Sediment Transport and Morphodynamics

Physics-Based Graded-Sediment Modelling for Room for the River 2.0

This project develops a state-of-the-art, physics-based modelling framework for understanding sediment sorting and morphodynamic evolution in mixed sand–gravel riverbeds. The research responds directly to the knowledge needs identified within Rijkswaterstaat's Room for the River 2.0 programme, particularly the need to improve prediction of persistent riverbed erosion in the River Rhine.

Extensive studies conducted during the first Room for the River programme provided a strong basis for system-scale morphodynamic analysis. However, they also demonstrated that conventional operational modelling concepts do not represent the underlying sorting processes in mixed sand–gravel beds with sufficient physical realism. This limits their ability to resolve critical questions concerning grain-size-selective transport, spatial variations in bed composition and long-term riverbed erosion.

Room for the River 2.0 therefore requires more detailed, physics-based approaches that go beyond the processes represented explicitly in conventional operational models. My research addresses this requirement by extending a high-resolution LES–particle– morphology framework from uniform sediment to graded sediment containing multiple grain-size fractions.

• Three-dimensional Large Eddy Simulation
• Discrete Lagrangian sediment particles
• Multiple sediment grain-size fractions
• Mixed sand–gravel riverbeds
• Grain-size-dependent sediment entrainment
• Size-selective transport and deposition
• Hiding and exposure effects
• Particle interlocking and grain-scale interaction
• Vertical and horizontal sediment sorting
• Active-layer and subsurface-layer administration
• Bed armouring, coarsening and fining
• Dynamic dune and bedform development
• Feedback between turbulence, sorting and morphology
• Validation against graded-sediment laboratory experiments
• Application to sediment mixtures representative of the River Rhine

A Digital Flume for Fundamental Process Research

One of the principal objectives is to use the model as a high-resolution digital flume. This creates a controlled numerical laboratory in which sediment-transport and sorting processes can be isolated, observed and quantified in considerably greater detail than is possible in operational river-scale models.

The digital flume can be used to test the physical assumptions and empirical formulations currently employed in conventional morphodynamic models. Existing relations can be evaluated against detailed grain-scale simulations to determine where they remain valid, where they require modification and which physical processes are missing from their present formulation.

The detailed simulation results can also be used to derive new parameterised relations for processes such as hiding and exposure, size-selective entrainment, differential deposition, vertical exchange, bed sorting and armouring. These relations can subsequently be incorporated into large-scale operational models without requiring those models to resolve every sediment grain explicitly.

Upscaling on the Snellius Supercomputer

The present model has demonstrated promising graded-sediment behaviour at laboratory scale. The next major step is to upscale the physical and computational domain and investigate increasingly realistic sediment mixtures and river configurations.

These simulations require national high-performance computing resources because the framework combines three-dimensional LES, discrete particles, several sediment fractions, dynamic morphology and long morphodynamic integration times. The model is therefore being prepared for large, tightly coupled MPI simulations on the Dutch national supercomputer Snellius.

Snellius provides the computing capability needed to move beyond laptop- and laboratory-scale simulations, conduct systematic validation and sensitivity studies, and explore the computational requirements of semi-realistic River Rhine applications.

Long-Term Exascale Ambition: JUPITER

The longer-term ambition is to develop the framework towards simulations on JUPITER, Europe's first exascale supercomputer. Reaching this level requires further development of MPI scalability, memory-efficient particle administration, GPU acceleration, parallel input and output, and algorithms suitable for very large computational domains.

JUPITER represents a pathway towards simulations that retain detailed grain-scale and turbulence-resolving physics over substantially larger spatial scales. The objective is to investigate how far the current framework can be advanced towards realistic river-scale modelling without losing the physical processes responsible for sorting, armouring and morphodynamic adjustment.

Two Complementary Innovation Pathways

1. High-performance upscaling: develop and run progressively larger simulations on Snellius, with JUPITER as the longer-term exascale ambition for extending the framework towards realistic spatial scales.
2. Digital-flume research: use the detailed model to investigate grain-scale physics, validate existing empirical formulations, identify missing processes and derive improved parameterised relations for operational models.

The purpose is therefore not simply to replace operational river models with a much more computationally expensive research model. The detailed model serves as a scientific instrument for discovering, testing and quantifying the processes that determine sediment sorting and long-term riverbed evolution.

