Historic records and research have suggested that the 1783–1784 eruption of the Laki fissure volcano in Iceland impacted Northern Hemisphere climate significantly, probably as a result of the direct injection of volcanic materials into the stratosphere where the volcanic aerosols would linger for years to cause surface cooling across the Northern Hemisphere. However, recent modeling work indicates the Laki climatic impact was limited to the Northern Hemisphere and only in the second half of 1783. We measured sulfur-33 isotope excess (Δ33S) in volcanic sulfate of historical eruptions including Laki found in Summit, Greenland ice cores. No Δ33S excess is found in sulfate of apparently tropospheric eruptions, while sulfate of stratospheric eruptions is characterized by significant Δ33S excess and a positive-to-negative change in Δ33S during its gradual removal from the atmosphere. Because the same characteristics have been previously found in volcanic sulfate in Antarctica snow, the results from Greenland indicate similar global processes of stratospheric chemical conversion of SO2 to sulfate. The isotopic composition of Laki sulfate is essentially normal and shows no characteristics of sulfate produced by stratospheric photochemical reactions. This clearly indicates that the Laki plume did not reach altitudes of the stratospheric ozone layer. Further, the short aerosol residence time (<6 months) suggests that the bulk of the Laki plume and subsequent aerosols were probably confined to the middle and upper troposphere. These conclusions support the hypothesis of D'Arrigo and colleagues that the unusually cold winter of 1783–1784 was not caused by Laki.
 The 1783–1784 (June 1783 to February 1784) eruption of the Lakagígar (Laki) fissure is well-known for the devastation it caused [Thordarson and Self, 2003]. The acidic dry fog produced by the eruption and the acid deposition killed livestock and vegetation across Iceland that resulted in widespread famine and starvation. Similar, though less severe, impacts were observed in other places in the Northern Hemisphere including the British Isles, continental Europe, and beyond. The unusually cold winter of 1783–1784 was attributed by many to the climatic aftermath of the Laki eruption [Lamb, 1970; Sigurdsson, 1982; Thorarinsson, 1981], as volcanic sulfuric acid aerosols tend to cool the troposphere and the surface by reducing the surface receipt of solar radiation. Estimates of the Laki emissions are as high as 100 Tg SO2 [Metrich et al., 1991; Oman et al., 2006; Palais and Sigurdsson, 1989; Thordarson and Self, 2003]. In comparison, the great Tambora eruption of 1815, with well-documented extraordinary climatic impact [Stothers, 1984], is believed to have emitted 50 Tg SO2 [Palais and Sigurdsson, 1989].
 The Laki sulfuric acid fallout has been found in ice cores from across Greenland [Mosley-Thompson et al., 2003; Zielinski et al., 1994] and other Arctic locations [Kekonen et al., 2005; Yalcin et al., 2006]. In most ice cores, the Laki acid signal is the largest or the second largest of all volcanic eruptions in the last 1,000 years [Clausen et al., 1995; Zielinski et al., 1994]. The prominence and the pan-Arctic appearance of the Laki signal, along with the apparently significant climatic impact, has led to suggestions that the Laki eruption was large enough, despite being a fissure eruption, to have introduced volcanic ash and gases significantly above the tropopause [Chenet et al., 2005; Fiacco et al., 1994]. Volcanic eruptions that inject materials directly into the stratosphere tend to have much stronger global or hemispheric impact on climate and atmospheric chemistry than tropospheric eruptions, due to the facts that stratospheric eruptions are usually more massive, the volcanic materials are transported and distributed more widely by stratospheric circulation and persist longer due to increased atmospheric residence time, and the stratospheric chemistry is more susceptible to atmospheric perturbations [Cole-Dai, 2010; Robock, 2000].
 Despite the documented devastation and apparent influence on the meteorological conditions during 1783–1784, some evidence suggests that the climatic impact of the Laki eruption was far less than that of other large eruptions such as Tambora, because of the high latitude location of Laki and of the possibility that the Laki aerosols were confined to the troposphere and washed out quickly after the eruption ceased. Recently, D'Arrigo et al.  put forward the hypothesis that the unusually cold winter of 1783–1784 was probably the result of a rare combination of a negative phase of the North Atlantic Oscillation (NAO) and an El Niño warm event. The Laki contribution to the unusual climatic conditions that winter appears to be minimal or insignificant [D'Arrigo et al., 2011].
