The 87 m “Summit-2010” ice core was collected in 2010 close to Summit Station, Greenland (72∘20′ N 38∘17′24′′ W; Fig. 1). The average snow accumulation at Summit, as determined from the ice core record, is ∼ 0.22 m year-1 water equivalent, with few instances of melt. Due to the relatively high snow accumulation rate, seasonal analysis of the sea salt species concentrations was feasible. The 213 m Tunu core was collected in 2013 (78∘2′5.5′′ N, 33∘52′48′′ W, Fig. 1), approximately 3 km east of the Tunu-N automatic weather station, part of the Greenland Climate Network. The average snow accumulation at Tunu, as determined from the ice core record, is ∼ 0.11 m year-1 water equivalent.

The Summit-2010 and Tunu cores were first dated using volcanic horizons in sulfur (S) from well-dated historic eruptions (e.g. 1815, 1835, 1846, 1854, 1873, 1883, 1912). To aid the volcanic horizon assignment in the Tunu core the non-sea-salt S record was synchronized to the NEEM-2011-S1 volcanic record (Sigl et al., 2015). The dating of both cores was then refined by annual layer counting using a combination of seasonal cycles in Na, Ca, and the ratio of non-sea-salt S / Na for each entire core. In addition, hydrogen peroxide was used as a winter marker in the upper section of the Summit-2010 core. January was defined as the minimum value in the ratio of non-sea-salt S / Na. Bigler et al. (2002) demonstrated that the input of anthropogenic sulfate during the Arctic haze slightly shifted the seasonal cycle of sulfate preserved in a north-eastern Greenland ice core ice toward spring by, on average, 1 month. The ratio of the Na to S was thus used to define the formal annual layer boundaries because using the ratio of two tracers with opposing seasonality (winter Na maximum and summer S maximum) reduces the sensitivity of the series to any small temporal fluctuations. Comparison with weekly surface snow samples collected from Summit (from 2007–2013; GEOSummit project) confirms the assignment of a non-sea-salt S / Na minimum in December/January. Monthly values were calculated assuming a constant distribution of snowfall within each year. Due to the lower accumulation rate and strong katabatic winds at the Tunu site, constraints from volcanic synchronization played a more important role in developing the depth-age scale for the Tunu core compared with Summit-2010. This dating technique is described in more detail for another Greenland ice core (NEEM-2011-S1) by Sigl et al. (2013, 2015). The annual layer dating for these ice cores resulted in a plutonium record that is consistent with other ice cores from Greenland between 1950 and 1970 and with the emission histories from nuclear weapon testing in the Northern Hemisphere (Arienzo et al., 2016). The error in the dating of the ice core records was estimated as ±0.33 years for the Summit-2010 record and ±1 year for the Tunu record.

The ice cores were sampled from 33×33 mm cross-section sticks using a continuous melter system (McConnell et al., 2002). The silicon carbide melter plate provides three streams from concentric square regions of the ice core sample: an innermost stream (with a cross sectional area of 144 mm2), an intermediate stream (340 mm2), and an outer stream that was discarded along with any contaminants obtained from handling of the ice core. The innermost melt stream was directed to two inductively coupled plasma-mass spectrometers (ICP–MS, Thermo Scientific Element 2 high-resolution with PFA-ST concentric Teflon nebulizer (electrospray ionization – ESI)) run in parallel. All calibrations and runtime standards were run on both instruments and several elements were also measured in duplicate (Na, Ce, Pb) to ensure tracking between both ICP-MSs. In addition, an internal standard of yttrium flowed through the entire analytical system and was used to observe any change in system sensitivity. The instrument measuring bromine was run at medium resolution and there were no mass interferences observed at the bromine isotope mass monitored (79 amu). The sample stream was acidified to 1 % HNO3 to prevent loss of less soluble species, degassed just prior to analysis to minimize mixing in the sample line and sampled at a rate of 0.45 mL min-1 (McConnell et al., 2002; Sigl et al., 2013). The following elements were measured by ICP–MS: Br, Cl, Na, Ca, S, Ce, and Pb. Calibration of the ICP–MS was based on a series of seven mixed standards measured at the start and end of each day for all elements except for the halides. Due to the high volatility of acid halides, a set of four bromine and chlorine standards were made individually in a 1 % ultra high purity HNO3 matrix from fresh, non-acidified intermediate stock solution (Inorganic Ventures) every day. The intermediate melt stream was directed to a continuous flow analysis (CFA) system on which nitrate ion (NO3-) and snow acidity (sum of soluble acidic species) were measured using the technique described by Pasteris et al. (2012) in addition to other atmospheric species of interest (Röthlisberger et al., 2000). Stable water isotope records were also collected using the CFA system according to the method described by Maselli et al. (2013).