By combining frontier-scale simulation with systematic parameterisation, the research creates a direct route from grain-scale physics to improved predictive capability for the River Rhine and Room for the River 2.0.

Modelling Radioactive Caesium Transport in the Abukuma River

Following the Fukushima Daiichi nuclear disaster, I contributed to a Japanese national research project investigating the transport and redistribution of radioactive contamination through river systems. The project was funded by the Japanese government and the Japan Atomic Energy Agency (JAEA).

The research addressed an issue of major societal importance for Japan: understanding how radioactive caesium deposited on the landscape is mobilised, transported with sediment through rivers, deposited within the river system, and ultimately transferred towards the Pacific Ocean.

Radioactive caesium such as ¹³⁷Cs can become strongly associated with fine surface-soil particles. Consequently, predicting its redistribution requires modelling not only water movement but also the erosion, transport and deposition of the fine sediment carrying the contaminant.

For this project, I worked with a conventional process-based sediment-transport and river-morphology framework rather than the discrete Lagrangian particle model used in my high-resolution sediment research. The computational model combined two-dimensional shallow-water hydrodynamics with conventional washload and suspended sediment transport and an advection–diffusion formulation for sediment-associated ¹³⁷Cs.

• Fukushima environmental contamination
• Radioactive caesium (¹³⁷Cs) transport
• Abukuma River
• Two-dimensional shallow-water modelling
• Conventional sediment-transport modelling
• Suspended-load transport
• Washload transport
• Advection–diffusion modelling
• Sediment-associated contaminant transport
• Erosion, transport and deposition
• Transport of contaminated fine sediment
• River-to-ocean contaminant pathways
• Environmental risk and radiological assessment
• Comparison with observed river behaviour

Environmental and Societal Importance

Understanding the movement of radioactive contamination after the Fukushima accident is important for assessing its long-term distribution in the environment. River systems provide an important transport pathway because contaminated soil and fine sediment can be mobilised by rainfall and runoff and redistributed downstream.

Numerical modelling complements field monitoring by helping to trace these pathways, identify where contaminated material may be transported or deposited, and evaluate its longer-term redistribution through the river system. Such information contributes to environmental and radiological assessment and supports the broader scientific basis for protecting communities affected by the Fukushima accident.

Application to the Abukuma River

The numerical framework was applied to the lower reach of the Abukuma River, one of the major river systems affected by fallout from the Fukushima Daiichi accident. The calculations examined how ¹³⁷Cs associated with washload and suspended sediment is transported through the river and towards the Pacific Ocean.

The research demonstrated how established hydrodynamic and sediment-transport modelling techniques can be extended to investigate a fundamentally different environmental problem: the redistribution of radionuclides associated with mobile sediment.

Location of a) Abukuma River and b) the targeted river reach in present computations. The 3rd Airborne Monitoring Survey Data was used to show the inventory of ¹³⁷Cs on the land surface in a). KP represents the distance from the river mouth in kilometers.

Publication

The work resulted in the peer-reviewed publication: “Computational modeling of ¹³⁷Cs contaminant transfer associated with sediment transport in Abukuma River”, published in the Journal of Environmental Radioactivity.

Physics-Based Modelling of Scour beneath Cohesive Clay

This research addresses a bed-stability problem in the River Meuse in the Netherlands. In parts of the river, the underlying sandy bed is naturally protected by a cohesive clay layer. Under normal hydraulic conditions, this layer provides substantial resistance against erosion and protects the more erodible non-cohesive sediment underneath.

Local changes in hydraulic conditions can, however, compromise this natural protection. Increased flow velocities associated with changing discharge conditions and human interventions along the river can locally damage or remove portions of the cohesive clay cover. Once this happens, patches of the underlying sandy material become exposed directly to the flow.

This creates a fundamentally different erosion mechanism. The exposed sand is considerably more mobile than the cohesive material above it. Turbulent structures can entrain sediment from the exposed patch, initiating a local scour hole that progressively deepens and expands.

Scour beneath the Protective Layer

Of particular importance is the horizontal development of the scour hole beneath the surrounding cohesive layer. As sand is removed from below the edge of the clay cover, the unsupported cohesive material can become unstable. Continued undermining may eventually cause portions of the protective layer to collapse, exposing additional sand and allowing the erosion process to propagate further.

