Climate changes modulated the history of Arctic iodine during the Last Glacial Cycle

Iodine has a significant impact on promoting the formation of new ultrafine aerosol particles and accelerating tropospheric ozone loss, thereby affecting radiative forcing and climate. Therefore, understanding the long-term natural evolution of iodine, and its coupling with climate variability, is key to adequately assess its effect on climate on centennial to millennial timescales. Here, using two Greenland ice cores (NEEM and RECAP), we report the Arctic iodine variability during the last 127,000 years. We find the highest and lowest iodine levels recorded during interglacial and glacial periods, respectively, modulated by ocean bioproductivity and sea ice dynamics. Our sub-decadal resolution measurements reveal that high frequency iodine emission variability occurred in pace with Dansgaard/Oeschger events, highlighting the rapid Arctic ocean-ice-atmosphere iodine exchange response to abrupt climate changes. Finally, we discuss if iodine levels during past warmer-than-present climate phases can serve as analogues of future scenarios under an expected ice-free Arctic Ocean. We argue that the combination of natural biogenic ocean iodine release (boosted by ongoing Arctic warming and sea ice retreat) and anthropogenic ozone-induced iodine emissions may lead to a near future scenario with the highest iodine levels of the last 127,000 years.

A tmospheric iodine, primarily emitted from oceans, forms new aerosol particles in the atmosphere [1][2][3][4][5][6][7] and efficiently destroys ozone in the troposphere and lower stratosphere [8][9][10][11] , which reduces ozone radiative forcing [12][13][14][15] . In the Arctic environment, iodine has recently been identified as a significant source of cloud condensation nucleii (CCN), with the potential to influence clouds formation 3 . Iodine in the atmosphere is increasing globally (3-fold since the 1950s), as independently evidenced by polar and alpine ice core and tree ring measurements, following anthropogenic ozone pollution and global warming [16][17][18] . In the Arctic, a coastal ice core from Greenland revealed that ocean primary productivity controlled atmospheric iodine variability during the Holocene (i.e., last 11,700 years) 19 . Before the climatically stable Holocene, the Arctic underwent abrupt environmental changes featuring vast ice sheet growth and marked climate events during the Last Glacial Period (LGP) (115-11.7 kyr before 2000 CE (b2k)) and experienced temperatures warmer than today during the Last Interglacial Period (Eemian, 130-115 kyr b2k). However, to date, the natural evolution of Arctic iodine before the Holocene, in the absence of anthropogenic forcings, and its coupling with climate variability remain unknown.
In this work, we report the iodine levels in the Arctic during the Last Glacial Cycle (LGC) (last 127 kyr) using records from two Greenland ice cores (NEEM and RECAP) drilled in northwestern and eastern coastal Greenland, respectively (Fig. 1). Both ice cores provide continuous paleoenvironmental records back to the Eemian interglacial period [20][21][22] . We provide evidence that ocean-ice-atmosphere exchange of biogenic iodine in the Arctic Ocean dominated the atmospheric iodine budget during periods of rapid climatic changes. The results show that abrupt warming and sea ice retreat in the Arctic preceded maximum iodine levels, reaching peak concentrations during interglacial periods.
Results and discussion Present-day iodine emission sources and depositional fluxes over Greenland. Atmospheric iodine in Greenland is controlled by a complex interplay involving iodine emissions from the oceans, as well as particle-bound iodine compounds related to dust and/or sea spray aerosol (ssa) variability 23 . The oceans, which are the largest reservoirs of iodine on Earth 24 , are known to be the dominant source of atmospheric iodine [25][26][27] . The major oceanic sources of gas-phase iodine are (i) inorganic, ozone-induced, hypoiodous acid and molecular iodine emissions and (ii) organic iodine from the metabolic activity of primary producers (phytoplankton and macro-and microalgae). Furthermore, atmospheric iodine can also be related to mineral dust by promoting iodate stability in the ice matrix and/or favouring the adsorption of iodine gas-phase molecules to dust particles during atmospheric transport 10 . On the other hand, ssa expelled from the ocean surface during wave breaking incorporate iodine in their composition and are a substrate where gaseous iodine species undergo heterogeneous recycling reactions [28][29][30] .
