Sea-ice reconstructions from bromine and iodine in ice cores

As the intricacies of paleoclimate dynamics are explored, it is becoming understood that sea-ice vari- ability can instigate, or contribute to, climate change instabilities commonly described as “ tipping points ” . Compared to ice sheets and circulating ocean currents, sea-ice is ephemeral and continental- scale changes to sea ice cover occur seasonally. Sea-ice greatly in ﬂ uences polar albedo, atmosphere-ocean gas exchange and vertical mixing of polar ocean masses. Major changes in sea ice distribution and thickness have been invoked as drivers of deglaciations as well as stadial climate variability described in Greenland climate records as “ Dansgaard-Oeschger ” cycles and atmospheric circulation, such as stadial/interstadial or glacial/interglacial climate variability. In this review, we describe the rationale and available evidence for linking the halogens bromine and iodine found in polar snow and ice to sea-ice extent. Reported measurements of bromine and iodine in polar snow and ice samples are critically discussed. We also consider aspects of halogen transport and retention in polar snow and ice that are still poorly understood. Overall, there is a growing body of evidence supporting the application of bromine to sea-ice reconstructions, and the use of iodine to reconstruct marine biological activity mediated in part by sea-ice extent. These halogens complement existing sea-ice proxies but most crucially, offer the capacity to greatly extend the temporal and spatial coverage of ice core-based sea-ice reconstructions. We identify knowledge gaps existing in the current understanding of spatial and temporal variability of halogen distributions in the polar regions. We suggest areas where polar halogen chemistry can contribute to a better understanding of the halogen records recovered from ice cores. Finally, we propose future steps for establishing reliable and constructive sea-ice reconstructions based on bromine and iodine as observed in snow and ice cores.


a b s t r a c t
As the intricacies of paleoclimate dynamics are explored, it is becoming understood that sea-ice variability can instigate, or contribute to, climate change instabilities commonly described as "tipping points". Compared to ice sheets and circulating ocean currents, sea-ice is ephemeral and continentalscale changes to sea ice cover occur seasonally. Sea-ice greatly influences polar albedo, atmosphereocean gas exchange and vertical mixing of polar ocean masses. Major changes in sea ice distribution and thickness have been invoked as drivers of deglaciations as well as stadial climate variability described in Greenland climate records as "Dansgaard-Oeschger" cycles and described in Antarctic climate records as "Antarctic Isotopic Maxima".
The role of halogens in polar atmospheric chemistry has been studied intensively over the past few decades. This research has been driven by the role of bromine, primarily as gas-phase bromine monoxide (BrO), which exerts a key control on polar tropospheric ozone concentrations. Initial findings led to the discovery of boundary-layer self-catalyzing heterogeneous bromine reactions fed by sunlight and ozone, known as bromine explosions. First-year sea-ice and blowing snow have been identified as key components for this heterogeneous bromine recycling in the polar boundary layer. This understanding of polar halogen chemistry e supported by an expanding body of observations and modeling e has formed the basis for investigating quantitative links between halogen concentrations in the polar atmospheric boundary layer and sea-ice presence and/or extent.
Despite the clear importance of sea-ice in paleoclimate research, the ice core community lacks a conservative and quantitative proxy for sea-ice extent. The most commonly applied proxy, methanesulphonic acid (MSA), is volatile and has not been demonstrated reliably for ice core records extending beyond the last few centuries. Sodium has also been applied to reconstruct sea-ice extent in a semi-quantitative manner although the effects of meteorological transport noise are significant. Contrary to a priori expectations, the halogens bromine and iodine appear to be stable in polar snow and ice over millennial timescales, addressing the temporal limitations of MSA records. Unfortunately, transport and meteorological variability influence sodium deposition as well as the deposition of halogens and the many other ionic impurities found in ice cores. The atmospheric chemistry of halogens is more complex than those of sodium or MSA due to the mixed-phase (gas and aerosol) nature of halogen photochemistry. Thus the application of halogen records in ice cores to sea-ice reconstruction overcomes some challenges posed by existing proxies, but also opens new challenges specific to halogens. Challenges common to all sea-ice proxies include the deconvolution of changes in emission source locations and changes in transport efficacy, particularly those occurring during climate transitions combining changes in sea-ice and atmospheric circulation, such as stadial/interstadial or glacial/interglacial climate variability.
In this review, we describe the rationale and available evidence for linking the halogens bromine and iodine found in polar snow and ice to sea-ice extent. Reported measurements of bromine and iodine in polar snow and ice samples are critically discussed. We also consider aspects of halogen transport and retention in polar snow and ice that are still poorly understood. Overall, there is a growing body of evidence supporting the application of bromine to sea-ice reconstructions, and the use of iodine to reconstruct marine biological activity mediated in part by sea-ice extent. These halogens complement existing sea-ice proxies but most crucially, offer the capacity to greatly extend the temporal and spatial coverage of ice core-based sea-ice reconstructions. We identify knowledge gaps existing in the current understanding of spatial and temporal variability of halogen distributions in the polar regions. We suggest areas where polar halogen chemistry can contribute to a better understanding of the halogen records recovered from ice cores. Finally, we propose future steps for establishing reliable and constructive sea-ice reconstructions based on bromine and iodine as observed in snow and ice cores.
© 2021 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).

Introduction
Ice cores provide sensitive and detailed records of past climate variability, with the unique ability to directly archive atmospheric gases and aerosols. Snow deposition and retention is the fundamental requirement for ice caps to form and eventually grow to large ice sheets such as those found in Greenland and Antarctica. Smaller ice caps form on mountain ranges although many are currently threatened by anthropogenic warming. The polar ice sheets are up to 4 km thick, while most mountain ice caps are not more than approximately 100 m thick (Bradley, 2015). Ice cores record climate on seasonal to millennial timescales, with the oldest ice cores recovered from central Antarctica (EPICA community members, 2006). The time period archived in an ice core primarily depends on the rate of snow accumulation at the site, the amount of glaciological thinning due to ice flow and the thickness of the ice sheet. Ice core chronologies are essential to the accurate interpretation of ice core impurities, and are established by a variety of complementary techniques, including the counting of annual layers determined by seasonally-varying impurities (Sigl et al., 2016;Winstrup et al., 2019), identification of known or notable volcanic eruptions (Sigl et al., 2015;Svensson et al., 2020), determination of globally-synchronous atmospheric events such as methane uptake/release from tropical wetlands (EPICA community members, 2006) and fluctuations in the Earth's magnetic field (Raisbeck et al., 2017), as well as the application of glaciological ice flow models (e.g.: Parrenin et al., 2007).
Climate records are typically obtained from ice sheets by drilling ice cores, often striking a compromise between the amount of ice sample obtained and the ability to access deep ice far below the surface (Johnsen et al., 2007). An electromechanical ice core drill is a hollow tube with cutters at the bottom, which rotates and draws cuttings up into a chamber above the ice core cylinder. The drill is suspended on a cable, providing power and control data as well as lowering and raising the drill. Once a suitable core length has been drilled, the drill is raised and upward-pointing core-dogs dig into the ice to break the ice cylinder and hold it in position. Electromechanical ice core drills typically produce 75e108 mm diameter cores with up to 30 m drilled per day . Deep ice core drillings conducted in central Greenland and Antarctica may require 3 or more drilling seasons to reach bedrock . Electromechanical drills can recover ice from a few hundred metres' depth typically within a week or two.
In the absence of excessive melt, impurities are preserved in the snow/ice matrix and can be determined to investigate a wide variety of climate parameters. The impurities are employed in a quantitative or semi-quantitative manner as proxies of earth system parameters including desert dust, storminess, wind speed, biomass burning, volcanism, and many others (Schüpbach et al., 2018). Key to the employment of such proxies is their sensitivity to source emissions, lack of ambiguity/alternate emission sources, reliability of transport and stability in the snow/ice matrix. Atmospheric transport processes may be investigated by deposition studies, in-situ analyses and/or model-based studies. Surface melting occurs in sites where a significant number of consecutive positive degree days are experienced or liquid precipitation events may occur due to atmospheric phenomena such as 'atmospheric rivers' (Nghiem et al., 2012). Such sites may be located hundreds of metres above the calculated equilibrium line altitude (e.g. Holthedahlfonna in Svalbard is located at 1150 m asl whereas the equilibrium line is 600e700 m a.s.l.). Meltwater produced at the surface is then able to mix with deeper, cooled layers and eventually refreezes in dense and/or cold layers in snowpack, typically corresponding to the previous winter strata. Surface melting can volatilize some unstable molecules such as MSA, iodine and nitrate, leading to unreliability in their use for climate reconstructions (Pohjola et al., 2002). Many ionic species appear to be unaffected by surface melt (Bonne et al., 2015;Nilsson et al., 2015).
Sodium has been studied extensively in ice and snow as it is relatively easy to determine, stable in the snow matrix and primarily represents sea-salt deposition to the sampling site. Early studies of sodium deposition, conducted on snow pit sites at central Greenland and coastal Antarctic locations, revealed a seasonal cycle peaking in winter (Legrand and Mayewski, 1997). This peak has been traditionally attributed to a peak in storminess, generating waves and transporting marine aerosols efficiently to the polar ice sheets. Later studies identified a sodium/sulphate fractionation indicative of sea-ice based frost flowers: mirabilite crystals that grow from the salty brine on the sea-ice surface and can be transported inland during wind gusts (Rankin et al., 2002). While the importance of sea-ice sea-salt aerosol emissions was recognized, the idea that frost flowers were the main sea-salt aerosol source was challenged by laboratory and model studies (e.g. Legrand et al., 2016;Levine et al., 2014). To date, the most efficient substrate for sea-salt aerosol emissions in the polar regions is attributed to the snow layer that deposits onto fresh sea-ice surfaces, known as salty blowing snow, that becomes enriched in brine and can be easily lofted by the strong polar winds (Frey et al., 2020). Given that sodium can be emitted from the open ocean as well as the sea-ice surface, modeling studies have been employed to better constrain the circumstances in which sodium in ice cores may be attributed to sea-ice (Levine et al., 2014;Rhodes et al., 2018). Such studies indicate that sodium, and likely other sea-ice related elements such as Br and MSA, could be controlled by meteorological effects on annual to decadal scales. Sea-ice reconstructions (SIR) based on sodium are highly dependent on individual site characteristics (Rhodes et al., 2018). Sodium has been demonstrated to be an effective sea-ice indicator primarily at coastal sampling locations (Mezgec et al., 2017).
Methanesulphonic acid (MSA) is an oxidation product of dimethylsulfide (DMS) released by marine algae. The links between DMS and sea-ice were first investigated in the 1990s (Curran et al., 1998) and DMS is now identified as a credible indicator of sea-ice at some sites in Antarctica and the Arctic . Depending on the configuration of sea-ice cover and open water, different interpretations are applied to MSA records from different sampling sites, so MSA has been used variously to quantify open water extent (summer sea-ice absence), winter sea-ice extent or atmospheric transport conditions. Models overestimate the concentration of MSA in the snowpack of central Antarctica e most likely because of postdepositional loss of MSA from such zones of low snow accumulation e but do attribute the majority of MSA variability to sea-ice variability (Hezel et al., 2011). The stability of MSA in ice cores has been demonstrated to be problematic in some ice cores e MSA is volatile and will remobilize in the ice matrix with time (Curran et al., 2002;Osman et al., 2017). MSA is volatilized from ice core samples stored for several years but appears stable in samples that have been melted and then refrozen within a few months of ice core extraction . Studies of Antarctic and Arctic ice cores have shown that MSA migrates from summer deposition layers to impurity-laden winter layers, particularly those containing high levels of sea salt (Osman et al., 2017). MSA stability in ice cores is also promoted by the presence of dust . Recently, it has been proposed that MSA found in Greenland ice is predominantly controlled by DMS emission due to North Atlantic ocean productivity (Osman et al., 2019). For central Antarctic sites, there is evidence that MSA is destroyed in-situ by photochemical processes (Legrand et al., 2016) particularly during the summertime. The longest MSA-based SIR for the northern hemisphere is from Lomonosovfonna, Svalbard  and covers the last few centuries. For the Southern Hemisphere, the longest MSA-based SIRs cover approximately 300 years (Thomas and Abram, 2016;Xiao et al., 2015) although several shorter records have also been reported (Abram et al., 2013 and references therein;Bertler et al., 2018;Murphy et al., 2014). Sea-ice occurs in the polar regions, where atmospheric temperatures are sufficiently cold to freeze seawater (Thomas and Dieckmann, 2003). There are several different forms of sea-ice, but relatively stable water conditions are required for sea-ice to form as frazil and grease ice, and then conglomerate into larger forms such as pancake ice (Eicken, 2003). Sea-ice that forms during the autumn and winter is known as seasonal, first-year or fresh, sea-ice while sea-ice that remains frozen through the spring and summer is classified as multi-year sea-ice. Antarctica is primarily surrounded by first-year sea-ice, with pockets of multi-year sea-ice that remain in the Weddell Sea and some bays along the Antarctic coastline. In the Arctic, multi-year sea-ice is present as a mass located between Greenland and the North pole, with first-year seaice growing to cover the Arctic Ocean, Baffin and Hudson bays as well as often filling the Barents and Chukchi seas (Stroeve et al., 2012a). As sea-ice forms, salt is expelled into channels and pockets of brine collecting on the upper surface of the ice. Freeboarding of seawater onto the ice is another source of salt for the sea-ice surface. Sea-ice forms an important barrier to the exchange of heat, gas and aerosols from the ocean to the atmosphere and vice versa. Sea-ice is permeable to gases at temperatures above -5 C and is a vital habitat for algae in polar ecosystems (Vancoppenolle et al., 2013).
Despite being observed for hundreds of years, Arctic sea-ice variability is still a topic of considerable interest with substantial uncertainty regarding its variability and future. The earliest sea-ice observations are drawn from the written records of Iceland fishermen, going back over 400 years (Ogilvie, 1984). With increasing emissions of greenhouse gases, the impact of recent climate change has led to renewed interest in the understanding of sea-ice variability (Kinnard et al., 2011). Since the start of the satellite era, Arctic sea-ice has been observed to decrease rapidly in extent and thickness. Past models of sea-ice decline have failed to capture the rapidity of change in the Arctic predicting that the Arctic will be free of summer sea-ice (sea-ice area<1 million km 2 ) before the end of the 21st century (Stroeve and Notz, 2015). The latest generation models, however, are able to provide a more realistic estimate of the sensitivity of September Arctic sea-ice area to a given amount of anthropogenic CO 2 emissions and to a given amount of global warming. Most simulations from the last Coupled Model Intercomparison Project (CMIP6) indicate the occurrence of sea-ice free September events in the Arctic Ocean by 2050 (SIMIP Community, 2020). In this perspective, reconstructions of sea-ice extending back to the penultimate interglacial warm period (the Eemian as it is known in the Northern Hemisphere) are of vital importance to understanding the extent of sea-ice in warm climate periods.
The majority of paleo-scale sea-ice reconstructions are based on sea-ice proxies preserved in marine sediment cores. Marine basins typically feature very low sedimentation rates, resulting in relatively poor time resolution e compared to ice cores e but much longer temporal coverage, combined with superior spatial coverage. The direct sedimentation of sea-ice debris onto the seafloor is also strongly favourable to the reconstruction of sea-ice presence from sediment records. Sea-ice reconstructions from ocean sediments have been produced from siliceous diatoms, phytoplankton markers and most recently from IP 25 , a highly branched isoprenoid lipid biomarker associated with sea-icedwelling diatoms in the Arctic Ocean (Belt et al., 2007). IP 25 is specific to seasonal sea-ice, as algal communities are not sustained below relatively thick and impermeable perennial sea-ice, where light and nutrients are unavailable. As the absence of IP 25 may indicate either open water conditions, or perennial sea-ice, it is typically combined with a complementary phytoplankton biomarker to establish which extreme condition exists (i.e.: perennial sea-ice or open water). These are identified as P B IP 25 or P D IP 25 respectively, if Brassicasterol or Dinosterol phytoplankton biomarkers are employed. Reconstructions of Antarctic sea-ice extent are based on an equivalent biomarker present in the Southern Ocean, denoted as IPSO 25 , and are presently limited to the Atlantic and Indian Ocean sectors due to the problem of carbonate dissolution primarily in the Pacific Ocean sector and near to the Subtropical Front (Gersonde et al., 2005).
Although it is a halogen element, chlorine is not considered as a sea-ice proxy in this review due to key issues regarding the emission, transport and deposition of this ubiquitous marine element. The emission of chlorine (and sodium) into the atmosphere is overwhelmingly driven through wave breaking/bubble bursting processes occurring at the ocean surface. This feature has been used in past attempts to semi-quantitatively evaluate sea-ice extent through the fluxes of sodium and chloride to central Antarctica (Levine et al., 2014;R€ othlisberger et al., 2003;Wolff et al., 2006). The findings of this course of research indicate that at best, sea-salts may be used to infer sea-ice extent in a qualitative manner: meteorological (transport) "noise" exerts a strong influence on the sea-salt signal deposited to inland Antarctica, and this effect becomes stronger as the distance from the sampling site to the marginal ice zone increases. Furthermore, the stability of chlorine in the snow matrix is challenged by post-depositional dechlorination of snow due to acid reactions producing gas-phase HCl (R€ othlisberger et al., 2003). The combination of these factors inhibits the general reliability of Cl as an environmental proxy in ice cores; and specifically degrades any quantitative link between chlorine and seaice extent specifically. In comparison, bromine and iodine enrichment processes can be directly linked to sea-ice, specifically firstyear sea-ice. While still subject to meteorological noise (as are all aerosols), the post-depositional effects demonstrated by chlorine appear to be far less significant for bromine (as discussed further in section 3.4.2). Post-depositional effects are observed for iodine and are described in section 3.4.3.
In this review we will present the current state of understanding on the applicability of Br to track the past sea-ice extent and variability from ice core records. Although some uncertainties remain, such as a detailed understanding of the interactions ongoing within the mixed gas-and aerosol-based phases during bromine transport, this halogen shows promise as a possible tracer of past sea-ice extent. We will provide a brief introduction to the polar photochemistry associated with iodine and bromine and evaluate their stability in the polar snow/ice matrix. We will then present the available bromine data on timescales relevant to ice core archives, and summarize the findings of sea-ice reconstructions based on mainly bromine in polar ice. The locations of relevant studies of bromine and/or iodine timeseries are shown in Fig. 1 and summarized in Table 1. We conclude with a summary of the state of the art and outline future directions in this field of research, relevant for future deep ice core drilling as well as complementary sea-ice reconstructions from proxies found in marine sediment records.

