Reduced marine phytoplankton sulphur emissions in the Southern Ocean during the past seven glacials

Marine biogenic sulphur affects Earth’s radiation budget and may be an indicator of primary productivity in the Southern Ocean, which is closely related to atmospheric CO2 variability through the biological pump. Previous ice-core studies in Antarctica show little climate dependence of marine biogenic sulphur emissions and hence primary productivity, contradictory to marine sediment records. Here we present new 720,000-year ice core records from Dome Fuji in East Antarctica and show that a large portion of non-sea-salt sulphate, which was traditionally used as a proxy for marine biogenic sulphate, likely originates from terrestrial dust during glacials. By correcting for this, we make a revised calculation of biogenic sulphate and find that its flux is reduced in glacial periods. Our results suggest reduced dimethylsulphide emissions in the Antarctic Zone of the Southern Ocean during glacials and provide new evidence for the coupling between climate and the Southern Ocean sulphur cycle.

and Dronning Maud Land (EDML). a Flux of ssNa + at DF and EDC over the past 720,000 years. b δ 18 O 1 at DF over the past 720,000 years. c Flux of nssNa + at DF, EDC, and EDML over the past 150,000 years. d δ 18 O 1 at DF over the past 150,000 years. The EDC and EDML fluxes are plotted on the AICC12 timescale 2,3 using previously published ion data 4,5,7 and accumulation rates 2,3 . All data are averages over 1000 years. The similarity between ssNa + flux at DF and EDC suggests that both sites are affected by inland air masses, which are likely a mixture of air masses from different oceanic sectors 8,9 . 9 using equivalent seawater ratios of K + /Na + = 0.02 and Mg 2+ /Na + = 0.11. We use [ion] for the equivalent concentration/flux of each ion. Much lower equivalent concentrations of nssK + compared with nssCa 2+ (less than 5% for most of DF2 core, not measured for DF1 core) suggest that contributions of K-containing minerals are much lower than those of CaSO4. Equivalent concentrations of nssMg 2+ are correlated with those of nssCa 2+ , with the former being about 35% of the latter. This suggests that during cold periods in glacials, a major part of nssMg 2+ could also originate from evaporites in the form of MgSO4 or CaMg(CO3)2, as deduced from the strong correlation between the Mg 2+ and SO4 2− fluxes. To remove terrestrial MgSO4 contributions from the nssSO4 2+ flux, we should subtract the nssMg 2+ only associated with nssSO4 2− .

Supplementary Discussion
It should be noted that South American dust source areas include CaCO3-rich sources 10 , as well as gypsum-rich sources 11,12 . Because CaCO3 is not a major chemical form of Ca 2+ in ice cores from Antarctic interior sites 13 , previous studies assume that CaCO3 reacts with H2SO4 or SO2 and forms CaSO4 14,15 . Ca 2+ , however, has also been reported to be associated with NO3 − in the Antarctic interior during cold periods in glacials when dust concentrations increases [16][17][18] .
Supplementary Fig. 4a displays a strong correlation between the NO3 − flux and nssCa 2+ flux at DF, which indicates the presence of Ca(NO3)2 at DF. Recent studies show that CaCO3 reacts more readily with HNO3 or NOx than with SO2 or H2SO4 and that the reaction with SO2 or H2SO4 is very slow 16,[19][20][21][22] . We therefore deduce that primary continental gypsum is a major contributor to the CaSO4 flux and that secondary CaSO4, formed by the reaction between CaCO3 and H2SO4/SO2, is only a minor contributor. Here, we assume that CaCO3 completely reacts with HNO3/NOx and that CaSO4 originates solely from continental gypsum. Similarly, some nssMg 2+ would be associated with NO3 − because MgCO3 and CaMg(CO3)2, as well as CaCO3, react readily with HNO3/NOx and form Mg(NO3)2 19 . A strong correlation between nssMg 2+ flux and NO3 − flux at DF provides evidence of this reaction. Using the following equation, we can better estimate residual nssSO4 2− concentrations [nssSO4 2− ]res, a revised proxy for marine biogenic sulphate, because nssCa 2+ and nssMg 2+ are both associated with NO3 − .
Here we assume that the majority of terrestrial CaCO3, MgCO3, and CaMg(CO3)2 are converted to Ca(NO3)2 and Mg(NO3)2 by HNO3/NOx. This assumption is supported by a strong correlation between the [NO3 + nssSO4 2− ] flux and [nssCa 2+ + nssMg 2+ ] flux at DF during cold periods with increased dust flux and by the slope (~1) of the lower bound of the scatter plot ( Supplementary   Fig. 4b).
Although residual nssSO4 2− fluxes calculated with different assumptions give different r 2 values (Fig. 4), both decrease during cold periods in glacials. Interestingly, the first-order approximation, assuming that nssCa 2+ is associated only with nssSO4 2− and that nssSO4 2− is associated only with nssCa 2+ during cold periods in glacials, is similar to the estimation of residual nssSO4 2− flux using the equation above. Although nssMg 2+ and NO3 − data are not available for EDC and EDML cores, it is expected that the calculation of residual nssSO4 2− using only nssCa 2+ and nssSO4 2− would be fairly reasonable.