We used a global chemistry climate model 17 to understand present-day iodine levels in Greenland (see Methods). Present iodine depositional fluxes at the RECAP drilling site (71°18′ N; 26°43′ W; 2315 m a.s.l.) are~51% higher than at NEEM (77°45′ N, 51°06′ W; 2484 m a.s.l.) ( Supplementary Fig. 1). This is explained by the different ocean iodine emission strengths at the source regions of air masses that influence both sites (the Canadian Arctic, Baffin Bay and Hudson Bay for NEEM 31 and the North Atlantic and Arctic Ocean for RECAP 32 ). Note also that RECAP is closer to the coast (30 km from the North Atlantic Ocean) than NEEM (350 kms from the Baffin Bay). Our model results show higher iodine emissions in the RECAP source regions (50N-80N, 45W-12E; mean iodine emissions of 9.1 × 10 −14 kg m −2 s −1 ) than in the NEEM source areas (50N-80N, 120W-45E; mean iodine emissions of 3.5 × 10 −14 kg m −2 s −1 ) (Fig. 1).  ]) concentrations were also used as proxies of mineral dust in Greenland ice cores 33,34 . These geochemical measurements along with other ice cores and marine paleoceanographic archives allow the reconstruction of Arctic iodine variability and environmental sources during the LGC. Maximum iodine levels were found during the Eemian in the NEEM ice core, with a total mean iodine depositional flux (I flux ) of 9.6 µg m −2 yr −1 coinciding with the highest Arctic Ocean primary productivity inferred from the phytoplankton biomarker brassicasterol 35 (Fig. 2). Our analysis suggests that these high Eemian iodine levels resulted from the interplay of (i) enhanced biogenic iodine emissions driven by primary productivity; (ii) higher sea surface temperatures (SSTs) in the Arctic and the North Atlantic 36,37 , which facilitates sea-air transfer of volatiles; and (iii) dominant open water conditions in the Arctic, as reflected by the low P B IP 25 values. P B IP 25 has been used as a proxy of the intensity of sea ice cover in the Arctic 38 , as well as in Baffin Bay 31,39 since it accounts for the algal and phytoplankton (brassicasterol) biomarkers activity, which allows to assess the spatial and temporal extent of the sea ice cover 40 (Fig. 2). The Arctic sea ice extent was significantly reduced during the Eemian, and minimum sea ice concentrations towards almost ice-free summers occurred in wide areas as far north as the northern Barents Sea as a consequence of the strong inflow of warm Atlantic waters 38 , thus facilitating springtime blooms and oceanic iodine production. Indeed, summer North Atlantic SSTs during the Eemian were 2°C greater than at present 36,37 . Furthermore, a mean global sea level 6-8 m above the present level [41][42][43] (Fig. 2) triggered the flooding of most parts of the shallow Siberian marginal seas, eventually resulting in more productive coastal areas in the Arctic. At NEEM, the highest I flux was recorded at 122 kyr b2k concomitant with maximum levels of biological productivity, sea level rise and SST. The coupling of all these drivers during the Eemian would have eventually released large amounts of iodine into the Arctic atmosphere, resulting in the largest iodine emissions on record. The significant decrease in NEEM and RECAP I flux recorded during the last glacial inception at~120 kyr b2k is correlated with an increase in sea ice 31,38 and a decrease in both SST and brassicasterol (Fig. 2), as also indicated by independent modelling experiments that resemble preindustrial scenarios 41 .
Ice core iodine showed reduced values throughout the LGP, with mean NEEM and RECAP I flux of 3.8 and 5.1 µg m −2 yr −1 , respectively (Fig. 2). The LGP is characterized by abrupt climatic variability marked by Dansgaard/Oeschger (D/O) events. These climatic phases consist of warm Greenland interstadials and cold Greenland stadials that are well recorded in Greenland ice cores and marine sediments from the North Atlantic region and the Nordic seas [44][45][46][47][48][49] . To investigate in detail the iodine variability across the rapid D/O transitions, additional high-resolution iodine measurements were carried out in the NEEM ice core from 34 to 42 kyr b2k (D/O 7 to 10), with a 3-to 10-year resolution ( Fig. 3 and Supplementary Fig. 4). In order to differentiate the main sources of iodine that determine I flux variability, we have decoupled oceanic iodine emissions (I ocean ) from dust-related iodine (I dust ) ('Methods'). The sub-decadal resolution measurements of iodine during this period show the concomitant evolution of I ocean , Arctic Ocean primary productivity and sea ice variability 46,48,49 (Fig. 3). Each D/O event showed a similar sequence: (i) iodine emissions increased concurrently with a reduction in sea ice at the onset of the interstadials; (ii) iodine emissions decreased during the late interstadial periods as the climate cooled and sea ice gradually increased; and iii) negligible iodine was released from the oceans with the development of thicker sea ice conditions during the cold stadial periods.
