Unprecedented decline of Arctic sea ice outflow in 2018

The study investigates why sea ice export through Fram Strait, the dominant pathway for Arctic sea ice outflow and a key component of the Arctic–North Atlantic freshwater cycle, experienced a record decline in 2018. Sea ice export modulates freshwater delivery to the Nordic Seas and downstream North Atlantic, influencing surface stratification, marine ecosystems, European weather and climate, and deep-water formation. Historically, Fram Strait accounts for ~90% of the Arctic’s sea ice export, and 20th-century amounts were comparable to the liquid freshwater transport in the East Greenland Current (EGC). Recent decades have seen reduced sea ice export, primarily linked to declining mean ice thickness in Fram Strait (~15% per decade). The physical setting includes strong seasonal freshwater export in the western Fram Strait, interactions between recirculating warm, saline Atlantic Water (AW) and fresher Polar Water along the EGC/Polar Front, and the Transpolar Drift stream delivering sea ice from Siberian shelves toward Fram Strait. A persistent SLP gradient between Greenland and the Barents Sea typically drives northerly winds and southward ice motion through Fram Strait. The paper aims to document the magnitude of the 2018 decline, diagnose its causes using in situ ULS measurements, satellite remote sensing, backward ice-trajectory analysis, and atmospheric reanalysis, and assess implications for the Arctic–Atlantic system.

Prior work establishes the central role of Fram Strait in Arctic freshwater export (~90% of sea ice outflow) and its climatic importance for the Nordic Seas and North Atlantic. Historical studies showed that sea ice export and EGC liquid freshwater transport were comparable in magnitude in the 20th century. Over the last two decades, Fram Strait sea ice thickness has declined by ~15% per decade, contributing to reduced export, with models projecting continued declines and a changing Arctic freshwater budget. The regional oceanography of Fram Strait includes AW recirculation with strong mesoscale variability and subduction beneath Polar Water, and the Transpolar Drift transporting ice from Siberian shelves. Atmospheric circulation patterns such as the SLP gradient between Greenland and the Barents Sea maintain southward ice motion, while modes like the Arctic dipole anomaly can influence winter export. Observations have also documented AW reaching far north of Svalbard, contributing to ice melt north of Fram Strait, and increasing ocean heat transport to the Arctic, potentially preconditioning the system for anomalous events.