11
Causes of reduced biogenic sulphate flux during glacials A tracer transport model using the DMS concentration data of [23] shows that in austral summer, more than 60% (80%) of the geographic origin of DMS leading to sulphate at Vostok ( Supplementary Fig. 1), an East Antarctica interior site, is located south of 55°S (50°S) 24 .
Although the climatology of summer DMS concentrations has been recently revised, mean summer DMS concentrations in the SO are close to the original estimates 23,25 . The Antarctic Polar Front, the northern edge of the AZ, lies roughly between 50°S and 60°S 26 . These studies suggest that a major source region of DMS for the Antarctic interior is the AZ. A study of sulphur cycling at the LGM based on atmospheric general circulation and sulphur chemistry models indicates that biogenic sulphate flux on the Antarctic ice sheet is sensitive to both oceanic DMS concentration and spatial distribution 27 . The latter strongly depends on the extent of summer sea ice 27 . The model study also shows that the flux is almost proportional to oceanic LGM sea ice extent is about two times greater than the present day, the LGM summer extent appears to be only slightly greater, except for the Weddell Sea area, where the occurrence of sporadic sea ice is observed around the present-day winter sea ice edge 28 . DMS-derived sulphur species in EDC, EDML, and Siple Dome ice cores have been traditionally assumed to be dominated by those from the nearby Indian, Atlantic, and Pacific Ocean sectors of the SO, respectively [4][5][6]29 . Accordingly, DF should be mainly affected by the Atlantic and Indian Ocean sectors. The large difference in summer sea ice extent between the Atlantic and Indian Ocean sectors would lead to differences in the marine biogenic sulphate flux between DF and EDC.
However, this is not the case. DF and EDC show rather similar fluxes and variations in the marine biogenic sulphate flux (Fig. 3a).
According to recent back-trajectory studies, the Antarctic interior sites (e.g. DF and EDC) are affected by inland air masses, which are likely a mixture of air masses from different oceanic sectors, as well as by air masses from the ocean sectors they face 8,9 . Thus, the marine biogenic sulphate fluxes at DF and EDC reflect variations in the overall DMS emissions in the AZ. This result is supported by the similarity between ssNa + fluxes at DF and EDC, which likely originates from the sea ice surface 4,5,7,29 (Supplementary Fig. 7). Overall, the summer sea ice field around Antarctica at the LGM is estimated to be about 1.25 to 1.50 times that of today, although data remain sparse 28 . This finding points to similar or only slightly more distant source areas for DMS during glacials compared to interglacials. Hence, the large glacial/interglacial variability in the marine biogenic sulphate flux at DF and EDC would be dominated by glacial/interglacial variability in DMS emissions in the AZ, rather than changes in transport distance.
EDML, a site close to the Weddell Sea where summer sea ice during glacials seems to have expanded much farther compared to other ocean sectors 28 , shows higher biogenic sulphate fluxes than DF and EDC (Fig. 3c), as well as higher ssNa + fluxes ( Supplementary Fig. 7c). This can be 13 explained by more efficient transport of marine air masses to EDML than to the sites located farther inland (DF and EDC) 8,9 . Reduced marine biogenic sulphate flux at EDML during glacials could be a result of the longer transport distance from the summer sea ice edge. Reduced DMS emissions during glacials should also play a role considering the strong DMS emissions from leads and polynyas in the present-day summer sea ice zone 29 . Although we cannot yet conclude which of the two causes is more important, the coherent variability in marine biogenic sulphate fluxes at EDML, DF, EDC, and Siple Dome suggests reduced DMS emissions in the Atlantic Ocean sector, as well as in other ocean sectors.
To evaluate the effect of solar irradiance on DMS emissions, we calculated the integrated summer insolation at 55°S, the latitude of a major source region of DMS 24 . Since the modern daily insolation at the spring and autumnal equinox at 55°S is about 250 W/m 2 30 , we integrated daily insolation >250 W/m 2 over the year 31 . We calculated power spectra with the Blackman-Tukey method (30% lag) using the Analyseries software package 32 . Although strong powers were found in the 41-kyr and 98-kyr bands, similar to those for δ 18 O, the variability was less than 3%. Integrated summer insolations at 50°S, 60°S, 65°S, and 70°S give similar power spectra as 55°S. Variability of the integrated summer insolation from 50°S to 70°S was 3%-10%, depending on the latitude. This small variability would not be a major cause of the variability in DMS emissions 33 .