Analytical techniques
Bromine and iodine are found in meteoric ice at trace levels, typically from tens of picograms per gram (pg/g) to tens of nanograms per gram (ng/g). Although the measures are not strictly identical, ng/g (weight/weight) concentrations are often also reported as ppb (part per billion), which is a more generic concentration metric. For consistency with the literature, here we will consistently report ng/g concentrations as ppb, and pg/g as ppt. The highest concentrations have been reported for ice cores from low altitude, coastal locations whereas the lowest concentrations have been found far inland on the high-altitude East Antarctic plateau (Table 2).
Bromine concentrations are typically determined in polar snow and ice by Ion Chromatography (IC) or Inductively-Coupled Plasma Mass Spectrometry (ICPMS). Both instruments accept liquid samples, so the snow/ice sample is melted prior to analysis. The melting may take place as a dedicated sampling or, more commonly, will form part of a Continuous Flow Analysis (CFA) setup. CFA is a technique in which firn/ice sticks are sequentially melted onto a melthead which separates the pristine inner section of the sample from the contaminated exterior (e.g.: Kaufmann et al., 2008). Meltwater from the inner section is pumped into various flowthrough sensors or sent to optical flow cells for determination of soluble ionic impurities, or collected in sampling tubes or beakers for later analysis. Typically, these discrete samples will be refrozen and stored until ready for analysis by IC or ICPMS. These techniques will be further discussed below.
Inductively Coupled Plasma Mass Spectrometry is a standard technique for trace analysis, with good sensitivity in the ppb to ppt range. ICPMS comprises three fundamental steps: sample ionization, separation and detection. Sample ionization is conducted using an argon-fed plasma, operating at 6000e10000 K. For signal stability, a fine aerosol of uniform size is produced by the introduction system and fed into the plasma using a nebulizer with a typical pore size of 50e200 mm. The introduction system may consist of a Scott-type spray chamber or a cyclonic spray chamber, with cooling to assist the drying of the aerosols. Once introduced into the plasma and ionized, the target ions are separated from other ions by either a quadrupole (ICP-QMS) or sector-field magnet (ICP-SFMS). The quadrupole has a faster rate of scanning but lower mass resolution compared to the sector-field instrument. In both instruments the detector is usually a secondary electron multiplier (SEM), which absorbs the ion and in turn produces a photon that can be amplified by a sequence of cascading anodes. Typically, a helium-filled collision cell is added to the ICP-QMS instrument to boost measurement sensitivity and signal stability. ICP-QMS is suitable for ppb concentrations while ICP-SFMS is recommended for ppt concentrations. However, ICP-QMS capabilities have improved in recent years also allowing determination of halogens at sub-ppb level, closer to the performance obtained from the ICP-SFMS.
Ion chromatography is a standard technique for the determination of soluble ions in snow and ice. The heart of the instrument is the separation column, a resin-based high-pressure ionexchange column through which the ions will pass at different rates, based on their charge and other properties determining their level of binding to the column resin. The IC will typically feature a guard column to remove particles from the sample stream, then a preparation column to conduct a broad removal of unwanted ions, and then the sample will be introduced into the analytical column. Once the column has been loaded, an eluent removes the ions from the column, which then pass through a sensitive conductivity meter. The concentration of a particular ion is determined as the corresponding area below the conductivity peak and is calibrated using standard solutions of the ions to be determined. Optimization of the elution time and separation of interferences is normally achieved by finding an ideal combination of temperature, eluent concentration(s) and flow speed(s) applied to the column. The sensitivity of the IC technique is suitable for determination of bromine and iodine species at ppb concentrations but is generally insufficient for determination of halogens at low ppb and ppt levels. Therefore, iodine species are usually determined by ICPMS systems or IC separation systems coupled to an ICPMS detection system.
For any analytical technique, contamination control and memory effects must be carefully evaluated to ensure accurate measurements are made. In laboratory settings, bromine may be found in some plastics and in fire control equipment (as BFRs, bromine fire retardants). Generally, these materials are easily isolated from sample preparation areas. Due to the volatility of halogens, samples should not be exposed to acids or acid vapour and ice samples should generally be kept frozen and in darkness until they are to be analyzed. Samples analyzed for bromine and iodine are therefore not readily prone to contamination, but a more common source of difficulty is the substantial memory effects associated with quartz glassware commonly used in MS introduction systems. Therefore a moderate level of regular cleaning with ammonium hydroxide is required once the sample introduction system has been cleaned to an acceptable level (Spolaor et al., 2012). An initial cleaning with ammonium hydroxide removes bromine and iodine from the quartz glassware of the ICPMS introduction system, however the ammonium hydroxide may lead to salt deposition within the introduction system that may potentially block the nebulizer. Therefore, after the ammonium hydroxide cleaning stage, the introduction system is cleaned with 2 % nitric acid for 90 s and then MilliQ-quality (>18.2 MU/cm) ultrapure water (UPW) for 90 s before sample analysis. The use of UPW before the sample analysis has two effects: firstly, to remove any nitrate salts that may be deposited by the nitric acid and secondly, to condition the instrument before introducing the unacidified samples for halogens analysis. Fig. 2 shows an example of the extensive memory effect that occurs when only nitric acid is used in the cleaning procedure e a technique that is typical for analysis of trace metals by ICPMS. In the case shown, the ICPMS introduction system is washed between analyses for 45 s using only a 2 % nitric acid solution. Each   calibration standard series results in a progressive coating of iodine and bromine onto the surface of the sample introduction system. During each sequence of Antarctic sample measurements (represented by black circles) the iodine and bromine in the introduction system is slowly and inefficiently removed. The memory effect is the main reason for the long cleaning step during the analysis between samples previously described. The duration of each analysis is approximately 5 min and each standard calibration sequence progressively increases from 0 to 600 ppt (10 À12 g/g). In the absence of the ammonium hydroxide cleaning step, halogens are inefficiently removed from the ICPMS introduction system and substantial sample carry-over occurs.
The substantial memory effect of halogen elements in ICPMS sample introduction systems also reduces the suitability of coupled CFA-ICPMS systems for high-resolution analysis of halogens in ice cores. CFA systems typically involve sampling "runs" in which a sequence of ice up to 20 m in length is continuously melted. Between melting runs, the analytical systems are allowed to return to their baselines and are typically fed by a supply of UPW. The memory effect of halogens in ICPMS glassware leads to a "smoothing" of the ICPMS data and can be observed as a reduced sample variance. This effect was observed in bromine dataseries' reported for the Law Dome ice core, where overlapping ice core records drilled at the same location were measured using different techniques (See section 3.1 in Vallelonga et al., 2017). Greater variance was observed in the core that was sampled using a discrete cutting technique typically applied to IC samples, with the samples then measured individually by ICP-SFMS. Reduced variance was observed in the core that was melted continuously in a CFA system and then measured using an online ICP-SFMS instrument. The mean results from the cores were in agreement, but interannual variability of bromine concentrations was not fully resolved by the continuous-sampling technique. This is a counterintuitive result, as one would expect data variance to increase as sampling resolution increases. Sodium e which does not suffer from memory effects in ICPMS sample introduction systems e was also measured in both cores, and showed the expected pattern of increased data variance with increased sampling resolution (See Table 1 in Vallelonga et al., 2017).
Interlaboratory comparisons are a valuable method of comparing analytical methods as well as investigating the longterm stability of halogens in samples. Fig. 3 shows the results of an interlaboratory measurement comparison for bromine and iodine. The samples were first melted as part of a CFA campaign, where discrete aliquots were collected for measurement of halogens (Maffezzoli et al., 2019). The refrozen samples were sent to the University Ca' Foscari of Venice (Italy, IT) where they were melted, sampled for measurement and then refrozen. The refrozen samples were then sent to Curtin University of Technology (Australia, AUS) to be melted once more for analysis of halogens using similar instrumentation. The laboratory intercomparison between sodium and bromine measurements was performed on a common set of samples (n ¼ 140) to investigate differences between the analytical techniques and laboratories, as described in Vallelonga et al. (2017). The correlations and the gradients between the measured concentrations in the two set-ups are r(Na IT-Na AUS) ¼ 0.99 (n ¼ 140; p < 0.01), mean gradient Na ¼ 1.084 ± 0.011 (1s) for sodium; and r(Br IT-Br AUS) ¼ 0.93 (n ¼ 140; p < 0.01), mean gradient Br ¼ 1.085 ± 0.011 (1s) for bromine. Fig. 3 shows that sodium and bromine measurements in the two labs are in good agreement and are close to a 1:1 line, whereas there has been some loss of iodine  during the refreezing, transport and remelting processes that occurred between the measurements. This loss is most evident for the samples containing the lowest concentrations of iodine. Bromine concentrations may be accurately measured from repeatedly melted samples as long as the samples are kept frozen and away from light and the sample container is in a good condition. Repeated measurements of iodine samples is inadvisable and likely results in under-reported values subsequent to the first measurement.

Element concentrations
The evaluation of halogen enrichment in ice and snow is primarily conducted by comparing bromine and iodine with corresponding sodium data. In the following discussion we will consistently refer to bromine but all equations are equally suitable to iodine. The concentration of bromine in ice is represented by [Br]. The standard approach is to calculate the concentration of seasalt (ss) bromine based on the sodium concentration in ice (assuming the sodium in the sample to be purely of marine origin) and the bromine/sodium mass ratio commonly found in seawater. The latter has been reported in key studies of mean ocean water composition (e.g. : Millero, 1974;Millero et al., 2008;Turekian, 1968). The sea-salt component of bromine is determined using the following relation ½ssBr ¼ ½Na * WðBrÞ=WðNaÞ where W(Br) is the sea water mass fraction of solute Br (grams of Br/grams of total solute), following the style of Millero et al. (2008).
If the site of investigation is located in a region affected by mineral dust input, sodium may contain some crustal input. In this case, the sea-salt sodium component (ssNa) must be calculated before it is used in the above sea-salt halogen calculation. Two different approaches can be used to evaluate ssNa. In general, when two elements have mixed crustal (non-sea-salt) and marine (seasalt) input, the two contributions can be separately calculated if their elemental ratios in the sea-salt and in the dust source are known (Vallelonga et al., 2004). In the case of sodium, its non seasalt contribution is usually calculated by using a crustal element such as aluminum (Boutron et al., 1991), calcium Schüpbach et al., 2018), cerium (McConnell et al., 2014, iron or direct dust measurements (Vallelonga et al., 2013). The source of uncertainty of this approach lies in the knowledge of the dust source elemental ratio, while the marine ratio can be assumed well known. To date, upper continental crust average elemental ratios have been widely used both in the Arctic and Antarctica, but in some ice records this has been shown not to be an accurate representation of the source dust (Maffezzoli et al., 2019). The second approach to evaluate ssNa makes use of departures from the sea-salt sodium-to-chlorine ratio (Hansson, 1994). However, chlorine concentrations in sea-salt aerosol are biased by dechlorination processes, which lead to loss of chlorine from the aerosol to the gas phase (R€ othlisberger et al., 2003). Overall, the chlorine-to-sodium ratio could be enhanced (and therefore mask the decreasing behavior of the Cl/Na ratio due to the crustal sodium factor) because of the longer residence time of gas phase chlorine (HCl) compared to sodium, which remains in the aerosol phase. We recommend the use of elemental dust ratios, in preference to chlorine-based evaluations, as the best available method to represent mineral dust inputs to ice core records. The selection of elemental ratios to be used should be made carefully and the results should be evaluated based on the range of outcomes possible from each of the available crustal markers.
The non-sea-salt (nss) component of bromine (or iodine) can be calculated as:

½nssBr ¼ ½Br À ½ssBr
Note that this measurement combines uncertainties in the determination of bromine and sodium, as well as their ocean water compositions, so nssBr may be negative if [Br] is particularly low. In such cases, it is common practice to round up to zero and interpret the result as indicative of a negligible non-sea-salt contribution.
Negative nss values may also be obtained if the element has been remobilized, re-emitted or otherwise lost from the snow surface.
The term nssBr expresses an abundance of bromine (or iodine) that cannot be attributed to the calculated sea-salt loading present in the sample. As such, it can be considered an excess (or depletion in case of a negative value) beyond the "direct" sea-salt input. The source of this excess must then be hypothesized. In the case of bromine, heterogeneous recycling over first-year sea-ice is a clearly demonstrated and well-constrained mechanism for enhancing bromine in the boundary layer (Abbatt et al., 2012). For iodine, we attribute non-sea-salt emissions to algal communities hosted by first-year sea-ice (Saiz-Lopez and Blaszczak-Boxe, 2016), particularly in the Antarctic. In the Arctic, an iodine-ozone feedback mechanism has recently been identified, in which oceanic iodine emissions are enhanced by ozone deposition to the sea surface . This process is most relevant to the Arctic, which is subject to perturbation of natural ozone levels due to industrial processes (Prados-Roman et al., 2015).

