Assessment of Mass, Dynamics and Environmental Effects of the Minoan Eruption of Santorini Volcano

Taking into account new evidence of fall-out in Turkey, Kos and Rhodes, a more easterly dispersal of tephra is inferred. The erupted mass from the Minoan event consists of 2 km³ DRE plinian fall (4 x 10¹² kg), 17 km³ DRE co-ignimbrite ash (3.8 x 10¹³ kg) and 20 km³ fines-depleted ignimbrite (4.2 x 10¹³ kg), or a total of 39 km³ DRE (8.4 x 10¹³ kg), i.e. three times the conservative estimate of Watkins et al. (1978), which only included the ash fall. Magmatic temperatures and oxygen fugacity calculated on the basis of Fe-Ti oxide composition are 850^o C and 10^-13 fO₂. The rhyodacite magma was probably water-undersaturated with P_total in the range 2.5 to 5 kbar, indicating a reservoir at 8 to 15 km depth. Peak column height during the plinian phase of the eruption is estimated as 36 km, corresponding to a mass eruption rate of 2.5 x 10⁸ kg/s. Petrologic studies of glass inclusions in phenocrysts of the tephra provide a basis for evaluating pre-eruption volatile concentrations in the magma reservoir and indicate a 5.5 x 10⁹ kg sulphur yield to the atmosphere, corresponding to a 1.7 x 10¹⁰ kg sulphuric acid atmospheric aerosol. This value is lower by a factor of 6 to 12 than estimates of the Minoan volcanic aerosol based on ice-core acidity layers. These results indicate problems with correlation of the Minoan eruption with either the 1390 BC or the 1645 BC ice-core acidity layers, and raise the probability of correlation of the latter with the large 3430 ± 100 ybp Aniakchak eruption in Alaska or the near-simultaneous eruptions of Mount St. Helens (Y_n eruption) and Vesuvius (Avellino eruption). The petrologic estimate indicates that sulphur yield from the Minoan event was somewhat higher than that from the Katmai 1912 and Krakatau 1883 eruptions, and an estimated northern hemisphere temperature decrease of 0.5^o C is inferred following the eruption. This is consistent with the occurrence of comparable frost-ring events in tree rings for these three eruptions.

INTRODUCTION

It is now recognized that the predominant impact of volcanism on the Earth's atmosphere relates to the effects of volcanic aerosols on the radiation budget of the Earth, whereas the atmospheric effects of volcanic ash are short-lived and minor. Studies of recent explosive eruptions have revealed that the atmospheric residence time of volcanic ash particles is short, in large measure due to the process of aggregation of micron-size silicate ash particles due to electrostatic attraction and by other cohesion forces, leading to the formation of large particle aggregates, which have a short atmospheric residence time and fall-out close to source (Carey and Sigurdsson 1982). On the other hand, volcanic gases such as SO₂ react with water in the eruption column and in the Hatmosphere, leading to a process of gas-to-particle conversion as the volcanic gas forms a stratospheric aerosol (Pollack et al. 1976). The best known volcanic aerosols from recent' eruptions, such as Agung 1963 and El Chichon 1982, are composed of sulphuric acid particles, which may have stratospheric residence times of several years. Interaction of this volcanic aerosol with the Earth's radiation budget leads to absorption and back-scattering of solar radiation in the stratospheric aerosol layer, which can lead to significant surface cooling and temporary climate change. The magnitude of the observed surface cooling events associated with several recent major volcanic eruptions correlates positively with the total mass of sulphur released (Devine et al. 1984; Palais and Sigurdsson 1989).

Several recent studies have proposed, on the basis of proxy data, that the Minoan eruption of Santorini may have led to significant global climate deterioration (LaMarche and Hirschboeck 1984; Baillie and Munro 1988). Furthermore, high-acidity layers in Greenland ice cores have been correlated with deposition of a volcanic aerosol allegedly from the Minoan eruption and inferences about the total volcanic aerosol from the eruption have been made (Hammer et al. 1980; 1987). Interpretation of such proxy data is, however, plagued by the uncertainty in temporal correlations of the observed tree-ring or glaciological events with a specific eruption. This correlation problem is even further confounded by the uncertainty in the absolute age of the Minoan eruption (Betancourt 1987), and by the fact that a major explosive eruption in the Aleutian volcanic arc has a ¹⁴C date of 3430 ± 100 ybp (Miller and Smith 1987), as well as the 3340 ± 30 ybp Avellino eruption of Vesuvius (Vogel et al. ms; Nelson et al. 1990) and the 3510 ± 230 ybp Y_n eruption of Mount St. Helens volcano (Mullineaux et al. 1975), which all have ages virtually indistinguishable from that of the Minoan event. In this paper we give new estimates on total erupted mass of magma during the Minoan eruption, based on revised isopachs of the fall-out deposit, which take into account discoveries of the tephra fall in Turkey and on other Aegean islands. The new magma mass estimate is in turn used in revising a petrologic estimate of the total Minoan volcanic aerosol mass, and in making an assessment of the atmospheric impact of the eruption.

