A new study published in The Cryosphere identifies the mechanism behind giant swirling plumes that have appeared for decades in airborne radar images of Greenland’s deep ice, structures that distort ancient internal layers at scales of hundreds of meters. The research concludes that these plumes form through a convection-like process driven by contrasts in basal heat and salinity beneath northern Greenland, overturning earlier explanations that attributed the features to freeze-on at the ice sheet’s base or migrating zones of low friction. Because disrupted ice stratigraphy feeds directly into models used to estimate future ice flow and sea-level rise, the finding carries consequences well beyond glaciology.
Why Disrupted Ice Layers Change Sea-Level Projections
Ice-sheet models rely on internal layers, visible in radar soundings, to reconstruct how ice has flowed and deformed over millennia. These layers behave like time stamps: each surface snowfall event is buried and compacted, forming horizons that, in an undisturbed column, appear as smooth, nearly parallel reflectors. When those layers are folded or overturned by plume-like structures, age–depth relationships break down, and modelers lose a key constraint on past and future ice dynamics. The plumes identified in northern Greenland cut across expected layer ages, meaning any projection that treats those layers as smooth and continuous will carry a built-in error.
Layer geometry matters because it controls how scientists infer strain rates and flow history. If layers are assumed to thin gradually with depth, models will attribute most deformation to broad-scale flow toward the coast. Swirling plumes instead indicate localized vertical motion and mixing, implying that some ice is being recycled upward or sideways rather than simply sliding toward the ocean. That, in turn, changes estimates of how quickly ice can transmit surface warming signals to depth, how fast outlet glaciers may accelerate, and how much grounded ice is vulnerable to future ocean or atmospheric forcing.
For years, two leading hypotheses competed to explain the disruptions. One proposed that basal freeze-on generates complex stratigraphy when supercooled water refreezes beneath the ice sheet, producing large folds visible in radar. The other suggested that traveling zones of reduced basal friction, sometimes called slippery patches, could produce thickness-scale folds as they migrate along the bed. Both mechanisms reproduce some features of the observed stratigraphy, but neither fully matches the geometry and spatial distribution of the swirling plumes concentrated in northern Greenland.
The 2026 study argues that convection driven by temperature and salinity gradients at the ice–bed interface explains the plume locations and shapes more accurately than either alternative. In the proposed scenario, pockets of relatively warm, saline water trapped in subglacial troughs enhance basal melting and create buoyant ice–water mixtures that rise into the overlying ice, dragging and overturning internal layers as they ascend. This matters because the choice of mechanism changes how modelers parameterize ice deformation. If convection is the dominant driver, then geothermal heat distribution beneath the ice sheet becomes a first-order input for ice-flow simulations, not a secondary correction.
Competing Mechanisms and the Radar Evidence
The debate over what creates these structures has played out across more than a decade of published work. A foundational synthesis of Greenland radiostratigraphy and age structure, built from airborne ice-penetrating radar, established the isochrone framework that later studies used to identify and classify disruptions. That framework drew on radar products from the Center for Remote Sensing of Ice Sheets at the University of Kansas, whose data repository holds geolocated echograms, ice-thickness picks, and related products collected over multiple survey campaigns. Within that dataset, most of the ice sheet shows well-behaved, gently dipping layers, but northern Greenland stands out for its clusters of vertically oriented, spiral-like features.
Against that observational baseline, a separate assessment found that subglacial supercooling-driven freeze-on has a limited ability to explain the prevalence and geometry of disrupted stratigraphy across Greenland. Modeled hydrology and refreezing can generate folded layers near major subglacial channels, yet they struggle to reproduce the repeated, columnar plumes observed away from obvious drainage pathways. Likewise, the slippery-patch hypothesis can generate transient folds where basal friction changes sharply, but those folds tend to align with flow direction and bed geometry rather than forming the quasi-circular patterns seen in the radar profiles of northern Greenland.
