We develop a physics-informed neural network (PINN) to significantly augment state-of-the-art experimental data of stratified flows. A fully connected deep neural network is trained using experimental data in a salt-stratified inclined duct (SID) experiment. SID sustains a buoyancy-driven exchange flow for long time periods, much like an infinite gravity current. The data consist of time-resolved, three-component velocity fields and density fields measured simultaneously in three dimensions at Reynolds number= O(10^3) and at Prandtl or Schmidt number = 700 [1]. The PINN enforces incompressibility, the governing equations for momentum and buoyancy, and the boundary conditions at the duct walls. These physics-constrained, augmented data are output at an increased spatio-temporal resolution and demonstrate five key results: (i) the elimination of measurement noise; (ii) the correction of distortion caused by the scanning measurement technique; (iii) the identification of weak but dynamically important three-dimensional vortices of Holmboe waves; (iv) the revision of turbulent energy budgets and mixing efficiency; and (v) the prediction of the latent pressure field and its role in the observed asymmetric Holmboe wave dynamics. These results mark a significant step forward in furthering the reach of fluid mechanics experiments, especially in the context of stratified turbulence, where accurately computing three-dimensional gradients and resolving small scales remain enduring challenges.

References
[1] L. Zhu, X. Jiang, A. Lefauve, R. R. Kerswell, and P. F. Linden. New insights into experimental
stratified flows obtained through physics-informed neural networks. J. Fluid Mech., 981:R1, 2024.

How to cite: Lefauve, A., Zhu, L., Jiang, X., Kerswell, R., and Linden, P.: New insights into experimental stratified flows obtained through a physics-informed neural network, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-18513, https://doi.org/10.5194/egusphere-egu25-18513, 2025.

In this contribution, we present 3D two-phase flow simulations of lock-release turbidity currents using sedFOAM. The Large Eddy Simulation is used for the turbulence modeling while the granular stresses are modeled using a frictional-collisional kinetic theory including interparticle friction (Chassagne et al., 2023). Simulations are performed for different bed slopes, initial volume fractions and particle diameter and density. The numerical results are compared with experiments in terms of front propagation and current shape (Gadal et al., 2023). The simulation results are further used to infer the mechanisms controlling the current attenuation, i.e. the current propagation speed reduction with time. We further use the model to analyse the influence of the flume geometry, free surface versus rigid roof.

How to cite: chauchat, J., sharma, M., Rastello, M., Bonamy, C., Gadal, C., Dossmann, Y., Mercier, M., and Lacaze, L.: 3D two-phase flow simulations of lock-release turbidity currents, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-20195, https://doi.org/10.5194/egusphere-egu25-20195, 2025.

Temperature-induced density variations within fluids drive gravity-driven flows. Such flows occur for example in geophysical processes such as differential cooling or stratification dynamics in lakes. Heat transfer through the boundaries, whether in a lake or in a laboratory flume, can impact flows. The objective of this study is to evaluate numerically and experimentally the effect of these boundary conditions on thermal convective mixing in two different configurations. The first configuration is the thermal convective mixing in a box of water with cooled boundaries. The second configuration is a lock exchange gravity flow.

Numerical simulations were performed using OpenFOAM with a Large Eddy Simulation solver and the Boussinesq approximation for density effects. The sensitivity of the solution to different boundary conditions for heat transfer were analyzed. Experiments rely on an innovative phosphor thermometry technique able to measure spatial patterns of fluid temperature instead of common pointwise measurements. Notably, we introduce a novel approach that combines the use of a laser sheet and high-resolution CMOS sensors operated in a multi gate accumulation mode to extract the temperature pattern.

The choice of appropriate parameters in the boundary condition enabled the accurate representation of the temperature evolution and convective flow patterns observed in the experiments.

How to cite: Pons, M., Rousseau, G., Adelerhof, H., Carde, B., Giesbergen, M., Fond, B., Borisov, S., and Blanckaert, K.: A numerical and experimental investigation of the influence of heat transfer boundary conditions on convective mixing, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-4429, https://doi.org/10.5194/egusphere-egu25-4429, 2025.