Element ratios
Bromine excess may also be expressed as a relative metric, rather than an absolute abundance. We consider this a relative enrichment, rather than the abundance excess mentioned in the previous section. Thus, fractional bromine (or iodine) enrichment may be calculated as a ratio: Where a value of 1 indicates no enrichment of Br beyond the ss contribution (equivalent to nssBr ¼ 0) and values greater than 1 indicate an additional quantity of Br that cannot be attributed to ss (equivalent to nssBr > 0). Values of Br enr less than 1 indicate depletion, due to loss of Br from ss either during transport to the site or due to post-depositional processes. A value of Br enr ¼ 2 indicates that only half of Br in the sample can be explained by seasalt, thus ssBr ¼ nssBr. A Br enr value greater than 5 indicates a substantial Br enrichment process has occurred. Br enr can only approach zero if the concentration of Br is below the limit of determination or if ssBr [ Br.
As mentioned previously, values of Br enr less than 1 indicate depletion of bromine relative to sea-salt inputs. The first published interpretation of bromine as a possible sea-ice proxy quantified this depletion metric as an indicator of sea-ice extent (Spolaor et al., 2013b). For clarity, we show that this depletion metric is effectively the inverse of the aforementioned Br enr enrichment metric, and the two can be used equivalently. The depletion calculation employed by Spolaor et al. (2013b) was: %Br fi ¼ 100 * ðð½Na * 0:006 À ½BrÞ = ½Na * 0:006Þ where %Br fi was called the bromide fractionation index. Rather than a percentage, this can be expressed as a fraction and more accurately labeled as Bromine depletion (Br dep ): Br dep ¼ ð½Na * 0:006 À ½BrÞ = ½Na * 0:006

¼ ð½ssBr À ½BrÞ = ½ssBr
In section 4.4.1 we will further explore the data presented by Spolaor et al. (2013b). The interpretation of the data and its use to reconstruct sea-ice extent are described and considered further in section 5.1.1. At this point, we reiterate that the plotting of bromine depletion, rather than enrichment, should not alter the findings arising from the study and most importantly, arise from the same two data series of bromine and sodium concentrations.
The applicability of iodine enrichment metrics is still a matter of debate and uncertainty. For iodine enrichment (I enr ) and nss-I the same calculation can be applied as for Br but using the iodine/sodium ratio in seawater of 5.93 Â 10 À6 (Turekian, 1968). Unlike Br enr and nssBr, iodine enrichment is infrequently used due to the many biological processes associated with iodine emission, in addition to direct sea spray emission (Saiz-Lopez and Blaszczak-Boxe, 2016). Thus while I enr may, in principle, be used to apportion biotic and abiotic iodine emission pathways, in practice the complex heterogeneous reactions that take place above seasonal sea-ice invalidate any simple consideration of two independent biotic and abiotic pathways describing iodine in the polar environment. We refer the reader to  for a more detailed description of iodine atmospheric chemistry.

Element fluxes
In general, upon measurement, ice core impurities are detected using instruments that provide concentration values: c ice . Such concentrations are the parameter we referred to in the above sections 2.2.1 and 2.2.2 as [Br] in the specific case of bromine concentration determination. From the ice concentration, the air-tosnow depositional flux (J) can be calculated as: if the snow accumulation rate A is known (from experimental data or model results). A is expressed as the annual mass of deposited snow per m 2 .
In general, the deposition of an impurity onto ice sheets or glaciers proceeds via both wet and dry deposition: where c air is the concentration of impurity c in the air, v dry is the dry deposition velocity, k is the snow scavenging factor and A is the accumulation rate. In Eq. (1), the first sum (c air * v dry ) is the dry component of the deposition which is independent of A. The second sum, the wet component, is (k*A*c air ). Using the definition of J, we can express the ice concentration in Eq. (1) as: When interpreting ice core records, we would like to consider a parameter which is linearly related to the air concentration c air and ideally does not suffer from the effects of temporal changes in the other parameters. In general, all parameters in Eq. (2) can be a function of time, in particular across glacial-interglacial timescales. The particular effect of the accumulation rate A on c ice can be investigated in two case limits: In the case of dominant wet deposition, v dry /A ≪ k, so from Eq.
(2) c ice /k*c air . In this case the ice concentration will be effectively linearly related to the air concentration and independent of accumulation changes. This case is most suitable for climatic interpretation as the potentially large glacial/interglacial variability in snow accumulation rate will not influence the analyte concentration in ice. In case of dominant dry deposition, v dry [ kA, so from Eq. (1), J/v dry *c air . In this case the flux is the parameter linearly related to the air concentration and independent of accumulation changes. Therefore fluxes are most suitable for climatic interpretation of predominantly dry-deposited analytes.
Generally, for low accumulation sites and/or during periods of reduced accumulation (e.g. glacial periods), when dry deposition is believed to be the dominating mechanism, J is used to express the atmospheric loading of the analyte of interest. In contrast, if wet deposition is believed to be the dominating mechanism of deposition, c ice is the parameter linearly related to c air and independent of A. This general case works well for analytes which are reasonably conservative and do not undergo phase changes or excessive chemical changes in the atmosphere (Alley et al., 1995;Legrand, 1987) In the case of an analyte which may be subject to a combination of wet and dry deposition, the relative strengths of each mode can be tested. This is done by assembling a suite of c ice and accumulation data from snow/ice sampling locations encompassing a broad range of A values. If the flux deposition J can be modeled accurately by Eq. (1), one can estimate the dry deposition component v dry *c air from the y-intercept of a linear regression of the data. From the same linear regression, the wet deposition component will be represented by the gradient k*c air .
Antarctic records have been historically investigated using fluxes, as dry deposition is believed to be the dominating presentday mechanism at continental sites (e.g. Dome C) and even more so during glacial periods (Wolff et al., 2006). Other studies, however, have indicated that sodium is mainly wet deposited at more coastal If Antarctic records have shown that for some species, like sodium, dry deposition is probably dominant at inland sites while, at least for interglacial periods, wet deposition is dominant at the coast, Greenland records are even more challenging. Across Greenland, accumulation rates are typically greater than those found in Antarctica, hence wet deposition has a consistently greater influence on aerosol deposition, compared to Antarctica. Recent efforts based on analysis of the NEEM ice core, in northwestern Greenland, have employed a direct calculation of c air from Eq. (2) (Schüpbach et al., 2018). Unfortunately, a limited knowledge of the variables, especially v dry and k, result in a poorly constrained reconstruction of c air . Such uncertainties particularly influence bromine and iodine, as these two impurities are emitted and transported in both gas and aerosol phases as well as different chemical species. Detailed field-based investigation campaigns e encompassing locations featuring a range of accumulation rates, and combining direct sampling of snow and atmospheric BrO and particulate loading -will be of great benefit for constraining bromine dry and wet deposition characteristics.

Considerations
In this section we describe some standard methods for determining and expressing deviations from the Br/Na ratio found in sea-salt. The interpretation of a data set should not change greatly if a deviation is expressed as an absolute excess or a relative enrichment, although there may be advantages to using one over the other, such as in the circumstances described below: When considering the concentrations of any aerosol deposited in snowpack, the relevant processes of deposition are critical to an accurate evaluation. It has been demonstrated that sea-salt aerosol (containing sodium and bromine) is transported as fine aerosol, whereas the transport of bromine includes chemical recycling processes occurring in mixed aerosol and gas phases. The relative contribution of the dry and wet components controlling bromine and sodium deposition are currently poorly constrained, and are dependent on snow accumulation as well as the chemical parameters of the many bromine species involved. With respect to Br enr calculations, however, we can assume that dry deposition mechanisms are much more important for bromine than for sodium. When evaluating bromine recycling over long timescales involving significant snow accumulation changes (such as glacial/interglacial transitions and, in the case of Greenland, Dansgaard/Oeschger events) it is important to consider the effect that such accumulation changes may have on the relative dry and wet contributions of the analyte deposition, which may or may not influence the measured concentration. Considering element ratios rather than absolute abundances can minimize the effect of accumulation changes. Hence, bromine enrichment (i.e., Br enr ) is supposedly less biased by accumulation rate changes than bromine excess (i.e., nssBr), as the former incorporates element ratios rather than absolute quantities. Although Br enr is less influenced by changes in accumulation over time, it is strongly influenced by changes in sodium concentration. For example, sodium concentrations display a strong gradient across Antarctica as aerosols travel inland (Bertler et al., 2005). Variability in Br enr observed at an inland site may be artificially inflated due to the very small quantities of sodium deposited at that site. Furthermore, Br enr variability observed along an inland traverse will be induced by the different deposition characteristics of Br and Na. These considerations are discussed in section 3.6. Hence, the geographical setting and absolute sodium concentration found in ice core records should be considered when interpreting Br enr variability across sites. The same interpretation of a Br and Na data set should arise irrespective of whether the data are plotted as nssBr or Br enr , as long as the sodium fraction is predominantly derived from seasalt. The Br enr metric incorporates sea-salt as the primary determinant of Br and therefore can be biased if a substantial proportion of sodium originates from non-sea-salt origins, such as mineral dust. The usage of Br enr as indicator of past sea-ice variation employs a sea-salt correction based on sodium. The assumption behind this approach is that there is no fractionation of Br and Na prior to the emission "starting point", that is, the sea-ice surface. Therefore, the use of a "standard" sea-salt composition relies on this sea-salt composition being identical in the ocean and on sea-ice. First-year sea-ice is rich in sea-salt and available data indicates this sea-salt is not fractionated from its seawater composition . Although there is no evidence of heterogeneous recycling of halogens over multi-year sea-ice, which is effectively drained of sea-salts, the content and composition of sea-salts in multi-year sea-ice should be comprehensively investigated.
Accurate analytical techniques are essential to the determination of iodine and bromine in polar snow and ice samples and to ensure they are reliably compared between sampling sites. It is also crucial to understand the various chemical and physical processes leading to the halogen concentrations found in the snowpack and in the ice core record. The following chapter will summarize the current state of knowledge regarding processes controlling halogens, in particular Br, in the polar atmosphere. These processes are divided into emission, transport, deposition and postdepositional processes.

Halogens in the polar boundary layer
Despite initial measurements conducted in the 1950s, it was not until the 1980s that Arctic ozone depletion events (ODEs) were associated with high levels of brominated compounds (Barrie, 1986). ODEs alter the oxidizing capacity of the atmosphere: minimizing the atmospheric oxidizers for which ozone is a precursor and promoting halogen oxidation pathways. Since the 1980s, several field measurement campaigns have been conducted in the Arctic and Antarctic, in conjunction with laboratory studies and model development necessary for parameterizing multiphase halogen chemistry in the polar boundary layer (PBL). The PBL, and its marine equivalent, the marine boundary layer (MBL), is the layer of the atmosphere directly impacted by processes or emissions originating from the corresponding ground/ice/seawater surface. The PBL or MBL may be quantified as the airmass within which a certain molecule [e.g., Na, bromine monoxide (BrO)] or parameter (temperature, humidity) falls to a specified percentage of its surface value. We provide here a brief overview of the current state of knowledge of emission sources, transport and deposition processes relevant to halogen-based sea-ice reconstructions from ice cores. Two previous review articles provide a comprehensive description of the atmospheric dynamics of the PBL (Anderson and Neff, 2008) and halogen chemistry of ODEs (Simpson et al., 2007b). The reader is also referred to Saiz-Lopez and von Glasow (2012) and Simpson et al. (2015) for overviews of the global context and broader relevance of tropospheric halogen chemistry. Saiz-Lopez et al. (2012) provide a thorough review of atmospheric iodine chemistry.

Sources of halogen emissions
Atmospheric bromine primarily originates from the open ocean surface, although volcanism may be a significant secondary source in some regions such as West Antarctica. The concentration of bromine in seawater is 67.3 ppm by weight (Turekian, 1968), and it is released to the atmosphere via sea-salt aerosol (SSA) or organobromine species produced by ocean biota. SSA is released through the common processes of ocean wave-breaking and bubble bursting, whereas organobromine species are released from the ocean through sunlight-driven processes. Although bromine originates from the oceans, the level of bromine in the PBL is controlled by heterogeneous halogen recycling processes which have been observed over the majority of polar substrates: first-year sea-ice, frost flowers, blowing snow and open-water leads (Abbatt et al., 2012). The activation of halides over sea-ice and snow is discussed in the following sections. The few available measurements of bromine emissions from volcanoes located at high latitudes indicate that volcanism may be a locally significant source of halogens to the polar atmosphere. The volcanoes that have been studied are Mt Erebus in Antarctica (Zreda-Gostynka et al., 1993 and Redoubt in Alaska (Kelly et al., 2013).
Similar to bromine, iodine in the PBL and MBL originates from the ocean surface (Carpenter et al., 2013;, although biological processes are much more critical to the spatial and temporal distribution of iodine in the atmosphere. Many aspects of iodine emission and distribution still need to be better understood. It has not yet been conclusively explained why iodine oxide (IO) is observed at tens of ppt levels throughout coastal Antarctica, but in the Arctic IO is observed at levels in the low-ppt range near to detection limits: the so-called polar iodine paradox (Saiz-Lopez and Blaszczak-Boxe, 2016). The difference has been tentatively attributed to the different proportions and extents of MY and FY sea-ice in each polar region, in additional to potentially different biological communities in each polar region. In the Arctic, an inorganic process also appears to be relevant to atmospheric iodine emissions. A feedback mechanism involving ozone-driven emission of iodine from the ocean surface has also recently been postulated (Prados-Roman et al., 2015), which may be of relevance to the large expanses of the North Atlantic and North Pacific oceans exposed to industrial emissions of NO x pollutants . This ozone-iodine emission feedback loop is discussed further in section 4.3.2.

Bromine
Substrates relevant for halogen chemistry in the polar regions include sea-ice, blowing snow, snowpack and aerosols. Gaseous bromine reservoir species such as hypobromous acid (HOBr), bromine nitrite (BrNO 2 ) and bromine nitrate (BrONO 2 ) are reactively taken up by these substrates leading to a sequence of reactions that liberate halogens (e.g. Br 2 ) from the condensed phase. This reaction cycle leads to the exponential growth of gas phase bromine as the uptake of one Br atom is followed by the release of two Br atoms. The photolysis of emitted precursors such as Br 2 and the subsequent oxidation by ozone of the bromine atoms leads to the formation of BrO, which is the most common bromine species measured in the atmosphere. Further reaction of BrO with hydrogen and nitrogen oxides leads to the formation of HOBr, BrNO 2 , BrONO 2 and hydrobromic acid (HBr) reservoir species, thereby closing the cycle of polar bromine chemistry.
The importance of these substrates for halogen activation has been evaluated through a series of observation-and/or modelbased studies (Abbatt et al., 2012;Pratt et al., 2013;Simpson et al., 2018). Primary criteria include salt concentration, activation surface area and micro environmental conditions such as acidity, humidity and wind speed. As sea-ice grows, the surface salt content is increased through brine rejection processes. Frost flowers may form and both seawater and blowing snow may contribute to the distribution of salts over the sea-ice surface. Sea-ice that persists through summer to become multiyear sea-ice is typically depleted of salts and therefore becomes less effective for halogen recycling. Satellite-based observations of BrO indicate that heterogeneous recycling occurs over freshly formed sea-ice, although more detailed studies have begun to elucidate the specific conditions contributing to such recycling (Abbatt et al., 2012). In particular, the role of multi-year sea-ice for bromine as well as for sodium emissions is still to be explored. Sea-ice salinity decreases with age, although some multi-year sea-ice snowpack measurements suggest that multi-year sea-ice regions have the potential to play an active role in Arctic boundary layer bromine chemistry (Peterson, 2019). Throughout the rest of the review we discuss the source of bromine and sodium in terms of first-year sea-ice surfaces although we acknowledge that multi-year sea-ice, particularly second-year sea-ice, could play an as-yet unquantified role in polar halogen chemistry.
Early studies of bromine recycling in polar regions proposed frost flowers as both a mechanism for enriching bromine and a substrate for heterogeneous recycling processes (Rankin et al., 2002). The enrichment of bromine and other salts in frost flowers was proposed as a means of ensuring fractionation of bromine beyond concentrations found in sea water. Furthermore, the complex structure of frost flowers was proposed to enhance the active surface area available for halogen chemical reactions to occur. Frost flowers are more prevalent in Antarctica than in the Arctic, and do not account for observed Arctic ozone depletion events (Simpson et al., 2007b). Analogous studies of sodium transport in Antarctica have demonstrated that frost flowers are relevant to the salt loading in areas of coastal Antarctica  but can also influence sodium flux across continental Antarctica (Legrand et al., 2016).
As the role of frost flowers in bromine recycling became less clear, attention turned toward the influence of blowing snow and saline snow on sea-ice (Frey et al., 2020;Pratt et al., 2013;Yang et al., 2010). Blowing snow may be generated from moist marine air containing high quantities of marine aerosol. Furthermore, it may take up salts from the sea-ice surface, either from salty brine or seawater that has been rafted onto the sea-ice surface. Snow cover over sea-ice has been observed to be rich in sea-salts, particularly in the deeper layers that interact with open brine channels emerging from the sea-ice matrix. Both model and observational studies have linked blowing and saline snow to bromine heterogeneous recycling and ozone depletion events.
Currently, it is understood that blowing snow over sea-ice is a suitable precondition for bromine activation, but it is still necessary to separate the various factors that co-exist under those conditions. It has been observed that low pH conditions promote halogen recycling but are not an essential component. Overall, abundance of marine salts in a mobile matrix, such as enriched brine channels found in first-year sea-ice, is a strong precondition to bromine activation. These factors promote the importance of first-year seaice as a critical substrate for bromine recycling.
The complex conditions of blowing snow do allow for additional processes of bromine recycling. For example, bromine recycling appears to occur between the gas and aerosol phase within the airborne blowing snow. As bromine recycling occurs within a moving airmass, additional factors such as deposition and remobilisation of snow may come into play. Some of these considerations are discussed in sections 3.5 and 3.6.

