Introduction

The effect of volcanic emission of aerosols into the atmosphere on climate is well documented (Lamb, Reference Lamb1970; Reference Pollack, Toon, Sagan, Summers, Baldwin and van CampPollack and others, 1976; Reference Self, Rampino and BarbaraSelf and others, 1981; Reference Rampino and SelfRampino and Self, 1982, Reference Rampino and Self1984; Reference Sear, Kelly, Jones and GoodessSear and others, 1987, Reference Self and RampinoSelf and Rampino, 1988). Large amounts of volcanic debris introduced into the stratosphere between 1500 and 1900 may have played a causative role in the “Little Ice Age” of this period (Lamb, Reference Lamb1970). Major explosive volcanic events such as Tambora (1815) and Krakatau (1883) produced a consistent but small temperature decrease on a hemispheric scale for periods up to five years (Reference Self, Rampino and BarbaraSelf and others, 1981). Even smaller eruptions such as Agung (1963) produced similar temperature perturbations (Reference Self, Rampino and BarbaraSelf and others, 1981). It has been suggested that volcanic emissions of sulfur and halogen aerosols may have more effect on climatic change than that of volcanic dust (Reference Pollack, Toon, Sagan, Summers, Baldwin and van CampPollack and others, 1976; Reference Rampino and SelfRampino and Self, 1982). The presence of acid droplets in the Stratosphäre can reduce transmissivity and hence decrease surface temperatures. Thus, smaller sulfur- and halogen-rich eruptions may have more pronounced climatic effects than larger less halogen-rich eruptions (Reference Rampino and SelfRampino and Self, 1984). This may be especially true in the Northern Hemisphere (Reference Sear, Kelly, Jones and GoodessSear and others, 1987). Since the amount and chemical composition of erupted material have an important effect on regional and global climate, knowledge of past volcanic events and of the geochemical signature of the aerosol is of extreme importance. Detailed glaciochemical records provide the best milieu wherein the glaciochemistry of paleovolcanic events can be documented.

Although other types of information are available to establish the chronology and the volume of the past volcanic events (Lamb, Reference Lamb1970; Reference Simkin, Siebert, McClelland, Bridge, Newhall and LatterSimkin and others, 1981) these data sets are incomplete (Reference Self, Rampino and BarbaraSelf and others, 1981). In fact, Sedlacek and others (1983) have shown through an in-situ stratospheric sampling program that, as late as the 1970s, many volcanic eruptions that influence the chemistry of the stratosphere went unreported. Therefore, past records of volcanic events affecting stratospheric chemistry and hence the climate are undoubtedly incomplete. In addition, there have been very few eruptions from which sufficient data exist to develop quantitative estimates of the mass of materials introduced into the atmosphere (Reference Self, Rampino and BarbaraSelf and others, 1981; Devine and others, 1984).

It has long been acknowledged that volcanic emissions are an important source of several chemical species to the atmosphere. Yet it has been extremely difficult to evaluate the qualitative let alone quantitative role of volcanic emissions. Although volcanic injection of acidic anions and sulfur dioxide into the atmosphere is episodic, Sedlacek and others (1983) have shown that the volcanic contribution to the stratospheric sulfate concentration over the period 1971–81 was ∼60%. In addition, stratospheric volcanic Cl⁻ emissions may be greatly under-estimated (Johnson, Reference Johnson1980). Both Neftel and others (1985) and Barrie and others (1985), utilizing ice core data from southern Greenland and Ellesmere Island, respectively, have argued that there is a strong background acid contribution to Arctic snow during the past ∼80 years that cannot be accounted for by anthropogenic emissions. Previous work has demonstrated that volcanic events can be documented in Arctic ice cores (Table I).

