Sulfur isotopes quantify the impact of anthropogenic activities on industrial-era Arctic sulfate in a Greenland ice core

Anthropogenic sulfate aerosols are estimated to have offset 60% of greenhouse-gas-induced warming in the Arctic, a region warming four times faster than the rest of the world. However, sulfate radiative forcing estimates remain uncertain because the relative contributions from anthropogenic versus natural sources to total sulfate aerosols are unknown. Here we measure sulfur isotopes of sulfate in a Summit, Greenland ice core from 1850 to 2006 CE to quantify the contribution of anthropogenic sulfur emissions to ice core sulfate. We use a Keeling plot to determine the anthropogenic sulfur isotopic signature (δ34Santhro = +2.9 ± 0.3 ‰), and compare this result to a compilation of sulfur isotope measurements of oil and coal. Using δ34Santhro, we quantify anthropogenic sulfate concentration separated from natural sulfate. Anthropogenic sulfate concentration increases to 67 ± 7% of non-sea-salt sulfate (65.1 ± 20.2 µg kg−1) during peak anthropogenic emissions from 1960 to 1990 and decreases to 45 ± 11% of non-sea-salt sulfate (25.4 ± 12.8 µg kg−1) from 1996 to 2006. These observations provide the first long-term record of anthropogenic sulfate distinguished from natural sources (e.g. volcanoes, dimethyl sulfide), and can be used to evaluate model characterization of anthropogenic sulfate aerosol fraction and radiative forcing over the industrial era.


Introduction
Sulfate aerosols have a net cooling effect on climate by scattering incoming shortwave radiation and influencing cloud microphysical and macrophysical properties (Szopa et al 2021). Anthropogenic emissions of SO 2 , a precursor to sulfate, increased the abundance of sulfate aerosols over the industrial era, thereby imparting a significant negative radiative forcing (Shindell et al 2013). This increase in sulfate abundance is estimated to have offset some of the greenhouse-gas-induced warming over the industrial era, but the magnitude of this offset is the most uncertain aspect of climate modelling (Szopa et al 2021). A large portion of this uncertainty is caused by the unknown abundance of sulfate aerosols from natural sources because the magnitude of anthropogenic radiative forcing from aerosols is highly dependent on the preindustrial (natural) aerosol abundance (Carslaw et al 2013, Gettelman 2015. In other words, although inventories of historic anthropogenic SO 2 emissions are relatively certain compared to natural emissions (Carslaw et al 2013, McDuffie et al 2020, natural emissions of sulfate aerosol precursors are unknown and thus the fractions of total sulfate from natural vs. anthropogenic sources throughout the industrial era are uncertain (Carslaw et al 2013).
In the Arctic, which is warming four times as quickly as the rest of the world (Rantanen et al 2022), anthropogenic aerosols are estimated to have offset 60% of greenhouse-gas-induced warming over the industrial era, mostly due to sulfate aerosols resulting from midlatitude anthropogenic SO 2 emissions (Najafi et al 2015). Size-resolved sulfate aerosol measurements from Greenland ice cores confirm that anthropogenic SO 2 increased the number of small sulfate particles contributing to cloud condensation nuclei over the industrial era (Iizuka et al 2022). In the regions where emissions strongly affect Greenland and the North Atlantic sub-Arctic region, including North America and Europe, anthropogenic aerosol precursor emissions increased from the preindustrial through the mid-1970s, and then decreased following implementation of clean air policies in the source regions. Greenland ice core sulfate concentrations qualitatively align with these emissions trends (Moseid et al 2022). As a result of these trends in anthropogenic emissions of aerosol precursors, aerosol radiative cooling increased from the preindustrial into the mid-1970s and then decreased in the late 20th and early 21st centuries. The decrease in aerosol abundance contributed 1.09 • C of the 1.48 • C observed warming in the Arctic from 1976 to 2007 (Shindell and Faluvegi 2009).