Physical Flume Experiments

Laboratory experiments at Delft University of Technology were designed to investigate this process under controlled conditions. The protected portions of the bed were represented by rigid metal plates, while selected openings exposed the underlying sand directly to the flow.

This configuration isolates the fundamental mechanism: sediment can initially be entrained only through the exposed sandy region, after which the resulting scour hole can develop vertically and horizontally beneath the protected part of the bed.

Why Conventional Morphodynamic Models Are Insufficient

This problem is particularly demanding for conventional sediment- transport and morphodynamic models. The rate and spatial development of scour are controlled by strongly three-dimensional and highly unsteady turbulent structures close to the transition between the protected and exposed bed.

Accurate prediction therefore requires considerably more information than a depth-averaged velocity or mean bed shear stress. Coherent turbulent structures, local flow separation, pressure gradients and vortex dynamics can determine where sediment grains are mobilised and how the scour hole develops.

High-Resolution Particle-Based Modelling

I applied my high-resolution, particle-based sediment model to this problem. The approach combines turbulence-resolving flow simulation with a Lagrangian description of individual sediment grains. This makes it possible to connect the instantaneous turbulent structures directly to particle entrainment, transport and deposition.

The objective is to reproduce not only the total amount of erosion, but also the physical development of the scour process: initiation at the exposed patch, deepening of the hole, horizontal expansion beneath the protected bed and the conditions leading to progressive undermining.

• River Meuse bed stability
• Cohesive clay cover over non-cohesive sand
• Local failure of the protective bed layer
• Exposed sandy patches
• Three-dimensional local scour
• Large Eddy Simulation (LES)
• Coherent turbulent structures
• Vortex-driven sediment entrainment
• Lagrangian sediment-particle modelling
• Physics-based sediment pickup
• Vertical scour-hole development
• Horizontal scour propagation beneath the cover
• Undermining of the cohesive layer
• Progressive bed instability
• Comparison with laboratory flume experiments

Scientific and Engineering Relevance

The project provides a direct connection between detailed sediment- transport physics and a practical river-management problem. Rather than parameterising local scour solely through empirical transport relationships, the model resolves the turbulent structures responsible for sediment mobilisation and follows the resulting sediment motion explicitly.

This provides a framework for investigating why local failures of a naturally protective bed layer can develop into much larger scour features, and for assessing the hydraulic conditions under which such local disturbances may lead to progressive instability of the riverbed.

Ripples and Dunes

My research examines small-scale river morphodynamics, particularly the development of ripples and dunes under turbulent flow. The bedforms were not imposed as predefined geometries. They emerged from the interaction between turbulence, sediment pickup, particle motion, deposition and bed evolution.

• Spontaneous bedform initiation from a flat bed
• Ripple and dune growth
• Dune migration, merging and splitting
• Bedform wavelength and height
• Uniform grain size
• Turbulence over non-flat beds
• Flow separation behind bedforms
• Coherent turbulence–sediment interaction
• Bedform-induced hydraulic resistance
• Form drag and changing discharge

The work provided a physics-based connection between grain-scale transport and river-scale hydraulic behaviour and was validated against multiple laboratory flume experiments.

Publication

This work resulted in the peer-reviewed publication: “Detailed simulation of morphodynamics: 3. Ripples and dunes”, published in Water Resources Research, Volume 49, Issue 9, pages 5930–5943, in 2013.

Three-Dimensional Morphodynamics in a Curved Channel

This research work investigates the development of bed morphology in a strongly curved channel using a detailed three-dimensional, process-based numerical model. Channel curvature generates secondary circulation and strongly modifies the distribution of velocity, turbulence and bed shear stress, producing characteristic cross-channel sediment transport and morphological adjustment.

The numerical framework couples three-dimensional turbulent flow with Lagrangian sediment transport and dynamic bed evolution. Sediment particles are entrained from the bed, transported through the curved flow field and subsequently deposited, allowing the morphology to develop naturally from the interaction between hydrodynamics and sediment motion.