The analysis of MSA by batch analysis using electrospray triple-quad mass spectrometer (ESI–MS–MS) has been reported previously (Saltzman et al., 2006). A portion of the debubbled CFA melt stream (150 µL min-1) was subsampled for continuous online analysis of methanesulfonate by ESI–MS–MS (Thermo-Finnigan Quantum). This subsample was mixed with pure methanol (50 µL min-1) delivered using an M6 pump (syringe-free liquid handling pump, Valco Instruments Company Inc.). The methanol was spiked with an internal standard of deuterated MSA (CD3SO3-; Cambridge Isotopes) at a concentration of 52 nM. The internal isotope standard was used to correct for any changes in instrument response due to variations in water chemistry (such as acidity). The isotope standard was calibrated against non-deuterated MSA standards prepared in water from non-deuterated MSA (CH3SO3-; Sigma Aldrich). MSA was detected in negative ion mode using the CH3SO3- / SO3- transition (m/z 95/80) and CD3SO3- / SO3- (m/z 98/80). The concentration of MSA in the sample flow was determined from the ratio of the non-deuterated and deuterated signals after minor blank corrections. This study is the first use of the technique for ice core MSA analysis in a continuous online mode. The uncertainty in the MSA intensity as calculated from the standard calibrations is 1 %.

A second portion of the debubbled CFA melt stream was directed to an autosampler collection system to collect a discretely sampled archive of the melted ice cores. The collected samples were frozen at the end of each day and later analysed for MSA again using ion chromatography and ESI–MS–MS.

The Pb derived from anthropogenic sources (exPb) was calculated as the difference between total lead measure in the ice core, [Pb]obs, and that from dust sources. The Pb from dust was calculated as a fraction of the dust proxy cerium, ([Ce]obs) as follows: exPb=[Pb]obs-Ceobs×PbCedust, where the relative amount of Pb in dust, ([Pb] / [Ce])dust, has the constant mass ratio of 0.20588 (Bowen, 1979).

Similarly the amount of non-sea-salt sulfur (nssS) was calculated relative to the sea salt sodium (ssNa): nssS=[S]obs-[ssNa]×SO42-Naseawater, where the amount of sulfur relative to Na in sea water, ([SO42-] / [Na])seawater has the constant mass ratio of 0.252 (Millero, 1974). The ssNa was calculated by comparison with calcium as both have sea salt and dust origins (Röthlisberger et al., 2002): ssNa=Naobs×Rt-CaobsRt-Rm, where Rt and Rm are the Ca / Na mean crustal and mean marine mass ratios of 1.78 and 0.038, respectively (Millero, 1974).

Bromine enrichment factors relative to sea water concentrations were calculated using the following: enrBr(Na)=BrNaobs/BrNaseawater, where the ([Br] / [Na])seawater mass ratio is 0.00623 (Millero, 1974).

To identify the likely sea ice source regions of MSA and Br deposited at the ice core sites, we perform 10-day air mass back trajectories of boundary layer air masses from each ice core site using the GDAS1 archive dataset in the Hysplit4 software (Draxler and Hess, 1998). The starting height of the back trajectories was 500 m to ensure that the monitored air masses travelled close enough to the surface at the ice core site to potentially deposit aerosols. The vertical velocity field was taken from the meteorological data files. Air mass back trajectories were started every 12 h and allowed to travel for 10 days (total number of trajectories hours = 14 400 h per month). The number of hours that the trajectories spent in a 2∘ × 2∘ grid was summed over all of the trajectories for that month between the years 2005 and 2013. Previous work showed that the rapid advection of MBL air was the likely source of reactive halogens at Summit (Sjostedt et al., 2007).