Iodine
Due to its biological utility, iodine activation is strongly linked to algal communities that inhabit sea-ice. Consequently, organic forms of iodine, such as iodocarbons and methylates, are primarily identified in the vicinity of sea-ice. Biological activity primarily occurs on the underside of the sea-ice, therefore mechanisms for the transfer of iodine through the sea-ice to the upper surface is an essential precondition to activation and atmospheric emission of iodine . An efficient mechanism for such transport is the formation of liquid brine channels resulting from brine rejection, which is favored in FYSI and in ice that is warmer than À5 C. It has been proposed that the very low concentrations of iodine in the Arctic are due to the prevalence of thick, older multiyear sea-ice, which inhibits the transmission of light required to sustain photosynthetic organisms in the water column below the ice (Arrigo and van Dijken, 2011). The present trend of thinner and younger Arctic sea-ice (Stroeve et al., 2012b) will be an important test to this proposal, to see if atmospheric iodine levels increase in correspondence with the development of a more seasonal sea-ice environment (Arrigo et al., 2012;Horvat et al., 2017).

Halogen activation over land-ice
There have been few studies of halogen activation over snowpack from which definitive findings can be concluded. This lack of research is due to the geographical dispersal of the study locations and the low bromine concentrations encountered in polar snow. In some cases, bromine concentrations are at or below detection limits, which places limitations on the study of deposition or emission processes. Two such studies will be described here.
The Greenland Summit Halogen-HO x Experiment (GSHOX), conducted in 2007 and 2008 at Summit station in central Greenland, studied surface photochemical processes at the site and considered mechanisms for transport of reactive bromine to central Greenland . Samples were collected every 12e24 h: in the gas-phase from the boundary layer and the snowpack (firn-air); in the particulate-phase by ground-based aerosol samplers; and surface snow was also sampled. The bromine results were presented by Dibb et al. (2010): trace levels of bromine were found in boundary layer air (0e1 ppt with a median of 0.3e0.6 ppt) and snowpack (approximately 10 nmol/kg) although bromine/sodium ratios in both matrices were consistently enriched with respect to seasalt. In air, the median bromine concentration was enriched compared to the seawater Br/Na mass ratio by 16 and 8 times (respective results from 2007 to 2008 measurement campaigns) and in snowpack the median enrichments were 33 and 47 times greater (again for the respective 2007 and 2008 measurement campaigns). Bromine enrichment only approached oceanic seasalt values during winter sodium spikes.
The GSHOX experiments were conducted during spring/summer and demonstrated active bromine photochemistry occurring in the snowpack and boundary layer at an altitude of more than 3 km above sea level and hundreds of kilometers from the coast. Importantly, it was observed that surface concentrations of BrO could only be accounted for by emission of bromine from the snowpack. Three mechanisms for transport of reactive bromine to the Summit site were proposed: "Rare rapid transport events" were observed during GSHOX. These are analogous to 'atmospheric rivers' in which there is a rapid transport of air masses from the MBL up to the central Greenland ice sheet. These events are not sufficiently frequent to account for the majority of bromine observed at Summit site. "Leapfrogging" boundary-layer transport, involving surface deposition and re-emission of reactive bromine, as well as possible reactive halogen chemistry occurring on aerosols during transport. This process was considered to be of particular importance during springtime, when there is a stable boundary layer and bromine photochemistry is particularly active over sea-ice (Fernandez et al., 2019). Transport via the free troposphere, with surface-generated bromine lofted into the free troposphere where it is available to be later deposited at distant sites such as Summit.
Importantly, the results revealed that despite diurnal variability, bromine concentrations in the snowpack accurately represented boundary-layer BrO concentrations. Building on the initial findings of Simpson et al. (2007a), the GSHOX study suggests that photochemical bromine recycling processes can occur far from coastal settings and can lead to enhanced bromine concentrations at those sites. Legrand et al. (2016) investigated sea-salt and bromine content in aerosols and snowpack at Dumont D'Urville (DDU) and Concordia stations in Antarctica. DDU is a coastal station, rich in seasalt aerosol, whereas Concordia has a weak sea-salt input due to its location approximately 1100 km from the East Antarctic coast and 3233 m in altitude. Combining state-of-the-art chemical models, aerosol and snowpack measurements, they found that bromine was depleted in marine aerosol at DDU, suggesting overall loss of bromine from the aerosol phase to gas phase. The very low concentrations of aerosols at Dome C precluded any interpretation of the bromine content relative to sodium. Based on depletion of sulphate/sodium ratios indicative of mirabilite (i.e., frost flower) production, the sea-ice surface was identified as a dominant source of marine aerosol to DDU (70 %) and a significant source to Dome C (50 %). Similar to the findings of Dibb et al. (2010), concentrations of nitrate and bromide were greatest in snowpack surface layers, indicative of photochemical activation, although photochemical activation of bromine was not found to be sufficient to account for near-surface air concentrations. Instead, Legrand et al. (2016) concluded that marine aerosol sources were the only credible source of bromine found across the Antarctic plateau.
3.4. Seasonality of halogen activation and post-depositional effects 3.4.1. Satellite observations The spatial and temporal distribution of reactive halogens in the atmosphere is effectively captured by sensors on polar-orbiting satellite platforms. Polar-orbiting satellites typically achieve total coverage of the Earth surface within a 48-h period through overlapping observation swathes. Total atmospheric column concentrations of BrO and IO are currently determined by the GOME-2 and SCIAMACHY sensors, respectively (Burrows et al., 1995(Burrows et al., , 1999. As these sensors determine BrO and IO column concentrations from sunlight reflecting off the Earth surface, observations are not possible during the polar winter nor from within 5 latitude of the poles. As mentioned, these sensors determine the concentrations of BrO and IO in the total atmospheric column, hence the altitude of any detected halogens must be inferred. Nonetheless the observations derived from these sensors present a clear picture of enhanced atmospheric bromine concentrations above sea-ice, ice shelves and coastal areas during the polar spring and summer. Enhanced IO is only observed in the Antarctic, again over sea-ice, ice shelves and coastal areas. In the Arctic, IO has not been observed above the detection limit of approximately 10 12 molecules/cm 2 .

Bromine
Seasonality of bromine activation has been investigated through a variety of methods, from satellite-based observations to aerosol sampling and snowpack monitoring. Saiz-Lopez et al. (2007) were the first to report seasonal variability of BrO and IO in the boundary layer at Halley station, Antarctica, observing a similar behavior in both molecules e a springtime peak with elevated summer and low winter values (Fig. 4). Sch€ onhardt et al. (2012) produced maps of BrO across Antarctica and the Arctic, identifying enhanced concentrations of BrO over sea-ice throughout the springtime (August to December) and lower concentrations during summer (Fig. 5). This result was confirmed and expanded upon by Bougoudis et al. (2020) who provided a 22-year-long merged time series (1996e2017) of Arctic tropospheric BrO vertical column densities (VCDs) derived from four different ultravioletevisible (UVeVIS) satellite instruments (GOME, SCIAMACHY, GOME-2A and GOME-2B). From this multidecadal dataset, Bougoudis et al. (2020) determined an increasing trend of about 1.5 % per year in the extent and magnitude of tropospheric BrO VCDs during polar springtime, temporally and spatially related to the increase in firstyear ice extent in the Arctic north of 70 N (Fig. 6). Similarly, in the Southern Ocean, the highest tropospheric BrO VCDs are observed during austral polar springtime (Sept-Oct-Nov, SON) consistent with the presence of seasonal sea-ice (Fig. 7). During late-spring/ summer, maximum BrO concentrations are observed to shift from the sea-ice zone toward ice shelves and the coastal margins of the continent. Validated through a variety of chemical transport models, the importance of bromine activation for ozone depletion events has been demonstrated thoroughly (von Glasow and Crutzen, 2014). Photochemical reactions are critical to bromine explosions; hence they are associated with the polar springtime and arrival of solar activity. Although the photolysis of bromine compounds has been identified as a critical component to bromine recycling in Polar Regions, a variety of halogen activation processes have been identified that do not require solar radiation and hence may occur through the polar winter (Abrahamsson et al., 2018;Oum et al., 1998). Furthermore, nighttime production of bromine precursors ready for activation in the presence of sunlight is also a topic of consideration. At this stage there are only a few observation campaigns that have attempted to address this issue (Abrahamsson et al., 2018;Foster et al., 2001;Simpson et al., 2018) and a thorough evaluation using a coupled chemistry-general circulation model (GCM) is necessary for greater understanding of these processes.
A number of factors have been suggested to contribute to postdepositional loss or remobilisation of bromine from snowpack. The most thoroughly demonstrated factor is that of photochemical oxidation, as described previously in the studies by Dibb et al. (2010) and . A recent study indicates that surface reactions may also play a greater role than previously thought e leading to greater oxidation of bromide by ozone at the aqueous solution-vapour interface (Artiglia et al., 2017). Analogous to earlier work conducted on photo-oxidation of nitrate, the snow accumulation rate may be a critical component to limiting the exposure of snow surfaces to sunlight. That is, regions of high snow accumulation should be less sensitive to photo-oxidation of bromine, on account of the relatively shorter time that a deposited snow layer is exposed to sunlight. Coastal regions of Antarctica and Greenland generally feature greater accumulation rates than inland regions, hence this effect should be most apparent on the central Antarctic plateau, and to a lesser extent in central Greenland. The effect of acidity, particularly nitrate, in snowpack has also been considered important to bromine activation (Pratt et al., 2013). Maselli et al. (2017) suggested that increased acidification of Arctic sea-ice may account for increasing bromine concentrations in Greenland surface snow since the 1950s, although it will be interesting to see if this trend follows recent reports of a decrease in acidity of Greenland surface snow layers (Kjaer et al., 2016).

Iodine
The seasonality of iodine in Antarctica is distinct from that of bromine, despite early indications of a similar annual profile (Saiz-Lopez et al., 2007). Sch€ onhardt et al. (2012) identified widespread but low concentrations of IO across Antarctica in early spring, suggesting a possible winter emission mechanism. Annual cycles of Antarctic IO are shown in Fig. 8, indicating strong seasonal and spatial variability. As the atmosphere warmed and spring developed, it was observed that IO would peak above Antarctic sea-ice, later and more briefly than the "bromine explosion" occurring from September to December. IO is produced by algal communities hosted below sea-ice, and it is only when these algal communities have been exposed to light for some time, that the sea-ice warms  (Fernandez et al., 2019;G alvez et al., 2016;Kim et al., 2016) to be potential sources of active iodine in the polar boundary layer. There is sufficient evidence to conclude that iodine photochemistry is reasonably active in snow, with satellite-and insitu-  2010) proposed that interstitial air is a mechanism by which iodine can be converted into the gas phase and released from snowpack, as only the upper 10 cm of snowpack receives sufficient sunlight to enable iodine photochemistry. Presenting a four-year time series of sodium, iodine and bromine, Spolaor et al. (2014) showed that iodine was not present in century-old summer snow strata, although concentrations of up to 0.3 ppb of iodine were present in winter strata. In the same work, the authors presented sub-annual iodine records from Greenland and Svalbard although concentrations were lower and seasonal variability was less clear. Such findings are consistent with satellite observations of below-detection-limit-levels of IO in the Arctic (Sch€ onhardt et al., 2012).
Retention of iodine is much more effective on rock and mineral dust surfaces than within the ice matrix. Early studies of Antarctic meteorites identified high concentrations of iodine in surface layers and stability of metallo-iodide molecules has been reported. Spolaor et al. (2013b) found that iodate is only present in the Talos Dome ice core during the dust-rich periods of the last two glacial maxima. Iodine was present throughout the 200 thousand-year record, but iodate was only found during dust-rich periods, implying the formation of stable metallo-iodate complexes on the mineral dust surface.

Re-emission of halogens from Antarctic surface snow: case studies
As both iodine and bromine are highly reactive and participate readily in oxidation reactions, it is important to consider their behavior in surface snow after deposition. Results obtained from Law Dome ice cores and Neumayer surface snow demonstrate that iodine peaks are found in winter strata whereas bromine peaks are found in spring and summer strata (Frieb et al., 2010;Spolaor et al., 2014;Vallelonga et al., 2017). With iodine emissions attributed to spring/summer biological activity, winter iodine peaks found in Law Dome ice indicate post depositional mobility of iodine (Fernandez et al., 2019). Evidence of iodine reactivity in the snow pack is also highlighted in more recent work by Spolaor et al. (2018), where the diurnal variation of iodine in surface snow is shown to be driven by changing incoming solar radiation. In contrast, bromine seasonality at the salt-rich, high-accumulation site of Law Dome features higher concentrations in spring and summer, consistent with spring/summer heterogeneous chemical recycling over first-year sea-ice (Spolaor et al., 2014;Vallelonga et al., 2017). The post depositional behaviour of iodine and bromine further inland is a topic of necessary study, particularly considering the high levels of solar irradiance on the dry and elevated East Antarctic plateau. In the following paragraphs we present experiments reporting halogen partitioning in Antarctic snow surface covering a range of geophysical parameters: coastal vs inland, depletion vs deposition zones, and high vs low accumulation. With an altitude of 3233 m a.s.l., very little snow accumulation (0.03 m ice equivalent/yr, mostly consisting of diamond dust, i.e. "clear sky" precipitation) and located approximately 1100 km from the East Antarctic coast, Dome C may be considered an extreme location for halogen stability in snowpack. In 2015/16, surface snow samples (0e3 cm sampling depth range) were collected with approximately 12-h resolution (10 a.m. and 6 p.m. local time) for more than a month . The results are shown in Fig. 9. Surface snow concentrations of Na, Br and I were observed to vary after a precipitation event, consistent with a change in the wind source: higher concentrations of Na, Br and I between 18 and 25 November 2015 were driven by precipitation from air masses originating from the Ross Sea region; and lower concentrations of the elements from 10 to 29 December 2015 resulted from limited precipitation, primarily diamond dust, of sea-salt-depleted (i.e., non-coastal) air mass origins. Following the precipitation event, iodine concentrations in surface snow declined greatly with a loss close to 80 %. Bromine concentrations decreased by approximately 50 % and sodium decreased about 35 %. The decrease in sodium is most likely due to physical processes such as wind-driven snow remobilisation, whereas iodine (and bromine, to a lesser extent) demonstrate susceptibility to photochemical remobilisation. The findings of this study are consistent with others pointing to remobilisation and loss of bromine from snowpack in lowaccumulation zones of central Antarctica (Legrand et al., 2016;McConnell et al., 2017).
The Italian/Korean GV7 ice core drill site is located in North Victoria Land, approximately 200 km from the coast with an elevation of 1950 m a.s.l. and a moderately high snow accumulation. The topmost 3 cm of the snow surface was sampled daily for 38 days (28 November 2013 to 6 January 2014), after which a 2.5 m snow pit was sampled at the same location. No accumulation was observed over the first half of the experiment, followed by 6 cm of steady precipitation from mid-December to early-January (Fig. 10). The left panel of Fig. 10 displays the 3 cm surface sampling, with the 2.5 m snow pit results shown on the right side of the figure. The figure indicates that Br and Na are well preserved after snow deposition at this location, as there is no appreciable difference between the surface snow (left panel) and the snow pit (right panel) concentrations. Iodine concentrations are distinctly lower in the snow pit (45 ppt) compared to surface snow (70 ppt), suggesting an iodine loss of approximately 30 % at this location. These results are consistent with previous observations of iodine loss and/ or remobilisation in snowpack at coastal, high-accumulation sites (Frieb et al., 2010;Spolaor et al., 2014).
Deposition characteristics of sodium and bromine have also been investigated in a compilation of sodium and bromine fluxes reported for 11 ice core drill sites across West Antarctica and Dronning Maud Land (McConnell et al., 2017, Supplementary section 6). The accumulation rates covered by the compilation vary from 0.027 m/yr (NUS7-2) to 0.40 m/yr (DIV), comparable to conditions at Dome C and Talos Dome, respectively. The bromine and sodium flux data for each site are plotted against snowfall (i.e., accumulation) in Fig. 11. As described in section 2.2.3, the arrangement of flux vs accumulation creates a linear array in which the gradient represents the wet-deposition component of flux and the y-intercept represents the dry-deposition component. The authors found sodium fluxes to be almost entirely driven by wet deposition; whereas for bromine dry deposition is comparable to or greater than wet deposition, particularly at low accumulation sites. Furthermore, the study demonstrated a negative y-intercept for bromine, 'indicating substantial release of bromine from the snowpack back to the atmosphere after deposition' . It is clear from the compilation that bromine fluxes are consistently and anomalously low at sites with very low accumulation rates less than 0.04 m/y, although the effect is less relevant at higher accumulation sites.
The surface snow experiments described here provide insight into the behaviours of Br and I in different accumulation, irradiation and environmental conditions. We can conclude that halogen postdepositional remobilisation and loss is active on the high Antarctic plateau, with close to 100 % of iodine and at least 50 % of bromine lost from the snowpack at low-accumulation sites. These results are difficult to quantify because confounding processes, such as snow remobilisation, also have a strong influence at lowaccumulation sites. At sites with moderate snow accumulation, located nearer to the coast, bromine re-emission appears to be minor although iodine re-emission is still substantial. Additional experiments should be planned with the aim of evaluating the competing influences of solar irradiation and snow burial on the photochemical activation of halogens in snowpack. Further quantification studies should also be carried out in areas with snow accumulation rates between 0.05 and 0.3 m/y. Finally, it is possible that enhanced solar irradiance due to the destruction of stratospheric ozone since the 1970s (the "hole in the ozone layer") could play an important role in halogen photochemical activation. Such questions may be addressed by more widespread sampling of the snowpack on the central Antarctic Plateau.