TOTAL ERUPTED MASS

On the basis of studies of the Bronze Age Minoan tephra layer in deep-sea piston cores from the Eastern Mediterranean, the volume of fall-out from the eruption within the 5 mm isopach was initially estimated as 13 km³ dense-rock equivalent, or 3.3 x 10¹³ kg magma (Watkins et al. 1978). The marine deposit indicated an east-south-easterly dispersal axis, but a more easterly dispersal axis could not be ruled out as only the southern half of the fall-out area was available for study on the Mediterranean floor. It was recognized that the deep-sea fall deposit consists of two major fall units which correlate with the eruptive sequence of plinian and ignimbrite activity exposed on land. Thus the distal cores contain a coarser, lower unit of plinian fall, which thins rapidly from source, and an upper unit of fine-grained co-ignimbrite ash, which forms the bulk of the deep-sea ash layer. The total Minoan deposit, therefore, consists of three main components: the plinian fall, ignimbrite deposits on and around Santorini, and the co-ignimbrite fall.

Since the work of Watkins et al. (1978), the Santorini tephra fall layer has now been found on the islands of Kos (Keller 1980), Rhodes (Doumas and Papazoglou 1980) and in Turkey (Sullivan 1988). The new findings (Fig. 1) indicate that the ash dispersal was even more easterly than proposed by Watkins et al. (1978) and provide much better constraints for making a new mass estimate of erupted products. The available data for the thickness of the Minoan ash fall on land and in deep-sea cores are plotted in Fig. 2a, showing a systematic decrease as a function of distance from the volcano. We have contoured the total tephra fall deposit based on the new evidence, and calculated a new volume estimate on the basis of the relationship between tephra thickness and area within the isopachs, shown in Fig. 2b. The isopach contours are drawn at the upper limits of the available data points and thus yield the upper limit of the erupted mass. We have, for example, taken the 12 cm ash thickness reported in a Turkish lake as representative (Sullivan 1988), although it may be over-thickened due to secondary transport within the lake basin.

On a plot of isopach thickness versus the square root of the isopach area (Fig. 2b) the proximal and distal data points do not lie on a single exponentially-decaying curve. Most well-constrained plinian fall deposits exhibit at least two linear trends on such a plot. We have thus fitted the limited proximal data to a decay curve based on the average slope of 55 plinian fall deposits. The distal data define a more gently-sloping linear trend (Fig. 2b). The volume of the deposit can be determined by integration of the areas defined by the two decay profiles (e.g. Pyle 1988). Using this method the total Minoan tephra fall is now estimated as 42 km³. Adopting a deposit density of 1 g/cm³ and magma density of 2.2 g/cm³ (calculated from the chemical composition, using equations of Bottinga and Weill 1970), we calculate a total fall mass corresponding to 4.2 x 10¹³ kg or 19.1 km³ dense-rock equivalent (DRE).

Although up to 7 m thick at source, the plinian layer is a minor component of the distal ash fall. Sparks and Huang (1980) estimated that the co-ignimbrite ash represents about 90% of the total fall deposit, in agreement with the 3 to 5 km³ plinian tephra estimate of Bond and Sparks (1976). On the basis of crystal/glass partitioning and mass balance, Sparks and Huang (1980) have suggested that the co-ignimbrite ash fall represents about 46% of the magma erupted during the ignimbrite phase, and that an additional 54% must occur as fines-depleted pyroclastic flows on Santorini and in adjacent marine basins. The erupted mass from the Minoan event, therefore, consists of 2 km³ DRE plinian fall (4 x 10¹² kg), 17 km³ DRE co-ignimbrite ash (3.8 x 10¹³ kg) and 20 km³ fines-depleted ignimbrite (4.2 x 10¹³ kg), or a total of 39 km³ DRE (8.4 x 10¹³ kg), i.e. three times the conservative estimate of Watkins et al. (1978), which only included the ash fall. Pyle (1990) has independently assessed the total erupted Minoan magma volume on the basis of similar constraints used by us. His magma volume estimate ranges from 24 to 35 km³, with a 'most likely scenario' of 28 km³. The 28% difference between Pyle's and our estimate is an appropriate indication of the volume error.

INTENSIVE PARAMETERS OF THE MINOAN MAGMA

Petrologic evidence can provide information on the intensive parameters and physical properties of magma in the pre-eruption reservoir, such as total pressure, PH₂O, oxygen fugacity, viscosity, density and magma temperature, which are critical in modelling ascent and eruption of the magma. We have studied petrologic features of the Minoan tephra and determined intensive parameters and phase compositions from microprobe studies of glasses and mineral phases (Carey and Sigurdsson in prep.). The Minoan magma is dominantly a rhyodacite, with approximately 5% phenocrysts of plagioclase, hypersthene, augite, magnetite and ilmenite. Plagioclase phenocrysts show complex zoning, but compositions in equilibrium with the rhyodacite magma are in the range An_38-40. The composition of co-existing magnetite and ilmenite crystals can be used to estimate magmatic temperatures and oxygen fugacities, and indicate temperatures of 850^o C and fO₂ of 10^-13 in both plinian and pyroclastic flow deposits, indicating oxidation conditions in the magma comparable to or even more reducing than that of the NiO-Ni buffer curve. Temperatures calculated from orthopyroxene-clinopyroxene geothermometry are in the range 886 to 912^o C.

