis. This sediment size distribution analysis included both dry sieve analysis (according to ASTM C136/C136M-19 2019) and hydrometer analyses (according to ASTM D7928 standard test method 2020), conducted by a South African National Accreditation System accredited laboratory ensuring the accuracy of the results.

After drawdown flushing, deposited BG samples were collected to determine the median sediment particle sizes deposited across the trap's length using hydrometer analysis. At chainage 1,225 m (within the effective settling zone), no sediment sample could be collected due to only twigs being present at this location. An integrated sample was taken at locations with a high sediment accumulation (chainage 1,235 and 1,245 m) to represent a more accurate median sediment size.

During the low flow of 5 m³/s observed in the field, the trap effectively deposited sediment particles with a median diameter of approximately 0.32–0.95 mm within the sediment deposition area. Sediment particles with a median size of 0.129 mm were deposited directly upstream of the outlet gates, while particles of 0.106 mm were deposited downstream of the sand trap outlet gates in the feeder canal. These results indicate that the sand trap successfully deposits sediment with a as fine as 0.129 mm under low-flow conditions, though particles with a of 0.803 mm escaped the sand trap, approaching the design limit of 1 mm. Thus, at the maximum flow rate of 15.5 m³/s, particles larger than 1 mm are expected to pass through the sand trap without being trapped.

In Southern Africa, sediment particles larger than 0.075 mm are classified as coarse-grained sediments and are non-cohesive, while those smaller than 0.075 mm are fine-grained sediments (silt and clay particles) and are cohesive. Clay-sized cohesive sediments are finer than 0.004 mm, and silt-sized cohesive sediments can be finer than 0.075 mm (Shrestha & Blumberg 2005). The cohesion forces of deposited sediments containing silt and clay can influence the physical properties of sediment transport, such as flocculation, sedimentation and compaction rate, erosion resistance, and the angle of repose (φ) (Simons & Senturk (1977) and Grabowski et al. (2011)). Cohesive sediments are characterized by their inherent ability to bind together due to the dominance of attractive forces over repulsive forces. Typically, sediment is considered cohesive if its clay fraction exceeds approximately 10% (Shrestha & Blumberg 2005).

Sediment particles with a of 0.97 mm were found inside of the sediment excluder conduit, just upstream of the flushing gate, demonstrating the excluder conduit's efficiency in flushing these particles. Samples taken at chainage 1,255, 1,265 m, and the downstream feeder canal contained more than 10% silt and clay, indicating that these sediment samples are cohesive. Once deposited, these cohesive sediments compact, making them difficult to be eroded and flushed out of the sand trap through the distributed scour outlet ports.

The designed scour outlet ports at the bottom of the trap are 0.1 m in diameter and spaced 1.5 m apart from chainage 1,220–1,255 m. According to the design of trash racks as outlined in Basson et al. (2020), the largest sediment particle capable of entering the flushing conduit should not exceed 2× the openings of the trash racks at the diversion works. However, the organic plant debris entering the sand trap through the trash racks (vertical bars with 50 mm spaced openings), located at the river diversion works, were too large to flush through the scour outlet ports, causing blockage and subsequent sediment build-up (Figure 5). Consequently, the deposited sediment reduces the cross-sectional area, hindering the trap's ability to deposit fine sediment due to increased velocities, and ultimately causing sediment to escape the sand trap through the downstream gates into the feeder canal.

Agricultural practices in the Usuthu River basin influence the types and quantities of debris entering irrigation systems. Vegetation contributes to organic debris, including crop residues, branches, and plant matter. During land preparation, harvesting, or post-harvest activities, debris from fields can wash into the river during flood events and subsequently into sand traps. Installing more robust trash racks with smaller gaps or horizontal bars can help capture larger debris before it enters the sand trap. However, this may require frequent maintenance and cleaning to avoid blockages.

The application of computational fluid dynamics (CFD) methods holds significant promise for predicting sediment transport in a sand trap, especially when turbulence effects pose challenges (Van Wachem & Almstedt 2003). In this study, the sand trap was numerically analyzed using a fully three-dimensional CFD model developed by Sawadogo (2015), implemented in ANSYS Fluent 2021. This numerical model is coupled in terms of hydrodynamics and sediment transport.