In order to assess the relationships between sea ice conditions and ice core chemistry, correlation maps were generated between annual MSA concentrations and monthly sea ice using the HadISST1 ICE dataset at 1∘ latitude–longitude monthly resolution (Rayner, 2003). Pre-1979 sea ice datasets were interpolated from sea ice extent maps compiled by Walsh (1978) which incorporate a variety of empirical observations. The data were later bias-corrected using modern satellite data (Rayner, 2003). Correlations were performed separately for the satellite period (1979–2012) and for the extended record (1900–2012), excluding the period 1940–1952 when the record has no variability due to scarcity of data (Rayner, 2003). Because strong DMS emissions occur in marginal sea ice zones (Sharma et al., 2012), we considered both sea ice concentration (SIC) and the area of open water in the sea ice pack (OWIP) which represents the size of the marginal sea ice zone. OWIP is defined as the difference between sea ice area (calculated from sea ice concentration over the area of the grid cell) and sea ice extent (NSIDC). A SIC of 15 % was used as the threshold for a grid cell to contribute to sea ice extent. The area of OWIP was calculated within the coastal areas as defined by the results of the air mass back trajectories (Sect. 3.4).

Outliers were removed from the MSA time series (see Fig. 2) before the correlations were performed. The outliers were removed using the technique described by Sigl et al. (2013) for identifying volcanic signals using a 25-year running average filter. Correlations were performed on an annual rather than seasonal basis because the seasonality of ice core MSA is distorted due to post-depositional migration of MSA signal at depth in the snowpack (Mulvaney et al., 1992) (Fig. 3 in the article and Fig. S1 in the Supplement).

Ice core measurements of bromine at Summit and Tunu covering the period 1750–2010 are shown in Fig. 2. Ice core Br levels at each site were stable until ∼ 1820 at Summit and ∼ 1840 at Tunu when they both decreased by ∼ 1 nM, establishing a new baseline that was stable until the mid-1900s. Both ice cores also show a Br peak in the late 20th century. The concentration values and the timing of inflections in concentrations were determined by a three-step linear regression of the dataset. The analysis was performed by simultaneous linear least squares fitting of three straight lines joined by “inflection points” to the dataset. The variables of the fitting procedure were the slopes and intercepts of each line as well as the x axis locations at which the total function switched from one linear section to the next (the inflection points). Initial guess values were supplied for each variable to help the fitting procedure reach reasonable values. A summary of the regression results can be found in Table S1 in the Supplement.

Sea salt transport onto the Greenland ice sheet occurs predominantly during winter. Historically the wintertime sea salt maximum was believed to be due to increased cyclonic activity over the open oceans (Fischer and Wagenbach, 1996) though more contemporary studies show that blowing snow from the surface of sea ice may be a significant source (Rankin et al., 2002; Xu et al., 2013; Yang et al., 2008, 2010). At Summit, a wintertime maximum is observed in the most abundant sea salts, Na and Cl (Fig. 3). Bromine also shows a significant wintertime signal, however the annual maximum appears in midsummer – at concentrations ∼ 70 % above winter levels (Fig. 3a). Comparison with Br measured in weekly surface snow samples collected from Summit (from 2007 to 2013; GEOSummit project) confirms that this summer signal is real and not a result of post-depositional modification of seasonality of the bromine signal (Fig. S2). The results from that study confirm that total Br concentrations peak in summer on the ice sheet closely following the Br cycle observed in the Summit-2010 ice core. In addition to the comparison with the GEOSummit data, in the ice cores studied here there are routinely more than 10 measurements made within a yearly layer of snow, giving confidence to the allocation of a summer maximum in bromine at Summit. Analysis of the annual cycle of bromine in the Tunu ice core also shows a summer maximum when averaged over the entire ice core time series but with significantly larger error than observed at Summit. The timing of this peak suggests a predominant summertime deposition of bromine that dwarfs that from winter sea salt sources.

The shape of the annual bromine cycle does change slightly over the course of the Summit record (see Fig. 3). Starting in the early 1900s, the annual bromine cycle slowly becomes broader. A slight shift in the maximum from a solely summer peak in the pre-industrial era towards a broad spring–summer peak by 1970 is observed (Fig. 3 lower panel). Comparison with the sea salt tracer, sodium, which does not undergo the large temporal shift and broadening of its seasonal cycle shows that this change in bromine seasonality is not linked to changes in production or transport of sea salt aerosols or even dating uncertainties in the ice core but perhaps the introduction of an additional, smaller bromine source in the springtime during the industrial era.