Halogen and sea-salt transport from sea-ice to land ice
Considering that seasalt is present in the polar atmosphere only as aerosol, whereas halogens exist in both aerosol and gas phase (as mentioned in sections 2.2.3 and 3.2.1), one would expect a priori that seasalts and halogens will respond differently to the various transport and deposition processes present in the polar boundary layer. Marine aerosols are micron-scale particles (Legrand et al., 2017;O'Dowd et al., 1997) easily mobilised long distances (several hundreds of kilometres) and subject to wet deposition processes. The atmospheric lifetime of gas-phase halogens (HOBr, BrO and Br 2 for bromine; I 2 , IO, OIO and I 2 O 2 for iodine) is limited only by its reactivity and is therefore driven primarily by dry deposition processes. Bromine undergoes multiphase (aerosol-gas) chemical exchanges in the course of its transport from the sea-ice surface to a polar ice cap, leading to the photochemically-driven "leapfrogging" proposed by the GSHOX experiment (section 3.3).
In this section we revisit two important evaluations of bromine and seasalt transport and critically investigate inconsistencies between the two. We go on to review observational data allowing further evaluation of these studies and propose a framework integrating the two.
3.6.1. Models of halogen and sea-salt transport Simpson et al. (2005) presented the first evaluation of spatial variability of bromide in snowpack. Snow sampling was conducted in 2004 within a grid pattern consisting of a north-south sampling line from Barrow station to 280 km inland (Douglas and Sturm, 2004). Ion chromatography (Br À , Cl À ) and Atomic Absorption Spectroscopy (Na þ ) were applied to determine bromide, chloride and sodium concentrations, respectively, in three distinct snowpack layers. The main results from the study were: 1) Minimal to no enrichment of Br À relative to Na þ in frost flowers; 2) a wide range of Br À /Na þ ratios in snowpack from approximately 0.1 to 100 times the ratio found in sea-salt; and 3) Na þ concentrations decrease strongly with distance from the coast whereas Br À concentrations decrease gradually or not at all. Fig. 12 displays the primary findings of the study, indicating different coastal concentrations and rates of change for Na þ and Br À as snow is sampled at greater distance from the coast. This study demonstrated effective transport of recycled bromine inland, with a clear distinction between the aerosol-phase transport of sodium and the mixed gas/aerosol phase of bromine further complicated by the likelihood of heterogeneous bromine recycling during transport. Spolaor et al. (2013b) investigated the transport and deposition characteristics of sodium and bromine in East Antarctica, in the first application of halogens data to sea-ice reconstruction (see sections 4.4.1 and 5.1.1). A 1-D chemical transport model was constrained to a small number of surface snow measurements of Br and Na from Talos Dome (assumed 7 h from sea-ice edge) and Dome C (assumed 24 h from the sea-ice edge), based on the transport of an air mass inland at a constant velocity. The model used deposition velocities of 0.2 cm/s for Na and 1 cm/s for HBr, leading to a faster deposition of bromine relative to sodium. These results ran counter to the findings of Simpson et al. (2005) as well as the a priori physicallybased argument described earlier.
To unravel the contradiction between the schema of bromine transport presented by Simpson et al. (2005) and Spolaor et al. (2013b) we compare representative values of sodium and bromine at the coastal zone (<30 km from the coast), the inland zone (250e300 km from the coast) and a "central ice sheet" zone (>700 km from the coast) from each study, as shown in Table 3. Data equivalent to the "central ice sheet" are not applicable to the  Table 3 clearly demonstrates that the "coastal zone" sodium concentrations used by Spolaor et al. (2013b) are too low by one to two orders of magnitude. For the inland and central ice sheet zones, sodium values are not unreasonable (e.g.: Bertler et al., 2005;Schüpbach et al., 2013). Bromine concentrations reported by Spolaor et al. (2013b) are consistently lower than those reported by Simpson et al. (2005) but do not seem as strikingly problematic as the sodium concentrations. Although the deposition velocity assigned for HBr (1 cm/s) is consistent with other independent modelling studies (eg. Legrand et al., 2016), the most significant bias in the Spolaor et al. (2013b) study is the inaccurately assigned sodium concentrations. As a result, the model used by Spolaor et al. (2013b) was incorrectly constrained and their findings require reassessment.

Compilation of spatial variability: bromine and sodium
Spatial maps of bromine and sodium concentrations have been compiled for Antarctica and Greenland. All available data are included irrespective of their temporal coverage, varying from days or weeks to millennia of deposition. In all cases, ice core data have been selected to represent Holocene (ie, interglacial) values. Fig. 13   The differences between these two datasets may be due to the time periods covered (decades vs centuries, respectively) and/or the analytical methods (IC vs ICP-SFMS, respectively). There is not a strong consistency between bromine concentrations and atmospheric BrO concentrations observed by the satellite-based GOME-2 sensor, particularly across the Antarctic plateau. Fig. 14 shows a composite of bromine data and associated sodium concentrations reported for Greenland. Two transects are included: a 2015 NEEM-EastGRIP traverse and a 2017 Windsled traverse from Kangerlussuaq to EastGRIP, both of which will be discussed in detail in sections 4.1.1 and 4.1.2. Ice core drill sites include TUNU and Summit , RECAP (Maffezzoli et al., 2019), NEEM (Spolaor et al., 2016b) and DYE-3 (this work). The DYE-3 data were obtained by resampling archive pieces of the original core drilled in 1981 (Johnsen et al., 1992) corresponding to the period 10e11 ky b2k. The NEEM samples cover the late Holocene period (0.03e3 ky b2k) and the Renland samples cover the period 0e11 ky b2k. Overall patterns of Br, nssBr and sodium are similar in Greenland and Antarctica: higher concentrations of sodium and bromine at the coast and lower concentrations on the inland plateau. Sodium does not show a consistent north-south gradient along the Greenland ice sheet, although there is a dearth of data corresponding to the southern half of the ice sheet. A strong sodium concentration gradient is observed from the northwest toward the central/eastern sector of the ice sheet, consistent with previously reported atmospheric circulation patterns (Fischer and Wagenbach, 1996;Rhodes et al., 2018) and accumulation gradients (Fischer et al., 1998). Bromine enrichment values increase from low values (<6) at sites located nearer the coast to higher values (>15) in the central and northern part of central Greenland. There are large data gaps in coastal and southern Greenland e these could be ideally filled by traverses from the coast to inland stations. Such traverses could include: Thule to NEEM; EastGRIP to station Nord (ideally travelling via the TUNU drill site); and EastGRIP to Zackenberg station. A better understanding of bromine variability across northern Greenland would also be important to the interpretation of existing (Maffezzoli et al., 2019;Spolaor et al., 2016b) and projected (Rhodes et al., 2018) reconstructions of Arctic sea-ice extent.

Synthesis of halogen transport models
Here we combine the spatial data presented in Figs. 13 and 14 into one graph, showing the variability of bromine and sodium as a function of distance from the coast. The data (Fig. 15, Table 4) show that spatial variability of sodium and bromine are consistent across both Antarctica and Greenland, and therefore suitable models should be applicable to both locations. We plot Br and Na concentrations, Br enr and nssBr as a function of the average distance of each site from the sea-ice edge from which airmasses are likely to travel to the site. The calculated distance is an approximation considering the main atmospheric circulation pattern for the area as well as variability in sea-ice extent during the Holocene. In both the Arctic and Antarctic, the sea-ice edge location coincides with the continental coastline since multi-year sea-ice is present mainly in the Canadian Arctic and in only a few sectors of the Antarctic coast. In addition, the average value for Br, Na, Br enr and nssBr for each site is calculated for the entire temporal coverage of the record. Fig. 15 demonstrates the different behaviours of sodium and bromine concentrations as airmasses travel from the coast to central Greenland and Antarctica. Sodium demonstrates a faster  decline than bromine, as it travels from the coast. At approximately 400 km inland, sodium deposition is effectively complete and concentrations remain stable further inland. Bromine shows a quasi-linear decreasing trend from about 100 km inland to about 1000 km inland, after concentrations are extremely low and are very likely present only as a remnant of re-emitted bromine. The different rates of decrease of sodium and bromine suggests different rates of efficiency in the deposition processes moving away from the coast: wet deposition driven by precipitation ensures that sodium deposits faster into the coastal snowpack and then reaches a stable concentration across the very dry central areas of Antarctica. Over the Antarctic plateau, bromine concentrations effectively drop to detection limits of the analytical techniques used. The different Na and Br spatial snowpack concentration patterns are also reflected in the resulting Br enr and nssBr patterns, which are again very similar for the Arctic and Antarctic. Br enr starts with low values near the coast and gradually increases to the highest values approximately 300e600 km inland.
Br enr values then quickly decrease further inland and are less than 1 for all sites greater than 800 km from the coast. The model advanced by Spolaor et al. (2013b) is of use once it is properly located e that is, once the bias caused by inappropriate constraints has been mitigated. The increase of Br enr from the coast to 300e600 km inland is consistent with the findings of Simpson et al. (2005) while the trend determined further inland (>600 km from the coast) aligns with the transport model proposed by Spolaor et al. (2013b). The suitability of these models for interpreting bromine concentrations (and particularly Br enr values) for sea-ice reconstructions will be described in the following section.

The role of halogen transport models in sea-ice reconstructions
Based on the integration of the Simpson et al. (2005) and Spolaor et al. (2013b) models of halogen transport, we can draw the following initial conclusions regarding the suitability of bromine as a metric for reconstructing sea-ice: 1) The increasing pattern of Br enr vs distance from the coast supports the use of Br as a metric related to sea-ice, specifically the sea salt release from the sea-ice surface, for broad swathes of coastal Antarctica (to approximately 600 km inland) and the majority of the Greenland ice sheet. This band of distance from the coast reflects the zone in which both Br and Na are decreasing with distance. 2) The zone of Antarctica located more than 800 km from the coast e ie, the vast expanse of the Antarctic Plateau e is potentially of limited value for quantitative sea-ice reconstruction. This zone features very low concentrations of bromine, which may be subject to photolytical reactivation and/or snow remobilisation processes. Variability in bromine concentrations and Br enr at such inland sites could reflect changes in solar irradiation, changes in meteorological patterns (influencing deposition of sodium and bromine) and changes in wind fields, influencing both deposition and surface snow remobilisation. 3) During glacial climate periods, the increased extent of Antarctic ice would lead to an outward expansion of the location of the sea-ice edge, thereby increasing the zone of low-accumulation across central Antarctica. Thus, locations which are currently within 800 km of the Antarctic coast will likely fall outside this zone and therefore cease to be reliable sites for sea-ice proxy reconstruction. This may be the case for Talos Dome, the site for which the Spolaor et al. (2013b) model was first proposed. In such a case, the site may be a qualitative indicator of sea-ice extent e useful for demonstrating the timing of transitions between 'large' and 'small' sea-ice extents rather than continuously quantifying the sea-ice extent over glacial/interglacial cycles. 4) Considering glacial-interglacial transitions, tracts of multi-year sea-ice are likely to form at the coast during glacials which will lead to a displacement of the first-year sea-ice edge away from the coast. For sampling sites located within 100 km of the coast, a relative increase in the Br enr signal is therefore likely occur independent of any change to the first-year sea-ice extension. For ice cores collected further inland (i.e. >300 km inland) the extension of the multi-year sea-ice should cause a decrease of the signal due to the increased distance of the bromine gas phase source. 5) The above considerations have taken into account only the distance from the sea-ice edge to the sampling site. Any potential reconstruction of sea-ice extent will have to take into account two factors: firstly, the distance from the sampling site to the sea-ice edge, and secondly, the area of sea-ice contributing sodium and bromine to the sampling site. Where the general climate conditions have not changed (i.e., during an interstadial), one would expect the first factor (distance) to be approximately constant, and hence bromine variability would primarily reflect changes in the area of FYSI available for photochemical bromine recycling. Where the change in the location of the sea-ice edge is substantial with respect to the distance to the sampling site, the transport models proposed by Simpson et al. (2005) and Spolaor et al. (2013b) will have to be appropriately deployed to accurately interpret the results. 6) The models of Simpson et al. (2005) and Spolaor et al. (2013b) should be further refined by a suitably designed process-study experiment. The experiment design employed by GSHOX should be revisited and ideally applied to East Antarctica, where the full extent of the Simpson et al. (2005) and Spolaor et al.
(2013b) models can be tested and an array of monitoring stations can be established e both at the coast and at inland ice core drilling sites.

Transect data
In this section we consider studies presenting data from surface transects in Antarctica and Greenland. Typically, such data covers a short time period e less than a decade e due to the limited time available at each sampling location. Nonetheless, the spatial coverage offered by such datasets compensate for their potential lack of temporal representativeness. The transects discussed here will be divided into two sub-groups: datasets that were not considered in the previous chapter due to their limited temporal coverage (<3 years); and those with a longer temporal coverage.
4.1.1. Transect data with sample coverage <3 years Li et al. (2014) reported impurities in surface snow and snow pit samples collected between December 2012 and January 2013 on a transect from Zhongshan Station to Dome A, Antarctica. The character of the sampling sites varied from coastal (40 km inland, altitude 832 m a.s.l., accumulation 260 kg/m 2 /yr) to the highest parts of the Antarctic plateau (1250 km inland, altitude 4093 m a.s.l., accumulation 23.5 kg/m 2 /yr). Sodium (Na þ ) and Br À were reported for surface snow samples collected every 10 km and 2 m deep snow pits sampled every 100 km along the traverse. The results were consistent with the findings of Simpson et al. (2005), but with a much smoother gradient of sodium decrease. Concerns regarding the representativeness and accuracy of these data, with respect to other measurements available from the Antarctic Plateau, have been discussed in section 3.6.4.
Surface snow was sampled every 10 km during a 120 km transect from Casey station to Law Dome summit in coastal East Antarctica . This region is strongly influenced by maritime airmasses and features a strong accumulation gradient, from 150 kg/m 2 /yr near Casey station to 800 kg/m 2 /yr at the summit. Despite a few outliers, sodium concentrations were usually high (average 180 ppb), bromine concentrations were usually low (average 1.3 ppb) and Br enr varied from 0.3 to 1.7. Although little can be concluded with regard to halogen transport due to the short distance covered by the transect, the data from this transect can be taken as indicative of halogen and seasalt concentrations representative of the coastal zone.
In May 2015, scientific sampling and observation programs were conducted in conjunction with a 460 km traverse from northwest (NEEM) to Northeast (EastGRIP) Greenland (Schaller et al., 2016). Surface snow samples were collected every 10 km, in addition to six firn cores drilled to approximately 12 m depth (Kjaer et al., 2021). The traverse followed a strong accumulation gradient from 220 kg/ m2/y in northwest Greenland (NEEM Community members, 2013) over the central ice divide into northeast Greenland which features accumulation rates of 110 kg/m 2 /y : the lowest reported for the Greenland ice sheet. The results are presented in Fig. 14 and Appendix A. Maffezzoli et al. (2017) investigated sodium and bromine in snow pits on a 200 km transect in Victoria Land, Antarctica, from Talos Dome (250 km from the Ross Sea) to core site 6 (85 km from the Southern Ocean). The location of the traverse enabled an investigation of the competing influences of seasalt arriving from the Ross Sea and Southern Ocean sectors, but for the purpose of this review we can consider the transect data as representative of the "inland sector" located a few hundred km from the coast. These results indicate the model constraints applied by Spolaor et al. (2013b) (Table 3) for Talos Dome site were accurate for sodium but about an order of magnitude too low for bromine. Extending the Victoria Land transect from Talos Dome to Dome C, seven short (4 m) snow cores were collected in November/December 2016 (this work). The samples were analyzed for sodium and bromine at the University Ca' Foscari of Venice in mid-2017, and the results are included in Fig. 13 and Table 4. There is no clear spatial variability in sodium or bromine concentration over the sampled locations and thus the results may be considered typical for the Antarctic ice sheet plateau.