 Here the isotopic composition of Laki sulfate in Greenland ice cores is examined to determine if significant amounts of gaseous substances (i.e., SO2) of the Laki eruption entered the stratosphere and altered stratospheric chemical composition. The stratospheric impact, or the lack of such impact, may have direct implications on the extent of climatic effects of the Laki eruption. A number of other prominent volcanic eruptions in the ice cores are also investigated in comparison with Laki.
2. Ice Cores and Measurements
 Four ice cores (10 cm diameter) were drilled in a small area at Summit, Greenland (72.5°N, 38.5°W) in 2007 [Cole-Dai et al., 2009]. A 79 m ice core was rapidly analyzed along its entire length for ionic species including sulfate. The results were used to date the cores and to locate the Laki sulfate in a depth interval of about 20 cm in all cores (Figure 1). The interval was divided into 4 equal-length samples (Figure 1 and Table 1) representing the progressive deposition stages of the volcanic sulfuric acid. Samples of the same stage from all four cores were combined and the sulfate in each sample was extracted for isotopic analysis. It was determined prior to the isotope analysis that one sample (Laki Sample 3) did not contain sufficient sulfate mass, probably due to unexpected loss during the extraction process, for measurement of sulfur isotopic composition. Sulfate in the ice surrounding the volcanic signal was also extracted to serve as the non-volcanic background of tropospheric sulfate in isotope analysis. Sulfate of other prominent volcanic signals (Table 1), including those of several well-known large eruptions in the last 800 years, in the cores was similarly sampled, extracted, and analyzed.
Table 1. Sulfur Isotope Composition (in ‰) of Volcanic and Non-volcanic Sulfate in the SM07 Ice Cores
Samples of insufficient sulfate mass yielding no significant (NS) isotope data.
 The extracted sulfate/sulfuric acid was converted to SF6 gas in a chemical process, and was analyzed for sulfur isotope composition with isotope ratio mass spectrometry. The method and process of sulfur isotope composition measurement have been fully described previously [Cole-Dai et al., 2009; Rai and Thiemens, 2007; Savarino et al., 2001].
3. Results and Discussion
 The sulfur isotope (δ33S and δ34S) and the excess (Δ33S) data, along with the instrumentally determined Δ33S analytical uncertainties (2× standard deviation, or 2σ), of several of the volcanic events in the 2007 Summit cores are presented in Table 1.
3.1. Stratospheric Eruptions
 The 1815 Tambora eruption and 1259 Unknown Eruption (1259 UE) have been well-established as stratospheric eruptions.Stothers  documented the geological and atmospheric aftermath and the great human loss of the cataclysmic Tambora eruption. Tambora sulfate fallout has been found in ice cores from all across Antarctica and Greenland [Clausen et al., 1995; Cole-Dai et al., 2009]. It was first pointed out over two decades ago that the 1259 UE was an extraordinarily large stratospheric eruption by a volcano in the low latitudes because of the appearance of its fallout in both polar regions and the unusually large magnitude of the fallout signals [Langway et al., 1988; J. Emile-Geay et al., The volcanic eruption of 1258 A.D. and the subsequent ENSO event, unpublished manuscript, 2006].
 Measurement of the sulfur isotope composition [Baroni et al., 2007, 2008; Savarino et al., 2003] has shown that sulfate formed in the stratosphere from the oxidation of volcanic SO2possesses significant (i.e., non-zero) Δ33S excess, while sulfate formed in the troposphere does not. Further, Baroni et al.  demonstrated that the Δ33S of volcanic sulfate changes from positive excess to negative excess during the continuous deposition of the sulfate of a stratospheric eruption. Isotopic measurements of the sulfate of Tambora [Baroni et al., 2008] and of 1259 UE [Savarino et al., 2003] in Antarctica ice cores show that the Δ33S excess is significant in both cases.
 In this work, the Δ33S values of sulfate of the Tambora eruption, the 1259 UE, and the 1809 Unknown Eruption are significantly non-zero (Table 1). The data for the 1809 Unknown Eruption, along with other ice core evidence, have been used by Cole-Dai et al.  to establish that the eruption was a stratospheric eruption in the tropics. These isotope data from a set of Northern Hemisphere ice cores show the same pattern of shifting Δ33S values and sign change from positive to negative during the course of the deposition of the volcanic sulfate, as those of the same eruptions found in Antarctica ice cores [Baroni et al., 2008; Savarino et al., 2003]. Baroni et al.  studied the pattern of Δ33S of volcanic sulfate in Antarctica snow of stratospheric eruptions and hypothesized that the photolysis-induced oxidation of volcanic SO2 in the stratosphere occurs gradually, with the sulfate formed initially possessing positive Δ33S and that the removal of the initial sulfate from the stratosphere results in negative Δ33S in the remaining SO2 as required by mass balance. Thus, the Δ33S of sulfate deposited at the later part of the removal process would be negative. The similar Δ33S and the same sign change of stratospheric volcanic sulfate found in the Summit cores strongly indicates that the photochemical reactions and dynamic atmospheric processes proposed by Baroni et al.  in the Northern Hemisphere are the same as those in the Southern Hemisphere.
