Pumice Can Drift Against the Waves, Lab Experiments Reveal

When a submarine volcano erupts, it can seed the ocean with vast rafts of floating pumice that drift for months across entire ocean basins. The 2021 eruption of Fukutoku-Oka-no-Ba in Japan’s Izu–Ogasawara arc produced a raft that traveled more than 1,400 kilometers and clogged harbors, fishing gear, and ship engines across the Nansei Islands, with cleanup in Okinawa Prefecture alone costing more than 1.6 billion yen. Yet the physics of how individual pumice fragments ride the waves has remained poorly understood. A new laboratory study published in Ocean Dynamics by Yu-Lin K. Chang of the Japan Agency for Marine-Earth Science and Technology and colleagues shows that waves can push pumice not only forward, as expected, but in some circumstances backward, against the direction the waves travel. The finding could reshape how scientists model the dispersal of pumice rafts and back-track them to previously unobserved eruptions.

The team conducted controlled wave-tank experiments at the University of Tokyo’s Kashiwa campus, using a tank 10.2 meters long, 0.8 meters wide, and 1 meter deep, filled to a depth of 0.6 meters. A plunger-type wave generator produced regular waves at frequencies between 1.0 and 1.6 hertz, spanning wavelengths from tens of centimeters to about a meter, the scale of locally generated wind waves that are always present at sea when winds exceed a few meters per second. A plastic coil mesh beach absorbed the waves to prevent reflections, and video cameras recorded the pumice through the tank’s sidewall at roughly 30 frames per second, allowing the researchers to track each fragment’s position and orientation by following the brightness-weighted centroid of the body.

The experiments used natural rhyolite pumice quarried from the Osumi Plinian fall deposit of the 30,000-year-old Aira caldera eruption in southern Kyushu, cut into 5-centimeter cubes and 10-by-10-by-5-centimeter cuboids, alongside polystyrene analog cubes weighted with aluminum, bronze, or lead sheets to span densities from about 20 to 500 kilograms per cubic meter. Because pumice is riddled with gas-filled vesicles, water steadily infiltrates its connected pores while it floats, raising its density over time. The team measured this evolution directly: after just five minutes afloat, densities rose by roughly 140 to 470 kilograms per cubic meter, and after 24 hours the samples spanned 658 to 980 kilograms per cubic meter, still floating but sitting progressively lower in the water.

In the first experiment series, pumice cubes with positive freeboard, meaning they protruded above the mean water surface, consistently drifted in the direction of wave propagation. Drift speeds increased with both wave frequency and amplitude: at 1 hertz the cubes averaged 0.0028 and 0.0048 meters per second for plunger strokes of 1 and 2 centimeters, while at 1.5 hertz the means climbed to 0.017 and 0.047 meters per second. Strikingly, drift velocity showed no clear dependence on pumice density; the lightest cube at 658 kilograms per cubic meter moved no differently from the densest at 958. The analog cubes behaved similarly across a much wider density range, suggesting that wave properties, not buoyancy alone, set the forward drift of floating fragments.

To interpret these speeds, the researchers normalized the measured drift by the theoretical Stokes drift, the net forward transport experienced by fluid particles beneath a propagating wave. When normalized this way, velocities from different pumice densities collapsed into a similar range, confirming that Stokes drift provides the right order of magnitude. But the normalized velocity was not constant: it systematically decreased as wave steepness increased within each frequency group. In linear wave theory, drift should scale with amplitude and be independent of steepness, so this dependence points to nonlinear wave effects and to the fact that a centimeter-to-decimeter rock is not a passive tracer but a finite body with its own translation and rotation.

A key diagnostic here is the Keulegan–Carpenter number, which compares the wave-induced orbital excursion of water particles to the size of the body. In the cube experiments, KC values ranged from about 1.0 to 3.4, meaning the orbital excursion was comparable to the 5-centimeter body dimension. In this regime, flow separates around the sharp edges of the fragment, and the asymmetric velocity field beneath wave crests and troughs can generate a net form-drag force that inviscid, linear theories cannot capture. Previous laboratory work on floating bodies such as model icebergs has documented exactly this kind of wave–body interaction, and the pumice results fall squarely within that regime.

The second experiment series produced the study’s most surprising result. Three larger natural pumice cuboids, designated C1, C2, and C3, were monitored as water infiltration changed their buoyancy. C1, the densest, repeatedly approached neutral buoyancy and was nearly fully submerged, and it began moving toward the wavemaker, opposite the wave direction. At 1 hertz it still drifted forward regardless of steepness, but at 1.2 and 1.4 hertz it stalled or reversed as steepness increased to 0.1 and 0.15, and at 1.6 hertz it moved backward under every steepness tested. C3 later showed the same reversal after a week of immersion raised its density to roughly 1,000 kilograms per cubic meter, while the lighter, partially emerged C2 always drifted forward. Backward drift thus occurred only within a narrow density range near neutral buoyancy, when the fragment was almost fully submerged but still brushing the free surface.

Video analysis of the backward-drifting fragment revealed a distinctive subharmonic signature: its horizontal velocity and tilt angle oscillated with a dominant period approximately twice that of the incident wave. The tilt angle changed in step-like jumps at this doubled period, producing alternating large and small forward excursions, so the fragment effectively skipped a large displacement every other wave cycle, yielding a net stepwise drift toward the wavemaker. The researchers suggest that the weak and strongly nonlinear hydrostatic restoring force of a nearly submerged body couples its rotation to its translation, though they caution that the precise mechanism, including whether parametric resonance plays a role, remains unresolved. Tank-scale circulation was ruled out as an explanation, since comparable forcing drove different specimens in opposite directions and the analysis window ended before reflected waves could return.

The team frames its findings in a three-stage conceptual model of raft evolution. In the early stage, decimeter- to meter-scale clasts form densely packed rafts much larger than the wind-wave wavelength, and transport is dominated by currents and wind, with wave drift a secondary correction. In the intermediate stage examined here, fragmentation produces centimeter-scale fragments whose size matches the wave orbital excursion, and strong wave–body interaction makes drift sensitive to wave frequency, steepness, and buoyancy, capable of weakening, vanishing, or reversing. In the late stage, sub-centimeter fragments behave increasingly as passive tracers governed by Stokes drift and currents. Because dispersal simulations of the 2021 raft required best-fit windage values of 2 to 3 percent, higher than the typical 1 percent, the authors argue that wave-induced drift was likely folded invisibly into that empirical parameter. Explicitly representing wave–pumice interaction in the intermediate stage, they conclude, is essential for predicting where drifting pumice will strand and for tracing mysterious rafts back to their volcanic sources.

Subject of Research: Wave-induced drift of floating pumice and its reversal near neutral buoyancy

Article Title: Wave-induced drift of pumice: transitions from forward to backward transport

Article References: Wave-induced drift of pumice: transitions from forward to backward transport. (n.d.). https://doi.org/10.1007/s10236-026-01866-9

Image Credits: AI Generated

DOI: 10.1007/s10236-026-01866-9

Keywords: pumice raft, submarine volcano, Stokes drift, wave-body interaction, Keulegan-Carpenter number, Ocean Dynamics, Fukutoku-Oka-no-Ba, laboratory experiments, neutral buoyancy, volcanic hazards, pumice dispersal, physical oceanography