Both ice cores show a predominantly positive Br enrichment throughout the year (Figs. S3, S4) relative to both sea salt elements chlorine and sodium. This enrichment reaches a maximum in mid-to-late summer at Summit (Fig. 3). We assume that this enrichment reflects Br enrichment in the aerosol transporting Br to the ice sheet. In a comprehensive review of global aerosol Br measurements, Sander et al. (2003) concluded that in general, aerosols which showed positive Br enrichment factors were of sub-micrometre size. These small aerosols can travel further (lifetimes of around 5–10 days) and due to their larger surface / volume ratio may experience more atmospheric processing than larger aerosols, resulting in the positive enrichment. However, post-depositional reduction of the bromine concentration is a possibility during the summer months due to photolytic processes at the snow surface. This may be the cause of the noisiness of the bromine signal within the lower accumulation Tunu core. However, the increased snow accumulation that occurs during the summer months in both central and northern Greenland (Chen et al., 1997) should act to minimize these bromine-depleting effects driven by increased insolation in summer, and indeed Weller (2004) has shown that accumulation rates of this size are large enough to prevent the post-deposition loss of other species such as nitrate and MSA.

Both sites also show a (small) positive enrichment of chlorine relative to sodium, which is amplified at small sodium concentrations. Chlorine-containing aerosols are expected to undergo similar chemical processing to bromine-containing aerosols but the enrichment factors of bromine (relative to sodium) are much larger, which is likely due to the high solubility of bromine species such as HBr (Sander et al., 2003). Alternatively, the chlorine enrichment could be interpreted as a sodium depletion of the aerosols particularly in those of small diameter where both concentrations are low; this would amplify the bromine enrichment (relative to sodium) but would not explain the bromine enrichment relative to chlorine. It is likely that both halogens undergo some degree of enrichment and the sodium undergoes some depletion in the aerosols though it is difficult to determine this from the data.

A summertime maximum in Br enrichment was also observed by Spolaor et al. (2014) in short segments of Antarctic Law Dome ice core as well as two Arctic ice cores. Spolaor et al. believe that the main source of the inorganic bromine originated from springtime bromine explosion events above sea ice and the summertime maximum could possibly be an indication of lag time between bromine-containing particles becoming airborne and their deposition. Further investigation is needed to definitively establish the seasonality of bromine deposition at the poles. However, the results of the Arctic ice cores studied here suggest that the summer maximum in bromine deposition is indeed real.

In the Tunu ice core, 11 % of the monthly bromine enrichment measurements relative to Na were negative (less than the Br / Na seawater ratio, Fig. S3) and 12 % were negative relative to Cl. It is possible that the negative enrichment values observed in the Tunu ice core are therefore a result of larger aerosols (> micrometre) reaching the site due to its proximity to the coast (and thus the likely sea ice aerosol source region) in comparison to Summit.

The Summit-2010 MSA record (Fig. 2) replicates that measured by Legrand in 1993 (Legrand et al., 1997) and extends it an additional 17 years (see Fig. S5). The mean Summit-2010 MSA measurements over the period 1984–1992 (2.0 ± 0.7 (1σ) ppb) also compare well with the results of the sub-annually sampled Summit snow pit study performed by Jaffrezo et al. (1994); 2.1 ± 1.8(1σ) ppb. Both the Legrand and Jaffrezo studies measured MSA using ion chromatography of discretely sampled snow and ice. The similarity between the Summit-2010 measurements and the results of these studies demonstrates that the new, continuous technique is able to achieve a comparable accuracy in MSA measured concentrations to the traditional, discrete technique. It also demonstrates that negligible amounts of MSA are being lost by using the continuous melt method.

The Tunu measurements represent the first MSA profile at this location. Replicate measurements of the entire Tunu ice core were performed with the online, continuous technique by melting a secondary stick of ice cut from the original Tunu ice core. The replicate measurements closely followed the original MSA measurements demonstrating the reproducibility, stability, and high precision of the continuous MSA technique (Fig. S6). The Tunu MSA record was also reproduced using discrete samples collected from the CFA system (Fig. S7).