The flow of a hyperpycnal river plume into a lake with an inclined bed and without any lateral confinement is a common occurrence in nature.  A characteristic example of such geophysical fluid flow system is the one of  sediment-laden dense river inflows in freshwater lakes (e.g. river Rhone - lake Geneva). In this case, the plunging process close to the river mouth, as well as its dependence on the properties of the inflow, affect critically the subsequent hydrodynamics in the lake. A better understanding of these flow processes is therefore significant for the effective management of the lake water quality and overall ecology. In this talk, we focus on the complex near-field plume dynamics and establish novel criteria for plunging, based on Large Eddy Simulations (LES) of an unconfined hyperpycnal saline plume over an idealized bed. More specifically, we focus on the effect of the variation of the inflow densimetric Froude number, Frd, which is the non-dimensional parameter describing the ratio between inertial and buoyancy forces acting on the plume. In particular, three simulations were performed at different values of Frd, within the range of values encountered in the plunging of river Rhone in lake Geneva. It is found that the near-field dynamics of the hyperpycnal plume is different in the unconfined plunging scenario compared to the case where the flow is confined by lateral walls and that it critically depends on Frd. Indeed, it is a well-established result that in the confined case the plunging of the hyperpycnal plume occurs at the location downstream where a balance between dynamic pressure forces (inertia) and the resisting hydrostatic pressure forces (gravity) is obtained. However, in absence of any lateral confinement the plunging begins immediately upon the entrance of the dense river water in the lake, due to lateral slumping. The slumping takes place at both lateral sides of the plume in the form of a collapse of the dense water column followed by a spread across the lake bottom that is very similar to a lock-release flow configuration. This results in an earlier departure of the plume from the lake surface in the unconfined case compared to the confined case under similar inflow conditions. We are able to determine the effect of Frd on the plunge curve, as well as the extend of the plunging zone on the lake bed, before the plume turns into a gravity current and continues its propagation down the slope. In addition, an explanation is provided for the field observations of surface leakage, where sediment-rich water is detected at the lake surface even downstream of the plunge curve. This explanation focuses on the effect of Frd on the dynamics of the turbulent mixing layers that develop at the interface between the incoming river water and the surrounding lake water. These mixing layers facilitate a substantial transfer of mass and momentum from the inflowing river water to the ambient lake water.

How to cite: Giamagas, G., Bonamy, C., Blanckaert, K., and Chauchat, J.: Near-field plunging dynamics of laterally unconfined hyperpycnal plumes over inclined beds, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-17770, https://doi.org/10.5194/egusphere-egu25-17770, 2025.

Hyperpycnal (denser) river inflows into lakes bring sediments, nutrients, oxygen and contaminants, which are crucial for the water quality. Due to the higher densities of hyperpycnal inflows, they abruptly descend toward the lake bottom upon entering the lake, a process called plunging, and subsequently continue flowing along the lake bottom as a gravity-driven underflow. This plunging is accompanied by mixing and entrainment of ambient lake waters, which causes dilution of the initial density excess of the river inflow. The mixing is parameterized by the plunging mixing coefficient E_p  and is of critical importance as it conditions the fate and final destination of the contaminants carried by the river inflows. A recently proposed conceptual model (Thorez et al. 2024) for the plunging into an unconfined lake configuration highlights the importance of the lateral slumping motion of the river plume and secondary flow cells on each side of the plume with respect to the plunging mixing.

This study builds on previous research that suggests that E_p  is affected by the geometry of the river mouth. We investigate with Particle Image Velocimetry (PIV) in laboratory experiments how the width-to-depth ratio at the river mouth (W₀H⁻¹) influences the plume hydrodynamics and E_p. Four different ratios, W₀H⁻¹ = 5.4, 9, 13.5 and 27, were investigated in a configuration that mimics the Rhône inflow into Lake Geneva. The laboratory experiments were performed in the Coriolis platform at LEGI (Laboratoire des Ecoulements Géophysiques et Industriels) at a scale of 1:60 that allows minimal scaling effect.

The distance of the plunge location from the river mouth, x_p, and the corresponding depth, h_p, were found to decrease with the aspect ratio. In addition, the size of the secondary flow cells on each side of the slumping river plume decreased with aspect ratio which tentatively explains the observed variations in E_p.

How to cite: Eze, K. O., Ammendola, A., Giamagas, G., Thorez, S., Negretti, E., Chauchat, J., and Blanckaert, K.: Effect of inflow channel aspect ratio on the plunging dynamics of an unconfined hyperpycnal plume over a sloping bed, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-9640, https://doi.org/10.5194/egusphere-egu25-9640, 2025.

The dispersal of meltwater discharged from the St. Lawrence Valley is investigated from numerical experiments with a regional configuration of the MIT general circulation model representing the western North Atlantic during the last ice age. These experiments assume a horizontal resolution of 1/20^o and a vertical grid with 21 levels in the upper 100 m, so that both the mesoscale eddy field and the vertical structure of the meltwater plume could be simulated in detail. Our goals are to identify possible mechanisms responsible for the offshore spreading of meltwater in the ocean during the deglaciation and to help the interpretation of paleoceanographic observations from the seafloor and sediment cores.

We find that meltwater discharged from the St. Lawrence Valley forms a buoyant plume which turns to the southwest along the continental slope under the action of the Coriolis force. Part of the meltwater is entrained away from the slope by meander crests and warm-core rings of the Gulf Stream (GS) between the St. Lawrence Valley and Cape Hatteras. The other part is diverted offshore by the opposing GS near Cape Hatteras, where the GS leaves the continental margin. In one experiment, meltwater is incorporated into a meander trough that pinches off and produces a cold-core ring, leading to meltwater transport into the subtropical gyre, or it flows southward along the slope inshore of the GS to the South Atlantic Bight. Sensitivity tests show that the buoyant plume spreads at a consistent rate of O(10⁵ m² s^-1). A reduced-gravity two-layer model suggests that the spreading of the plume is governed by (i) the net ageostrophic motion produced by the total acceleration and the upwelling-favorable winds along the front of the plume and (ii) the advection of the front of the plume by the ambient geostrophic flow. In our experiments, meltwater in turn alters the upper part of the GS through meltwater-induced changes in cross-stream density gradients. Our results put constraints on the interpretation of ice-rafted debris found in (de)glacial sediments from the Sargasso Sea and of iceberg scours observed on the slope south of Cape Hatteras.