Interestingly, NEEM and RECAP I flux and I ocean do not follow the same trend during D/O events (Fig. 3). In the context of strongly reduced iodine emissions during the LGP due to the extensive development of perennial sea ice in the Canadian Arctic 31 , a significant fraction of iodine registered in the NEEM ice core was strongly controlled by dust and, to a lesser extent, by ssa. This is corroborated by the robust statistical correlations between [I] and [Ca] and between [I] and [Na] from D/O 7-10 (r > 0.75 and 0.78, respectively). These correlations are also seen in RECAP during the LGP (r > 0. 88 10 . This Antarctic ice core reported a correlation between iodate, the most stable iodine species in the atmosphere, and atmospheric dust, indicating uptake of gasphase iodine molecules into fine dust particles during atmospheric transport 10 . The largest iodine variability occurred during the LGP-Holocene transition when iodine emissions strongly fluctuated following the main climatic phases of the Last Glacial Termination. NEEM and RECAP I ocean showed very similar evolution patterns, with minimum I ocean values during Heinrich event 1 (H1, 16.8 kyr b2k) (Fig. 4), when surface waters cooled and a thick perennial sea ice cover reached both the Norwegian Sea and Baffin Bay 31,46,52 . The I ocean increased in both ice cores at the onset of the Bølling (14.7-14.1 kyr b2k) warming in the Arctic, slightly decreased during the Intra-Allerød Cold Period (14.1-13.9 kyr b2k) and increased again during the Allerød (13.9-12.9 kyr b2k) warm period (Fig. 4). The increase in I ocean values during the Bølling-Allerød period was synchronous with an abrupt increase in ocean productivity as far north as the Fram Strait 53 . NEEM and RECAP I ocean moderately decreased during the last cold spell of the Last Glacial Termination, the Younger Dryas (12.9-11.7 kyr b2k).
Oceanic iodine emissions increased in both ice cores at the onset of the Holocene (~11 kyr b2k; Fig. 4) resulting from the interplay of different factors: (i) enhanced marine primary productivity in the Arctic 46,54-56 ; (ii) dominant open water conditions in the subpolar North Atlantic 46 ; (iii) abrupt sea level rise (~120 m) during the Last Glacial Termination 57,58 resulting in more shallow oceanic domains increasing iodine emission from coastal areas; and iv) maximum levels of solar irradiance in the Arctic that would have also increased oxidative stress in algae, leading to an enhancement of biological iodine production in the ocean and subsequent release into the Arctic atmosphere 59 . I flux strongly increased in RECAP from 11 to 9 kyrs b2k, while the NEEM I flux increase was significantly more gradual at that time (Fig. 2), most likely due to the lower productivity and delayed sea ice retreat in Baffin Bay during the early Holocene 60 (Fig. 4). Indeed, this area remained densely sea ice covered until~7.4 kyr b2k, when warmer North Atlantic waters penetrated Baffin Bay, increasing regional bioproductivity 61 (Fig. 4). The mean Holocene iodine flux in NEEM is 57.9% of that in RECAP (mean iodine fluxes of 5.1 and 8.8 µg m −2 yr -1 , respectively), in agreement with present-day reactive iodine deposition at both sites obtained from atmospheric chemistry transport model results ( Supplementary Fig. 1).
Environmental implications. Ocean-emitted iodine leads to new particle formation (NPF) in the atmosphere 5,6 . Recent field observations have demonstrated that iodine is a significant driver of NPF in the central Arctic 3 and is currently considered a potentially relevant source of CCN 3 -from which marine clouds originate-that scatter incoming radiation and can contribute a cooling effect to the Earth´s radiation budget 62 . The significant present-day contribution of iodine to NPF leads us to speculate that, in the absence of other human-induced NPF nucleation drivers during pre-industrial times 63 , oceanic iodine may have played an even larger relative role in total NPF nucleation in a pristine Arctic during the LGC.  38,46 ); e North Atlantic summer sea surface temperature (SST) 36 ; f North Atlantic relative sea level changes 88 ; g Greenland air temperature reconstruction from NGRIP 64,65 ; h solar irradiation reconstruction in July at 77°(NEEM) and 71°N (RECAP) 89 .