At Summit, MSA concentrations averaged 48 nM in the late 18th century, compared with just 27 nM at Tunu. From 1878 to 1930 MSA concentrations at Summit plateaued at 36 nM after which they began to drop rapidly, at a rate of 0.27 nM year-1, reaching 18 nM by 2000 CE. Large fluctuations in the MSA record after this time make it difficult to assess the most recent trend in Summit MSA concentrations. MSA concentrations in the Tunu core showed a similar temporal variability to those in the Summit record, and until the mid-20th century, were consistently lower in magnitude. MSA concentrations only began to decline consistently at Tunu after 1984, almost 50 years after the rapid decline observed in the Summit record. After 2000 CE, large fluctuations in concentration were again observed, making the modern-day trend in MSA concentration at Tunu difficult to establish.

Comparison with the total sulfur record (Fig. 4) reveals that during the pre-industrial period, MSA contributes to ∼ 12 and ∼ 7 % of the total sulfur signal at Summit and Tunu, respectively, compared with < 2 % at the height of the industrial period (1970 CE) at both sites.

The low-frequency, pre-industrial trend in MSA concentrations seen in these ice core records closely follows that of bromine; particularly distinct is the decrease in both MSA and bromine at both sites in the early to mid-1800s (Tables S1 and S2). In the 1900s, however, both sites show a divergence between the MSA and Br records – as MSA begins to decline, Br concentrations increase.

A dramatic shift in the “timing” of the annual MSA maximum in the Summit-2010 ice core is illustrated in Figs. 3c and S1. The signal shifts gradually and continuously along the length of the entire Summit-2010 record from a spring to winter maximum (Fig. S1). This phenomenon has previously been observed in several Antarctic ice cores and has been attributed to post-depositional migration within the ice due to salt gradients (Mulvaney et al., 1992; Weller, 2004). At very low-accumulation ice core sites, post-depositional loss of MSA (and nitrate) must also be considered. Extrapolation of data collected by Weller (2004) from a series of East Antarctic ice cores predicts that sites with annual average accumulations of greater than 105 kg m-1 year-1 (0.105 m year-1) will not show post-depositional loss of MSA (or nitrate). Both ice cores in this study have sufficient average annual accumulation that post-depositional loss of MSA (and nitrate) is predicted to be negligible and so is not discussed further.

In winter, with the collapse of the polar vortex, polluted air masses enter the Arctic region as the phenomenon known as the Arctic haze (Barrie et al., 1981; Li and Barrie, 1993). SO2 and NOx from the haze are adsorbed onto aerosols or deposited directly on the ice/snow and oxidised to sulfuric (H2SO4) and nitric acid (HNO3). There are also natural sources of SO2 (biomass burning, volcanic eruptions, oceans – Li and Barrie, 1993; McConnell et al., 2007; Sigl et al., 2013) and NOx (microbial activity in soils, biomass burning, lightning discharges – Vestreng et al., 2009) as well as other snow/ice acidifiers including MSA, hydrogen chloride, and organic acids released from biogenic or biomass-burning sources (Pasteris et al., 2012).

The annual cycle for nitrate (NO3-) is shown in Fig. 3d. Before 1900 CE the nitrate shows a seasonal maximum in late summer/early autumn after which the maximum shifts to late spring/early summer. Although there are biological sources of NO3- in the ice core aerosol source regions, in a recent study focused on the NO3- and δ15N-NO3- record in the Summit-2010 ice core, Chellman et al. (2016) concluded that the pre-industrial (1790–1812 CE) NO3- seasonal cycle was driven by biomass-burning emissions. However, in the modern era (1930–2002 CE) oil-burning emissions became the dominant source of NO3- in the snowpack. The change in the dominant NO3- source due to industrialization is the cause of the shift in timing of the seasonal cycle.

Total snow acidity was stable at both sites from 1750 through to ∼ 1900 CE, except for sporadic, short-lived spikes due to volcanic eruptions. The average pre-industrial acidity was the same at both sites (∼ 1.8 µM). Both records also show two distinct maxima in acidity centred on 1920 and 1970 CE (Fig. 4), with Tunu displaying higher acidity than Summit over the entire industrial period. Overlaid with the acidity is the total S record for both ice cores. The high correlation between the acidity and S records illustrates that the sulfur species are the dominant natural and anthropogenic acidic species in the ice cores. The trend in acidity closely follows the global SO2 emissions with maxima from coal (∼ 1920 CE) and coal plus petroleum combustion (∼ 1970 CE) (Smith et al., 2011). After 1970 the records of acidity and S deviate. This deviation can be attributed to the presence of nitric acid that remains at a relatively high concentration in the late 20th century whilst sulfur species reduce in concentration (Fig. 4).