Farquhar et al.  found that Δ33S excess in sulfate is generated in the UV photolysis of SO2 and that the Δ33S excess is linearly dependent on the fractionation of the 34S isotope, with a slope that is characteristic of the wavelength of the UV light used in the photolysis. The Δ33S vs. δ34S data of the Tambora and 1259 UE appear to show (Figure 2) slopes similar to those obtained in lab experiments using a Xe lamp (<200 nm) and a KrF laser line (248 nm) [Farquhar et al., 2001]. Baroni et al.  found a similar Δ33S – δ34S slope in sulfate of stratospheric eruptions in Antarctica ice cores. Unfortunately, it is not possible to estimate the variability of the slopes of the volcanic sulfate in the Summit cores due to the fact that only two time-resolved sulfate samples were obtained for each volcanic event.
3.2. Non-stratospheric Eruption
 A volcanic event was found at 1477–1478 in the 2007 Summit ice cores. This event is also present in previously published volcanic records from Greenland ice cores [Clausen et al., 1997; Zielinski et al., 1994]. However, no volcanic signals have been found during these years in any Antarctica cores [Ferris et al., 2011]. This suggests that the erupting volcano was probably located in mid- to high northern latitudes and may be a small or moderate tropospheric eruption. The Δ33S excess (+0.05‰) of the sulfate associated with this event in the 2007 Summit cores (Table 1) is clearly within the range of the measurement uncertainty (±0.08‰), consistent with background sulfate and volcanic sulfate formed in the troposphere [Baroni et al., 2008]. Zielinski et al.  suggested that the sulfate signal in Greenland ice cores is from the 1478 eruption of Bárdarbunga in Iceland. If the sulfate signal is that of Bárdarbunga, the lack of significant Δ33S excess appears to confirm the volcanological evidence pointing to a tropospheric eruption [Briffa et al., 1998; Clausen et al., 1995].
3.3. The Laki Eruption
 Two of the Laki samples (Samples 2 and 4) show definitively no Δ33S excess (no valid isotopic data were obtained for Sample 3). The Δ33S excess (−0.13‰) in Sample 1 is only slightly outside the 2σ uncertainty range (±0.08‰). This is in contrast to the large Δ33S excess values (+0.18 to +1.60) of early sulfate of Tambora, 1809 Unknown Eruption and 1259 UE (Table 1), all stratospheric eruptions. Also apparent is the negative Δ33S in Laki Sample 1, in contrast with the positive Δ33S excess in the early sulfate of a stratospheric event.
 Clearly, the sulfur isotope composition of the Laki sulfate is inconsistent with the characteristic positive-to-negative shift in Δ33S excess established for stratospheric eruptions found in both Antarctica and Greenland ice cores. For the Laki samples, this does not hold true for all four data points. This strongly suggests that no significant fraction of the large amount of Laki SO2 was directly injected into the stratosphere where photochemical reactions played a key role in its conversion to sulfate. Thordarson and Self  estimated, based on the ash dispersal pattern, that the highest altitude reached by Laki eruption columns is approximately 13 km (lower stratosphere), which is below the 18–20 km altitude of maximum ozone concentration at Laki's latitude [Brasseur and Solomon, 2005]. The shielding of the shortwave UV solar radiation by the ozone layer would prevent the Laki SO2from being photochemically oxidized to sulfate. This appears to be similar to the case of the 1991 Cerro Hudson eruption in Chile (45.9°S, 73.0°W), an upper troposphere-lower stratosphere eruption with a maximum altitude of approximately 17 km for the bulk of its 2 Tg SO2 [Schoeberl et al., 1993]. Baroni et al.  found that the Cerro Hudson sulfate in Antarctica snow did not contain Δ33S excess.