We find that the time lag between temperature changes and iodine maximum emission values recorded in Greenland ice cores throughout the LGC strongly depends on the atmosphere-iceocean interactions driving warming SST, sea ice thinning rate and algae metabolic activity that eventually control biogenic iodine emissions. Indeed, the highest iodine emission periods during the LGC were preceded by the highest air temperatures that would have first enhanced SST and ocean biological activity and later Together with the efficiency of iodine to destroy ozone and activate CCN in the polar atmosphere, this points to a possible contribution of iodine to ozone loss and new particle formation, and to their associated radiative impacts, during past abrupt climate change periods, such as D/O events and the onset of interglacials. However, this potential contribution of iodine to past ozone loss and atmospheric NPF is currently unknown and warrants further research. We now turn to the use of past warm intervals recorded in Greenland ice cores as analogues to evaluate potential feedback and thresholds in the Arctic climate system and their implications for future warming impacts. The rate of ongoing atmospheric warming resembles only the abrupt climate changes and sea ice retreat that occurred during the D/O events 64 , with climate warming of 5-16.5°C occurring within a century 65 . The high resolution NEEM iodine dataset during D/O 7-10 show empirical evidence of a fast response of iodine emissions synchronously with the rapid North Atlantic sea ice retreat preceding abrupt Greenland warming during D/O events 48 . Therefore, one question arising is if Arctic iodine levels are facing this scenario in the near future?
Iodine emissions and depositional fluxes during the Eemian and/or Holocene Thermal Maximum shed light on the possible environmental conditions in an expected ice-free Arctic Ocean and thus remain useful as an observational constraint on projections of future impacts. Thus, while atmospheric iodine depositional fluxes recorded in RECAP during the Holocene Thermal Maximum 19 mirror present-day conditions 17 , iodine levels recorded in NEEM during the mid-Eemian tripled presentday values at this location. Similar trends might be expected in near future environmental scenarios including (i) predicted icefree summertime conditions in the Arctic Ocean by 2050 CE 66 and (ii) 3-4°C warmer global temperatures with a three-to four-fold amplification in the Arctic by 2100 CE 67 . However, one important point to consider is that the predicted iodine emissions in the Arctic in the near future based on past analogues represent only the biogenic fraction of the total foreseen atmospheric iodine concentrations in the Arctic. Ozone-induced inorganic iodine emissions during these past warm periods could be considered negligible in the context of the very low ozone concentrations before industrialization 19,68 . In contrast, at present, ocean release of inorganic iodine is estimated to account for 75% of the total source of atmospheric iodine, with the remaining 25% coming from biogenic emissions 69 . Furthermore, ozone-related inorganic iodine emissions are predicted to increase during the 21st century 15 . Therefore, we argue that the future combined (inorganic + biogenic) ice-free Arctic Ocean iodine emissions might not have analogues in the past, which could potentially lead to a near future scenario with the highest iodine levels of the last 127,000 years.
In summary, this study uncovers how past climate changes modulated the variability of the ocean-atmosphere exchange of iodine in the Arctic during the LGC. We show that ocean productivity was the dominant source of Arctic iodine during interglacial periods, leading to highest levels during the Holocene and the Eemian, while dust-related iodine constituted a significant source of iodine in the Arctic atmosphere during the coldest phases of the Last Glacial Period. The mirrored trends of iodine emissions and ocean primary productivity proxies during different past environmental scenarios highlight the key role of ocean biology and sea ice cover in driving iodine variability in the Arctic region. Finally, we conclude that understanding the past variability of iodine as a key environmental element is important to further comprehending the evolution of Arctic atmospheric chemistry on centennial to millennial timescales and its coupling to climate changes.

Methods
Ice core chronologies and sampling resolution. NEEM age-depth model was achieved by transferring the annual layer counting from the NGRIP to the NEEM ice cores using 787 tie points (mainly volcanic layers) 21 . NEEM samples were collected at 1.10 m mean intervals of melted ice where 10 mL of meltwater was collected in an acid-cleaned polyethylene bottle and immediately frozen. Chronological resolution between samples (n = 84) ranges from 4 to 267 years during the Holocene. There is a sampling gap between 3-9 ky b2k where the ice was not analysed due to the poor quality of the core in the "brittle ice" section 31 . The mean sampling frequency during the Glacial Period (n = 247) is~400 years, where the highest resolution achieved was between 25 and 50 kyrs b2k with sampling resolution ranging from 60 to 150 years during warmer interstadial periods and colder glacial stadial periods, respectively. Additional sampling was carried out during the intervals 34.5-35.8, 37.6-38.8 and 39.4-41.6 (n = 1304) providing a time resolution ranging from 3 to 10 years during those intervals to better resolve the rapid stadialinterstadial transition. RECAP age-depth model is based on (i) annual layer counting from the last four millennia using the StratiCounter algorithm 70 constrained by volcanic eruption markers and synchronized to the GICC05 Greenland Ice Core Chronology framework 71 ; (ii) a modified Dansgaard-Johnsen ice flow model 72 constrained to well-dated age markers from 4048 to 11,703 ka b2k and (ii) linear interpolation between 73 GICC05-modelext age markers between 11,703 and 120,215 years b2k 32 . Each RECAP sample integrated a 55 cm depth interval, where the resolution in the Holocene (n = 1035) ranged from sub-annual in the upper metres to decadal or centennial resolution during the mid to early Holocene (average resolution of 12 years per sample). Due to the thinning of annual layers according to depth, each sample integrates approximately 470 years during the glacial period (n = 218) 32 . For details on the age model we refer the readers to Simonsen et al. 22 .