However, the estimated temperature increase due to a reduction in aerosol abundance is highly dependent on unknown natural Arctic and sub-Arctic sulfate abundance, which is sourced from volcanic sulfur emissions, marine phytoplankton dimethyl sulfide (DMS) emissions, and sea-salt aerosols (Norman et al 1999, Patris et al 2002, Abbatt et al 2019, Jongebloed et al 2023. Biomass burning is not a significant source of Arctic sulfate (Breider et al 2014, Abbatt et al 2019. Sulfur isotopes can be used to determine relative contribution of anthropogenic and natural sources to total sulfate when the isotopic signatures are known and sufficiently distinct from each other (Norman et al 1999, 2000, Patris et al 2002, Wadleigh 2004, Wasiuta et al 2006, Jongebloed et al 2023. In this study we present observations of sulfur isotopes of sulfate in Greenland ice core samples and use them to quantify anthropogenic and natural sulfate over the industrial era.

Methods
In an ice core from Summit, Greenland collected in 2007, we measure sulfate concentration and sulfur isotopic composition of sulfate (δ 34 S(SO 4 2− )) in samples selected in years between 1202 and 1980 CE without influence from large volcanic eruptions (Cole-Dai et al 2013, Gautier et al 2019 and every year from 1980 to 2006 CE. Results and discussion of anthropogenic sulfate trends in sections 3.2 and 3.3 exclude the years 1976, 1980, 1982, 1986, 1991, and 1992, which are influenced by major volcanic eruptions. Years with major volcanic eruptions prior to 1980 were not measured. Detailed methods for sample analysis are described in Jongebloed et al (2023). We analyze a total of 135 samples representing 184 sampled years, including one sample per decade from 1200 to 1750 CE at 1-to 2-year resolution, one sample every 4 years from 1750 to 1980 CE at 1-year resolution, and one sample each year at 1-year resolution from 1980 to 2006 CE. The uncertainty for sample SO 4 2− concentration measurements is ±1.0 µg kg −1 and the uncertainty for δ 34 S(SO 4 2− ) measurement determined by replicate analysis of whole-process standards is ±1.2 ‰.
The isotopic composition and concentration of ice core sulfate can be used to estimate the relative contribution from each of its main sources (Norman et al 1999, Patris et al 2002, Wasiuta et al 2006, Jongebloed et al 2023, as effects of oxidation and transport on sulfur isotopic composition are minimal (Uemura et al 2016, Ishino et al 2019. The fraction of sea-salt sulfate in each sample is calculated using the mass fraction of bulk sea water SO 4 2− /Na + = 0.25 (Holland 1978), and sea salt contributes 2.1% of total ice core sulfate on average. The non-sea-salt sulfate (nssSO 4 2− ) sulfur isotopic composition (δ 34 S(nssSO 4 2− )) is calculated by subtracting out the isotopic signature of sea salt sulfate (δ 34 S ss = + 21 ‰; Rees et al 1978). In the preindustrial samples (prior to 1850), the anthropogenic contribution to sulfate concentration is assumed to be negligible, and nssSO 4 2− originates from only DMS and volcanoes. Other natural sources of sulfate are thought to be negligible in the Arctic (Abbatt et al 2019, Patris et al 2002, Jongebloed et al 2023. The isotopic signature of biogenic sulfate (δ 34 S bio ) is well constrained at +18.8 ± 0.3 ‰ and the volcanic isotopic signature (δ 34 S volc ) was previously estimated to be +4.1 ± 0.5 ‰ (Jongebloed et al 2023). Consequently, preindustrial nssSO 4 2− is the sum of biogenic and volcanic sulfate (equation (1)) and δ 34 S(nssSO 4 2− ) is a weighted average of δ 34 S bio and δ 34 S volc (equation (2)): (1) where f bio is the biogenic fraction and f volc is the volcanic fraction of total non-sea-salt sulfate.
