🔬 condensed matter

Wind-driven collisions between floes explain the observed dispersion of Arctic sea ice

This study demonstrates that stochastic granular simulations accounting for wind-driven collisions between ice floes accurately reproduce observed Arctic sea ice dispersion and velocity distributions, resolving previous discrepancies with Brownian models and establishing a predictive framework based on local floe-scale processes.

Original authors: Bryan Shaddy, P. Alex Greaney, Bhargav Rallabandi

Published 2026-09-17

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Bryan Shaddy, P. Alex Greaney, Bhargav Rallabandi

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The Arctic Ocean is not a solid sheet of ice but a vast, shifting mosaic of broken ice sheets, known as floes, that drift across the water. These floes, ranging from a few meters to several kilometers across, are pushed by the wind and pulled by ocean currents. While the overall movement of this ice pack is crucial for regulating the Earth's climate and supporting polar ecosystems, the chaotic, jittery motion of individual floes has long puzzled scientists. When researchers tried to predict how fast these ice pieces spread out or how they move in response to gusty winds, standard models failed. These models, which treat ice floes like tiny particles drifting randomly in a fluid, predicted a speed of spreading that was ten times faster than what was actually observed in the real world. They also failed to explain why the ice moved with a specific, unusual pattern of speed that did not match the random fluctuations of the wind itself.

A new study by researchers at the University of Southern California and the University of California, Riverside, offers a clear resolution to these discrepancies by focusing on a single, often overlooked factor: the collisions between ice floes. The team built a sophisticated computer simulation that tracked thousands of individual ice floes as they were pushed by noisy, fluctuating winds. In their model, the ice floes were not isolated particles; they were crowded together, covering up to ninety percent of the ocean surface. As the wind pushed them, the floes constantly bumped into one another. The researchers found that these frequent, energy-draining collisions were the missing key. Every time two floes collided, they dissipated the energy injected by the wind, slowing the ice down and preventing it from spreading as wildly as the old models predicted.

By feeding the simulation with real-world wind data from the Fram Strait, a major gateway for sea ice between Greenland and Svalbard, the team achieved a remarkable match with actual observations. The simulation reproduced the exact rate at which the ice spread, the specific distribution of speeds, and the energy patterns of the movement over time scales ranging from hours to months. The results showed that in a crowded ice field, the ice behaves very differently than in open water. When the ice is sparse, the floes move freely and quickly, matching the random gusts of the wind. But as the ice becomes denser, collisions become so frequent that they dominate the motion. The ice floes bounce off each other so often that they cannot build up speed, leading to a much slower, more constrained drift than previously thought possible.

The study explicitly rules out other potential explanations that scientists had previously considered, such as the intermittent nature of the wind, the turbulence of the ocean currents, or long-range correlations in the ice field. The researchers demonstrated that none of these factors were necessary to explain the observed behavior. Instead, the complex, chaotic motion of the ice emerged directly and simply from the local interactions between the floes themselves. The team also developed a mathematical framework, based on the physics of granular materials like sand or grains, to describe these dynamics. This theory confirmed that the rapid collisions act as a brake, converting the chaotic energy of the wind into heat and friction, which narrows the range of speeds the ice can reach and flattens the energy spectrum of its motion.

The findings suggest that to understand the global transport of sea ice, scientists do not need to model the entire atmosphere or ocean in extreme detail. Instead, the behavior of the ice pack can be understood by looking at the local scale: how individual pieces of ice interact with their immediate neighbors and the wind. The researchers used only directly measured quantities, such as wind speed and ice thickness, to drive their model, proving that the complex, large-scale patterns of ice movement are a direct consequence of simple, local collisions. This approach provides a new, more accurate way to predict how sea ice will move in a changing climate, offering a clearer picture of a critical component of the Earth's climate system.

Technical Summary: Wind-driven collisions between floes explain the observed dispersion of Arctic sea ice

Problem Statement
The transport of sea ice over polar oceans is a critical component of local ecology and global climate systems. This transport is driven primarily by turbulent winds, resulting in the stochastic motion of ice floes. While the mean drift of ice is well-understood, the fluctuations governing ice spreading (dispersion) present significant modeling challenges. Observations reveal that sea ice exhibits "fat-tailed" velocity distributions and anomalous dispersion that deviate by orders of magnitude from standard Brownian motion models. Specifically, observed diffusivities (20–100 m²/s) are an order of magnitude lower than theoretical predictions for isolated floes (~1000 m²/s), and velocity distributions are narrower and non-Gaussian compared to the normally distributed wind fluctuations that drive them. Previous attempts to explain these discrepancies have attributed them to wind intermittency, turbulent structures, long-range spatial correlations, or intermittent ice fractures, but a unified, quantitative explanation has remained elusive.

