The unquantified mass loss of Northern Hemisphere marine-terminating glaciers from 2000–2020

Marine-terminating glaciers lose mass at the ice–ocean interface through frontal ablation, which includes iceberg calving, submarine frontal melting, and subaerial melting/sublimation. Quantifying frontal ablation is essential to partition total glacier mass loss and to refine sea-level budgets, freshwater fluxes to the ocean, impacts on marine ecosystems, and iceberg hazards. Prior to this study, comprehensive observation-based estimates for the entire Northern Hemisphere were lacking due to limited, consistent datasets of ice thickness, surface velocity, and terminus position, especially at high latitudes. The authors aim to produce the first hemispheric, observation-based estimates of frontal ablation for all Northern Hemisphere marine-terminating glaciers (excluding the Greenland Ice Sheet proper) by combining ice discharge across near-terminus flux gates with mass change due to terminus retreat or advance for two decades: 2000–2010 and 2010–2020.

Modeling suggested Northern Hemisphere frontal ablation of ~39 Gt a−1 (1980–1999) and projected 50.6 ± 23.8 Gt a−1 for 2020–2040. Previous global glacier ablation estimates indicated that frontal ablation comprised about 10% of the 882 Gt a−1 global glacier ablation rate, with roughly half occurring in the Southern Hemisphere. Observation-based frontal ablation estimates existed comprehensively only for Alaska and Svalbard and were not updated since 2013. Numerous regional studies documented total mass changes via glaciological and geodetic methods, satellite gravimetry, and numerical modeling, but none provided a complete Northern Hemisphere frontal ablation rate due to prior lack of consistent ice thickness, velocity, and terminus datasets.

Study scope and periods: Identified all Northern Hemisphere marine-terminating glaciers (distinct from the Greenland Ice Sheet) that contacted the ocean during at least part of a tidal cycle from 1999–2020 via manual inspection of Landsat and ASTER imagery. Total: 1496 glaciers. Analyses split into two decadal periods: 2000–2010 (termini ~2000 and 2010; velocities 2000–2009) and 2010–2020 (termini ~2010 and 2020; velocities 2010–2019). Frontal ablation components: Frontal ablation Af is estimated indirectly as the sum of ice discharge Dice through a near-terminus flux gate and the mass change due to terminus position change Mterm (positive for retreat, negative for advance), while accounting for climatic mass balance below the flux gate. Submarine melt beneath floating tongues is excluded and considered negligible for the Arctic. Flux gates: Manually delineated 5016 km of flux gates (3802 km after perpendicular correction) generally near termini without intersecting lost area (median 630 m up-glacier in 2010). Flux gates constrained by RGI v6.0 outlines (edited to one terminus per glacier); unconstrained ocean-terminating ice caps (e.g., Russian Arctic) used encircling gates. Gates were oriented perpendicular to flow using velocity components, medial moraines, and surface features, and subdivided into ~25 m segments for sampling. Formulation: Af = Dice + Mterm. Dice = ρ Σ(Vn Hn dn) − (Sf − Bclim tm), where ρ=900 kg m−3; Vn is the vertically averaged normal velocity (assumed 95% of surface velocity); Hn thickness along the gate; dn segment width; Sf area below the gate not involved in terminus change; Bclim climatic mass balance below the gate over tm. Mterm = (ρ ΔSterm Hterm + (ΔSterm/2)·Bclim tm)/Δt, with ΔSterm the area lost/gained between mapped termini and Hterm mean thickness of that area. Velocities: Primary source ITS_LIVE annual displacement mosaics (240 m). Supplemented with MEASURES InSAR for Greenland periphery and Sentinel-1 for the Russian Arctic. Where ITS_LIVE sparse (Svalbard, Russian Arctic, early 2000s), used SAR offset tracking (JERS-1, ERS-1, ALOS PALSAR, TerraSAR-X) at 100 m to compute winter velocities. For >82.7°N or data-poor sites, produced AutoRIFT-derived velocities from Landsat or Sentinel-2 (2016–2020). A few very small/slow glaciers assigned 5 ± 5 m a−1. Decadal means computed per glacier; seasonal variability not resolved. For 3.9% of 2000–2010 gates without data, used 2010–2020 means. Ice thickness: Used radar-derived measurements (GlaThiDa 3.0.3; NASA OIB CReSIS swaths in Arctic Canada North; additional RES in Russian Arctic). Adjusted thicknesses to epoch (2005 and 2015) with elevation changes from Hugonnet et al. Where only centerline or partial coverage existed, assumed a U-shaped cross-section to distribute thickness; minimum 10 m at margins. A total of 268 glaciers had at least one observation, contributing 69% of frontal ablation (2010–2020). For gaps, used Millan et al. modeled thickness, found to overestimate by ~135 m at gates; applied empirical, thickness-dependent debiasing. Imposed minimum average thickness of 30 m where modeled values were unrealistically small, with 20 m uncertainty. For terminus-gained/lost areas without observations, used debiased modeled thickness or 60 ± 30% of mean gate thickness; corrected unrealistic low values in 74 glaciers (primarily East Greenland). Terminus positions: Digitized summer, cloud-free termini circa 2000, 2010, 2020 using Landsat 5/7/8 (30 m) and ASTER (15 m); where unavailable, used Radarsat-1 (2000) and ALOS PALSAR (2010). Mapped in QGIS; area measurements in glacier-centered orthographic projections to minimize distortion. Climatic mass balance below gate: Modeled with PyGEM using ERA5 temperature and precipitation, calibrated to geodetic mass balance. For select glaciers with robust observations, used direct estimates (e.g., Columbia Glacier, Alaska: 8 ± 2 m w.e.; Austfonna Basins 2–3: −0.6 ± 0.3 m w.e.). For a few small glaciers lacking inputs, applied RGI-region averages. Addressed cases where |Bclim| exceeded discharge by adjusting discharge upward and Bclim downward within uncertainties to maintain physical plausibility; if not possible, neglected Bclim but assigned 100% uncertainty for discharge. Such cases comprised ~2% of total discharge in each decade. Surge-type classification: For glaciers with advancing termini (2000–2010 and/or 2010–2020), assessed morphological indicators and literature to classify surge/pulse likelihood (RGI scale 0–3) and adjusted a few inconsistent discharge–terminus mass budgets within uncertainties. Sea-level equivalents: Converted mass to sea-level using global ocean area 362.5×10^6 km2. Accounted for displacement by submarine ice via two scenarios for the ice fraction below sea level (best estimate 50%, high 90%). Noted that frontal ablation is only one mass-balance component. Uncertainties: Propagated independent uncertainties from velocity (including assumption of 95% of surface speed), thickness, gate width/geometry, Bclim, and mapped areas. Cross-validated using independent datasets and bias corrections.