Observational array and thickness: An array of four ocean moorings (F11 ~3°W, F12 ~4°W, F13 ~5°W, F14 ~6.5–7°W) at ~79°N (1990–2001) and 78.8°N (2002–2018) in the EGC recorded sea ice draft via Upward Looking Sonars (ULS; ES300) and Ice Profiling Sonars (IPS; IPS4/IPS5). Draft was processed (screening, sound speed corrections) and converted to thickness using a Fram Strait draft-to-thickness ratio (1.136). Daily and monthly mean thickness and effective thickness (thickness weighted by sea ice concentration to include open-water fraction) were computed. Sea ice concentration and drift: Sea ice concentration was taken from EUMETSAT OSI SAF climate data records (OSI-409 v1.2 to April 2015; OSI-430 v1.2 thereafter). Ice drift used NSIDC Polar Pathfinder Daily 25 km sea ice motion vectors, version 4.1 (1979–2018). For validation, a combined product merged OSI-405, Kimura, and NSIDCv2 for post-2002 periods. Satellite thickness products: For trajectory analysis and validation in freezing seasons (October–April), three CryoSat-2/SMOS-derived thickness products were used independently: AWI CS2SMOS (weekly), Univ. Bristol CryoSat-2 Baseline-C product, and NERC-CPOM CryoSat-2 product. Envisat-based thickness was used for 2002–2010 winter validation. Volume transport calculation: The Fram Strait section was discretized into 0.2° zonal segments. Monthly sea ice volume flux at location x was f(T,A,V)=T(x)·A(x)·V(x), where T=thickness (from interpolation/extrapolation of mooring-derived monthly effective thickness), A=ice concentration (monthly from daily OSI), and V=ice drift normal to the section (monthly from daily NSIDCv4). The total flux F was obtained by zonal integration from x_w=13°W (no motion year-round) to a time-varying eastern edge x_e set by ice extent. Interpolation within the mooring array used linear schemes supported by high inter-mooring correlations; extrapolation outside used the outermost mooring thickness and a monthly mean zonal gradient α estimated from 2010–2018. Annual means were computed from October–September to capture full seasonal cycles. Uncertainty estimation: Monthly flux uncertainty at each segment was computed by propagation of relative uncertainties in T, A, and V (including systematic and random components). A monthly 2.37% uncertainty was assumed for concentration; drift uncertainty followed an empirical function that increases with higher drift and lower concentration; thickness uncertainties reflected instrument accuracies and added terms for extrapolation using the standard error of α. Integration over the section yielded monthly flux uncertainty; annual uncertainties were derived from monthly uncertainties via error propagation. The largest contribution to monthly uncertainty was from zonal thickness extrapolation (49–73%), with concentration and drift contributing 13–24% and 13–26%, respectively. Backward trajectory analysis: Pseudo ice floes were initialized on the Fram Strait section (8 points from 0° to 10°W) and advected backward daily up to 4 years using NSIDCv4 drift, with Gaussian-weighted interpolation (e-folding 25 km). Trajectories were terminated if no drift data were within 25 km or local concentration fell below 15% (concentration interpolated with 12.5 km e-folding). Backward starts were on the 15th of each month (1990–2018). For 2011–2018, CS2/SMOS thickness products reconstructed monthly thickness evolution along trajectories (Oct–Apr), and ERA5 provided 2 m air temperatures along paths. Trajectory positional uncertainty was quantified against 83 IABP buoy tracks; mean errors were ~100 km after 3 months and ~230 km after 1 year, used as Gaussian kernels to average fields along trajectories. Atmospheric/oceanographic context and hydrography: ERA5 monthly/daily SLP, 10 m winds, and 2 m temperatures characterized the 2017–2018 anomalies. A dipole index (Barents Sea minus Greenland SLP) followed Tsukernik et al. (2010). Hydrographic impacts were assessed from spring CTD sections in western Fram Strait (2002, 2005, 2007, 2008, 2018), computing 2018 salinity anomalies (0–30 m and 0–70 m) relative to the multi-year mean and comparing with potential anomalies from the observed sea ice thickness reduction.

- Record-low sea ice volume export in 2018: Annual southward volume transport was 590 km³/yr, the lowest since records began in 1990, representing a 60% reduction relative to 2000–2017 and a 75% reduction relative to the 1990s. It was 36% below the previous minimum in 2013. - Thickness and concentration drop: Annual mean effective ice thickness across the mooring array decreased by 0.74 m from the preceding year, yielding a 55% reduction in ice volume over 3°–6.5°W at 78.8°N. A pronounced reduction in sea ice concentration occurred, especially in the eastern EGC. - Drift slowdown: Southward ice drift from January to July 2018 was 66% slower than the 1990–2018 average, producing the weakest annual mean drift since 1992. - Upstream origins/pathways: Backward trajectories show 2018 ice had similar origins/pathways to 2011–2017, but 3 months prior to arrival floes were positioned 200–300 km closer to Fram Strait (north of 82°N), prolonging residence in a region with strong AW heat flux to the ice. - Rapid pre-arrival thinning: Along 2018 trajectories, sea ice was thicker than in 2011–2017 for 13–30 months prior, but thinned rapidly in the last 3 months before arrival (Feb–Apr). Estimated additional thinning due to extended AW exposure was 20–50 cm in winter and 50–100 cm in spring. - Atmospheric anomalies: ERA5 shows repeated east–west SLP dipole anomalies over the Atlantic Arctic in 2017–2018, with a Barents Sea high and Greenland/Canadian Archipelago low, strongest in Feb 2018. These reversals generated southerly winds that reduced or halted southward ice motion, warming the region and increasing residence time north of the strait. - Thermal forcing: Along trajectories, February 2018 2 m air temperatures were nearly 10°C above 2011–2017 means, limiting ice growth but remaining below −10°C (insufficient alone to cause drastic thinning), indicating ocean-driven melt as key. - Hydrographic signal: In May 2018, the upper EGC (0–30 m) exhibited a −0.35 psu salinity anomaly versus prior springs, less than one-third of the potential freshening (−1.21 psu) expected from the estimated 1.22 m ice thickness reduction, implying much meltwater remained upstream, mixed downward, or was exported differently in time/space.

