Marine snow and biological particles play a fundamental role in the ocean's biological pump, transporting organic carbon from the surface to the deep sea. Their dynamics influence global nutrient cycles, carbon sequestration, and the functioning of marine ecosystems. Computational fluid dynamics (CFD) offers a powerful lens to study these particles, but modeling their complex shapes, densities, and interactions with turbulent flows remains a formidable challenge. This article explores state-of-the-art strategies for simulating marine snow and biological particles in CFD, from discrete methods to continuum approaches, while highlighting emerging computational techniques that promise to deepen our understanding of oceanic processes.

Understanding Marine Snow and Biological Particles

Marine snow is a continuum of organic and inorganic particles that aggregate as they sink through the water column. It consists of fecal pellets, dead plankton, detritus, mineral dust, and microbial colonies, often bound together by transparent exopolymer particles (TEP). Biological particles, such as phytoplankton cells, zooplankton carcasses, and bacterial flocs, contribute to this heterogeneous mixture. Their size ranges from micrometers to several centimeters, with densities that vary with composition and degradation state.

The behavior of these particles in the ocean is not simply a function of Stokes settling. Turbulent eddies, shear layers, and vertical density gradients cause differential settling, aggregation, and fragmentation. Accurately capturing these interactions is essential for predicting how much carbon reaches the seabed versus being remineralized in the water column. CFD models bridge the gap between small-scale physics and basin-scale biogeochemistry, but require careful handling of particle properties and multiphase interactions.

Biological particles further complicate simulations because they are often non-spherical, deformable, and can actively swim or adjust buoyancy. Additionally, their surfaces are colonized by microbial biofilms that alter aggregation efficiency and degradation rates. Any simulation strategy must account for such biological influences to produce realistic particle fluxes.

Simulation Strategies

CFD practitioners have developed several complementary approaches to model marine snow and biological particles, each with advantages depending on spatial scale, desired detail, and computational budget.

Discrete Element Method (DEM)

The Discrete Element Method treats each particle as an individual entity, resolving contacts, collisions, and cohesion. For marine snow aggregates, which can be considered clusters of smaller primary particles, DEM is particularly powerful. The method tracks forces such as drag, buoyancy, lubrication, and van der Waals interactions, allowing the model to capture how aggregates form, deform, and break under hydrodynamic stress.

DEM simulations have been used to study the fractal structure of marine snow and its effect on settling velocity. However, the computational cost scales with the number of particles, making it suitable only for small-scale domains or for developing empirical parameterizations used in larger models. Coupling DEM with a CFD solver for fluid flow enables a fully resolved approach, but this remains computationally intensive for oceanic-scale problems.

Eulerian-Lagrangian Framework

In the Eulerian-Lagrangian (EL) approach, the fluid phase is solved on an Eulerian grid while particles are tracked individually (Lagrangian points). This is the most common method for simulating particle-laden flows in environmental sciences. For marine snow, each Lagrangian particle can represent an aggregate with a defined size, density, and shape factor. The code calculates particle trajectories using a force balance that includes drag, lift, virtual mass, and gravity.

EL frameworks are well-suited for dilute suspensions where particle-particle interactions are negligible. They can handle millions of particles with reasonable computational cost, especially if parcels (representative clusters) are used. Many ocean models, such as the MITgcm or ROMS, employ Lagrangian particle tracking modules for studying sinking particle fate. Limitations arise when particle concentrations become high or when aggregation and breakup need to be resolved explicitly, as these processes require models for collision kernels and stickiness probabilities.

Multiphase Flow Models

When marine snow density is high enough to influence the bulk fluid rheology, multiphase flow models become necessary. In an Eulerian-Eulerian framework, both the fluid and the particle phase are treated as interpenetrating continua, governed by conservation equations with exchange terms for momentum, mass, and energy. This approach is efficient for simulating sedimentation clouds or particle plumes in the ocean.

For marine snow, the particle phase can be represented by a population balance equation that tracks the number density distribution of aggregate sizes. Sub-models for aggregation, breakage, and settling are embedded into the transport equations. Such models have been applied to study the dispersal of oil-mineral aggregates after spills or the vertical flux of organic matter in upwelling regions. The main challenge is closure: the exchange terms rely on empirical drag laws and turbulence models that are often derived for rigid spherical particles, not for fluffy marine snow.

Porous Media Approximation

Large, porous aggregates exhibit internal flow and reduced drag compared to solid particles. To simplify their simulation, many researchers approximate marine snow as porous spheroids with a prescribed permeability. The fluid flow through the aggregate is governed by Darcy's law or Brinkman's equation, while the external flow is solved via the Navier-Stokes equations. This porous-medium approach captures the increased settling velocity and altered wake structure without resolving every intra-aggregate pore.

This method is particularly useful when simulating the resuspension of aggregated material in benthic boundary layers or the interaction of marine snow with suspended sediment. The porosity and permeability parameters can be calibrated from measurements or direct numerical simulations of fractal aggregates. However, the approach assumes a homogeneous internal structure, which is often a simplification since aggregates have radial density gradients.

Incorporating Particle Aggregation and Breakup

Marine snow is not static; it undergoes continuous formation, fragmentation, and dissolution. Realistic CFD simulations must include population balance models that describe how particle size distributions evolve. Aggregation is typically modeled with a kernel that depends on particle encounter rates due to Brownian motion, fluid shear, and differential settling. Breakage kernels account for turbulent stresses or collisions with larger particles.