Geochemical analyses. Samples from NEEM and RECAP ice cores were collected using continuous flow analysis systems. NEEM samples were sent to the Environmental Analytical Chemistry laboratory of the CNR-ISP and Ca' Foscari University of Venice for iodine and sodium analyses. Total iodine and sodium concentrations were determined by inductively coupled plasma mass spectrometry (ICP-MS). Sodium and iodine detection limits were 1 ppb and 0.005 ppb, respectively. RECAP samples were sent to ISP-CNR and to Curtin University of Technology (Perth, Australia). Total iodine and sodium concentrations in Perth were determined by inductively coupled plasma-sector field mass spectrometry (ICP-SFMS) with sodium and iodine detection limits of 1.1 ppb and 0.002 ppb, respectively. The analytical procedure for both records was performed following Vallelonga et al. 73 and Maffezzoli et al. 32 . Instrumental errors for iodine and sodium concentrations were 5%.
I flux, I dust and I ocean calculation. Total iodine depositional fluxes (I flux ) in the ice cores were calculated following Corella et al. 19 . To estimate I flux in the RECAP ice core, we re-scaled NGRIP accumulation rates 65 by a factor of 436/174 since it is the modern accumulation ratio at the NGRIP and RECAP drilling sites [74][75][76] . Dustrelated iodine (I dust ) and oceanic gas-phase iodine emissions (I ocean ) were calculated according to Eqs. (1) Fig. 1). CAM-Chem includes a comprehensive benchmark chemistry scheme to simulate the evolution of trace gases and aerosols in the troposphere and stratosphere 79,80 . The model implements a halogen chemistry scheme for chlorine, bromine and iodine 12,81,82 . This includes the photochemical breakdown of five very short-lived bromocarbons (VSL Br = CHBr 3 , CH 2 Br 2 , CH 2 BrCl, CHBrCl 2 , CHBr 2 Cl) and four iodocarbons (VSL I = CH 3 I, CH 2 ICl, CH 2 IBr, CH 2 I 2 ), which are naturally emitted from the ocean into the atmosphere 83 . Additionally, abiotic oceanic sources of HOI and I 2 were included in the lowest layer of the model 26 based on laboratory studies of the oxidation of aqueous iodide by atmospheric ozone deposited on the ocean surface 25,84 . In this simulation, the model was configured with a horizontal resolution of 1.9°latitude by 2.5°longitude and 26 levels, from the surface to ∼40 km (with eight levels above 100 hPa) 17 Monthly mean time variations observed for sea surface temperature and sea ice distribution are also prescribed. To obtain a reasonable representation of the overall stratospheric circulation, the integrated momentum that would have been deposited above the model top is specified by an upper boundary condition. We used the output of a previous work 17 for CE 2010. This simulation was run in free-running mode 79 with prescribed sea surface temperatures, sea ice and meteorological fields from 1950 to 2010 86 . The simulated dynamics and transport therefore represent the daily synoptic conditions of the observations in CE 2010, allowing the online coupling between the ocean, ice and atmospheric modules of the CESM model 79 . Abiotic oceanic sources of iodine (HOI and I 2 ) are modelled according to the meteorology (temperature, winds, sea surface temperature and surface pressure) in CE 2010 based on MacDonald et al. 84 parameterization. The authors would like to remark that a quantitative assessment of the role of iodine in aerosol radiative forcing (through its influence in NPF in the past) is yet not feasible mainly because there are still remaining chemical mechanistic gaps in our knowledge of the iodine particle formation and growth 4,87 .

Data availability
The ice core iodine data generated in this study have been deposited in the Zenodo database (https://doi.org/10.5281/zenodo.5721369).

Code availability
The software code for the CESM model is available from http://www.cesm.ucar.edu/ models/.