The coupled fully three-dimensional numerical model was used to investigate the flow velocity and suspended sediment concentration distribution within the trap under various conditions, including the design inflow, flushing, and outflow discharges. The model does not simulate changes in the bottom level during the simulation due to sediment settlement, which would alter the consequent flow area. Instead, the model provides a concentration distribution of sediment throughout the trap. The velocity and sediment concentration distribution throughout the sand trap were simulated for both the maximum design discharge and the discharge measured during the field measurements (low flow). The velocity magnitude results of the numerical model during low flow were then qualitatively compared to the field observations. Furthermore, the numerical model was used to investigate possible modifications in the design of the sand trap aimed at enhancing the performance of sediment deposition and flushing by improving the flow velocity distribution.

ANSYS Fluent, a CFD simulation software, uses the finite volume method to model fluid flow in complex geometries (Fluent 2013). The interaction between fluid and sediment represents a coupled problem, as sediment transport alters the flow and affects the sediment bed in terms of roughness, elevation, and slope (Ahadi et al. 2020). Fully coupled numerical models that integrate hydrodynamic flow and sediment transport properties can simultaneously simulate the velocity field and sediment concentration. ANSYS Fluent includes a default hydrodynamic model and a User Defined Function (UDF), which allows users to modify material properties and boundary conditions through inputting customized code (Fluent 2013). The UDF can also define the sediment transport equation and sediment boundary conditions at the bed.

Sawadogo (2015) developed a coupled fully three-dimensional numerical model based on the Navier–Stokes equation, incorporating both hydrodynamic and suspended sediment transport parameters, to investigate the scour pattern caused by bottom outlet sediment flushing. He adopted the same approach that was modeled by Schneiderbauer (2010) for the aeolian snow transport. In this approach, rather than the default Fluent multi-phase model, solids transport is calculated as a passive scalar or volume of sediment concentration.

The turbulent Schmidt number for very fine sand was determined by Celik & Rodi (1988) to be 0.5.

The UDF within ANSYS Fluent was used to define the sediment transport equation and the sediment boundary conditions at the bed, using formulas that link both the rate of entrainment and deposition with the settling velocity and the reference concentration.

Sawadogo (2015) validated this model in ANSYS Fluent by simulating a range of suitable experimental cases involving net entrainment from a fixed and loose bed, as well as net deposition on fixed beds, based on laboratory flume experiments conducted by Jobson & Sayre (1970), Van Rijn (1981), and Ashida & Okabe (1982). The model's validation involved prescribing both velocity and concentration profiles at the inflow boundary and predicting the downstream development of concentration. The numerical model developed by Sawadogo (2015) effectively simulated sediment concentration and flow velocities within a case study conducted on a sand trap with a downstream gate outlet (Mc Leod 2024). As a result, the proposed numerical model can be used to predict both flow velocity distribution and turbulent suspended sediment transport processes in sand traps.

The sediment particle density, concentration, and settling velocity were specified for the sand trap model, while standard material properties were sourced from ANSYS Fluent's database. Operating settings included activating gravitational acceleration and referencing the atmospheric pressure at the outlets. The default solution methods suggested by ANSYS Fluent were selected to ensure computational stability for the transient pressure-based solver for incompressible flows. This includes the Phase Coupled SIMPLE scheme for pressure-velocity coupling alongside the first order implicit transient scheme and the second order upwind spatial discretization schemes. A standard initialization relative to the cell zone was chosen. Each simulation was configured with a number of time steps sufficient for a water particle to traverse through the trap, and 10 iterations per time step were selected to ensure convergence.

For each simulation, the velocity distribution across different planes in the trap and sediment concentration distribution were investigated. The settling velocity of a specific sediment particle size was calculated following Van Rijn's equations (1987), and an inlet concentration was introduced to the model by using UDFs.

Due to the absence of site-specific suspended sediment concentration or load data for the study area during this investigation, the recommendations by Bosman et al. (2002) were utilized. They suggest suspended sediment concentrations of 10,000 mg/L for normal flow conditions in Southern African rivers without available records, and up to 50,000 mg/L for non-cohesive sediment during extreme flood events. Additionally, Basson (2006) recommends that a sand trap should be designed to handle 1% of the total river sediment concentration during normal operations and up to 10% during flood events. Consequently, a conservative inlet concentration of 1,000 mg/L, representing typical non-cohesive sediment levels under normal conditions, was adopted for this investigation.