How to cite: Marchal, O. and Condron, A.: On the Spreading of Meltwater Plumes in the Ocean during the Last Deglaciation, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-2573, https://doi.org/10.5194/egusphere-egu25-2573, 2025.

The Mediterranean outflow from the Bosphorus advances through channels and delta features, reaching the shelf edge, spreading across the mid-shelf slope, and cascading down the steep continental slope. During the July 2024 cruise, intensive sampling along transects using CTD and Scanfish provided high-resolution data to better understand the pathways of the Mediterranean plume along the continental slope. This study explores the steering of Mediterranean streams through a complex channel network on the shelf, the evolution of distinct water properties—including dissolved oxygen content—and the eventual transition of the Mediterranean flow into neutrally buoyant layers.

To complement the observational data, the dynamics of the gravity current along the shelf-slope region of the Black Sea, characterized by anomalous temperature and oxygen levels were modeled. Model parameters were optimized and validated against the collected cruise data. These new hydrographic observations and modeling efforts shed light on the Mediterranean gradient flow's penetration into the Black Sea interior, advancing our understanding of its pathways and influence on regional water properties.

How to cite: Cokacar, T., Altıok, H., Yücel, M., and Örek, H.: Tracking the Mediterranean Outflow from the Bosphorus to the Continental Shelf Edge, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-19603, https://doi.org/10.5194/egusphere-egu25-19603, 2025.

The Strait of Gibraltar is the site of a rich and complex physics where flows of different densities interact with a strongly rugged topography. These interactions produce many small-scale processes such as the development of hydraulic jumps, shear instabilities, internal solitary waves, which induce strong mixing and impact the Mediterranean outflow water masses. The dynamics of this gravity current emerging from the Strait and flowing along the canyons of the Gulf of Cadiz is conditioned by its prior evolution inside the Strait. In turn, this dense gravity current plays a crucial role on the large-scale oceanic circulation as it mixes with overlying North Atlantic waters thus modifying the properties of the deep-water masses and therefore, likely the regional and basin-scale circulations. Better understanding and describing these small-scale processes is essential to improve operational oceanic models and climate models since they occur below the grid scale of these models and therefore require parameterizations. Recently, the LEGI team has implemented a realistic setup of the Strait of Gibraltar with the adjacent Gulf of Cadiz and Alboran Sea on the Coriolis Platform at Grenoble, including the barotropic forcing (tide), the baroclinic one (lock-exchange), the Earth’s rotation and a realistic topography. The aim is to bring a better understanding of the small-scale physics which take place in this area. In addition to this experimental approach, centimetric resolution of hydrostatic and non-hydrostatic numerical simulations at the laboratory scale were carried out using the numerical code CROCO (Coastal and Regional Ocean COmmunity model – https://www.croco-ocean.org). A specificity of this code is to be able to efficiently resolve sub-mesoscale processes and to relax both the hydrostaticity and Boussinesq assumptions using a non-hydrostatic and compressible Navier-Stokes solver. The purpose of this numerical approach is manifold: explore ranges of parameters that could not be studied experimentally, develop diagnostic tools, investigate the impact of non-hydrostatic effects, evaluate numerical schemes and parameterizations, investigate more specifically each physical process. However, setting up such a realistic numerical configuration is challenging. For example, if we are interested in the purely barotropic forcing, it is essential to represent the tidal forcing identically as in the experiment. Likewise, when we focus on the purely baroclinic forcing, the inherent steep slopes related to the experimental setup put strong constraints on the numerical code. From these numerical simulations, non-hydrostatic effects on the gravity current dynamics are estimated and analyzed at the laboratory scale. They significantly modify hydraulic jumps formation, overflow transport and deep waters circulation in the Gulf of Cadiz. Boundary conditions, vertical resolution, explicit and implicit vertical mixing or viscous effects are all numerical factors that influence gravity current dynamics. All these features must be carefully studied since they impact both the fate of the Mediterranean waters when they flow into the Atlantic Ocean and the fate of the Atlantic waters flowing into de Mediterranean basin. The aim of this presentation is to present the work carried out up to date in the development of these numerical configurations and to present the associated preliminary results.

How to cite: Gouhier, B., Bordois, L., Auclair, F., Nguyen, C., Tassigny, A., Bardoel, S., Gostiaux, L., and Negretti, M. E.: Towards a numerical configuration of the Strait of Gibraltar at the laboratory scale: The HERCULES project, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-18582, https://doi.org/10.5194/egusphere-egu25-18582, 2025.