NO3- concentrations show no trend during the pre-industrial era in either ice core records, averaging 1.1 (±0.02) µM and 1.3 (±0.03) µM for Summit and Tunu, respectively. The higher signal-to-noise ratio in the Summit-2010 record reveals a small peak in NO3- concentrations centred on ∼ 1910. The Tunu record also shows elevated NO3- concentrations over this period. However, the large variability in the signal makes it difficult to establish a higher-resolution temporal trend. Both records clearly show a large increase in NO3- after 1950, peaking in ∼ 1990 and followed by a general decreasing trend with the average NO3- levels still double that of pre-industrial concentrations: 2.1 and 2.3 µM at Summit and Tunu, respectively.

The nitrate records from both sites follow the trend in Northern Hemisphere NOx emissions with a peak in ∼ 1910 and 1990 CE – a result of emissions from increases in both Northern Hemisphere fertilizer usage and biomass and fossil fuel combustion (Felix and Elliott, 2013).

Air-mass back-trajectory results demonstrate that air masses reaching the Summit-2010 site between March and July originate primarily from the south/south-east of the ice core site (Fig. 5a). Previous back trajectory analyses by Kahl et al. (1997) also linked individual spikes in their Summit MSA record to air masses that had passed over this same region of coast (SE Greenland) within the previous 1–3 days. Similar back trajectories were calculated for Summit-2010 up to heights of 500 and 10 000 m (total column trajectory, Figs. 5a, S8a) illustrating that air masses that travel in the free troposphere and lower troposphere follow similar back trajectories and likely share the same source regions.

The results for Tunu indicate that air masses arrive primarily from the west coast of Greenland, passing over the Baffin Bay area, but there is also significant contribution from both the SE and NE (in May) coastal areas (Figs. 5b, S8b). Of these two secondary areas it is likely that aerosols transported from the NE would have a greater influence on the ice core concentrations due to proximity to the ice core site. Aerosol deposited at Tunu therefore represents a mixture of source regions, but are likely dominated by the NW Greenland Baffin Bay coastal region.

Locations which showed a SIC variability greater than 10 % (the average estimated range of uncertainty in the satellite measurements) and have a significant correlation to MSA (t test, p < 0.05) are displayed in Figs. S9 and S10 for the months of March–July. A greater emphasis must be placed on the post-1979 sea ice concentration maps as these were derived from passive microwave satellite data and, where available, operational ice chart data. The likely air mass source regions, as defined by the results of the air mass back trajectories, are indicated by the black bordered regions. Within these areas there is generally a negative correlation between SIC and MSA, particularly in the spring months and only small patches that show large correlation (> 0.4). The large areas of positive correlation along the east coast and in the western Barents Sea are striking for the Summit-2010 record, however, these areas are outside of the defined air mass source region and thus are unlikely to be contributing to the ice core aerosol records. The positive correlation is likely an artefact of the negative autocorrelation between sea ice conditions in this region and the SE coast source region (Fig. S11).

The effect of the estimated error in dating of the MSA records on the SIC correlation maps is explored in Fig. S12. By shifting the dating of the MSA records to either extreme of the dating error estimate and replotting the SIC correlation panels, it is clear the error in the dating of the MSA records does not affect the sign of the correlations displayed on the maps but can have an affect on the magnitude of the correlation found in different locations. This is likely a result of the peaks in the MSA record being shifted in or out of temporal coherence with peaks in SIC at the different locations.

Over the period 1900–2010 CE, highly significant correlation (t test, p < 0.001) is found between the annual ice core MSA and the amount of OWIP (representing the area of the marginal sea ice zone, Figs. 6a and 7a lower panels) in these aerosol source areas. For both ice cores the source region OWIP trend is followed by the MSA. In the Summit-2010 ice core the highest correlation between annual MSA and monthly OWIP occurs in May (r=0.58, p < 0.001) though the following months through to July all show highly significant correlations (July r=0.53, p < 0.001). For comparison, the May SIC correlation map is also shown as the upper panel in Fig. 6a. Figures 3f and S13 demonstrate that this time period (May–July) corresponds to the peak and then rapid decline in the amount of annual OWIP within the Summit-2010 aerosol source area because of the decreasing extent of sea ice. Rapid loss of sea ice reveals areas of biological activity previously capped by the ice, allowing surface–atmosphere exchange of DMS, resulting in the seasonal peak in atmospheric MSA correlation with the peak in the area of OWIP.