Transect data with sample coverage >3 years
Bromine and sodium fluxes have been reported for two Norwegian-US collaborative traverses through Dronning Maud Land conducted in /8 and 2008/9 (McConnell et al., 2017. Seven ice cores were collected over the two traverses, which represent high-altitude plateau conditions. Details of the sampling and ice core specifications (McConnell et al., 2014) are reported separately to the halogen data .
In May/June 2017, a windsled traverse was conducted from southwest Greenland (inland from Kangerlussuaq) up to the East-GRIP ice core drilling site, along which a number of 2 m deep snow cores were collected. The snow cores each cover the period 2014e2017 CE and were subsampled at the University of Copenhagen with aliquots sent to the University Ca' Foscari of Venice for sodium and bromine determination. We report here the final six cores collected along the traverse, from the central ice divide (north of Summit station) to EastGRIP site (Fig. 13, Table 4). The sampling Fig. 15. Compilation of sodium and bromine concentrations from Arctic and Antarctic firn and ice records as a function of distance from the coast. The figure shows the halogen transport range investigated by Simpson et al. (2005) as well as the transport range relevant to the dataset described by Spolaor et al. (2013b) when their transport model is appropriately constrained. See Table 4 for data and references. locations are all far inland and at high altitude (<2900 m a.s.l.) and therefore are important for investigating halogen deposition to the central Greenland ice sheet between EastGRIP and Summit station. There is a strong accumulation gradient also featured along this traverse, essentially halving from the ice sheet divide (accumulation approximately 230 kg/m2/y) to EastGRIP (Karlsson et al., 2020;Nielsen et al., 2015).

Seasonal timescales
Only a few studies have reported the variability of halogens in ice on seasonal timescales. These include three for Antarctica (Dronning Maud Land, Law Dome, Northern Victoria Land) and four for the Arctic (Severnaya Zemlya, Svalbard and two sites in Greenland). For most of these locations, only bromine seasonality has been reported. The results are summarized in Tables 1, 2 and 4 and also described in the following sections.

Antarctica
Seasonal variability of halogens in Antarctic snow and ice have been reported from coastal locations where snow accumulation is sufficiently high to allow annual cycles in snowpack to be resolved. The first measurements of iodine in Antarctic snowpack were by Frieb et al. (2010) who reported seasonal strata near Neumayer station, Dronning Maud Land, in a 2 m snowpit covering two annual cycles. High iodine concentrations in winter strata, and low concentrations in summer strata, were interpreted as evidence of photochemical remobilisation of iodine from snowpack during the Antarctic summer. This pattern of low summertime iodine concentrations was confirmed by Spolaor et al. (2014) who reported a four-year sequence (1910e1914 CE) of iodine from a Law Dome ice core. The same article was the first to report the seasonality of bromine in ice in Antarctica, finding a strong summer peak in both nssBr and Br enr . A smaller Br enr peak in late-summer/autumn was also found and tentatively attributed to the deposition of gas-phase boundary-layer bromine reservoir species at the austral dusk. At the same site, the seasonal pattern of Br enr was confirmed and extended by Vallelonga et al. (2017), presenting a high-resolution study of a Law Dome firn core covering the period 1987e2012 CE. For this more detailed study, the summer peak was confirmed but the late-summer/autumn peak was not clearly observed. Maffezzoli et al. (2017) further confirmed the summer peak in Br enr , from a study of six short firn cores collected across Northern Victoria Land in 2015. The highest average Br enr values were found in early summer (October/November). In these samples, iodine concentrations were consistently very low although the few observed concentration peaks corresponded to winter strata. The iodine peaks in Northern Victoria Land (>0.1 ppb) were much lower than those reported for Neumayer (0.6 ppb) and Law Dome (0.3 ppb), consistent with satellite observations of very low atmospheric column amounts IO above Northern Victoria Land, compared to the lower-altitude coastal regions of Antarctica (Spolaor et al., 2014).

Arctic
Arctic iodine concentrations are much lower compared to Antarctica and few measurements are available. Spolaor et al. (2014) reported the first profiles of iodine in the Arctic from a 3year snow pit from the NEEM ice core drill site in northwest Greenland and from a 5-year snow pit in Holtedahlfonna, Svalbard. In both profiles, surface iodine concentrations (0.2 ppb maximum) quickly decreased to lower values (0.02 ppb for Holthedalfonna and 0.05 ppb for NEEM) within two years of deposition. Only the Severnaya Zemlya ice core (Spolaor et al., 2016a) shows some longterm retention of an iodine signal going back to the 1950s. For the Severnaya Zemlya record, iodine concentrations were highly variable (0.01e3 ppb) but consistent peaks were observed in April (late winter/early spring) and lower concentrations in October. Such behavior is quite different to that observed in Antarctica, perhaps due to the proximity of Severnaya Zemlya to the marginal sea-ice zones of the Laptev and Kara Seas.
Bromine seasonality in the Arctic has been found to be consistent with the behavior shown in Antarctica, with higher concentrations in spring/summer (Fig. 16). Together with iodine, Spolaor et al. (2014) also reported the first Arctic profiles of bromine in snowpack from NEEM and Holtedahlfonna. Despite the short timeseries, consistent summer peaks of bromine and Br enr were observed. These findings have since been confirmed by longer timeseries from Severnaya Zemlya (1950e1999 CE) reported by Spolaor et al. (2016a) and central Greenland Summit2010 ice core (1750e2010 CE) reported by Maselli et al. (2017).

Antarctica
Only one study of halogen variability over decadal-centennial timescales is available from Antarctica, that being a composite record of bromine, iodine and sodium from Law Dome and covering the period 1927e2016 CE . The record is a Fig. 16. Seasonal variability of bromine enrichment (Br enr ) from Arctic ice cores. The period covered by each record is shown on the figure. The mean of the data is indicated by a solid line with 1s error bars and the median is indicated by a dashed line. The data are from ice cores drilled at Summit  and NEEM (Spolaor et al., 2014) in Greenland, Severnaya Zemlya (Spolaor et al., 2016a) and Svalbard (Spolaor et al., 2013a). composite of two firn/ice cores and a snowpit, respectively covering the time periods 1927e1986, 1987e2012 and 2015e2016 (Fig. 17). The Law Dome halogen records confirm earlier measurements of median concentrations of sodium (81 ppb), bromine (2.1 ppb) and iodine (0.056 ppb), further finding that these median values were 5e20 % lower than their corresponding arithmetic means due to the 'spiky' nature of the data. No significant trends were identified in the three data series over the sampled time period.

Arctic
A few records have been published reporting halogen variability over decadal-centennial timescales in the Arctic. These range from one decade of iodine and bromine at Svalbard (Spolaor et al., 2013a) to a half-century of iodine from the Alps  and nearly three centuries of iodine  and bromine  in ice core records from Greenland. A longer record of bromine and iodine covering the period 1950e1999 was reported from Severnaya Zemlya (Spolaor et al., 2016a). Bromine and iodine displayed a high level of interannual variability, with mean (1s) concentration values of 5.2 (6.1) ppb for bromine and 0.23 (0.34) ppb for iodine. Bromine showed an increasing trend from 1950 to the late 1970s, followed by a decreasing to 1999. Iodine showed no apparent trend until the late 1970s, after which concentrations decreased. Sodium variability (mean 187 ppb, standard deviation 197 ppb) was comparable to that of bromine and iodine but no trend was apparent from 1950 to 1999. Variability of bromine and iodine were linked to changes in sea-ice cover in the adjacent Laptev Sea, as discussed further in section 5.2.1.
An iodine concentration record covering the period 1890e1990 CE was reported from ice cores recovered from the Col Du Dome (CDD) saddle in the French Alps, displaying a periodically increasing trend in iodine since 1950 CE . Col Du Dome, located at 4250 m asl, is an ideal location for sampling tropospheric air arriving from the North Atlantic Ocean and from western Europe, and anthropogenic pollutant emissions records have been previously reported from the location (Barbante et al., 2011;Preunkert et al., 2000). Combining the ice core record with an iodine emissions inventory and the GEOS-Chem chemical transport model, the authors model a pervasive and steady increase in iodine emissions from the surfaces of the North Atlantic Ocean and Mediterranean Sea. This increase in iodine is due to increasing anthropogenic NOx emissions, which lead to higher atmospheric concentrations of Ozone (O 3 ), which promotes the emission of iodine from the ocean surface. The authors note also a change in the primary atmospheric species of iodine, from a predominance of HOI in the pre-industrial period to the more-soluble IONO 2 after 1950 CE. The higher solubility of IONO 2 results in a shorter atmospheric lifetime as well as a greater efficiency of being captured in the snowpack.
The longest iodine record currently available from the northern hemisphere covers the last 250 years and has been extracted from the RECAP ice core in eastern Greenland . The RECAP ice core (Fig. 1) was drilled in 2015 on the Renland ice cap, a coastal ice cap located 2300 m above sea level and isolated from the main Greenland ice sheet. Renland features a relatively high annual snow accumulation rate of approximately 50 cm ice equivalent and receives its precipitation primarily from the North Atlantic Ocean and Nordic Seas. The iodine record shows relatively stable concentrations of 0.02 ppb from 1750 to 1940 CE, then lower values from 1940 to 1970, and a strong increase after 1970 CE. This pattern is also seen in iodine fluxes, but not in sodium concentrations that remained stable around 15 ppb for the whole period. The tripling in iodine concentrations observed from 1950 to 2015 CE was attributed to increasing emissions of oceanic inorganic iodine due to anthropogenic ozone increases, as well as increased emissions of oceanic organic iodine from thinning and retreating Arctic sea-ice .
From two sites in Greenland, a well-constrained 250 year highresolution record of bromine, MSA, sodium and acid species was reported by Maselli et al. (2017). The two core sites investigated were Summit, a medium-accumulation site in central Greenland, and TUNU, a low-accumulation site located in northeast Greenland (Fig. 18). Both records displayed similar trends for MSA and bromine, although concentrations of Br and MSA were consistently lower in TUNU compared to Summit. The authors found a common preindustrial pattern between bromine and MSA, with a stable period of higher values between 1750 CE and 1835 CE and a second stable period of lower values between 1860 CE and 1940 CE. The durations of each stable period varied with respect to the core and analyte considered (MSA or Br). After 1940 CE, bromine and MSA trends diverged, with a sharp decrease in MSA and a sharp increase in bromine leading to a peak in the 1970s. Pre-industrial concentrations of bromine were 0.43 ppb at Summit and 0.34 ppb at TUNU. For the 1970's peak, bromine concentrations were approximately 0.48 ppb at Summit and 0.8 ppb at TUNU. The divergent behaviours of Br and MSA after 1940 CE were thoroughly investigated, as both species displayed similar trends and both have been employed in various manners to reconstruct sea-ice. The decreasing MSA trend was in good agreement with earlier observations of MSA trends in central Greenland (GRIP ice core;  and Svalbard . A recent study e including the Summit and TUNU MSA records -has proposed that the primary MSA emission source influencing Greenland over the industrial period is North Atlantic Ocean primary productivity rather than sea-ice algae (Osman et al., 2019).
In contrast, the bromine record was strikingly similar to the well-known record of acid deposition to Greenland resulting from anthropogenic emissions of sulfur-and nitrate-acid (NO X ) species associated primarily with the combustion of coal, leaded gasoline and diesel. Through a systematic investigation, Maselli and coauthors identified that sulphate had a negligible covariation with bromine concentrations, whereas nitrate species covaried strongly with bromine (Fig. 19). Other potential bromine sources, such as coal and 1,2-dibromoethane (DBE, used as a scavenging agent to remove lead from engine combustion chambers), were found to be of minor importance. The increased concentrations of bromine in recent Greenland snow strata were thereby attributed to increasing acidity at the sea-ice surface and in Greenland snowpack, enhancing bromine mobility and promoting transport of bromine toward central Greenland. While these results call into question the reliability of using bromine to reconstruct sea-ice since the 1950's, they also confirm the good agreement between bromine and MSA in the two centuries preceding the wide-scale acidification of the Arctic. The reported Summit/TUNU bromine pattern (low values in the 1950's, leading to a 1970's peak and subsequent decline) was similar to that observed in Severnaya Zemlya, even though bromine concentrations in Severnaya Zemlya were ten times greater than those reported for Greenland. The application of this data to sea-ice reconstructions in the North Atlantic Ocean and Nordic Seas is discussed further in section 5.2.3.1.

Antarctica
Only one record of bromine and iodine, covering millennial timescales, has been reported from the TALDICE ice core recovered from Talos Dome in northern Victoria Land (Spolaor et al., 2013b) and is shown in Fig. 20. The TALDICE ice core was drilled between 2005 and 2007, covering the past 220 kyr of climate and covering the last two full glacial-interglacial cycles (Buiron et al., 2011). Concentrations of bromide (Br À ), iodide (I À ) and iodate (IO 3 À ) were determined using a coupled HPLC-IC-ICPMS technique, which split the elements into their major species using HPLC-IC and then detected them using ICPMS. Similar to the concerns raised in section 3.6.4, the reported TALDICE Br concentrations were conspicuously low compared to total Br concentrations later reported by ICP-SFMS for comparable sites in Victoria Land (e.g., Maffezzoli et al., 2017). Although the reported concentration of bromine may be incorrect e and these samples should be remeasured as a matter of priority e it is likely that the qualitative pattern of Br fi% (and Br enr ) over time is correct even though the quantitative values may be incorrect. This study presented the first evidence of long-term (multimillennial) stability of halogens species in polar ice. The highest bromide concentrations corresponded to interglacial periods 0e12 kyr, 120e130 kyr and 210e220 kyr ago. Bromine/sodium ratios were found to be enriched during interglacials and depleted during the glacials. A rather unwieldly bromine fractionation metric was used, bromine fractionation index (Br fi% ), although this is shown in section 2.2.2 to be equivalent to the more commonly-used Br enr metric. Both bromine metrics are shown in Fig. 20. Regarding the iodine species, iodide was found throughout the record with lower values during interglacials and higher values during the glacials e a pattern matching that of sodium e while IO 3 À was only present during the coldest climate periods with high dust fluxes (17e34 kyr and 134e143 kyr ago). The findings reported by Spolaor et al. (2013b) and their applicability to Antarctic sea-ice Fig. 18. Annual record of bromine (thin blue) and MSA (thin red). Annual record of bromine (thick blue) and MSA (thick red) with outlying spikes removed using a 25-year running average filter described by Sigl et al. (2013). All records were fit with a three-step linear regression (black) and the results of the fits which identify the timing of inflection points are summarized in Maselli et al. (2017, Table S1). The time series have been plotted to match the signal variability in the pre-industrial era (1750e1850 CE). Reprinted from . (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.) reconstructions are discussed further in section 5.1.1.