A 2024 modeling effort broadened the picture by exploring large-scale fold formation mechanisms including ice anisotropy, flow convergence, and perturbations from bed roughness. Under certain conditions, those processes can indeed produce plume-like distortions in synthetic radiostratigraphy, particularly where ice streams accelerate over bumps or ridges. However, the resulting folds typically fan out downstream and remain attached to topographic features, whereas the observed Greenland plumes often appear as isolated, near-vertical columns that persist over tens of kilometers of flight lines.
The convection model advanced in the 2026 paper fills a gap left by these earlier efforts. By linking plume locations to basal thermal and salinity contrasts rather than to surface velocity fields or friction patterns alone, it offers a testable prediction: plume positions should correlate with independently mapped geothermal anomalies visible in bed-echo strength data. Where radar returns from the bed are unusually bright or diffuse-signatures consistent with warm, possibly water-rich substrates-the model expects stronger convective overturning and more intense layer disruption. If that correlation holds when existing radar lines are reprocessed with geothermal priors, it would strengthen the case that convection, not freeze-on or anisotropy, is the primary driver in northern Greenland.
Another strength of the convection framework is its ability to explain the vertical extent of the plumes. Freeze-on and basal sliding anomalies tend to produce deformation that decays upward, leaving deep layers distorted but shallower horizons relatively intact. In contrast, the observed plumes often reach more than halfway to the surface, suggesting a process that actively transports deformation upward. Buoyant convective cells rising from the bed provide a natural explanation for this pattern, consistent with laboratory analogs of viscous overturning in stratified fluids.
Open Questions About Plume Distribution and Model Integration
Several significant gaps remain. The 2026 study’s convection framework has not yet been tested against every region where disrupted stratigraphy appears. Large-scale folding from bed roughness and flow convergence still offers a viable explanation in parts of the ice sheet where geothermal gradients are weak, and no published comparison has quantified how much of Greenland’s total disrupted stratigraphy each mechanism accounts for. The raw radar data needed for such a comparison exists in publicly available archives, but the specific plume coordinate tables and disruption metrics used in the new study have not been released alongside the paper, limiting independent replication.
There is also no published record of how the convection model interacts with the anisotropy parameters explored in the 2024 folding study. Ice crystal fabric orientation affects how easily ice deforms, and convection-driven overturning would itself alter fabric structure, creating a feedback loop that current models do not capture. If convection tends to randomize fabric within plumes, it could locally weaken the ice column, making it more responsive to future changes in stress and temperature. Conversely, if rising plumes align crystals in preferred directions, they might channel flow along specific pathways, reshaping the internal plumbing of the ice sheet over time.
Integrating these processes into predictive models will require new parameterizations that link basal thermal conditions, hydrology, and fabric evolution. Most large-scale ice-sheet models currently treat the bed as a boundary with prescribed friction and melt rates, while internal layers are used primarily for validation rather than as dynamic constraints. The emerging picture from northern Greenland suggests that internal stratigraphy should instead be an active component of the modeling framework, with convective plumes represented as zones of enhanced vertical mixing and altered rheology.
For sea-level projections, the immediate implication is not an abrupt revision of global numbers but a widening of uncertainty for sectors of Greenland where plumes are common. If convection accelerates the transmission of basal conditions upward, surface mass-balance changes could propagate more quickly through the ice column than models assume, potentially hastening the response of outlet glaciers that drain the affected regions. Conversely, if convective mixing distributes strain more evenly, it might delay localized thinning and buttress some fast-flowing ice streams, at least temporarily.
Resolving these competing possibilities will depend on targeted radar surveys and coordinated modeling experiments. Denser flight-line spacing over known plume clusters, combined with improved bed-echo analysis, could sharpen estimates of geothermal flux and subglacial water distribution. At the same time, ice-sheet models that explicitly include convective overturning and evolving fabric could be run in parallel with more traditional freeze-on and sliding scenarios, allowing researchers to bracket the range of plausible futures. Until then, the swirling plumes beneath northern Greenland will remain both a striking radar curiosity and a reminder that the ice sheet’s interior is more dynamic-and more consequential for sea level-than smooth layers alone would suggest.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