These processes are highly nonlinear and often require solving the Smoluchowski coagulation equation alongside the fluid dynamics. Recent advances use quadrature methods (e.g., QMOM, DQMOM) to track moments of the size distribution efficiently. For biological particles, additional terms for growth, death, and aggregation induced by microbial activity can be integrated. The coupling between aggregation kinetics and flow dynamics remains an active area of research, as even small errors in stickiness or fragmentation rates can drastically alter predicted carbon fluxes.

Validation and Observational Constraints

Models are only as good as their verification against real-world data. Simulating marine snow requires validation using in situ measurements from sediment traps, underwater video profilers (e.g., CPICS, RDI), and optical sensors. Settling velocity distributions, aggregate size spectra, and concentration profiles provide ground truth for CFD predictions. However, field data are sparse and often limited to certain regions or seasons, making extrapolation uncertain.

Laboratory experiments in settling columns and rotating tanks offer controlled environments to test model parameterizations. For example, shear-driven aggregation studies using diatom cultures have helped refine coagulation kernels. Similarly, microfluidic devices now allow visualization of single aggregate breakup in controlled flows. CFD models that replicate these lab conditions can be validated directly before scaling to oceanic simulations. Open benchmark cases, such as the Marine Snow Simulator intercomparison project, help standardize evaluation.

Computational Considerations and High-Performance Computing

Simulating marine snow across relevant scales—from millimeters to kilometers—demands advanced computing. High-performance computing (HPC) enables high-resolution LES (large-eddy simulation) domains with millions of grid cells and billions of virtual particles. Many ocean CFD codes, such as OpenFOAM, Nek5000, and FVCOM, have been adapted for particle-laden flows with custom libraries for aggregation and sintering.

Parallelization strategies must handle load balancing when particles move across processor boundaries. Adaptive mesh refinement (AMR) can concentrate grid points near settling particles or aggregation zones, significantly reducing computational cost. GPU acceleration is increasingly used for Lagrangian particle tracking, as particle-level calculations are highly data-parallel. Cloud computing platforms are also making HPC accessible to smaller research groups, allowing more simulations of marine biological particles.

Machine learning (ML) has begun to complement traditional CFD approaches. Neural networks can be trained to approximate the drag coefficient of fractal aggregates or to predict aggregation kernels based on flow and particle properties. Data-driven surrogate models can replace expensive sub-routines, enabling longer simulations of patchy marine snow distributions. However, trustworthiness and extrapolation ability remain concerns—ML models often fail outside their training data distribution, and ocean conditions are highly variable.

Challenges in Simulating Marine Snow and Biological Particles

Despite progress, several fundamental challenges persist. First, the inhomogeneous and time-varying nature of marine snow makes it difficult to define a single representative particle class. Aggregates are not uniform; they contain channels, filaments, and remnants of exopolymers. Second, biological activity introduces feedbacks: living particles can excrete sticky substances, change buoyancy, or be grazed by zooplankton. These biological processes occur on timescales that interact with transport timescales, requiring coupled biogeochemical-CFD models that are computationally expensive.

Third, turbulence affects both the encounter rates of particles and the breakup of aggregates. The interaction between small-scale turbulence and particle stickiness is still poorly parameterized. Many models use a constant turbulence dissipation rate, but in reality, marine snow sinks through a stratified water column with strongly varying turbulence levels. Resolving these gradients requires a fine vertical grid and sophisticated turbulence closure schemes.

Fourth, the sparsity of in situ measurements for validating particle properties (e.g., density, porosity, fractal dimension) limits model reliability. Remote sensing cannot directly see marine snow at depth, and sediment traps provide only integrated flux, not detailed size distributions. New autonomous profiling floats with high-resolution cameras, such as the MINION system, are beginning to fill this gap, but comprehensive datasets remain scarce.

Future Directions and Emerging Research

Future simulations will likely couple CFD with resolved biological models that include phytoplankton physiology, bacterial degradation, and zooplankton swimming. The integration of agent-based models (ABMs) for individual plankters within a CFD domain can simulate patchy feeding and aggregation hotspots. This multiscale approach will better predict the formation of organic aggregates and their subsequent fate.

Another promising direction is the use of physics-informed neural networks (PINNs) to solve inverse problems—e.g., inferring particle properties from observed settling velocities. Combined with CFD, PINNs can learn the effective parameters of aggregate behavior directly from experimental data, bypassing some of the traditional modeling uncertainties. However, the oceanographic community must ensure that these new tools are validated against standard benchmarks.

Global climate models are increasingly incorporating explicit particle dynamics, but at coarse resolution they rely heavily on parameterizations. CFD simulations can provide the process-level understanding needed to improve these large-scale models. For instance, the relationship between turbulence, aggregation, and carbon export is critical for refining the biological pump in Earth system models. Recent work has shown that even small changes in stickiness or fragility can alter carbon sequestration rates by 10–20%—enough to affect future climate projections.

Finally, collaboration between fluid dynamicists, oceanographers, and computer scientists will be essential. Open-source simulation frameworks like OpenFOAM and community-driven databases of aggregate properties can accelerate progress. Standardizing formats for particle data and simulation outputs will facilitate intercomparison studies, much like the Coastal Ocean Model intercomparison efforts. The path forward lies in combining physical realism with computational efficiency, ensuring that marine snow simulations become a reliable tool for understanding and protecting our ocean.

In summary, simulating marine snow and biological particles in CFD demands a suite of strategies tailored to different scales and particle characteristics. From the discrete element method to porous media approximations, each approach contributes a piece of the puzzle. Validation against observations and advances in computing will continue to push the boundaries of what is possible. As the ocean's role in climate regulation grows more urgent, these simulations will provide crucial insights into the hidden snowstorm that links the surface to the deep sea.