At Tunu, the highest correlation over the 1900–2012 CE period is found between annual MSA and annual OWIP (r=0.59, p < 0.001), though the July OWIP shows the highest monthly correlation and is also highly significant (r=0.41, p < 0.002). For comparison, the July SIC correlation map is also shown as the upper panel in Fig. 7a. Due to the more northerly location of the Tunu aerosol source region, the sea ice pack in this region is generally less fractured and break-up occurs later in the year, with a sharp peak in OWIP occurring in July (Fig. S13). The higher stability of the ice pack throughout the year compared to that in the Summit-2010 source region is the likely reason the Tunu MSA shows highest correlation with the annual average of the OWIP. However, like Summit-2010 the highest monthly OWIP correlation occurs between the annual MSA and the timing of the maximum in annual OWIP (July).

Over the shorter, satellite era (1979–2012 CE) again Tunu shows strongest correlation between annual MSA and annual OWIP though at a much lower significance (r=0.32, p < 0.05), and the highest monthly correlation occurs in March (r=0.2, p < 0.1) albeit with low significance. The significance of the Tunu correlation over this period can be dramatically increased (annual OWIP r=0.54, p < 0.001; March OWIP r=0.63, p < 0.001) if the closer, secondary aerosol source region (NE Greenland, 80–73∘ N, 20–0∘ W) is assumed to also influence the site in equal proportion. March corresponds to the timing of increased insolation and thus the rapid increase in ice algal production (Leu et al., 2015). The shift from a July to March peak in the correlation of OWIP with annual Tunu MSA may be a result of the reduced overall sea ice extent (SIE) (and thus OWIP), influencing the timing of MSA production. Unfortunately, the post-depositional migration of the MSA signal within the ice cores masks any evidence of true seasonal MSA shifts. Summit-2010 also shows a much less significant monthly OWIP correlation with the annual MSA signal over this time period, with the most significant correlation again occurring in March (r=0.4, p < 0.02). The greater significance of both the SIC–MSA and OWIP–MSA correlations at both sites over the longer time period is likely a result of the averaging of any MSA production or transport variability as well as the dominance of the low-frequency variability of both time series on the overall correlation.

In an era where climate is driven by only natural forcings, chemical species that share a common source should show broadly consistent variability. This is evident in the pre-industrial section of both ice core records where the relationship between MSA and Br (monitored as Br / MSA) remains constant over the entire period (Fig. 4) despite individual records going through step function changes. Using a 25-year running average on all records, the correlation between MSA and Br over the pre-industrial period was calculated as Summit-2010: r=0.282 (p=0.0008); Tunu: r=0.298 (p=0.0004), n=138. After ∼ 1930 CE, relative increases in Br concentrations cause the Br / MSA ratio to increase above the stable pre-industrial levels by more than 160 %, reaching a peak in ∼ 2000 CE at both sites.

Bromine in excess of what is expected from a purely sea ice source (non-sea-ice bromine, nsiBr) was calculated by comparison to the other sea ice proxy, MSA. A linear regression of MSA vs. Br was performed with the pre-industrial data (1750–1880 CE) to establish the relationship between the two proxies during an era free of anthropogenic forcing (Fig. S14a, b). This relationship was then extrapolated into the period after 1880 CE in order to estimate the amount of bromine sourced only from sea ice sources during the industrial era. The MSA record was smoothed with a ninth-order polynomial function before being used in the extrapolation to reduce the noise in the resultant record whilst maintaining the low-frequency trends (Fig. S14c, d). The nsiBr is thus the difference between the total bromine measured and the calculated, natural sea ice bromine (Figs. 8 and S14e, f); in contrast to Brexc defined by Spolaor et al. (2016) as the amount of bromine in excess of the Br / Na seawater ratio.

An estimate of the nsiBr is shown in Figs. 6, 7, and 8. By definition, nsiBr is essentially constant during the pre-industrial period, but during the industrial period nsiBr peaks, reaching a broad maximum between 1980 and 2000 CE of ∼ 3.4 and 1.9 nM at Summit and Tunu, respectively.