- Identified 1496 Northern Hemisphere marine-terminating glaciers (excluding Greenland Ice Sheet proper). Between 2000–2010, 49 became land-terminating; an additional 120 did so in 2010–2020. - Hemispheric frontal ablation rates: 44.47 ± 6.23 Gt a−1 (2000–2010) and 51.98 ± 4.62 Gt a−1 (2010–2020). The difference suggests a modest increase but remains within uncertainties. - Ice discharge contributed 80% (2000–2010) and 78% (2010–2020) of frontal ablation; the remainder arose from terminus retreat. - Sea-level contribution from ice discharge (2000–2020): 2.10 ± 0.22 mm sea-level equivalent; an additional 0.11 ± 0.11 mm from terminus mass loss above sea level. - Regional patterns (2010–2020): Russian Arctic had the highest frontal ablation, followed by Svalbard and Alaska; Greenland periphery and Arctic Canada North were similar and lower; Iceland and Arctic Canada South were lowest. Svalbard ice discharge nearly tripled due to the surge of Austfonna Basin-3 (glacier rate ~6.10 ± 0.24 Gt a−1), while in Alaska terminus mass loss increased nearly tenfold and discharge decreased by ~15%. - Disproportionate contributions: 1% (15/1496) of glaciers account for 45% of hemispheric frontal ablation; 2% (30) account for 55%. Most glaciers (87%) each contribute <0.04 Gt a−1 but collectively only 14% of the total. - Hotspots (intensity index, 2010–2020): Northeastern Svalbard (Austfonna surge), strait between October Revolution and Komsomolets islands (Severnaya Zemlya), Hubbard and Columbia glaciers (Alaska), west coast of Novaya Zemlya, and Franz Josef Land. - Role of terminus change: Terminus retreat dominated frontal ablation in Arctic Canada South (67%), Arctic Canada North (48%), and Greenland Periphery (41%); it accounted for only 8% in Alaska, where discharge dominates. - Event examples: Matusevich Ice Shelf tributaries increased frontal ablation from 0.20 ± 0.17 Gt a−1 (2000–2010) to 1.88 ± 0.23 Gt a−1 (2010–2020) due to rapid retreat (18.91 ± 0.28 km2 a−1). The surging Vavilov Ice Cap outlet had discharge 1.24 ± 0.01 Gt a−1 but frontal ablation 0.29 ± 0.10 Gt a−1 due to advance (mass storage at the front). - Terminus mass budget (2010–2020): Net terminus mass loss 11.57 ± 3.81 Gt a−1 = 14.54 ± 5.97 Gt a−1 retreat − 2.98 ± 0.70 Gt a−1 advance. - Comparison with regional mass balances: Accounting for ice lost/gained below sea level (5.16–9.29 Gt a−1) implies Northern Hemisphere glacier net mass loss (223.7 ± 14.6 Gt a−1 for 2000–2020) may be underestimated by ~2–4% in geodetic assessments (this component does not affect sea level). Climatic-basal balance inferred positive in the Russian Arctic (+6.38 ± 5.27 Gt a−1 higher frontal ablation than net loss suggests positive climatic-basal), near-zero in Svalbard (+1.71 ± 3.45 Gt a−1), and negative elsewhere. - Hemispheric ice discharge from peripheral glaciers is an order of magnitude smaller than recent Greenland Ice Sheet discharge estimates, though comparisons are limited by differing coverage and omission of terminus-change mass in GIS studies.