In the post-1850 samples, we assume that nssSO 4 2− is the sum of anthropogenic, biogenic, and volcanic sulfate (equation (3)) and that δ 34 S(nssSO 4 2− ) results from a weighted average of δ 34 S anthro , δ 34 S bio , and δ 34 S volc (equation (4)): where f anthro is the fraction of anthropogenic sulfate in the sample and δ 34 S anthro is the anthropogenic sulfur isotopic source signature. We estimate the anthropogenic sulfur isotopic source signature (δ 34 S anthro ) using two methods: a Keeling plot over the post-1850 ice core samples (Keeling 1958, 1989, Pataki et al 2003 and statistical analysis of direct δ 34 S measurements of coal and oil. Results of both methods are discussed in section 3.1. The distinct isotopic composition of biogenic sulfur compared to that of other sources (Norman et al 1999, Patris et al 2002, Wasiuta et al 2006, Ghahremaninezhad et al 2016 and prior quantification of volcanic sulfate concentrations during the preindustrial era (Jongebloed et al 2023) quantify how much anthropogenic emissions perturbed sulfate aerosol abundance from 1850 to 2006 in years without major volcanic eruptions. Due to the similarity in δ 34 S anthro (+2.9 ± 0.3 ‰, see section 3.1) and δ 34 S volc (+4.1 ± 0.5 ‰), volcanic and anthropogenic sulfate cannot be separated using δ 34 S(nssSO 4 2− ) measurements. Additionally, the likely underestimate of global volcanic sulfur emissions from satellite observations (Carn et al 2015, 2017, Fischer et al 2019, Jongebloed et al 2023 preclude the use of volcanic sulfur emissions to constrain ice core volcanic sulfate concentrations over the industrial era. We therefore assume volcanic sulfate concentrations during the industrial era are equal to the median of preindustrial volcanic sulfate concentration (figure S1) and subtract the contribution of volcanic sulfur from δ 34 S(nssSO 4 2− ) based on this assumption. We also assume volcanic sulfate concentrations ±2 standard deviations around the mean preindustrial volcanic sulfate concentration (figure S1) to provide a range of anthropogenic sulfate concentrations considering the unknown volcanic sulfate contribution. It is difficult to use other measurements to examine volcanic emissions because compounds emitted by volcanoes (e.g. sulfur, carbon dioxide, trace metals) are also emitted during anthropogenic activities (e.g. fossil fuel burning and metal smelting). With these assumptions, we estimate f anthro from equations (3) and (4), and the concentration of anthropogenic sulfate is calculated in each sample (equation (5)): where [SO 4 2− ] anthro is anthropogenic sulfate concentration. The range in f anthro based on the 2.5th-97.5th percentile of volcanic sulfate concentrations in the preindustrial is also estimated in each ice core sample.

Results and discussion
3.1. Determination of anthropogenic sulfur isotopic signature Figure 1 shows a Keeling plot of the post-1850 ice core samples in years without large volcanic eruptions. The resulting isotopic source signature for anthropogenic sulfur emissions is δ 34 S anthro = +2.9 ± 0.3 ‰, where the signature is determined by the geometric mean intercept of 1/[nssSO 4 2− ] vs. δ 34 S(nssSO 4 2− ) to correct for the negative slope bias in ordinary linear regression (Angleton and Bonham 1995, Laws 1997, Pataki et al 2003 and the uncertainty is determined by the standard error of the intercept as recommended by Pataki et al (2003).
We compare δ 34 S anthro = +2.9 ± 0.3 ‰ determined by the Keeling plot in figure 1 to direct measurements of coal and oil sulfur isotopic composition from prior studies summarized in table S1 and figure S2. We do not include δ 34 S measurements of precipitation or aerosols because these samples include influence from non-anthropogenic sources (e.g. DMS). The δ 34 S measurements are organized by location (country or region), fossil fuel type (coal vs. oil), and fossil fuel source. In general, when multiple measurements represent the same fossil fuel source (e.g. δ 34 S measurements of sulfur in oil from the same well), they are combined into one 'sample.' We compile 1969 measurements of fossil fuel δ 34 S into 1012 samples, including 258 coal samples and 754 oil samples. Figure S2 shows the distribution of δ 34 S of the 1012 samples summarized in table S1. The sample δ 34 S measurements range from −35 to +33 ‰ with a mean of +3.4 ‰ and a standard deviation of ± 9.6 ‰. We assume that the δ 34 S anthro in the ice core is a mass-weighted mean of many isotopic signatures from many individual anthropogenic sources and therefore analyze the mean statistics of the compilation shown in figure 1 to estimate δ 34 S anthro . We estimate that the mean δ 34 S anthro = +3.4 ± 0.3 ‰, where the uncertainty is the standard error of the mean determined by a bootstrapping method (re-sampling the 1012 samples 1000 times with replacement and taking the mean of each re-sample (Wasserman 2004)), which is an appropriate method for determining the variance of a population that cannot be assumed to be normally distributed. Assuming the measurements are normally distributed, the standard error of the mean is also estimated to be ±0.3 ‰.