As the simulation develops, the curvature-induced secondary flow redistributes sediment across the channel. This produces erosion towards the outer part of the bend and deposition towards the inner region, together with the development of transverse bed slopes and complex three-dimensional bedforms.

• Three-dimensional flow through a strongly curved channel
• Curvature-induced secondary circulation
• Large Eddy Simulation (LES)
• Lagrangian sediment-particle transport
• Sediment pickup, transport and deposition
• Cross-channel sediment transport
• Outer-bank erosion and scour
• Inner-bank deposition
• Development of transverse bed slopes
• Three-dimensional and oblique bedforms
• Interaction between secondary flow and sediment particles
• Dynamic evolution of the channel bed
• Immersed-boundary treatment of curved geometry

The study demonstrates how river-bend morphology emerges from the coupled interaction between primary flow, secondary circulation, turbulence and individual sediment-particle motion. It provides a direct physical connection between detailed three-dimensional hydrodynamics and the characteristic erosion and deposition patterns observed in curved channels.

Publication

This work resulted in the conference publication: “Computational modelling of bed morphodynamics in curved channels”, published in the proceedings of River Flow 2016 – International Conference on Fluvial Hydraulics, pages 1346–1351, in 2016.

Dune Evolution under Discharge Hydrograph

This numerical experiment investigates the dynamic response of river dunes to a gradually varying discharge. The simulation starts from a relatively low-flow condition in which dunes are present on the sediment bed. The discharge is then progressively increased, allowing the morphological response of the bed to develop naturally as the hydraulic conditions move through different bedform regimes.

As the discharge increases, the dunes evolve in height, length and migration behaviour. At sufficiently high flow conditions, the simulation reaches the upper-stage plane-bed regime. Under these conditions, the existing dunes are progressively washed out and the bed approaches a nearly plane configuration.

The high discharge is subsequently maintained for a period to allow the upper-stage plane-bed condition to develop. The discharge is then gradually reduced. During the falling-flow stage, the plane bed becomes unstable again and dunes spontaneously re-form and continue to evolve as the hydraulic conditions decrease.

• Time-varying discharge
• Rising- and falling-flow conditions
• Dynamic bedform adaptation
• Dune growth and migration
• Transition between bedform regimes
• Dune washout at high discharge
• Upper-stage plane-bed development
• Re-formation of dunes during decreasing discharge
• Non-equilibrium morphological response
• Hysteresis in bedform behaviour
• Coupled turbulent flow and sediment transport
• Comparison with laboratory experiments
• Comparison with experimentally derived bedform-regime diagrams

An important aspect of this study is that the bedforms are not prescribed. Their growth, disappearance and subsequent regeneration emerge from the coupled interaction between turbulent flow, sediment transport and the evolving bed.

The numerical results are compared with experimental observations and the corresponding bedform-regime diagram. This provides a demanding test of the model because it requires the same numerical framework to reproduce both the transition from dunes towards an upper-stage plane bed during increasing discharge and the renewed development of dunes when the discharge is reduced.

Publication

This work resulted in the peer-reviewed publication: “Computational modeling of dissipation and regeneration of fluvial sand dunes under variable discharges”, published in Journal of Geophysical Research: Earth Surface, Volume 120, pages 1390–1403, in 2015.

Local Scour around Adjacent Cylindrical Piers

This research investigates the three-dimensional development of local scour around cylindrical hydraulic structures. The work focuses particularly on two adjacent cylinders arranged in side-by-side and tandem configurations and examines how the interaction between the cylinders influences the flow field, sediment transport and resulting scour geometry.

Local scour is a strongly coupled turbulence–sediment problem. Flow acceleration around the cylinders, separation, wake dynamics and coherent turbulent structures generate highly non-uniform hydrodynamic forces at the bed. These processes control where sediment is entrained, transported and deposited and consequently determine the evolution of the scour holes.

High-Resolution Numerical Modelling

The simulations combine three-dimensional Large Eddy Simulation (LES) with a Smagorinsky subgrid-scale model, a ghost-cell immersed-boundary technique for representing the cylinders, and detailed sediment-transport and morphodynamic modelling.

Sediment transport is represented using a process-based framework, allowing bed erosion and deposition to emerge from the interaction between the resolved flow field and sediment motion rather than imposing the scour-hole geometry in advance.