This work delivers the first observation-based, hemispheric estimate of frontal ablation for all Northern Hemisphere marine-terminating glaciers outside the Greenland Ice Sheet, thereby filling a key gap in partitioning glacier mass loss mechanisms. By integrating decadal ice discharge with terminus change while accounting for climatic mass balance below the gate, the study quantifies the magnitude and variability of ice delivered to the ocean. The findings show that a relatively small subset of outlets governs hemispheric frontal ablation, highlighting priority targets for monitoring, modeling, and hazard assessment. Regional contrasts (e.g., discharge-dominated Alaska vs. retreat-dominated Arctic Canada) reveal differing controls on mass loss and help refine regional sea-level and freshwater flux projections. The analysis also identifies coastal hotspots where cumulative frontal ablation elevates iceberg hazard and potential ecosystem impacts. Comparisons with regional geodetic mass balances suggest that standard approaches may slightly underestimate total net loss by excluding mass changes below sea level. Overall, the results provide critical benchmarks for calibrating glacier models and contextualize the relative role of peripheral glacier discharge compared to the Greenland Ice Sheet.

The study compiles a complete Northern Hemisphere inventory of marine-terminating glaciers and provides decadal, observation-based estimates of frontal ablation for 2000–2010 and 2010–2020. Frontal ablation averaged 44.47 ± 6.23 Gt a−1 in the 2000s and 51.98 ± 4.62 Gt a−1 in the 2010s, with ice discharge comprising about four-fifths of the total. The discharge contributed 2.10 ± 0.22 mm of sea-level rise over 2000–2020. A small number of glaciers dominate hemispheric totals, and several coastal regions are identified as hotspots of iceberg production. These results improve understanding of glacier mass loss partitioning, inform sea-level budgets, and support hazard and ecosystem assessments. Future work should prioritize sustained observations of key high-contribution outlets, improved ice thickness and velocity datasets (including seasonal variability), refined climatic-basal balance below flux gates, better treatment of submarine melt under any remaining floating termini, and extension of the record beyond 2020 to capture evolving dynamics (e.g., surges and rapid retreats).

- Submarine frontal melting below floating tongues was excluded; while considered minor due to the scarcity of floating tongues, it introduces potential underestimation where present. - Seasonal variability in ice motion was not resolved; decadal means may mask seasonal or interannual extremes affecting discharge. - Ice thickness required empirical debiasing of modeled products and assumptions about cross-sectional shape where observations were sparse; some minima/maxima thresholds (e.g., 30 m) were imposed. - Climatic mass balance below flux gates required modeling (PyGEM) with calibrations and, in cases of unphysical outcomes (|Bclim| > discharge), adjustments within uncertainties or neglect with 100% uncertainty were applied. - Some velocity gaps (especially 2000–2010 at high latitudes) were filled using later-decade means or SAR-derived winter velocities, potentially biasing decadal averages. - Assumptions for sea-level conversion regarding the fraction of ice thickness below sea level (50% best, 90% high) introduce a range in SLE estimates. - Surge-type classifications are based partly on morphological indicators and available literature, which may not capture all dynamic behaviors. - Regional comparisons to Greenland Ice Sheet discharge are limited by incomplete GIS outlet coverage and differing methodologies (e.g., omission of terminus-change mass in GIS studies).