The value of δ 34 S anthro = +3.4 ± 0.3 ‰ determined by statistical analysis of the compiled direct measurements of fossil fuel δ 34 S is similar to δ 34 S anthro = +2.9 ± 0.3 ‰ from our Keeling plot in figure 1. δ 34 S anthro = +2.9 ± 0.3 ‰ and δ 34 S anthro = +3.4 ± 0.3 ‰ both align with prior estimates of δ 34 S anthro = +3 ± 3 ‰ (Nriagu et al 1991, Barrie et al 1992, Norman et al 1999, Li et al 2018. This estimate for δ 34 S anthro is similar to measured δ 34 S of aerosols in regions dominated by anthropogenic SO 2 emissions (Fan et al 2020, Nriagu et al 1991, Barrie et al 1992, Li et al 2018, suggesting that minimal fractionation due to transport and oxidation. However, as discussed in Text S1, the compilation in table S1 and figure S2 is not an unbiased representation of anthropogenic sulfur in the Summit air mass source region. Thus, we perform calculations with the signature δ 34 S anthro = +2.9 ± 0.3 ‰ determined by the Keeling plot because this method is more likely to be representative of the anthropogenic sulfur reaching the ice core. Figure 2 shows ice core measurements of nssSO 4 2− and δ 34 S(nssSO 4 2− ) in samples from 1200 to 2006 CE. Prior to 1980, years with major volcanic eruptions are not measured. After 1980, years with major volcanic eruptions are marked with blue symbols and excluded from discussion of statistics (e.g. post-1980 mean anthropogenic sulfate concentrations). The concentration of nssSO 4 2− more than triples from the preindustrial mean of 29.0 ± 7.7 µg kg −1 to a peak concentration of 114.7 ± 15.8 µg kg −1 between 1970 and 1980 CE (figure 2(b)), aligning with observations from Mayewski et al (1986). The opposite pattern is seen in δ 34 S(nssSO 4 2− ), which decreases from +9.2 ± 1.4 ‰ in the preindustrial to a low of +4.9 ± 0.6 ‰ from the 1970s through the late 1990s, and then increases to +7.3 ± 0.6 ‰ from 2002 to 2006 ( figure 2(a)). The pattern in δ 34 S(nssSO 4 2− ) reflects a shift from natural sulfate sources in the preindustrial (δ 34 S bio = +18.8 ± 0.3 ‰ and δ 34 S volc = +4.1 ± 0.5 ‰) to natural sulfate plus variable contribution of anthropogenic sulfur (δ 34 S anthro = +2.9 ± 0.3 ‰; figure 1), causing a decreasing and then increasing trend in ice core  Figure 3 shows ice core anthropogenic sulfate concentrations from 1850 to 2006 CE and anthropogenic SO 2 + SO 4 2− emissions from North America and the European Union (McDuffie et al 2020), which are the regions contributing ∼90% of anthropogenic sulfate at Summit over the past century (Moseid et al 2022). Emissions from North America and Europe increased from a mean of 8 Tg S yr −1 from 1850 to 1900 to 20 Tg yr −1 in 1930. North American and European SO 2 emissions increased to a mean of 69 Tg S yr −1 from 1960 to 1990. North American SO 2 emissions decrease after 1973, earlier than the European decrease starting in 1979. Clean air policies in North America and Europe decreased emissions to a mean of 31 Tg S yr −1 from 1995 to 2006. Anthropogenic ice core sulfate concentrations increase from a mean of 12.8 ± 8.9 µg kg −1 from 1858 to 1900 (we disregard samples from 1850 to 1854 because anthropogenic sulfate values are indistinguishable from zero) to a mean of 65.1 ± 20.2 µg kg −1 from 1962 to 1990. A decrease in anthropogenic sulfate concentration from 1930 to 1940 and subsequent renewed growth reflects a decline in fossil fuel burning during the Great Depression and renewed growth during and after World War II. As emissions decreased in North America and Europe after 1970, concentrations decreased to a mean of 25.4 ± 12.8 µg kg −1 from 1995 to 2006 (near the top of the core). Due to volcanic eruptions in the early 1980s and 1990s, the beginning of the decrease in anthropogenic sulfate