The primary function of the LUSIP sand trap is to settle suspended particles greater than 1 mm in diameter. To obtain an indication of the sediment concentration distribution within the sand trap, a uniform sediment particle size of 1 mm (with a settling velocity of 0.14 m/s, calculated according to the equations recommended by Van Rijn (1987)), a particle density of 2,650 kg/m³ (Simons & Senturk 1977), and an inlet concentration of 1,000 mg/L, was introduced to the model using the UDFs.

The inflow discharge was set to 15.5 m³/s to represent the maximum discharge for which the sand trap was designed to settle suspended particles greater than 1 mm in diameter. The downstream outlet gates of the trap were opened, releasing a total of 13.5 m³/s, with 2 m³/s flowing through the scour outlet holes.

The observed spiral flow configuration likely results from the Avio gate's floatation chamber partially deflecting the central inflow sideways, as confirmed by the plan view's flow lines. The velocity becomes reasonably uniform halfway along the settling trap length, ranging between 0.3 and 0.5 m/s. Additionally, Figure 10 highlights high surface velocities contributing to turbulence in the flow.

The velocity through the distributed scour holes within the flushing conduits, seen in the Trap plane (2) in Figure 9, increases from the downstream side to the upstream side of the trap due to the negative slope of the flushing conduits. This effect is also evident in Figure 10, Section B, showing high velocities in the flushing conduit toward the flushing pressure outlet.

The sediment concentration displays an interesting pattern within the trap. High sediment concentration accumulates along the side walls due to upward flow (as depicted in Figure 11), whereas lower concentrations are observed where downward flow predominates in the middle (indicated in Figure 11) and where spiral flow occurs in between (inferred from the flow lines in the vertical plane above the settling trench in Figure 9). At the outflow of the trap, higher concentrations are also evident along the sides, indicating sediment concentration escapes through the downstream outlet gates of the sand trap where the velocity increases sharply as observed in Figure 9.

The numerical model's simulation results, as shown in Figure 11, reveal that the effective settling length for the sediment in the higher concentration zones (near the side walls) extends to the end of the settling trench where concentrations decrease close to zero. This suggests that turbulence induced by the Avio gate's floatation chamber creates a complex three-dimensional flow pattern, contributing to the trap's inefficiency in sediment capture. Despite these limitations, the sand trap demonstrates an ability to retain a significant portion of the sediment, although some sediment bypasses the system through the downstream outlet gates. Quantitatively, the trap retains 35% of the sediment, while 9% escapes through the outlet and 57% is flushed out. Overall, the results show that the sand trap performs effectively for 1 mm particles with the given inlet concentrations. However, the sediment build-up observed in the field may be attributed to the deposition of cohesive sediments in the trap and debris obstructing the scour holes.

The velocity pathlines for the 5 me³/s flow exhibit a similar pattern to those observed at 15.5 m³/s, albeit with lower velocity values and more pronounced upward flow on the side plane (3). The average velocity throughout the settling trap length is 0.2 m/s. The spiral flow pattern is also evident in both the plan view (Section C) and cross-sectional views (Sections A and B) shown in Figure 14, consistent with observations from the field observation in Figure 4. The velocity magnitudes at the top of Sections A and B align closely with the PIVlab results obtained in Figure 4, being in the order of 0.3–0.5 m/s.

Further investigation revealed a design limitation for the sand trap for flows below 10 m³/s. While the trap effectively reduces flow velocity to facilitate sediment deposition, high concentration accumulating within the flushing conduits hinders effective sediment removal from the sand trap. As a result, sediment builds up within the trap due to inadequate flushing and ultimately escapes through the downstream outlet.

Removing the Avio gate at the inlet reduces the spiral flow and turbulence, concentrating the velocity more uniformly along the centreline. The velocity gradually decreases through the inlet area and maintains uniformity as it progresses toward the settling trench. Within the flushing conduits, the velocity remains sufficiently high for efficient sediment flushing. Straightening of the velocity distribution could be achieved by adding buffer plates at the inlet area.