Analytical Methods and Procedures

In June 1984 we obtained a ∼115 m core from site 20 D, ∼40 km southwest of Dye 3 (65.01°N, 44.87°W, 2615 m a.s.l.). The top 71m of this core has been analyzed in detail to produce an anthropogenic deposition record for

and in southern Greenland (Reference MayewskiMayewski and others, 1986). Measurements of Cl⁻, Na⁺ and δ¹⁸O concentrations were also made while selected core sections were also analyzed for and F₋. Anions were analyzed using a Dionex^TM Model 2010 ion Chromatograph with a AS-4 column. Sodium was analyzed via stabilized temperature platform furnace atomic absorption spectrometry using a Perkin Elmer^TM Model 2280 system. The chronology assigned to the δ¹⁸O record was calibrated using known bomb layers in the total β-activity record from the upper 18m of the core. In addition, cross-correlation of δ¹⁸O from a core located 4 km to the northeast of our core site produced by the University of Copenhagen group was also used to determine depth-age relations. This glaciochemical record is one of the most detailed ever obtained with 6-8 analyses per year for a continuous 115 years. Less frequent sampling would probably not allow the observations of many of these signals.

Results of Volcanic Inputs to 20 D — 1869 to 1984

Herron and others (Reference Herron, Herron and Lang1981) have shown that at Dye 3 from 1960–79 ∼8% of the annual snow accumulation had melted and refrozen. However, the work of Herron (1982) and Koide and Goldberg (Reference Koide, Goldberg, Langway, Oeschger and Dansgaard1985) indicates that Dye 3 snows are good preservers of volcanic records as well as of nuclear weapon testing records.

In order to reduce the potential effects of acid anion redistribution via melting/freezing we attempted to avoid ice lenses during core processing. However, the sampling of ice lenses could not always be avoided, so that the samples corresponding to years in which the percentage of ice lenses (determined from our visible observations) was greater than 10% have been eliminated from the volcanic assessment. These years are listed in Table II.

“Potential” volcanic events from site 20 D core were chosen by the criteria listed in Table III. In addition, time series decomposition and locally weighted scatter plot smoothing analysis (Cleveland, Reference Cleveland1979; Reference Cleveland and TerpenningCleveland and Terpenning, 1981) of the total excess

record with the historic trend removed provide a clear view of the signals

above the background (i.e. residuals). This approach allows a visible observation of these strong deviations from the historical trend and supports our more qualitative approach at picking volcanic peaks. Potential volcanic events are tabulated in Tables IV and V. The

concentrations associated with many of these potential events (Tables IV—VI) will not be discussed here. In addition to the total Cl⁻ values presented in Table IV we have also presented “excess” Cl⁻. This parameter is calculated like excess using the Na⁺ data and the Na⁺:Cl⁻ ratio of seawater to calculate the amount of Cl⁻ present in the sample that is not associated with sea salt. The volcanic Cl⁻ question is complicated due to the fact that the volcanic Cl⁻ present could be in the form of HCl or NaCl (Woods and others, 1985). There is little doubt that certain volancoes may emit more Cl⁻ than during their explosive stage (Reference Devine, Sigurdsson, Davis and SelfDevine and others, 1984).

Volcanic loading is dependent on, in addition to other variables, source strength and the duration of the eruption. Hammer (1977) has suggested that it is possible to have a one-year lag between a volcanic eruption south of 50°N and the subsequent deposition of its aerosol products in Greenland. Two- and three-year lags have been shown for near-equatorial volcanic events (Herron, Reference Herron1982). In addition to these considerations, volcanic events south of 20°S are unlikely to be observed in Greenland ice (Hammer, Reference Hammer1977; Reference HammerHammer and others, 1980).

Although volcanic emissions of Cl⁻ and F⁻ can be quantitatively important to the global atmosphere (Cadle, 1980; Reference Symonds, Rose and ReedSymonds and others, 1988), only through stratospheric emission and/or through tropospheric transport from a North American volcano is it likely that substantial amounts of

volcanic Cl⁻ and F⁻ would reach Greenland. Recent work using in-situ techniques have established volcanic stratospheric inputs of both HCl and NaCl (Reference Lazrus, Cadle, Gandrud, Greenberg, Huebert and RoseLazrus and others, 1979; Reference Woods, Chuan and RoseWoods and other, 1985). This is an important consideration in establishing volcanic Cl⁻ records in ice cores. In only two events is the excess