Methodology
To resolve these gaps, the authors developed a Stochastic Discrete Element Model (SDEM) that fully resolves the dynamics of interacting ice floes under noisy wind forcing. The model treats the ice pack as a collection of rigid disks (floes) of varying sizes, moving on a two-dimensional ocean surface.

Physical Forces: The motion of each floe is driven by wind stress on its upper surface and resisted by ocean drag on its lower surface. The wind velocity is modeled as a Langevin process, characterized by a mean drift, isotropic Gaussian fluctuations, and an exponential time correlation.
Interactions: Crucially, the model accounts for frequent collisions between floes. These interactions are modeled as energy-dissipating (inelastic) but momentum-conserving collisions. The coefficient of restitution is set to 0.1, reflecting the plastic deformation and energy loss typical of ice collisions (e.g., ridge formation).
Inputs: All model inputs, including wind statistics (mean speed, fluctuation standard deviation, and correlation time) and floe size distributions (a power-law distribution between 10 m and 4500 m), are derived directly from ERA5 reanalysis data and field observations from the Fram Strait. The only parameter not directly measured is the coefficient of restitution, which is fixed based on physical reasoning.
Validation: The model was validated against observational data from the FRAMZY/ACSYS dataset in the Fram Strait and IABP data from the central Arctic. Key metrics compared included the mean-squared displacement (MSD), velocity probability density functions (PDFs), and power spectral densities (PSD).
Theoretical Framework: To provide physical insight, the authors derived a coarse-grained kinetic theory based on the Boltzmann transport equation (BTE). This theory treats the ice field as a population of floes subject to wind noise, ocean drag, and a mean-field collision operator.

Key Results
The SDEM simulations successfully reproduced observed sea ice transport statistics with remarkable quantitative accuracy, resolving discrepancies that have persisted in the field.

Dispersion and Diffusivity: The simulations reproduced the transition from ballistic motion at short timescales to diffusive motion at long timescales. Crucially, the model predicted cross-stream diffusivities in the range of 40–100 m²/s for ice area fractions ( $\omega$ ) between 0.84 and 0.92. This matches observed values and is an order of magnitude lower than the theoretical prediction for isolated floes. The reduction in diffusivity is directly attributed to collisions, which shorten the mean-free path of the floes.
Velocity Distributions: The model captured the non-Gaussian nature of ice velocity fluctuations. At low ice concentrations, the velocity distribution is Gaussian (Brownian-like). However, at high concentrations ( $\omega \approx 0.9$ ), the distribution transitions to an exponential form ( $P(v) \propto e^{-q|v|}$ ), matching observations. The model correctly predicted the narrowing of the velocity distribution (RMS velocity $\approx$ 6 cm/s) due to energy dissipation during collisions.
Power Spectral Density (PSD): The simulated PSD matched observations across two decades of frequency. The spectrum exhibited a $f^{-1}$ scaling at lower frequencies, transitioning to $f^{-2}$ at higher frequencies (near the ice dynamical timescale). The model demonstrated that collisions flatten the spectrum by reducing low-frequency energy (diffusivity) while boosting high-frequency contributions due to shorter mean-free paths.
Kinetic Theory Validation: The derived kinetic theory provided analytical expressions for the RMS velocity and diffusivity that agreed closely with both the simulations and observations. The theory confirmed that in concentrated ice fields, collisions are the primary source of energy dissipation, occurring on timescales much shorter than wind decorrelation times.

Significance and Claims
The authors claim that the observed features of sea ice motion are a direct and robust consequence of the interplay between noisy wind forcing and frequent floe-floe collisions. The paper argues that:

Local Processes Drive Global Transport: The complex, large-scale transport statistics of sea ice can be understood through "local" floe-scale processes. The model recovers observations quantitatively without invoking large-scale spatial structures of the ice field or detailed atmospheric/oceanic turbulence models.
Resolution of Anomalies: The "puzzles" of anomalous dispersion and non-Gaussian velocity distributions are simultaneously resolved by a single physical ingredient: energy-dissipating collisions between floes.
Predictive Paradigm: The developed ideas and tools pave the way for a new predictive understanding of global sea ice transport. By linking macroscopic transport properties (diffusivity, rheology) to local environmental properties (wind statistics, collision rates), the framework offers a pathway to improve climate forecasts.
Rheology Implications: The authors suggest that the effective rheology of sea ice, a critical but poorly understood component of climate models, may not be an intrinsic property of the ice pack alone but rather a consequence of how the ice interacts with its noisy environment through collisions.

The paper concludes that these findings constitute a novel advance in understanding sea ice transport, emphasizing that accurate prediction relies on resolving the stochastic dynamics of interacting floes rather than treating the ice as a continuum or a collection of non-interacting particles.

Technical Summary: Wind-driven collisions between floes explain the observed dispersion of Arctic sea ice

More like this