• Three-dimensional local scour
• Single and adjacent cylindrical structures
• Side-by-side cylinder configurations
• Tandem cylinder configurations
• Clear-water scour conditions
• Large Eddy Simulation (LES)
• Smagorinsky subgrid-scale modelling
• Ghost-cell immersed-boundary method
• Physics-based sediment transport
• Sediment pickup and deposition
• Dynamic bed evolution
• Flow acceleration around cylinders
• Coherent turbulent structures
• Scour-hole development
• Influence of cylinder spacing
• Comparison with experimental measurements

Influence of Cylinder Arrangement

A central objective of the study was to understand how the spacing and arrangement of neighbouring cylinders influence scour. The simulations showed that the scour patterns and maximum scour depths depend strongly on cylinder spacing.

For the side-by-side configuration, the calculated maximum scour depth increased as the distance between the cylinders decreased. For tandem cylinders, the response was different: the maximum scour depth initially increased with increasing cylinder spacing before decreasing after reaching a peak.

Validation against Experiments

The calculated scour depths and their dependence on cylinder spacing were compared with experimental measurements reported in the literature. The numerical model reproduced the measured trends and maximum scour depths with good agreement.

This study demonstrates the capability of high-resolution turbulence and process-based sediment modelling to reproduce the complex feedback between flow structures, sediment motion and evolving morphology around hydraulic structures.

Publication

This work resulted in the peer-reviewed publication “Numerical investigation of local scour at two adjacent cylinders” by H. S. Kim, M. Nabi, I. Kimura and Y. Shimizu, published in Advances in Water Resources pages 131–147, in 2014.

Multi-Scale Modelling of River Dunes

This project develops a multi-scale morphodynamic modelling approach for simulating the generation and migration of river dunes at field scale. The central idea is that the characteristic spatial scale of the overall flow is considerably larger than the scale of the bedforms that control sediment transport and hydraulic resistance.

Resolving both processes on the same uniformly fine computational grid would substantially increase computational cost. I therefore developed a two-grid approach in which the hydrodynamics are solved on a relatively coarse grid, while sediment transport and bed evolution are calculated on a finer grid.

• Multi-scale morphodynamic modelling
• Two-grid computational approach
• Coarse-grid hydrodynamic solution
• Fine-grid sediment-transport solution
• Fine-grid bed-evolution modelling
• Two-dimensional depth-averaged flow equations
• Shallow-water approximation
• Bed-shear-stress modelling
• Slope effects on sediment transport
• Parameterized bedload transport
• Exner equation for bed evolution
• Dune generation and migration
• Field-scale river morphodynamics
• Comparison with experimental data
• Comparison with field measurements

Bridging Different Spatial Scales

River dunes occur at considerably smaller spatial scales than the large-scale hydrodynamic field. The multi-scale formulation exploits this difference rather than forcing both processes to use the same computational resolution.

The hydrodynamic variables calculated on the coarse grid are transferred to the finer grid, where the sediment transport is evaluated and the bed is updated. This enables the model to reproduce relatively small-scale morphological structures while maintaining computational efficiency at the larger river scale.

Application to the River Waal

The methodology was applied to the River Waal in the Netherlands using measurements associated with the 1997 flood event. This provided an important field-scale test of whether the multi-scale approach could reproduce realistic river dunes rather than only laboratory-scale bedforms.

In the reported Waal application, the computed mean dune length and height were approximately 36 m and 1.1 m, respectively. The corresponding measured values were approximately 38 m and 0.9 m, demonstrating good agreement between the simulated and observed bedform dimensions.

From Detailed Morphodynamics to River-Scale Modelling

The significance of this approach is that it provides a route for representing detailed bedform dynamics within computational domains relevant to practical river applications. Instead of applying the finest resolution everywhere, the model assigns computational resolution according to the characteristic scale of the physical process being represented.

This project therefore represents an early step towards computationally efficient multi-scale river modelling, connecting the relatively small scales of dune formation with the much larger scales required for engineering assessment of river systems.

Publication

This work resulted in the peer-reviewed conference publication: “Multi-scale modelling of river morphodynamics”, published in the Proceedings of the International Conference on FluvialHydraulics, RIVER FLOW 2014, CRC Press/Balkema, pages 1253–1259, in 2014.