concentrations is difficult to discern as aligning with the North American vs. European decrease in SO 2 emissions. Decreasing ice core anthropogenic sulfate concentrations (figure 3) align with decreasing sulfate aerosol concentration trends in sites around the Arctic (Schmale et al 2022). Anthropogenic sulfate concentrations are correlated with North American SO 2 + SO 4 2− emissions alone (r 2 = 0.60) from 1850 to 2006, European SO 2 + SO 4 2− emissions alone (r 2 = 0.68) from 1850 to 2006, and with the sum of North American and European SO 2 + SO 4 2− emissions (r 2 = 0.68). Figure 4 shows ice core f anthro from 1850 to 2006 CE. Ice core f anthro is on average 28 ± 15% from 1854 to 1900, 67 ± 7% from 1960 to 1990 (years of highest anthropogenic SO 2 emissions; McDuffie et al 2020), and 45 ± 11% from 1995 to 2006 (near the top of the ice core). The natural nssSO 4 2− fraction (f natural = f bio + f volc = 1 − f anthro ) is assumed to be 100% of nssSO 4 2− during the preindustrial, is 33 ± 7%  Figure 4 also shows that natural sulfate aerosols become increasingly important as anthropogenic emissions decline following implementation of clean air policies in North America, and natural sulfate begins to dominate over anthropogenic sulfate at the end of the ice core record (1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006).

Ice core anthropogenic sulfate concentration and fraction
The estimated f anthro in figure 4 varies by 41% on average depending on the assumed volcanic sulfate concentration, where we assume a conservative range of ±2 standard deviations around the mean preindustrial volcanic sulfate concentration. The estimated f anthro assuming the median preindustrial volcanic sulfate concentration (from here on 'medianbased f anthro ') is on average 12 ± 5% lower than the upper-end estimate of f anthro (assuming industrialera volcanic sulfate concentrations are equal to the 2.5th percentile from the preindustrial, figure S1), whereas the difference between the median-based f anthro and the lower end estimate for f anthro (assuming industrial-era volcanic sulfate concentrations are equal to the 97.5th percentile from the preindustrial, figure S1) is on average 29 ± 18%. The relatively small difference between the median-based anthropogenic concentration and the upper-end estimate is due to the positive skew of preindustrial volcanic sulfate concentrations around the mean and median. This positive skew in the preindustrial volcanic sulfate concentrations and corresponding positively skewed assumed industrial-era volcanic sulfate concentrations results in the median-based f anthro near the upper end of its possible range. In other words, this dataset provides an upper bound on f anthro .

Comparison of ice-core anthropogenic sulfate concentrations with emissions
Anthropogenic sulfate concentration trends qualitatively align with anthropogenic emissions from regions affecting Greenland, primarily North America and Europe (Moseid et al 2022), throughout the industrial era (figure 3). We note that North American and European SO 2 emissions from 1996 to 2006 are similar to emissions from 1918 to 1930 (35 Tg S yr −1 and 33 Tg S yr −1 , respectively), but anthropogenic sulfate concentration is 23.3 µg kg −1 from 1996 to 2006 compared to 51.8 µg kg −1 from 1918 to 1930. Similar discrepancies between emissions and anthropogenic sulfate concentrations occur earlier in the time series (e.g. 1930-1950; however, this period was not sampled annually), and the discrepancy between emissions and sulfate concentrations between 1996 and 2006 is not statistically significant (p = 0.013) due to the low number of samples from 1918 to 1930 (N = 5). We therefore do not discuss this discrepancy further.