of the total Cl⁻ measured. These events are the 1930 and the 1941 Cl⁻ spikes. We have interpreted the 1941 spike to be the eruption of Grimsvötn, Iceland, while the 1930 event is unknown. The 1930 spike represents a massive injection of excess Cl⁻ onto the southern Greenland ice sheet. It is curious that no extremely large eruption is documented in late 1929 or early 1930 to explain this signal. In addition, the Cl⁻ peak is preceded by a very large signal in late 1929 (Table VI). There is no associated excess event. It is possible that these peaks are due to the Komaga-Take eruption in Japan in 1929. However, 1947 Hekla and 1963 Surtsey have only ∼20% of their total Cl⁻ recorded at 20 D as excess Cl⁻. This would suggest that the 1930 event could have been a nearby Icelandic eruption, for it appears that the majority of volcanic HCl reaching southern Greenland is usually neutralized during transit. It is probable that much of the stratospheric gaseous Cl⁻ is also neutralized in transit to southern Greenland.

The major Cl⁻ events observed at site 20 D include the Hekla, Grimsvötn and 1929–30 eruptions mentioned above, as well as Cotopaxi (1878), Agrigian (1917), Chikurachki (1959) and the most recent El Chichón eruptions (Mayewski and others, 1987). There is also an undocumented 1950 event.

Fluoride measurements have also been undertaken on many of these samples. Only in the two core segments was F⁻ detected as greater than 1 μg kg⁻¹. Our results indicate that in only a very few limited cases does HF contribute to the volcanic input of acid to southern Greenland during the period 1869–1984. Our maximum F⁻ concentrations are much lower than were previously reported by Herron (Reference Herron1982). Only after the eruptions of Hekla (1947) and Katmai (1912), the two most obviously reconizable volcanic events in the 115 year record (Tables IV-VI), do we measure any F⁻. The major

signals include the 1894, 1941, 1967, 1968, and 1984 events. (The 1984 has been previously interpreted as El Chichón by Mayewski and others, 1987.)

Discussion

It is clear from this work and that of Holdsworth and Peake (1985), Delmas and others (Reference Delmas, Legrand and Holdsworth1985a) and Legrand and Delmas (1987) that only through detailed anionic analyses along with precise and accurate dating can volcanic records such as these be produced from ice cores. The events with both Cl⁻ and

peaks and in some cases Cl⁻, and peaks are indeed volcanic signatures. These eruptions are listed in Table VI.

Holdsworth and Peake (Reference Holdsworth and Peake1985) have also observed the eruptions of Katla (1918–19), Raikoke (1924–25) and Hekla (1947) via detailed anion analysis from an ice core from Mt. Logan, Canada. These workers have also detected eruptions that are not clearly evident from our data. They include Mt. Wrangell (1922), St. Augustine (1934–35) and Kliuchevskoi (1937–38).

Are there any other data to support our contention that these chemical signals are volcanic events? Although not explicit in their work, data sets of both Neftel and others (1985) and Barrie and others (1985) have corroborated many of our volcanic signals. In the Neftel and others (1985) data from Dye 3, years when the averaged

data lie well above their spline-smoothed curve are shown in Table VI. For the Barrie and others (Reference Barrie, Fisher and Koerner1985) data from a core from Ellesmere Island, Canada, we have taken their conductivity data for years that show two peaks per year instead of the usual single peak per year (i.e. one of these peaks is presumed to be volcanic), as well as years with extremely high single peaks. (This was done by comparing peaks to adjoining year data.) These years are also listed in Table VII. It cannot be circumstantial that there is good agreement among the different data sets. Although all the data are not always in exact agreement (i.e. ± 1 year), this comparison of data sets supports our contentions of volcanic events

[Table VII]

being recorded in glacier ice and snow throughout the Arctic region.

Other types of information such as atmospheric transmission measurements at Table Mountain, California (the Smithsonian Astrophysical Observatory) indicate that such Southern Hemispheric eruptions as Paluweh (1928) and Quizapu (1932) were easily observed in the mid-latitudes of the Northern Hemisphere (Hoyt, Reference Hoyt1979). These eruptions can also be documented in the 20 D core (Tables IV-V).

