Centennial response of Greenland's three largest outlet glaciers

The study addresses how Greenland’s largest outlet glaciers have responded to climate forcing over the past century and what this implies for future sea level rise. Despite recent acceleration in ice loss and modelled projections of further increase, the centennial-scale dynamic response to atmospheric and oceanic variability remains poorly constrained due to limited observations. Jakobshavn Isbræ, Kangerlussuaq Glacier, and Helheim Glacier together drain ~12% of the ice sheet and hold ~1.3 m sea level equivalent. Their differing bed geometries—retrograde basins for Jakobshavn and Kangerlussuaq versus a largely non-retrograde bed for Helheim—provide a natural experiment to assess the role of fjord/bed geometry versus climate forcing in driving retreat and mass loss. The purpose is to extend observational records back to the Little Ice Age, quantify mass loss partitioned into surface mass balance (SMB) and dynamic components, and assess potential stabilizing feedbacks (solid Earth uplift and local sea level fall) on tidewater glacier dynamics.

Prior work has documented 20th–21st century changes in Greenland’s outlet glaciers using satellite-era data and limited historical observations, showing links between fjord/bed geometry and retreat, and between frontal position, thinning, and discharge. Studies used aerial stereo-photogrammetry from mid-20th century and expedition records (e.g., Weidick, Koch) to map frontal positions and localized surface lowering at Jakobshavn, Kangerlussuaq, and Helheim. However, decadal-scale mass change prior to the satellite era remained poorly constrained, and previous century-scale estimates often neglected the full retreat from the Little Ice Age maximum to modern margins or lacked basin-wide mass change at high spatial resolution. Reconstructions of SMB (e.g., Box) provided surface-driven components but did not capture dynamic losses. Overall, existing models and observations suggested strong sensitivity of tidewater glaciers to geometry and ocean/atmosphere forcing but lacked century-long, basin-wide, dynamically partitioned mass change records needed to validate models and improve projections.

• Data sources: Historical and modern aerial stereo-photogrammetric imagery (years including 1902, 1913, 1933, 1944, 1953, 1959, 1964, 1985), orthophotos (2019), and trimlines/moraines delineating Little Ice Age (LIA) maximum extent. Elevations and front positions were compiled for Jakobshavn Isbræ, Kangerlussuaq Glacier, and Helheim Glacier.
• LIA-to-present elevation reconstruction: Elevations at LIA maximum were derived from moraines and trimlines. Surface elevation change fields were reconstructed on a 0.5 × 0.5 km grid using a scale-value approach (after Kjeldsen et al.), estimating scale values separately for each time interval and extrapolating to the ice sheet interior. Key time slices include 1875, 1902, 1913, 1931, 1946, 1959, 1964, 1987, 2002, and 2012, with linear interpolation between observation gaps. The analysis explicitly includes grounded-ice loss between early 1900s margins and the 2002 margin.
• Frontal position and geometry: Frontal positions from 1875/1880 to 2018 were digitized and analyzed against fjord and bed topography (BedMachine and other datasets) to assess the role of retrograde/prograde slopes and pinning points.
• SMB–dynamic partitioning: Surface mass balance anomalies (SMB) from the Box reconstruction (5 km grid, 1840–2012; RMSE ~0.45 m w.e.) were used after removing the 1961–1990 mean to estimate SMB-induced mass changes. Subtracting SMB components from total geodetic mass change isolated the dynamically induced component.
• Solid Earth and local sea level feedbacks: The local sea level change was computed as the sum of (i) elastic uplift from contemporary ice loss, (ii) sea level and geoid change due to reduced gravity, and (iii) viscoelastic glacial isostatic adjustment (GIA). Components (i)–(ii) were obtained by convolving mass loss fields with Green’s functions for elastic Earth model iasp91 with Crust 2.0 crustal structure; GIA used the GNET-GIA model. The combined relative sea level change near grounding lines was mapped for 1880–2012.
• Error characterization and validation: DEM and photogrammetric data RMS errors: 2–4 m (1944–1964, 1985) and ~25 m (1902–1933); trimline height error ~10 m. Weight-normalized bundle adjustments yielded rmsx, y ~1.6 m, rmsz ~3.5 m for example DEMs. Mass balance time series were compared with independent reconstructions based on thickness, velocity, and SMB (1972–2018), showing agreement within uncertainties.
• Outputs: Time series of basin-wide cumulative mass change (Gt) and sea level equivalent (mm) for each glacier, partitioned into SMB and dynamic components; spatial maps of relative local sea level change; retreat distances and timing.