A modeling study suggested that increasing East and South Asian anthropogenic SO 2 emissions began to influence Greenland in the late 20th and early 21st centuries in modelled sulfate deposition (Moseid et al 2022). However, anthropogenic ice core sulfate concentrations do not show an increase during the time period of increasing East Asian emissions (figure 3). Anthropogenic sulfate concentration decline faster than the decrease in North American and European emissions ( figure 3). However, the trends presented here represent a limited number of samples from only one ice core location and more investigation is required to quantify trends after 2006 in Summit, Greenland and the rest of the Arctic.
Due to the similar isotopic signatures of volcanic and anthropogenic sulfur, the assumption of constant volcanic sulfate concentrations based on the preindustrial distribution causes the appearance of high anthropogenic sulfate concentrations during years of anomalously high volcanic emissions. For example, in the years 1991-1993, anthropogenic sulfate concentrations are likely overestimated due to influence from the eruption of Mt. Pinatubo in June 1991, which injected about 9 Tg S into the atmosphere and increased atmospheric sulfate aerosol abundance for the following 2-3 years (Guo et al 2004). We mark years with volcanic eruptions that might cause a high bias in anthropogenic sulfate concentrations in figure 3 based on the Global Volcano Program archive of eruptions (Venzke 2013). Similarly, our ice core measurements cannot quantify the impact of volcanic eruptions on sulfate aerosol abundance during the industrial era because anthropogenic sulfate influences the quantity of total nssSO 4 2− and anthropogenic and volcanic sulfate have similar isotopic signatures.
The similarity in volcanic and anthropogenic sulfur isotopic signatures also creates a large range in estimated f anthro and anthropogenic sulfate concentration depending on the assumed volcanic sulfate concentration. However, as demonstrated in figure 4, the positive skew in the preindustrial volcanic sulfate concentrations causes our estimates of f anthro to provide an upper bound on anthropogenic contribution to ice core sulfate over the industrial era.

Conclusions
Our record of sulfur isotopic composition of sulfate in a Summit, Greenland ice core provides the first estimate of anthropogenic sulfate concentration separated from natural sulfate over the industrial era. Due to rising anthropogenic SO 2 emissions from the preindustrial to the 1970s, anthropogenic sulfate increased to 67 ± 7% of total nssSO 4 2− on average from 1962 to 1990 (excluding years with major volcanic eruptions). Following clean air policies and declining emissions in North America and Europe, anthropogenic sulfate declined in the late 20th and early 21st centuries, and anthropogenic sulfate concentration is similar to natural sulfate at the end of the record (1996-2006 mean f anthro = 45 ± 11% and f volc + f bio = 55 ± 11%; excluding years with major volcanic eruptions).
In this study, we provide the first long-term quantification of anthropogenic ice core sulfate fraction and concentration. Future combined measurements of oxygen and sulfur isotopes could provide information about changes in both the magnitude of anthropogenic sulfate and how it influences pHdependent sulfate formation chemistry (Alexander et al 2004, Hattori et al 2021. Quantifying anthropogenic sulfate fraction and abundance is critical for estimating sulfate aerosol radiative forcing. These results provide an upper-bound estimate for the anthropogenic fraction of industrial-era sulfate deposited in Summit, Greenland. This dataset can be compared to high resolution global modelling of the anthropogenic perturbation to sulfate deposition over Summit, Greenland to constrain model estimates of anthropogenic perturbation to sulfate aerosol abundance over the industrial era and implications for radiative forcing.

Data availability statement
All ice core data are available in the National Science Foundation (NSF) Arctic Data Center (https://doi.org/10.18739/A26T0GX7K) and here as Supplementary Data File 1. The coal and oil sulfur isotope measurement data compiled from previous studies are available in Supplementary Data File 2.