Notably there are eruptions that we cannot document at site 20 D which one might expect to have been recorded. For example, we do not detect Agung (1963). Interestingly enough neither Delmas and others (Reference Delmas, Legrand, Aristarain and Zanolini1985b) nor Koerner and Fisher (Reference Koerner and Fisher1982), respectively, observed it at Mt. Logan and the Agassiz Ice Cap, Canada. Eighty per cent of the volcanic debris injected by the Agung eruption remained in the Southern Hemisphere (Reference Delmas, Legrand and HoldsworthDelmas and others, 1985a; Reference Self and RampinoSelf and Rampino, 1988). Self and Rampino (1988) have recently argued that the estimate by Hammer and others (Reference Hammer, Clausen and Dansgaard1980) of Agung fallout in southern Greenland is too high. Our data support this contention. Our data interpretation also agrees with their idea that some of the acid measured by Hammer and others (Reference Hammer1980) in 1963 could be related to the Surtsey eruption (see Table IV, especially excess Cl⁻ Although we observe what could possibly be a small Cl⁻ excess due to the Krakatau eruption (1883) (Table V), its significance in the 20 D record is small. Delmas and others (Reference Delmas, Legrand and Holdsworth1985a) have argued that both Agung and Krakatau should probably not be detected in Greenland ice. Our data also show little to no effect in southern Greenland of the Mt. St. Helens (1980) eruption. We observe a small Cl⁻ signal (Table IV) but no

signal. This is not surprising because this eruption was very poor in H₂SO₄ but relatively rich in halogens. As Self and Rampino (Reference Self and Rampino1988) state, “in terms of climate Mt. St. Helens was a non-event”. Our data support this premise.

Climatic Implications

Probably the most significant aspect of this work is the climatic implications. Volcanic aerosols can be major contributors to regional and global temperature changes both in the long- and short-term (Lamb, Reference Lamb1970; Reference Pollack, Toon, Sagan, Summers, Baldwin and van CampPollack and others, 1976; Reference Bradley and EnglandBradley and England, 1978; Reference Self, Rampino and BarbaraSelf and others, 1981; Rampino and Self, 1982, Reference Rampino and Self1984; Stothers, Reference Stothers1984). Eruptions such as Tambora (1815), Krakatau (1883), Santa Maria (1902), Katmai (1912), and Quizapu (1932) have produced temperature decreases on the order of 0.2–0.5°C on a hemispheric scale for time periods up to five years (Reference Self, Rampino and BarbaraSelf and others, 1981). The temperature perturbations due to volcanic aerosol emission may reach as high as 1.5°C (in earth surface temperatures) in the high-latitude zones (Reference Self, Rampino and BarbaraSelf and others, 1981).

The average yearly ΔΤ° from 1875 to 1977 in the latitudinal zone 60°–90°N tabulated by Self and others (1981) is shown in Figure 1. Their arrows indicate the eruptions of Krakatau, Santa Maria + Peleé + Soufrière, Katmai, Hekla, Bezymianny and Agung were followed by rapid decreases in ΔΤ°. Using the data presented within, not only do the Santa Maria + Peleé + Soufrière, Katmai and Hekla volcanic signals documented in southern Greenland correspond to ΔΤ drops, but at least eight other volcanically associated ΔT decreases can also be observed (Fig. 1). The ΔΤ° decrease Self and others (1981) have associated with Bezymianny might actually be associated with the Mt. Spurr eruption of 1954. Many of the decreases in ΔΤ between 1884–85 to 1902 may also be volcanically induced but our data from this period are not as convincing (i.e. only Cl⁻ or

spikes, not both or the dates of ΔΤ° drops and Greenland ice core signals are offset by a year or more). The decrease of ΔΤ° after 1960 may also be volcanically related but the anthropogenic record probably makes it. This dramatic coincidence of ΔΤ° drop above 60 °N latitude and volcanic acid input observed in southern Greenland snow and ice is probably not coincidental. It is our supposition that the volcanic eruptions documented in the ice core indicate a possible cause and effect relationship resulting in small but significant temperature decreases.