The findings demonstrate that the unprecedented 2018 collapse in Fram Strait sea ice export arose not from changes in source regions or pathways, but from regional atmospheric circulation anomalies that reduced southward drift and prolonged floe residence north of the strait. This increased exposure to Atlantic Water heat flux led to rapid ocean-driven thinning just prior to export, compounded by reduced ice concentration and slowed drift, jointly minimizing volume transport. The results clarify that interannual variability and extreme reductions in Arctic sea ice outflow can occur through regional-scale atmospheric forcing and ocean–ice interactions, in addition to broad Arctic-wide thinning trends. The downstream implications include altered freshwater delivery to the Nordic Seas and North Atlantic, changes in surface buoyancy along the EGC, potential impacts on deep-water formation, enhanced winter air–sea interaction and water mass modification due to a more poleward ice edge, and ecological effects on primary production. The observed hydrographic anomalies and documented extreme mixed layers northeast of the strait in 2017/2018 are consistent with enhanced local melt and stratification changes, possibly signaling a northward shift of marginal ice zones and transformation regions. The study underscores the need for sustained, integrated observations (Fram Strait AOO and subpolar arrays) to quantify cascading impacts and to assess the frequency and drivers of such atmospheric dipole events in a changing climate.

This study documents a record-low Arctic sea ice volume export through Fram Strait in 2018 (590 km³/yr), amounting to a 60–75% reduction relative to recent decades. The minimum resulted from regional atmospheric dipole anomalies that induced southerly winds, stalled southward ice motion, and extended floe residence in a zone of strong Atlantic Water–driven ice melt north of the strait. Despite warmer air temperatures, ocean heat flux emerged as the dominant cause of rapid pre-arrival thinning. The event highlights that extreme reductions in sea ice outflow can arise from regional atmospheric/oceanic processes, not solely from basin-wide thinning trends, with significant ramifications for Arctic–Atlantic freshwater pathways, ocean stratification, deep-water formation, and marine ecosystems. Future work should focus on: sustained monitoring of sea ice and upper-ocean properties in the Atlantic sector of the Arctic; integrated analyses linking Arctic outflows to subpolar overturning arrays (e.g., OSNAP, RAPID-MOCHA); and investigation of the dynamics, teleconnections, and future likelihood of Barents–Greenland SLP dipole patterns, including links to NAO, blocking, and stratospheric events.

- Spatial representation: Thickness measurements are limited to four mooring sites across the EGC, requiring interpolation within and extrapolation outside the array; uncertainty is dominated by the extrapolated zonal gradient (49–73% of monthly flux uncertainty). - Product limitations: Satellite thickness products are available only in the freezing season and differ in absolute magnitude; merged drift products for validation can introduce artificial temporal variability. - Trajectory uncertainty: Backward trajectory positions diverge from buoy-derived tracks by ~100 km after 3 months and ~230 km after 1 year, affecting reconstructed upstream conditions. - Atmospheric attribution: While the 2018 SLP dipole and southerly wind anomalies are identified, the underlying causes of their repeated occurrence and links to larger-scale climate modes remain unresolved. - Hydrographic inference: Salinity anomaly estimates rely on a limited set of spring CTD sections and cannot fully resolve the timing, pathways, and vertical redistribution of meltwater.