Arctic
From the Arctic, two millennial-scale bromine records have been reported from ice cores: NEEM (Spolaor et al., 2016b) in northwest Greenland and RECAP (Maffezzoli et al., 2019) in central eastern Greenland.
The NEEM ice core (drilled within the Northwest Greenland Eemian Ice Coring project) covers the past 130 ky and contains the first complete Eemian interglacial climate record from a Greenland ice core (NEEM Community members, 2013). The location of NEEM is favourable to aerosol transport from the west and southwest, with atmospheric transport models indicating North America as a primary source of terrestrial aerosols (Schüpbach et al., 2018) and the Canadian Arctic, including the Baffin and Hudson Bays, as a primary source of marine aerosols (Barrie et al., 1992). Samples for halogen analysis were obtained discretely from the CFA system used to determine chemical impurities, with each sample integrating a 1.1 m length of ice (Schüpbach et al., 2018). The NEEM brittle zone was unsuitable for CFA analysis, thus approximately 400 m of ice was not measured, corresponding to the period 3e9 ky b2k. The bromine record from NEEM was measured at University Ca' Foscari of Venice by ICP-SFMS, with particularly dense sampling from the present back to 50 ky b2k (Spolaor et al., 2016b).
Bromine concentrations in the NEEM ice core generally varied in a similar manner to those of sodium, but with relatively more bromine during interglacials (Fig. 21). It should be noted that in this study the authors assumed that sodium was purely of marine origin without any crustal contribution: nssNa ¼ 0. Bromine concentrations ranged from Holocene values of 0.45 ± 0.15 (1s) ppb to glacial values of 1.45 ± 0.07 (1s) ppb. The sampling density was sufficient to observe bromine variability through the deglaciation and glacial interstadials (GIs) including the Younger Dryas/Bølling-Allerød and GIs 3e13. Bromine enrichment values varied from approximately 16 during the early Holocene to approximately 2.5 during the glacial stadials. A strong positive correlation was observed between oxygen isotope temperature proxy (d 18 O) and Br enr (r 2 ¼ 0.68), providing a basis for linking sea-ice extent to Br enr and Greenland temperature (Section 5.2.2).
The bromine record from RECAP (Maffezzoli et al., 2019) is shown in Fig. 22. RECAP bromine concentrations are low in the Holocene 0.3 ± 0.1 (1s) ppb, higher in the glacial 0.7 ± 0.3 (1s) ppb and slightly higher than Holocene during the late Eemian 0.4 ± 0.2 (1s) ppb. The RECAP record sodium crustal contribution was deeply investigated and was found to be negligible during the Holocene and up to 20e30 % the glacial period. Such sodium crustal contribution was subtracted to the total sodium concentrations to calculate the bromine enrichment values. Br enr values are highly variable, with Holocene values ranging from 2 to 6. Glacial values of Br enr vary from 2 to 10, with values weakly correlating with broad variations in Greenland temperature, and late-Eemian average 5. The sampling resolution of the bromine record (integrating 55 cm depth in the Holocene and 18 cm in the glacial) provided annual to centennial resolution during the Holocene, but centennialmillennial resolution in the glacial. This was a result of the extreme glaciological thinning of the Renland glacial section e 100 thousand years of climate was archived in less than 50 m of ice. The interpretation of RECAP Br enr as a sea-ice proxy is discussed in section 5.2.3.2.

Antarctica
Only two halogen-based sea-ice reconstructions (SIRs) have been reported from Antarctica, both from East Antarctica. The first, from Talos Dome in Northern Victoria Land, covers the last two glacial cycles (Spolaor et al., 2013b). The second, from Law Dome, covers the past century . One other SIR is available for comparison, that being an MSA record reported for Law Dome  covering the period 1841e1995 CE. For the 220 ky Talos Dome reconstruction, the EPILOG compilation Fig. 19. Comparison between the measured total sulfur (shown as sulphate) and acidity records from each ice core (top panels). The acidity record is dominated by the influence of the sulfur species until the early 21st century when the NOx pollution remains elevated whilst anthropogenic sulfur sources are depleted resulting in a slight relative elevation of the total acidity relative to total sulfur concentrations. The large spikes in the acidity and sulfur records are identified as volcanic events. The ice core records cover the period of the 1783 Laki eruption as well as the unknown 1909 eruption and Tambora eruption (Indonesia) in 1815 (Sigl et al., 2013). Comparison between Br/MSA and total acidity (centre panels) and nitrate (NO 3 À , bottom panels) measured in the ice cores. The Br/MSA ratio follows the total acidity record closely except where the record is dominated by the sulfur component (e.g. early 1900s). Of the two major acidic species the Br/MSA follows the nitrate most closely at both ice core sites. Reprinted from . P. Vallelonga, N. Maffezzoli, A. Saiz-Lopez et al. Quaternary Science Reviews 269 (2021) 107133 of marine-sediment-based SIRs provides an essential data set for comparison (Gersonde et al., 2005).

Victoria Land
The Talos Dome SIR (Spolaor et al., 2013b) was based on bromine fractionation relative to sodium (Br fi% , see sections 2.2.2 and 4.4.1 and Fig. 20) and employed a novel but poorly constrained transport model to evaluate glacial/interglacial movement of the marginal sea-ice zone (see section 3.6). The synthesis of bromine and sodium transport mechanisms presented in section 3.7 will be applied here to critically re-evaluate the findings of the Talos Dome SIR.
In summary, Spolaor et al. (2013b) reported low Br enr values (or positive Br fi% ) during the glacial periods from 16 to 80 ky BP and 134e205 ky BP and interpreted this as indicative of increased MYSI. The few data points corresponding to the last interglacial featured higher Br enr values (or negative Br fi% ) compared to the Holocene, interpreted as a sign of reduced FYSI extent during the last interglacial. The findings were comparable in timing and variability to an SIR based on fossil diatom assemblages in marine sediments . In particular, the sea ice record indicated relatively little sea-ice expansion during the period 80e125 ky BP, but a rapid increase in sea-ice through the later part of the glacial, 16e80 ky BP. Perhaps the most salient point regarding the interpretation applied by Spolaor and coworkers, in contrast to later studies, was that there was no attempt to characterise the area of sea-ice expansion, only the radial distance of the extension. Such an approach is apposite for East Antarctica, where there are no barriers to sea-ice expansion. This approach is not possible in the Arctic, where sea-ice expansion in the numerous ocean basins is substantially constrained by continental barriers.
Following critical reappraisal of the analysis technique (section 4.4.1) and interpretive model (section 3.6) reported in the original work, it is not possible to draw any definite conclusions from the dataset in its current form. Primarily, the samples will need to be remeasured with a more sensitive and reliable analytical method, such as ICP-SFMS. Once a quantitatively reliable dataset is available, it will be possible to determine the extent to which variations in Br enr reported at the site are linked to variability in sea ice extent, as opposed to other factors such as transport distance/efficiency or surface remobilisation/re-emisson effects.

Wilkes land
Law Dome is a coastal ice rise approximately 1200 m in thickness which features regular annual snow accumulation of 0.7 m/yr ice equivalent at the summit (Roberts et al., 2015), making it an ideal location for investigation of the past marine environment with sub-annual resolution. With respect to SIRs, the Wilkes coast is also favourable because there is almost no MYSI, ensuring that any detected sea-ice signal may be attributed to FYSI only. Curran et al. (2003) reported the first SIR for the region, covering the period 1841e1995 CE, based on the then-novel sea-ice proxy MSA. MSA results from the oxidation of DMS, which is produced by phytoplankton and in the sub-Antarctic Ocean algal communities hosted by sea-ice dominated the production of DMS. Hence, a link may be established between FYSI, which hosts DMS-producing algae, and the MSA deposited at Law Dome when the sea-ice melts and DMS is released to the local marine atmosphere. Several MSA records have subsequently been reported from Antarctica, with each site requiring a critical evaluation of the link between MSA and local/regional sea-ice conditions . Curran et al. (2003) employed the available satellite observation record (1974e1995 CE) as a calibration dataset, finding a correlation of r ¼ 0.6 (p < 0.002) between Law Dome MSA concentrations (annual mean) and the northernmost latitude of FYSI extent in the 80 E-140 E sector during the preceding winter (AugusteOctober). Applying this calibration to the full MSA record, a 20 % decline in sea-ice extent since 1950 CE was predicted, with no discernible trend for the preceding period of 1841e1950 CE. High decadal variability was observed, with an 11-year cycle discernible from 1950 to 1995 CE. The source of such variability was not established, although cyclicity of similar duration was also proposed following sea-ice observations by satellite (10 years, Zwally et al., 2002) and recordings of fast-ice extent (9 years, Murphy et al., 1995).
The availability of an existing multi-centennial SIR at Law Dome makes it an ideal location for evaluating new proxies for sea-ice extent, even though some level of discrepancy is to be expected when comparing proxies. Extending a preliminary dataset covering the period 1910e1914 CE (Spolaor et al., 2014), Vallelonga et al. (2017) reported a SIR from Law Dome covering the period 1927e2012 CE (Fig. 23). The SIR was a composite from two firn/ice cores analyzed using different methods in two laboratories (see section 4.2.1), with bromine reported for the period 1927e2012 CE and iodine for 1927e1989 CE. Hence, the satellite-calibration periods were different for bromine (35 years) and iodine (12 years). A significant correlation was found between annually-averaged (JanuaryeDecember) ln(Br enr ) and FYSI area in the 90 E-110 E  . Talos Dome d 18 O is also shown. The red areas indicate Br À enrichment (i.e. negative Br fi% ) while blue areas indicate bromide depletion. The figure has been modified from the original (Spolaor et al., 2013b). (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.) P. Vallelonga, N. Maffezzoli, A. Saiz-Lopez et al. Quaternary Science Reviews 269 (2021) 107133 sector, with r 2 ¼ 0.357 and p < 0.001. Such a correlation is surprisingly similar to that found for MSA (the r ¼ 0.6 coefficient of correlation quoted above equates a coefficient of determination r 2 ¼ 0.36), albeit with a slightly smaller dataset for MSA (22 years), which may suggest a common level of noise due to meteorological effects, transport variability and uncertainties associated with the satellite data. Similar to MSA, high decadal variability was observed in the I enr and Br enr records, with higher values observed in the 1940s and 1970s. Br enr shows a decreasing trend over the period 1927e2012 CE, consistent with that reported for MSA, but there are differences in the decadal patterns of the two proxies. Both records suggest a maximum of sea-ice extent over the periods 1940e1950 CE and 1975e1985 CE, although there is apparent disagreement over the relative extent during these two maxima: MSA suggests 12 % lower sea-ice extent in the latter period, whereas Br enr suggests similar sea-ice extent for both periods. It is worthy of note that Maselli et al. (2017) also observe a discrepancy between MSA and Br enr in the period 1950e2000 CE, which they attribute to the increasing influence of anthropogenic NO x , leading to enhanced emissions of bromine from the FYSI. This process is not apparent at Law Dome, where no increases in NO x or nssSO 4 have been observed in recent decades (pers. comm., Mark Curran) with sources of anthropogenic emissions in the southern hemisphere being fewer, less active and more distant. The dataset reported by Vallelonga et al. (2017) provides a preliminary comparison of SIRs based on MSA and halogens, but for a more thorough evaluation the halogen record should be extended back to at least 1841 CE (the start of the MSA record) and ideally to the onset of the Holocene, to allow comparisons with available SIRs from high-resolution marine sediment records (Denis et al., 2010;Hodell et al., 2001;Nielsen et al., 2004)  2) which showed similar trends -but different patterns of decadal variability -over the 50-year record. This preliminary study, in which only the upper 29 m of the 724 m core was presented, employed the HYSPLIT model to calculate seasonal airmass back trajectories for the period 1976e2000 CE and therefore identify FYSI areas over which sea-salt aerosols were arriving. The model identified the Laptev sea as the dominant and most consistent airmass source in springtime, although for some periods the Kara sea was an important summertime airmass source. Therefore Spolaor et al. (2016a) investigated correlations between Br and I in the ice core to satellite observations of FYSI extent in both the Kara and Laptev seas. They found Laptev sea FYSI area was significantly correlated with both Br exc (r ¼ 0.44, p < 0.02) and iodine concentrations (r ¼ 0.50, p < 0.009); and Laptev sea summer FYSI area also correlated with iodine concentrations (r ¼ 0.49, p < 0.011). Such correlations were consistent with the current understanding of reactive halogen chemistry as well as satellite observations of halogens in the Arctic: spring is the peak season for bromine explosions e heterogeneous recycling of bromine above FYSI e while summer coincides with the peak extent of iodine-emitting algal blooms. As the algal community is hosted below FYSI, the extent of springtime FYSI may be related to the size of the summertime algal population.
The halogen record was also compared to an Arctic SIR by Polyakov et al. (2003), for which SIE in both the Kara and Laptev seas was resolved. Significant correlations are found between Laptev sea summer sea-ice anomalies and Br exc (r ¼ 0.31, p < 0.009) as well as I (r ¼ 0.32, p < 0.002). Although inconsistent with airmass back trajectory reconstructions, significant correlation was also determined between Br exc and reconstructed summer SIE anomalies for the Kara sea (r ¼ 0.34, p < 0.015). The relatively brief P. Vallelonga, N. Maffezzoli, A. Saiz-Lopez et al. Quaternary Science Reviews 269 (2021) 107133 timeseries presented by Spolaor et al. (2016a) is insufficient to critically evaluate the suitability of either bromine or iodine for quantitative SIR for Laptev Sea but offers a basis for further study. An important test case would be during the Medieval warm period, approximately 800e1000 CE, when warm climate conditions occurred through much of the northern hemisphere. During this period, it would be expected that a quantitative sea-ice proxy would show a noticeable reduction in SIE.