The simulated concentration distribution of the uniform 1 mm grain size sediment decreases along the length of the sand trap, indicating sediment deposition. With the Avio gate removed, the disturbance is minimized, allowing the sediment to settle, resulting in higher sediment concentration at the bottom of the trap extending throughout its depth. The outlet weirs effectively retain the deposited 1 mm sediment within the trap, preventing it from escaping into the downstream feeder canal, and improving the trapping efficiency over the initial bottom outlet design. Minimal sediment concentration is observed at the outlet section of the sand trap, suggesting scouring through the larger distributed scour outlets into the flushing canal, improving the flushing efficiency of the sand trap.

The flushing conduits maintain a high velocity to efficiently transport the sediment load toward the upstream flushing gate. To enhance sediment flushing efficiency, the installation of fine screens at the river diversion works is recommended to prevent blockage of the 200 mm scour holes by debris. With the proposed modifications, the sand trap's trapping efficiency is projected to increase to 85%, while flushing efficiency is expected to reach 80% under low-flow conditions.

In conclusion, this study aimed to evaluate the trapping and flushing efficiency of the LUSIP sand trap through field investigations and advanced numerical modeling, with a focus on both maximum design discharge and low-flow conditions.

The results of the fieldwork measurements and observations show that during the low-flow season, turbulent flow at the trap's inlet, caused by the Avio gate, reduced the effective settling length of the sediment. As a result, the trap was only able to effectively deposit particles around 0.32 mm (⁠⁠) during low-flow conditions. Flushing efficiency was further compromised due to debris blockages. The sediment samples collected revealed a high percentage of clay, indicating the presence of cohesive sediment that compacted in the sand trap and did not flush out through the distributed scour outlet ports. Under these low-flow conditions, the sand trap exhibited a trapping efficiency of 27% and a flushing efficiency of 36%.

The numerical investigations show that the sand trap settles and flushes 1 mm sediment particles for the maximum design discharge, retaining 35% of the sediment, with 9% escaping through the outlet and 57% being flushed out. Turbulence from the Avio gate reduces efficiency, but under an inlet concentration of 1,000 mg/L, the trap achieves a trapping efficiency of 90% and a flushing efficiency of 57%. The numerical investigation of the sand trap during low-flow conditions shows that sediment is deposited along the settling trench, with concentrations increasing in the flushing conduits, hindering effective flushing. The trapping efficiency was 29%, and the flushing efficiency was 48%, which aligns with field observations. Further analysis revealed a design limitation for flows below 10 m³/s, where high sediment concentrations in the flushing conduits prevent efficient sediment flushing. As a result, sediment accumulates in the trap and escapes through the downstream outlet, highlighting the need for design improvements to enhance flushing efficiency.

Modifications made to the model, including removing the Avio gate and adjusting outlet components, notably improved the efficiency of the trap at lower discharges. With the proposed modifications, the sand trap's trapping efficiency is projected to increase to 85%, while flushing efficiency is expected to reach 80% under low-flow conditions.

Despite the absence of quantitative data for high flow velocity and suspended sediment measurements, the study leverages the predictive capabilities of the numerical model to simulate velocity distribution for the maximum design and low-flow discharges as well as suspended sediment concentration distribution patterns within the sand trap for the sediment concentration inlet conditions. These insights provide a foundation for proposing targeted design modifications to improve sediment flushing efficiency and optimize the trap's performance. This work contributes to advancing sediment management practices and supports the sustainable operation of hydraulic infrastructure systems in semi-arid regions in Southern Africa.

It is recommended to conduct further investigations into the trapping and flushing efficiency of the sand trap under maximum design discharge conditions, utilizing actual sediment concentration data measured in the field and advanced velocity measurements at the scour holes. Additionally, studies should focus on evaluating the efficiency of real sediment loads during maximum and low-flow discharge conditions. Furthermore, it is recommended that the sand trap be numerically investigated with smaller sediment particles to better understand the cohesive effects, as these particles are more difficult to flush, once settled and compacted, and may significantly impact the overall efficiency within the trap.

The authors would like to thank Ousmane Sawadogo for making their numerical model available for use in this study.

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

The authors declare there is no conflict.