Increased oceanic dimethyl sulfide emissions in areas of sea ice retreat inferred from a Greenland ice core

Dimethyl sulfide (DMS), produced by ocean phytoplankton, oxidizes to methane sulfonic acid (MSA) and non-sea-salt sulfate, contributing to marine cloud condensation nuclei and radiative forcing. Model results and satellite-based estimates suggest that recent Arctic sea ice decline could enhance DMS emissions, but long-term observational evidence is scarce. Aerosol MSA, a near-exclusive oxidation product of DMS, can archive changes in oceanic DMS emissions in ice cores, provided sites minimize postdepositional losses. The southeastern dome of the Greenland Ice Sheet (SE-Dome) features exceptionally high accumulation and low temperature, enabling seasonal-resolution reconstructions. This study aims to reconstruct seasonal to decadal variability of MSA flux from 1960–2014 at SE-Dome to infer changes in DMS emissions, assess links to phytoplankton (chlorophyll-a) and wind-driven gas transfer, and evaluate the impact of sea ice retreat on summertime DMS emissions.

Prior work established that DMS emissions depend on surface ocean DMS concentrations (linked to phytoplankton/DMSP production) and gas transfer modulated by wind speed, with transfer coefficients peaking at intermediate winds (approximately 8–11 m s−1). Satellite algorithms estimating surface DMS from chlorophyll-a and optical/mixed-layer parameters improved global predictions but show regional/seasonal biases due to sparse in situ DMS data. The Arctic has experienced rapid sea ice loss and increasing light availability, with model and satellite analyses indicating increased DMS emissions in recent decades. Direct long-term aerosol MSA observations in the Arctic are limited to a few stations (e.g., Alert, Barrow, Ny-Ålesund), which showed increases post-2000 potentially tied to marginal ice zone migration. Ice cores offer continuous aerosol archives, though postdepositional losses can affect volatile species; high-accumulation, cold sites mitigate this. Previous Greenland ice cores documented long-term declines in MSA over the 19th–20th centuries, but contemporary increases post-2000 had not been reported from inland cores.

Site and dating: The SE-Dome (67.18°N, 36.37°W, 3170 m a.s.l.) features high accumulation (~1.01 ± 0.22 m w.e. yr−1) and low mean annual temperature (−20.9 °C). A 90.45 m ice core was analyzed. Dating was achieved by matching δ18O from the core with isotope-enabled climate model simulations, yielding seasonal-scale accuracy (typically ~1 month; largest uncertainty ±2.4 months in Oct 2004). Age uncertainty impacts on interannual/seasonal MSA were assessed via Monte Carlo resampling (1000 realizations), showing negligible differences from the original MSA flux series. Sampling and chemistry: The core was cut into 100 mm sections in a cold clean room, decontaminated, and melted in cleaned polyethylene bottles. Methane-sulfonate (reported as MSA) was measured by ion chromatography (Thermo Scientific ICS-2100) with a Dionex AS-14A column and 23 mM KOH gradient eluent; 1 mL injection volume. From 1960–2014, 395 samples (~7 per year) were analyzed. MSA flux (MSAflux) for annual and seasonal periods (winter JFM, spring AMJ, summer JAS, autumn OND) was calculated as concentration multiplied by the water-equivalent thickness for each sample/season. Meteorology/transport: Seven-day backward air mass trajectories from SE-Dome were computed with NOAA HYSPLIT using NCEP reanalysis, initial heights 10–1000 m AGL, initialized every 6 hours for 1960–2014. Air masses above 1000 m AGL were excluded to focus on surface sources. Existence probabilities (1° grid) were weighted by daily precipitation at arrival (ERA5). Cumulative distribution function (CDF) regions (CDF40/60/80) of air mass probability were defined. Chlorophyll-a: Merged multi-sensor GlobColour Chl-a products (SeaWiFS, MERIS, MODIS-Aqua, VIIRS) were used (8-day and monthly; 4 km resolution). Spatial means were computed for the Irminger Sea box (55°–62°N, 22°–38°W) and within CDF regions as needed. Wind and sea ice: The frequency of intermediate winds was defined as the fraction of hours with 10 m wind speeds between 8–11 m s−1 from ERA5 (0.25°), excluding grids with sea ice concentration >10%. Sea ice retreat day was defined as the latest day in April–September when daily sea ice concentration fell below 10% at a grid cell (ERA5 hourly sea ice aggregated to daily). Open water extent ratios within CDF regions were computed by integrating sea ice concentration with grid cell area. Accumulation assessment: Seasonal accumulation rates were taken from previous SE-Dome reconstructions, showing no significant seasonal differences and no spring/summer trends over 1960–2014, to evaluate precipitation impacts on MSA deposition.