Fiacco et al. found in the GISP2 (Summit, Greenland) ice core that most sulfuric acid aerosols of the Laki eruption deposited in 1784, approximately 12 months after the Laki ash fallout on the Greenland ice sheet and concluded that a substantial portion (more than one-third) of the Laki aerosols were in the upper troposphere-lower stratosphere altitudes (9–13 km). On the other hand,Wei et al. found that nearly all Laki aerosols were removed from the atmosphere by the end of 1783 (with no lag between the deposition of Laki ash and sulfate aerosols) and suggested that the bulk (∼95%) of the Laki aerosols reached only relatively low altitudes. In addition, the very small width of the near-Gaussian distribution of the sulfate deposition in the 2007 Summit cores (Figure 1) and in other Greenland cores [Wei et al., 2008] indicates that the injection of volcanic gases into the atmosphere occurred over a such short time period, probably only in the months of June-August (therefore, subsequent eruption episodes injected little), that the resulting aerosols were rapidly removed from the atmosphere. The data of the 2007 Summit core (Figure 1) show that nearly all Laki aerosols were deposited in 1783, similar to the observation of Wei et al.  in ice cores from a number of other Greenland locations. Such a short residence time (about 6 months from the initial eruptions in June 1783) also indicates that little Laki SO2 reached the main stratospheric altitudes.
 The altitude of the bulk Laki aerosols is critical to the climatic impact of the eruption. Two scenarios have been suggested based on geological evidence [Thordarson and Self, 2003]: (1) the bulk of Laki aerosols remained entirely in the troposphere (<9 km); (2) a substantial portion (∼75%) of the volcanic aerosols was injected into the upper troposphere-lower stratosphere (9–13 km). In order to account for the prolonged and severe climatic impact of Laki, model simulations [Highwood and Stevenson, 2003; Oman et al., 2006] assumed that significant portions of the Laki SO2 were injected at relatively high altitudes (9–13 km). Our conclusion from isotope data that the Laki aerosols were entirely below the maximum ozone altitudes (18–20 km) cannot rule out either of the scenarios. However, the volcanic sulfate deposition data of ice cores from Summit (Figure 1) and other Greenland locations cores [Wei et al., 2008] indicating rapid removal (by early 1784) of Laki aerosols from the atmosphere favor the scenario of low-altitude injection used in the model simulations [Highwood and Stevenson, 2003; Oman et al., 2006]. Under such a scenario, any Laki impact on the Northern Hemisphere climate would be minimal and short-lived (i.e., limited to 1783), similar to the impact of the 1980 Mt. St. Helens eruption [Robock, 1981]. This supports the conclusion by D'Arrigo et al.  that the Laki eruption probably did not contribute to the severe winter of 1783–1784.
 Significant Δ33S excess has been found in sulfate deposited in Greenland snow by known stratospheric eruptions. The pattern of the Δ33S excess in stratospheric volcanic sulfate in Greenland ice cores has been found to be the same as that discovered in Antarctica ice cores. This indicates that the photochemical processes of the conversion of SO2 to sulfate and the subsequent gradual removal of the sulfate occur in similar fashion throughout the global stratosphere.
 The Δ33S excess of sulfate of the Laki eruption in Greenland ice core is insignificant and inconsistent with stratospheric oxidation of the Laki SO2. We therefore conclude that the Laki eruption injected no significant amount of SO2into the stratosphere above the maximum ozone altitudes. The deposition pattern of the Laki sulfate in Greenland ice cores indicates a short (<6 months) residence time for the bulk of the Laki aerosols, consistent with a tropospheric-only scenario of Laki aerosols. Consequently, no significant climatic impact may be expected from the Laki eruption beyond the initial effects and, as suggested byD'Arrigo et al. , other factors were probably responsible for the unusual winter of 1783–1784 in the Northern Hemisphere.
 Financial support was provided by NSF Office of Polar Programs via awards 0538553 and 0612461 to Jihong Cole-Dai, and 0612422 to Mark H. Thiemens. We thank the Ice Coring and Drilling Services, University of Wisconsin for field assistance in drilling the ice cores and personnel in the Stable Isotope Lab at University of California, San Diego for their assistance with the isotope measurement. The French Polar Institute, Institut Polaire Paul-émile Victor (IPEV) and Agence Nationale de la Recherche (ANR) via contract NT09-431976-VOLSOL are acknowledged for the financial support of J.S. and A.L.
 The Editor thanks Haraldur Sigurdsson and an anonymous reviewer for their assistance in evaluating this paper.