Sandy Beaches Behave as Chaotic Oscillators, Satellite Records Reveal

Sandy beaches have long frustrated the scientists who try to predict them. Shorelines advance and retreat with the seasons, sandbars migrate offshore and onshore, and yet two beaches exposed to seemingly similar waves can evolve in strikingly different ways. A new study published in Nonlinear Processes in Geophysics argues that this stubborn unpredictability is not simply a data problem. By reconstructing the internal dynamics of four contrasting beaches directly from a decade of satellite observations, a team of French and Mexican-affiliated researchers shows that the coupled shoreline–sandbar system behaves as a deterministic but chaotic oscillator, governed by internal feedbacks that impose hard limits on how far ahead anyone can forecast.

The research team, led by Marius Aparicio of the Institute of Fluid Mechanics of Toulouse together with Sylvain Mangiarotti, Salomé Frugier, Laurent Lacaze, Marcan Graffin and Rafael Almar, exploited the steady stream of images from the Sentinel-2 satellites, which revisit every location on Earth every five days at a resolution of ten meters. Using Google Earth Engine, the team downloaded optical imagery for four sites chosen to span the main morphodynamic regimes of wave-dominated sandy coasts: Torrey Pines in California, Ensenada Beach in Baja California, Duck in North Carolina, and the Gold Coast in Queensland, Australia. These sites range from single-barred, cliff-backed beaches to double-barred, high-energy Atlantic shores, providing a demanding test of whether a common dynamical structure underlies their apparent diversity.

Extracting the relevant features from the images required two complementary detection techniques. The shoreline was identified with a subtractive coastal water index combined with threshold refinement, achieving typical errors below ten meters. The submerged sandbars posed a harder problem, since they cannot be seen directly. Instead, the researchers exploited the foam produced when waves break over a bar, using a normalized sandbar index that highlights breaking zones across different satellite sensors, with reported errors of roughly twenty meters, about two Sentinel-2 pixels. Positions were measured along perpendicular transects spaced thirty meters apart, averaged robustly across each beach, and smoothed to yield continuous daily time series of shoreline and sandbar position, together with their first and second time derivatives.

The analytical core of the study is a technique known as global polynomial modeling, implemented in the R package GPoM. Rather than assuming a set of physical equations in advance, the method searches for the smallest set of polynomial ordinary differential equations that can reproduce the geometry of the observed dynamics in phase space. Building on Takens’ embedding theorem, it treats the shoreline position, its rate of change, and the sandbar position as state variables, and iteratively selects the minimal subset of nonlinear terms that preserves the variance and structure of the reconstructed trajectories. The team tested three- and four-dimensional models with polynomial degrees from two to four, retaining only models that remained numerically stable over eighty years of integration and were neither fixed points nor simple limit cycles.

The resulting equations, containing between fifteen and nineteen terms, reproduced the essential features of each beach’s behavior. At Torrey Pines and Ensenada, the shoreline and sandbar oscillate in a quasi-linear negative correlation at seasonal scales, consistent with a coupled retreat-and-recovery mechanism: when winter wave energy intensifies, the bar migrates offshore while the shoreline retreats, and during calmer months the bar drifts landward, promoting recovery. At the Gold Coast, the correlation inverts, indicating that the whole surf zone translates coherently rather than stretching and compressing. Duck, the double-barred Atlantic beach, required a four-dimensional model and displayed a far more intricate attractor, with hysteresis in which rapid offshore bar motion accompanies abrupt shoreline retreat while recovery is slow and gradual, a signature of morphodynamic memory.

Interpreting the dominant terms of the equations in terms of forces acting on the sand revealed distinct feedback architectures. At Torrey Pines, a cubic restoring term pulls the shoreline back toward equilibrium after large excursions, while the sandbar acts as an energy buffer whose protective role weakens when it drifts far offshore. At Ensenada, the sign of the shoreline–sandbar coupling depends on how fast the shoreline is moving: during rapid storm-driven retreat, the correlation can invert, driving the bar in the same direction as the shoreline rather than opposing it. At Duck, the shoreline dynamics are controlled entirely by velocity-dependent nonlinear interactions, with no autonomous restoring force, and the sandbar exhibits intrinsic instability, making it the most complex system of the four.

The evidence for chaos rests on three converging lines of analysis. First, the reconstructed equations guarantee determinism, since their solutions converge toward bounded attractors. Second, the leading Lyapunov exponent is positive at all four sites, meaning that trajectories starting from nearly identical states diverge exponentially, the hallmark of sensitivity to initial conditions. Third, the Kaplan–Yorke dimensions of the attractors fall between 2.31 and 2.85, non-integer values confirming their fractal geometry. First-return maps computed from Poincaré sections show the folding structures typical of chaotic flows, and at Duck a color-tracer analysis revealed strong stretching, folding and squeezing of trajectories across the section. Notably, the attractors are thick and only weakly dissipative, a regime rarely documented in real-world observational data.

The practical consequence of this chaos is a sharply finite predictability horizon. Ensemble forecasts initialized from slightly perturbed conditions show that the models retain useful skill for roughly half a year when judged by the mean error, but only about one month before ninety percent of runs exceed a twenty-five percent relative error. This dual structure mirrors the observed behavior of real beaches: the seasonal rhythm remains partially predictable because the annual wave cycle is embedded in the system’s memory, while the precise trajectory of erosion and recovery becomes irreducibly uncertain beyond weeks. The authors suggest that this intrinsic instability, rather than inadequate forcing data alone, may explain why shoreline prediction skill has saturated across a recent large-scale benchmarking of models of very different complexity.

The study also carries a conceptual message for coastal science. Because the reconstructed models are fully autonomous, with no external forcing prescribed, the influence of waves, tides and sea-level anomalies is implicitly encoded in the internal state variables, which integrate forcing over time through their own nonlinear structure. Shoreline and sandbar are therefore mutually coupled rather than hierarchically linked, participating in closed feedback loops whose strength and sign depend on the system’s state. This implies that part of the observed variability on sandy beaches is internally generated, and that process-based models neglecting these nonlinear feedbacks may misattribute it to stochastic forcing or parameter uncertainty. The framework offers a typology of shoreline–sandbar coupling, from the translation-dominated Gold Coast to the near-hyperchaotic Duck, and points toward regime-dependent reconstructions that could track how feedback structures shift as storm climates change under global warming.

Subject of Research: Chaotic dynamics of coupled shoreline and sandbar systems on sandy beaches inferred from satellite time series

Article Title: Sandy beaches' chaos: shoreline-sandbar coupling inferred from observational time series

Article References: Aparicio, M., Mangiarotti, S., Frugier, S., Lacaze, L., Graffin, M., & Almar, R. (2026). Sandy beaches' chaos: shoreline-sandbar coupling inferred from observational time series. Nonlinear Processes in Geophysics, 33(2), 197-231. https://doi.org/10.5194/npg-33-197-2026

Image Credits: AI Generated

DOI: 10.5194/npg-33-197-2026

Keywords: sandy beaches, shoreline, sandbars, chaos, nonlinear dynamics, morphodynamics, satellite remote sensing, Sentinel-2, Lyapunov exponents, global polynomial modeling, coastal forecasting, surf zone