• Cumulative mass loss (1880/1900–2012): Jakobshavn Isbræ 1518 ± 189 Gt; Kangerlussuaq Glacier 1381 ± 178 Gt; Helheim Glacier 31 ± 21 Gt. Combined total 2930 ± 322 Gt.
• Sea level equivalent from the three basins over 1875–2012: 8.1 ± 0.9 mm (table reports ±0.9 to 1.1 mm across text/tables).
• Temporal partitioning (Gt; 1875–1932 | 1932–1964 | 1964–1981/5 | 1981–2012): • Jakobshavn: 661 ± 199 | 503 ± 199 | 91 ± 199 | 322 ± 60. • Kangerlussuaq: 870 ± 205 | 178 ± 205 | 138 ± 205 | 195 ± 30. • Helheim: −3 ± 20 | −4 ± 20 | −9 ± 20 | 44 ± 8. • Total: 1528 ± 424 | 677 ± 424 | 220 ± 424 | 561 ± 98.
• Average mass loss rates (Gt/yr; 1875–2012): Jakobshavn 11 ± 1; Kangerlussuaq 10 ± 1; Helheim ~0. Recent rates since early 2000s for Jakobshavn reached ~20–30 Gt/yr.
• Dynamic dominance: About 91% of total mass loss is dynamically induced (table reports 89–96% across intervals, 91% overall). For Helheim, losses are more evenly split between SMB and dynamics.
• Retreat behavior and geometry: Jakobshavn and Kangerlussuaq retreated tens of kilometers (~40 km at Jakobshavn), often during episodic events in the 1900s, 1930s, and 2000s, linked to atmospheric/ocean warming and loss of floating buttressing. Both currently occupy retrograde beds that deepen and steepen inland, implying heightened future sensitivity. Helheim experienced limited net retreat (~5 km by 2012) due to a pinning point and lack of immediate inland deepening, but showed strong retreat in 2005 driven by warm deep ocean water and may cross a bedrock peak ~18 km inland, potentially initiating renewed retreat.
• Solid Earth and local sea level feedbacks: Relative local sea level fell by ~2.8 m near Jakobshavn and ~3.5 m near Kangerlussuaq grounding lines, and ~0.4 m near Helheim (1880–2012) due to elastic uplift, reduced gravity, and GIA. These feedbacks are insufficient to stabilize tidewater glaciers lacking significant floating tongues; thus, the Antarctic-style stabilizing feedback is not effective here.
• Model implications: RCP8.5-driven models project 9.1–14.9 mm SLR contribution by 2100 from these three glaciers. Given that ~1.5 °C Greenland-average warming in the 20th century already yielded 8.1 ± ~1 mm SLR from these basins, and RCP8.5 implies much larger future warming (global +3.7 ± 0.7 °C, amplified over Greenland), current projections likely underestimate worst-case dynamic mass loss.
• Consistency: Reconstructions align with independent mass balance estimates for 1972–2018 within uncertainties.

The century-long, basin-wide reconstructions demonstrate that dynamic processes dominate mass loss (~90%) for Jakobshavn and Kangerlussuaq, with substantial retreat into retrograde and steepening beds amplifying sensitivity to modest atmospheric and oceanic warming. Helheim’s relative stability underscores the critical role of local geometry and pinning points; however, recent ocean-forced retreats indicate vulnerability if pinning thresholds are crossed. The negligible stabilizing effect from local sea level fall and bedrock uplift confirms that tidewater glacier dynamics in Greenland are not significantly buffered by solid Earth–sea level feedbacks. These findings provide essential constraints on glacier sensitivity and reveal that present-day rates and magnitudes of dynamic loss have precedents earlier in the 20th century, implying that models must capture episodic, geometry-modulated responses to external forcing. Consequently, projections based on RCP8.5 forcing for these glaciers likely underestimate worst-case mass loss, with implications for Greenland-wide projections.

By extending observations back to the Little Ice Age and producing high-resolution, century-long, SMB-partitioned mass change records for Greenland’s three largest outlet glaciers, the study shows that dynamic processes drive the majority of mass loss and that bed/fjord geometry strongly modulates retreat. Solid Earth and local sea level feedbacks provide minimal stabilization for tidewater glaciers. Given that modest historical warming produced sea level contributions comparable to or exceeding some end-of-century projections, current worst-case scenarios likely underestimate potential losses. Future research should: integrate long-term observational constraints into model calibration/validation; improve representation of fjord geometry, grounding line processes, and ocean thermal forcing; expand analyses to additional Greenland basins; and refine coupling among atmosphere–ocean–ice–solid Earth systems to reduce projection uncertainties.

• Temporal gaps and limited imagery, especially early 20th century, restrict resolution of whether Kangerlussuaq’s losses were episodic versus steady.
• Reconstructions rely on linear interpolation between observation epochs and extrapolation from trimlines/moraines, introducing uncertainty.
• DEM and trimline height errors (up to ~25 m for early imagery; ~10 m for trimlines) propagate into mass change estimates despite error modeling.
• SMB partitioning uses reconstructed SMB with finite RMSE (~0.45 m w.e.) and assumptions about spatial representativeness.
• Solid Earth and sea level feedback estimates depend on chosen Earth models and GIA parameters, contributing model uncertainty.
• Results focus on three basins (~12% of Greenland); extrapolation to all Greenland outlets should be done cautiously.