Accelerating ice flow at the onset of the Northeast Greenland Ice Stream

Ice-sheet mass loss drives sea-level rise through imbalances between snowfall, melt, and calving. Between 1992 and 2020 the Greenland Ice Sheet lost the equivalent of 13.5 mm of sea-level rise, with about half due to dynamic mass loss. The Northeast Greenland Ice Stream (NEGIS) is the only Greenland ice stream extending far into the interior, initiating within 150 km of the ice divide and draining ~12% of the ice sheet into Nioghalvfjerdsfjorden (79°N), Zachariæ, and Storstrømmen glaciers. Unlike other ice streams, NEGIS geometry is not strongly controlled by topography; instead, factors such as hydrology or substrate, thermal feedbacks, or crystal orientation fabric may control shear-margin position. Recent decades saw large dynamical changes at NEGIS outlets, including speedups and retreat, with modeling indicating outlet retreat may continue for at least a century. Understanding how and how fast such changes propagate inland is essential to disentangle internal ice dynamics from boundary forcing and to improve sea-level projections. This study investigates whether, and how far, dynamical changes have propagated into the ice-sheet interior and characterizes changes in NEGIS velocities near its onset.

Prior work documents substantial dynamic mass loss from Greenland and highlights NEGIS’ unique extension into the ice-sheet interior. Studies suggest NEGIS shear margins may be governed by thermal feedbacks and fabric evolution rather than topography, implying non-fixed width and sensitivity to perturbations. NEGIS outlets (Zachariæ, 79N/Nioghalvfjerdsfjorden, Storstrømmen) have experienced recent accelerations and retreat; Storstrømmen has exhibited surge behavior. Modeling indicates outlet retreat will continue under ocean forcing for at least a century, with inland propagation expected but damped by ice dynamics. Work on ice-stream dynamics and kinematic wave theory suggests upstream signals decay with distance and frequency, while Antarctic Siple Coast observations show ice-stream reorganization potentially driven by basal processes. Accumulation trends in NE Greenland over recent centuries are small, suggesting limited surface-mass-balance forcing on interior dynamics.

The study quantified along-flow accelerations using two datasets. (1) Remotely sensed velocities: ITS_LIVE annual velocity maps (1985–2018) derived from Landsat optical feature tracking over Greenland were used. For each grid point with at least 10 years of data, separate linear trends were fit to the x and y velocity components over time, and the along-flow acceleration was computed by projecting these trends onto the local flow direction. An average cross-margin profile was constructed to summarize spatial patterns, focusing on the southeastern shear margin. (2) In-situ GPS: A strain net of 63 stakes near the EastGRIP drill site (75°38′N, 36°00′W; ~2700 m a.s.l.) on NEGIS was surveyed annually during 2015–2019. Accelerations were estimated for stakes with at least three yearly measurements, applying corrections for advective acceleration. Stake velocities and accelerations were compared with the ITS_LIVE-derived accelerations, with attention to uncertainties. Spatial masks and processing steps excluded regions lacking sufficient temporal coverage. To interpret mechanisms, an idealized 2D cross-sectional ice-stream model was used to test four scenarios affecting surface velocity responses: (i) thickening the ice stream, (ii) increasing the basal sliding zone width, and (iii) softening the ice column beneath the shear margin (i.e., reducing viscosity via thermal/fabric changes). Model outputs were compared to observed acceleration patterns, especially the location of peak acceleration relative to the shear margins. Data and models are documented in online methods; derived acceleration and model code are archived at the provided DOIs.

• Remote sensing shows large dynamical changes at the three main NEGIS outlets (Zachariæ, Nioghalvfjerdsfjorden/79N, Storstrømmen) that have propagated roughly 100 km inland. Land-terminating glaciers in the region generally decelerate, consistent with thinning theory.
• In the ice-sheet interior, along the NEGIS shear margins, there is a coherent acceleration signal extending for several hundred kilometers from the onset region to about 77°N, indicating that the ice stream is widening. The peak acceleration is located just inside the shear margin.
• GPS stake observations at EastGRIP detect accelerations on the order of a few cm/yr², generally consistent with ITS_LIVE-derived accelerations within uncertainties. One stake in a narrow, deep part of the southern shear margin deviates due to coarse spatial resolution affecting the advective correction.
• Remote-sensing acceleration uncertainties in the interior are several tens of cm/yr², limiting strict validation but showing no disagreement with GPS.
• An intervening zone with no clear acceleration pattern separates inland margin acceleration from outlet-driven changes, implying decoupling of interior and frontal dynamics.
• Idealized modeling reveals distinct spatial fingerprints: thicker ice yields acceleration concentrated in the core; expanding basal sliding peaks just outside the shear margin; softening of ice beneath the shear margin produces acceleration peaking just inside the margin. The observed pattern best matches the shear-margin softening scenario.

The spatial separation between outlet accelerations and interior marginal acceleration, with a region lacking a clear acceleration pattern, indicates that interior widening of NEGIS is unlikely to be a direct response to recent outlet changes. Kinematic wave theory supports that outlet perturbations should be strongly damped at EastGRIP distances and likely undetectable. Alternative explanations include a slow, ongoing adjustment of ice-sheet geometry to past climate/geometry changes (e.g., post-glacial) that modulate large-scale flow and drainage, but surface accumulation trends in this region are small over recent centuries, making sustained accumulation forcing unlikely. Modeled spatial fingerprints favor softening of the ice column beneath the shear margin (due to warming and/or evolving crystal fabric) over surface or basal forcing as the driver of the observed acceleration pattern. Given that NEGIS interior flow is not strongly topographically constrained, the widening may reflect a dynamical instability inherent to streaming flow. These findings imply the ice-sheet interior is more dynamically variable than previously recognized, with implications for projections of future mass loss and sea-level rise.

This study provides multi-decadal satellite and multi-year GPS evidence that NEGIS is accelerating along its interior shear margins, indicating an ongoing widening that is decoupled from recent outlet dynamics. The spatial pattern of acceleration—peaking just inside the shear margins over hundreds of kilometers—aligns with model predictions for shear-margin softening (via thermal or fabric evolution), rather than changes in surface forcing or basal sliding extent. The results suggest either a slow, long-term adjustment of ice-sheet geometry or a dynamical instability within the ice stream, highlighting the need to better constrain internal rheological evolution and its drivers. Future work should: (i) refine rheological and thermal state estimates of shear margins via borehole/ geophysical observations; (ii) improve resolution and duration of interior velocity and strain measurements; (iii) develop models that couple evolving fabric/temperature with 3D ice-stream geometry; and (iv) disentangle internal dynamics from external forcings to enhance sea-level projections.

• Remotely sensed acceleration estimates in the interior have uncertainties of several tens of cm/yr², limiting the strength of validation against GPS.
• GPS observations cover a relatively short period (2015–2019) and only stakes with at least three measurements could be used, reducing temporal resolution.
• Spatial resolution limitations and advective-acceleration corrections can degrade accuracy in narrow, deep parts of the shear margin.
• The modeling employs an idealized 2D cross-sectional setup that simplifies 3D geometry, basal conditions, and rheological complexity; thus, mechanistic attribution is based on pattern matching rather than full inversion.
• Potential long-memory effects (e.g., post-glacial adjustments) are not explicitly reconstructed, and accumulation trends are inferred from prior studies rather than directly constrained in this analysis.