- Mean annual MSA flux (MSAflux) over 1960–2014: 0.39 ± 0.31 mg m−2 yr−1. Period means: 1960–1971: 0.45 ± 0.18; 1972–2001: 0.21 ± 0.20 (−0.18 vs full period, p < 0.01); 2002–2014: 0.73 ± 0.32 (+0.34 vs full period, p < 0.01). - Interannual trend 1960–2001: significant decrease in annual MSAflux (−0.0061 mg m−2 yr−2, p < 0.05), consistent with declines in other Greenland cores; post-2002, MSAflux increased markedly at SE-Dome. - Seasonal patterns: 1960–1971 peaked in June; 1972–2001 peaked in July with rapid decline by September; 2002–2014 maintained high values July–September. Summer (JAS) MSAflux in 2002–2014 was 3–6 times higher than 1972–2001. - Accumulation rates showed no seasonal differences (spring 0.25 ± 0.07 m, summer 0.25 ± 0.09 m, autumn 0.27 ± 0.09 m, winter 0.25 ± 0.07 m) and no spring/summer trend, indicating flux increases reflect atmospheric MSA changes rather than precipitation. - Spring (AMJ): Trend not significant (+0.0002 mg m−2 yr−2, p = 0.8888). Strong correlation between spring MSAflux and Irminger Sea Chl-a (r = 0.69, p < 0.01). No correlation with frequency of intermediate winds (r = 0.04, p = 0.87). Transport pathways remained stable. - Summer (JAS): Significant increasing trend (+0.0037 mg m−2 yr−2, p < 0.05). Correlation maps with Chl-a, wind frequency, and sea ice concentration did not yield significant grid-scale correlations, but trajectory analyses highlight source regions along SE Greenland. - Sea ice retreat timing: In the summer CDF40 region, sea ice retreat occurred in July (DOY ~182–212) during 2002–2014, about 1 month earlier than 1972–2001 (August) and much earlier than 1960–1971 (September). Open water extent ratios increased across months, with complete absence of sea ice in July since 2002 within CDF40. - Biological and wind conditions: Chl-a concentrations in CDF40 were substantially higher in 2002–2014 compared to 1998–2001; frequency of intermediate winds showed no substantial difference between these periods. The frequency of higher Chl-a occurrences increased markedly in 2002–2014. - Interpretation: Earlier sea ice retreat increases light availability, enhances phytoplankton production, lengthens bloom duration, and elevates summer DMS emissions, recorded as higher MSAflux at SE-Dome.

The 55-year SE-Dome ice core record provides observational evidence linking sea ice retreat to increased oceanic DMS emissions in regions influencing southeast Greenland. The stable spring transport regime and strong correlation between spring MSA flux and Irminger Sea chlorophyll-a indicate that springtime atmospheric MSA reflects offshore phytoplankton variability, rather than wind-mediated gas transfer. In contrast, the pronounced summer increase in MSA flux after 2002 aligns with earlier sea ice retreat within air mass source regions and higher chlorophyll-a, consistent with enhanced primary production and DMS emissions. The lack of significant changes in accumulation and the stability of transport patterns strengthen the attribution to source-region ecological and sea-ice changes. These findings fill a key observational gap highlighted by previous modeling and satellite studies, supporting the hypothesis that Arctic sea ice decline elevates biogenic sulfur emissions, with potential implications for cloud condensation nuclei and regional radiative forcing. Differences from inland Greenland ice cores, which did not show post-2002 increases, underscore spatial heterogeneity tied to source regions proximal to the SE Greenland seasonal ice zone.

This study reconstructs seasonal MSA fluxes from a high-accumulation SE-Dome Greenland ice core over 1960–2014, providing a long-term proxy record of oceanic DMS emissions. Spring MSA flux covaries with Irminger Sea chlorophyll-a, demonstrating a robust phytoplankton linkage. Summer MSA flux increased sharply after 2002—by a factor of 3–6 relative to 1972–2001—coinciding with a roughly one-month earlier sea ice retreat and elevated chlorophyll-a in coastal source regions, indicating enhanced summer DMS emissions due to increased light and extended bloom periods. These results offer observational support for model-predicted increases in Arctic DMS emissions associated with sea ice loss. Future work should expand similar high-resolution records across the pan-Arctic, integrate in situ DMS measurements with satellite products to reduce regional biases, and assess impacts on cloud microphysics and radiative forcing within climate models.

- Proxy-based inference: MSA is an oxidation product of DMS and a widely used proxy, but conversion yields and atmospheric processing can vary; direct DMS flux measurements were not available. - Spatial representativeness: The SE-Dome record reflects air masses weighted by precipitation and transport patterns toward southeast Greenland; results may not generalize to the entire Arctic. - Correlation limits in summer: Despite strong physical reasoning, summer MSA flux did not show robust grid-scale correlations with Chl-a, wind frequency, or sea ice concentration, suggesting complex, possibly sub-grid or timing-related controls. - Dating and analytical uncertainties: Although dating uncertainty was assessed via Monte Carlo and found not to alter conclusions, residual uncertainties remain. Analytical uncertainties in ion chromatography and potential minor postdepositional effects cannot be entirely excluded. - Limited ancillary datasets: Sparse in situ DMS data and reliance on reanalyses/satellite products can introduce biases; wind-driven gas transfer was simplified to an intermediate wind frequency metric.