Canadian Arctic
From measurements of the northwest Greenland NEEM ice core, Spolaor et al. (2016b) reported the first multi-millennial halogenbased SIR for the Arctic (Fig. 24). The halogen record covered the past 120 ky but with low sampling resolution for the period prior to 50 ky ago. The sampling was sufficient to establish estimates of sea-ice extent at important climate extrema: the late Holocene (last 3 ky), the Holocene Climatic Optimum (HCO, 6e9 ky ago), and the Last Glacial Maximum (LGM, 20e30 ky ago). Note that with respect to the HCO, samples from the period 9e11.5 ky ago were used for comparison, as poor ice quality prohibited sampling of ice corresponding to the period 3e9 ky ago. Based on aerosol source evaluations, the reconstruction was applied to the marine sector to the west and northwest of NEEM: the Canadian Arctic, Baffin Bay and Hudson Bay. The source of aerosols to NEEM was evaluated from both atmospheric modeling work and satellite observations of bromine explosion events. Based on the results of a 1-D chemical transport model run, greater values of Br enr at NEEM were interpreted as a greater extent of FYSI. In contrast, the model indicated that both MYSI and Greenland surface snow were zones of deposition of Br. FYSI is considered the predominantly gas phase bromine source (Bougoudis et al., 2020;Simpson et al., 2007a) mainly because the recycling over MYSI occurs on salts deposited becoming halide depleted in much shorter time periods compared FYSI. In FYSI the halide content is in excess with respect to the gaseous halogen recycling, able to sustain the bromine recycling process longer. Bromine recycling over MYSI could occur but is limited compared to the recycling occurring over FYSI. Ice core observations suggest significant correlation only with FYSI and not when considering MYSI extension supporting the hypothesis that FYSI extension is the main driver for gas phase bromine in the polar spring. Satellite measurements also support that bromine explosions are less frequent over the central Arctic where the multiyear sea-ice is mainly present.
The hypothesis proposed by Spolaor et al. (2016b) to reconstruct sea-ice extent from the NEEM Br enr record was based on the model results showing FYSI being a source of bromine species, as also supported by several experimental campaigns and satellite sensing over the Arctic. In contrast, the same model suggests that both MYSI and Greenland surface snow are zones of deposition of bromine. We point out, however, that the FYSI-Br enr apportionment made by Spolaor et al. (2013b) in the Antarctic TALDICE record was originally based on different fractionations of bromine and sodium, hence on a transport mechanism, leading to increased bromine enrichment further inland (Fig. 20), rather than on a source effect. The differences in the interpretation of sea-ice extent in the two studies are based on the differing geographical settings between the Canadian Arctic and East Antarctica: Arctic sea-ice is strongly constrained by continents (even more so considering the lowered sea level during the glacial) and therefore the transport effect was assumed of secondary importance, Antarctic sea-ice grows radially with no continental constraints. Although the overall glacial/ interglacial variability is comparable between TALDICE and NEEM Br enr records, the absolute values likely reflect the different geographical settings of the two sites and their respective bromine source regions. Well-constrained chemical transport models are required to disentangle the influences of sea-ice source area changes and transport fractionation effects, if accurate interpretations are to be applied to ice core records of Bromine enrichment.
Following the assumptions made in Spolaor et al. (2016b) of higher Br enr indicating greater FYSI surfaces, the NEEM Br enr SIR indicates extensive FYSI in the Canadian Arctic during the Holocene and glacial interstadials, interspersed with periods of extensive MYSI cover during the LGM and glacial stadials. The data suggested reduced FYSI during the HCO, with respect to the last 3 ky, and also that MYSI during the Younger Dryas was less extensive than the preceding LGM period. For the first time, an ice core-based SIR indicated repeated and extensive sea-ice changes over stadial/ interstadial cycles. Although the NEEM results relate to the Canadian Arctic, not the Nordic Seas, the finding of repeated and rapid  (Spolaor et al., 2016b)); (g) RECAP Br enr series (Br enr,Cl , green and Br enr,Mg , black) calculated using chlorine and magnesium respectively. Reprinted from Maffezzoli et al. (2019). (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.) stadial/interstadial sea-ice change supports studies arguing that sea-ice is a key component to understanding the mechanisms of rapid climate change in the Arctic (Dokken et al., 2013). Despite the relatively low sampling resolution, there was also some indication that FYSI extent was less during the "cooler, shorter" glacial interstadials (GI-3, -4 and -6) in comparison to the "warmer, longer" interstadials (GI-8 and -12). Spolaor et al. (2016b) also provided a quantitative estimate of past sea-ice extent by combining satellite observations of sea-ice variability and physical constraints on glacial sea-ice extent imposed by lowered sea levels and the Laurentide ice sheet. The reconstruction was based on some key assumptions: 1. That the linear relationship observed between temperature proxy d 18 O and ln(Br enr ) may be similarly applied to FYSI extent and ln(Br enr ). 2. That the range of variability in FYSI area observed by satellites over the period 1979e2013 CE may be taken as representative of the range of FYSI area incorporated in the early Holocene (0e3 ky b2k) Br enr data. 3. That there was complete MYSI coverage (that is, no FYSI) in the Canadian Arctic during the LGM.
Considering these assumptions, perhaps the most critical is the second one e that the 30-year satellite data set may be used to represent 3000 years of climate variability e including such events as the little ice age and medieval climatic optimum. Considering the extreme reduction of Arctic MYSI (and corresponding increase in FYSI) in recent decades, the satellite data set may be biased toward greater values of FYSI. Kinnard et al. (2011) evaluated the Arctic MYSI reduction of the past decades as being anomalous with respect to the last 2000 years, which suggests that in the context of assumption 2 the high end of satellite FYSI observations should be treated with caution. In comparison, assumption 1 is reasonably tentative e a linear model is the simplest possible and the logarithmic scaling of Br enr reduces the relative effect of extremely high Br enr values. Of the three assumptions, the third is the most reliable: the exposure of continental shelves due to lowered sea levels at the LGM, and the presence of the Laurentide ice sheet, would have massively reduced the presence of FYSI in the Canadian Arctic during the late glacial, even before considering the influence of ice shelves on MYSI extent.
Even though the NEEM halogen SIR was the longest and most detailed available at the time, many questions remain unanswered. The NEEM halogen SIR implies that Arctic summer sea-ice extent was approximately 30 % lower in the Canadian sector of the Arctic Ocean during the HCO, relative to present values. But the last interglacial, the Eemian, was not sampled even though this period corresponds to approximately 6e10 m higher sea level and 5 C warmer global temperatures (NEEM Community members, 2013). Br enr values corresponding to the Eemian would have provided an extremely valuable context for evaluating the two Holocene time slices presented. In the context of a growing anticipation that Arctic summer sea-ice will disappear within the next decades, it would be extremely valuable to determine the upper limit of corresponding Br enr values for the NEEM site.
5.3.1. North Atlantic Ocean/Nordic Seas 5.3.1.1. Decadal-centennial time scales. Only in a few cases have multiple sea-ice proxies been directly compared within the same sample set or location. We discussed in section 5.1.2 the Antarctic Law Dome halogen record  and here we discuss the bromine and MSA records reported from Greenland Summit and TUNU ice cores . The two Greenland ice cores were measured on the same analytical system, with both bromine and MSA determined continuously. Bromine was determined by ICP-SFMS using the same system as for the Severnaya Zemlya halogens record (Spolaor et al., 2016a). MSA was determined using a novel electrospray triple-quadrupole mass spectrometer (ESI-MS-MS) technique, with discrete samples collected for verification of the accuracy of the continuous method. The two ice cores covered the period 1750e2010 CE, with the Summit 2010 core analyzed with seasonal resolution. The low snow accumulation rate at TUNU prohibited sub-annual data resolution. The Hysplit4 atmospheric circulation model was used to determine airmass back trajectories for the two ice core sites. On this basis, the representative origin of marine aerosols arriving to Summit core site was determined to be southeast Greenland, comprising Denmark Strait and the southern Nordic seas. The origin of marine aerosol to TUNU site was similarly evaluated as west Greenland, essentially Baffin Bay. In this manner, two complementary sea-ice records were produced, covering different ocean sectors. Correlations were also determined: MSA and Br enr were compared to the HadSST1 ICE dataset (1900e2012 CE) which incorporates satellite observations for the period 1979e2012 CE and uses this information to correct for potential bias in the pre-satellite data, primarily composed of ship-based observations. Both sea-ice concentration (SIC) and Open Water in Ice Pack (OWIP) were considered, where OWIP is the difference between sea-ice area and sea-ice extent.
Both of these metrics indicate the absence of sea-ice within the zone of study, but they do not distinguish between MYSI and FYSI.
The sea-ice reconstruction by Maselli et al. (2017) proposes gradually declining sea-ice over the period 1750e1900 CE, followed by a more pronounced decline in the late 20th century (Fig. 18). The reconstruction is also shown in Fig. 25, with other relevant metrics for Arctic sea-ice. The step-change model identified an initial "high" state for Br and MSA at Summit, from approximately 1750 to 1830 CE, after which a second, lower state occurred from 1879 to 1931 CE. From 1931 CE, the two Summit datasets diverge, with Br increasing and MSA decreasing. For TUNU, the initial state is from 1750 to 1830 CE, the second state from 1840 to 1960 CE, and the divergence in Br and MSA is not synchronous (Br rises from 1944 CE, MSA decreases from 1984 CE). Overall, the 1854e1878 CE step change in Summit Br is the most clearly identified by the statistical model, as is the increase of Br after 1932 CE. In TUNU, the clearest step change is the decrease in MSA after 1984 CE. These changes are interpreted as indicative of a gradual decrease in sea-ice extent in the Nordic seas commensurate with the end of the little ice age, followed by a more rapid decrease in the late 20th century: earlier in southern Greenland, later in northern Greenland.
As described in section 4.3.2, different trends are reported for MSA and Br variability in the Summit2010 and TUNU cores . Potential explanations are presented for these divergent behaviours, although a detailed modelling study is likely required to distinguish between the various possible sources of influence. Increased bromine concentrations observed in the cores after 1950 were attributed to the influence of anthropogenic emissions of acids, particularly nitric and sulfuric acid, causing seaice acidity to increase (see section 5.2.3.1). In contrast, the increasing emission of bromine may also be due to the steady decline in Arctic MYSI and its replacement by FYSI (Fig. 24). A recent study also calls into question the reliability of the Greenland MSA record as a proxy for sea-ice extent: Osman et al. (2019) extract a common trend from MSA records in 12 Greenland ice coresincluding those reported by Maselli et al. (2017) e and attribute this trend to changes in North Atlantic Ocean primary productivity over the recent centuries. The Greenland composite MSA record is correlated with North Atlantic Ocean sea surface temperature and Atlantic meridional overturning circulation, which play a strong influence on the sustenance of primary producers in the North Atlantic Ocean.
Ultimately, to better understand the behaviour of sodium, MSA and bromine e the leading proxies for sea-ice extent e a combination of ice core records and chemical transport models will be required. The availability of a more extensive array of ice cores, from locations other than central Greenland, will provide a richer dataset for validating the behaviour of sodium, MSA and bromine over recent centuries. Comprehensive chemical transport modelling should then be applied to such a dataset to accurately apportion the relative influences of sea-ice and oceanic circulation on the ice core MSA record, as well as the relative influences of sea-ice extent and anthropogenic acid emissions, on the ice core Br record.

Millennial time scales.
For the North Atlantic Ocean, the only halogen-based sea-ice reconstruction available is from Maffezzoli et al. (2019), reporting a 120 ky record of Br enr from the RECAP ice core recovered from the Renland ice cap located in coastal East Greenland. The RECAP ice core was drilled in 2015, updating and improving the climate record produced from an earlier Renland ice core drilled in 1989 (Hansson, 1994). The RECAP ice core is 584 m long, with a complete Holocene profile within the upper 530 m and a complete glacial profile within 30 m of ice. The extreme thinning of the glacial section limits the resolution of Br enr variability over stadial/interstadial timescales, although the Younger Dryas/Bølling-Allerød transition is clearly defined in the data (Fig. 22). Similar to the technique employed by , the sources of marine airmasses arriving to the Renland ice cap were identified using the Hysplit4 model. In this case, the authors studied the origins of 72 h-long back trajectories arriving at Renland every hour for the period 2000e2016 CE. In order to isolate the airmasses likely transporting marine aerosols, only the airmasses that were located within the MBL (defined as the 900 hPa isosurface, approximately 1000 m asl) for longer than 10 h were considered. From this analysis, 75 % of the marine trajectory endpoint distribution contours a portion of the North Atlantic Ocean extending from ca. 50 N to 85 N, thus including the Greenland Sea and the Denmark Strait region, the Icelandic Sea and the Norwegian Seas. Hence the RECAP Br enr record was interpreted as primarily representing changing marine conditions within the Nordic seas as north as the Fram Strait, although the ocean regions closer to Renland are expected to carry a more significant marine contribution.
The RECAP Br enr record shows low values during the Holocene and coldest periods of the glacial (MIS 2 and MIS 4), with intermediate/high values during the deglaciation and during the warmest periods of the last glacial period, MIS 3 and late MIS 5 (Fig. 22). Contrasting with the NEEM results, where the highest Br enr values were found during the Holocene and the lowest Br enr values during the cold glacial stadials, the RECAP Br enr record features low values during both the Holocene and cold stadials. This asymmetry was interpreted as a dual behavior of the Br enr indicator at Renland. Specifically, since reduced bromine explosions occur in the absence of seasonal sea-ice, a low Br enr value can indicate either open water conditions or multi-year sea-ice. Therefore, the authors suggested that RECAP low Br enr values during the Holocene indicated open water conditions, while the low values during cold glacial periods were indicative of extended multi-year sea-ice in the Nordic Seas. Their results are qualitatively consistent with the  (Bonnet et al., 2010), Arctic summer (Kinnard et al., 2011) and first-year (Stroeve et al., 2012a) sea-ice extents, ice core data from TUNU and Summit2010  and NGRIP (Legrand et al., 1997b) Greenland ice cores. P. Vallelonga, N. Maffezzoli, A. Saiz-Lopez et al. Quaternary Science Reviews 269 (2021) 107133 only existing sea-ice record extracted from a sediment core in the Norwegian Sea and covering the last 90,000 years (Hoff et al., 2016).

Conclusions
We have described the chemical and physical basis for the application of halogens, particularly bromine, in polar snow and ice as a proxy for sea-ice reconstruction. While the state of understanding of polar halogen chemistry has advanced to a sufficient stage that heterogeneous halogen recycling may be modeled effectively, there remain gaps in the understanding and accurate description of mixed-phase transport and deposition of bromine over the polar ice sheets. Retention of halogens in snowpack is an important consideration to the reliability of any proxy records: it has been consistently shown that iodine is poorly retained in the snowpack of high-accumulation sites, whereas bromine is retained reliably in sites with moderate levels of snow accumulation rates (greater than approximately 20 cm ice equivalent per year). For the purposes of reconstructing sea-ice extent over multiple glacial/ interglacial cycles, the reliability of bromine records from central Antarctica is still to be established.
Regarding the application of halogen proxy records to sea-ice reconstructions, caution must be applied in the interpretation of bromine enrichment records. The different deposition velocities of sodium and bromine, and particularly the mixed-phase nature of bromine transport, require a careful effort to accurately translate ice core bromine records to variability in sea-ice parameters. We demonstrate a method of: 1) collecting available spatial distributions of bromine and sodium concentrations, 2) developing appropriate transport models for describing the disparate transport of halogens and sea-salts, and 3) applying these models to distinguish between the competing influences of transport distance, seaice source area, and MYSI/FYSI variability. Published halogen-based sea-ice reconstructions for the northern and southern polar regions have been presented and critically assessed. Where possible, halogen-based SIRs have been compared to other available sea-ice proxy records. The halogen elements, particularly bromine, offer great potential for unlocking a wealth of knowledge regarding seaice variability during the late Quaternary period. It is vital that the available ice core records are accurately measured and appropriate chemical transport modeling is applied to the resulting records, to unlock the full potential of these valuable and rapidly diminishing climate treasures.

Recommendations
With regard the future development of halogens in ice cores as a quantitative proxy for sea-ice extent, we submit the following recommendations for future work: Surface sampling of snow should be conducted during Arctic and Antarctic traverses, with the aim of increasing the spatial coverage of Na and Br data presented in Figs. 13 and 14. This data will improve and inform models of halogen transport and retention in polar snowpack. Where possible, the sampling should be conducted along traverse routes where inland ice core drilling sites are supplied from coastal stations. Bromine data should be consistently measured by ICP-SFMS techniques, with a NH 4 OH washing step included between samples. Caution should be applied to Br records produced by a combination of Continuous Flow Analysis sampling and online ICPMS determination. Where possible, discrete samples should be collected and measured by ICP-SFMS. Where possible, previously reported studies employing Ion Chromatography should be repeated using ICP-SFMS. Interlaboratory comparisons will be of benefit in standardizing analytical setups and avoiding bias due to instrumentation or sampling techniques. A more complete understanding is required of halogen emission from FYSI and potentially MYSI. While it is known that MYSI is depleted in halides compared to FYSI, it is still necessary to characterise the extent to which MYSI may participate in heterogeneous recycling of bromine and otherwise augment or retard the enrichment of boundary-layer bromine with respect to sodium. An intercomparison project should be applied to the available chemical transport models incorporating halogen chemistry modules. There are currently a few halogen chemical transport models and these should be benchmarked for consistency. Particular attention should be placed on the deposition velocities used and the characterisation of mixed-phase transport over long distances. Ideally, the intercomparison project would include glacial and interglacial transport scenarios. Where possible, bromine, sodium and MSA should be consistently measured in ice core records, particularly those located 300e600 km from the present coastline. This is the zone offering elevated Br enr as well as being relatively close to the coast, with moderate snow accumulation and potential for multicentennial record duration. The record of multi-millennial Br records in Antarctica should be expanded from Talos Dome to other suitable ice core sites, such as WAIS Divide in West Antarctica, EDML in Dronning Maud Land, DSS/Law Dome in East Antarctica, and potentially the SPICE ice core from the South Pole. For ice cores located on the Antarctic plateau with very low snow accumulation, such as EPICA Dome C, Vostok, Dome Fuji and Dome A, preliminary sampling should be conducted on strata containing interglacial ice and glacial maxima ice, to determine representative glacial and interglacial values of bromine and iodine and investigate the potential for semi-quantitative evaluations of past sea-ice extent. More extensive ice core records of sodium and bromine are required across the Arctic. Two zones of immediate interest are 1) the available small ice caps circling the Arctic Sea, including northern Greenland, northeast Canada, Alaska, Siberia and Svalbard; and 2) Central and Southern Greenland. The circum-Arctic records should elucidate a record of Arctic sea-ice variability over the last millennia and possibly as far back as the Holocene Climatic Optimum approximately 8 ky ago. The records from Central and Southern Greenland should extend well into the last glacial period, providing a detailed record of sea-ice variability and timing related to Atlantic Meridional Ocean Circulation and other important mid-latitudinal climate forcings. An indication of Arctic sea-ice extent during the Eemian, the last interglacial, may possibly be obtained by sampling the ice strata close to the base of Greenland deep ice drilling sites including DYE-3, GISP2, GRIP, NGRIP and NEEM. A compilation of Br enr values from the Eemian sections of these ice cores would provide an important timeslice of Br enr variability across Greenland, relevant to future warming Arctic climate scenarios.
The consistent application of these recommendations will contribute to a successful and thorough development of the full potential for halogen-based proxy reconstruction of past sea-ice changes. The foundation for this development is ongoing discussion, communication and collaboration within the ice core and associated paleoclimate research communities.
Appendix A. Sample data for surface snow collected along scientific traverse from NEEM ice core drill site to EastGRIP ice core drill site