The argument is further substantiated by measured decreases in δ¹⁸O of snow after the deposition of volcanic

Fig. 1. Average yearly temperature change between 60–90°N from 1875 to 1977; from Self and others (Reference Self, Rampino and Barbara1981). Arrows present volcanic events observed in 20 D ice.

debris both at our 20 D site (Reference Mayewski, Spencer, Lyons and TwicklerMayewski and others, 1987) and at Mt. Logan (Reference Holdsworth, Krouse and PeakeHoldsworth and others, 1986). These more negative δ¹⁸O signals in the precipitation imply an immediate local to regional temperature decrease upon the “arrival” of the volcanic aerosol, such as that from the recent El Chichon eruption (Reference Mayewski, Spencer, Lyons and TwicklerMayewski and others, 1987). This “event” could have produced a 0.5 °C cooling event during the summer of 1983 in southern Greenland based on oxygen isotope-air temperature calibration (Reference DavidsonDavidson and others, 1987).

The previously undocumented 1941 event bears some detailed discussion. Although many frost tree ring events in the western USA have been associated with large volcanic events, one of the four major frost tree ring events in this century was in 1941. It has heretofore not been correlated to a known volcanic event (LaMarche and Hirschboeck, 1984). It is curious that one of the largest volcanic “events” observed by us at site 20 D was during this time period. Could this large event have gone unreported due to preoccupation with World War II, as previously suggested by Simkin and others (Reference Simkin, Siebert, McClelland, Bridge, Newhall and Latter1981)?

Handler (1986) has recently shown a strong association between stratospheric aerosols and Indian monsoon precipitation. He found that low-latitude aerosols precede below-average precipitation and high-latitude aerosols precede above-average precipitation. He showed that in 1941 there was below-average monsoon precipitation that has not been correlated to a specific volcanic event. Using his arguments and models, the volcanic event recorded by us at 20 D and in the frost tree ring records of the western USA by LaMarche and Hirschboeck (Reference LaMarche and Hirchboeck1984) was probably a low latitude (0–25 °N) volcanic eruption. In addition to the 1941 event, our previously unrecorded volcanic events can be utilized to support Handler’s (1986) contentions regarding monsoonal variations. Our 1917 and 1942 “events” (Table VI) are probably high-latitude eruptions because they coincide with above-average monsoonal precipitation (Handler, Reference Handler1986). This is not to say that all monsoonal variations can be interpreted in light of our volcanic data, but that in many cases they can be.

We feel strongly that the technique reported here and in Legrand and Delmas (Reference Legrand and Delmas1987) present the most significant means for detailed quantification of individual volcanic events in the past. In turn, this information can then be used to compare to other historic climatic data in order to assess accurately the role of volcanism on climate.

Conclusion

The results of our glaciochemical measurements conducted on firn and ice samples from southern Greenland indicate that many volcanic eruptions of Northern Hemispheric as well as of global concern are clearly recorded by elevated concentrations of Cl⁻,

and in some cases also . Although the majority of these acidic anions has been transported via the stratosphere, Delmas and others (Reference Delmas, Legrand and Holdsworth1985a) have argued that due to the closeness of many volcanic centers to Greenland the volcanically produced anions could also be transported in the troposphere. Our data show a strong coincidence of volcanic anion deposition and temperature drops within the latitude band 60–90° Ν for individual volcanic events.

More work of this type in various locations in the Arctic would provide valuable information concerning the type, strength and transport mode of volcanic eruptions. These records could then be better correlated to known climatic records and allow the development of better climatic models. This is particularly true in attempting to assess the role of relatively small but sulfur-rich eruptions (Reference Self and RampinoSelf and Rampino, 1988).

Acknowledgements

We thank J.V. James (Glacier Research Group) and T. Hinkley (USGS) and B. Koci, K. Kuivinen and S. Watson (Polar Ice Coring Office) for their companionship in the field. In addition, we are indebted to W. Dansgaard, H. Clausen, and N. Gundestrup (University of Copenhagen) for providing the oxygen isotope analyses and valuable insights. Several students in the Glacier Research Group provided assistance in the laboratory. We appreciate the thoughtful review of the original manuscript by J.M. Palais. We thank R. McGill for assistance in the statistical analysis, and A.E. Carey for her critical review of the second draft of the manuscript. This work was supported by an Environmental Protection Agency contract APP-0306-1903 administered through North Carolina State University.