On the basis of Fe-Ti oxide geothermometry and plagioclase composition, the PH₂O of the Minoan magma prior to eruption is estimated as ranging from 1.5 to 2.5 kbar (Carey and Sigurdsson in prep.). The high-pressure phase relations of the Minoan magma are not known, but it is compositionally comparable to some experimentally studied granitic systems (Wyllie 1977). As shown in Fig. 3, the vapour-saturated liquidus for these granite magmas in the pressure range 1.5 to 2.5 kbar is 700 to 750^o C, or about 100^o C below the temperature of the Minoan magma as deduced from mineral geothermometry. This evidence indicates that the Minoan magma could not have been water-saturated in this pressure range. A higher total pressure under water-undersaturated conditions would be required to have melt in equilibrium with crystals at temperatures greater than 850^o C. The position of vapour-undersaturated liquidi in Fig. 3 suggests that the water content of the melt phase should be between 3 and 4%. The absence of quartz in the mineral assemblage further indicates that total pressure in the magma reservoir was no greater than 5 kbar and that P_total > PH₂O (Fig. 3). From detailed comparison with experimental results of Brown and Fyfe (1970) and others on closely comparable melts, Carey and Sigurdsson (in prep.) have concluded that P_total for the water-undersaturated Minoan magma was in the range 2.5 to 5 kbar, corresponding roughly to magma reservoir depth of 8 to 15 km. Microprobe analyses of melt inclusions (Table 1) suggest total volatile contents of about 5.5%, or somewhat higher than the values inferred from the phase diagrams and geothermometry results. We consider this estimate to be the maximum value of H₂O in the melt and note that the lack of hornblende in the crystallizing assemblage may favour the lower estimates based on the temperature.

ERUPTION DYNAMICS

The mass eruption rate or intensity of plinian eruptions can be inferred from the characteristics of proximal fall deposits (Carey and Sparks 1986). Unfortunately, the intensity of the ignimbrite phase cannot be directly inferred, but is likely to be equal to or greater than the peak plinian intensity. Wilson (1980a) made a preliminary estimate of plinian peak intensity for the Minoan event as 2.1 x 10⁸ kg/s and inferred a column height of 29 km. The plinian column height has now been revised to 36 km and intensity to 2.5 x 10⁸ kg/s on the basis of the maximum lithic isopleth data (Bond and Sparks 1976; Carey and Sigurdsson 1989). The deposit is reversely graded, indicating that the peak intensity occurred towards the end of the plinian phase. Mass eruption rate is likely to have been somewhat higher during the ignimbrite phase, but the Minoan eruption would have lasted about four days with this intensity.

Unfortunately, our ability to model explosive eruptions from data collected in the deposits has not yet advanced to the point where the height of the eruption cloud during the ignimbrite phase can be estimated. Recent studies show, however, that co-ignimbrite eruption clouds may be considerably higher than plinian clouds. For example, Sparks et al. (1986) have shown that the initial 27 km high eruption cloud of the 1980 Mount St. Helens eruption was associated with the blast and pyroclastic flow phase of this event, whereas the plinian column averaged only around 16 km height. Similarly, recent studies indicate that the exceptionally high 45 km eruption cloud generated by Bezymmianny volcano in 1956 was also related to a blast and pyroclastic flow (Belousov and Bogoyavlenskaya 1988). As co-ignimbrite ash columns rise from active pyroclastic flows, it is likely that the column height is proportional to the surface area of the active flow and the mass eruption rate. By analogy with these historic events, we infer that the Minoan co-ignimbrite ash column may have exceeded the preceding 36 km high plinian column. Thus both the plinian and co-ignimbrite ash clouds extended well above the 17 km tropopause and lead to major stratospheric injection of fine tephra and aerosol-forming volcanic gases. 

The estimate of peak intensity during the plinian phase is based on field measurements of lithics dispersal near the source (Fig. 2; Bond and Sparks 1976). This well-constrained parameter can in turn be used to evaluate the magnitude (total erupted volume) based on the recently documented correlation of these two parameters (Carey and Sigurdsson 1989). The inferred intensity value (2.5 x 10⁸ kg/s) corresponds to a magnitude of 5 x 10¹³ kg, compared to 8.4 x 10¹³ kg based on the new isopach map (Fig. 4). We consider this as supportive evidence for the great magnitude of the Santorini event.

Our intensity and volatile estimates can be used to assess the dynamics of the eruption during the plinian phase based on the model of Wilson et al. (1980). Fig. 5 shows magma discharge rate and exit velocity of the plinian phase (as determined from the volatile content and the equations of Wilson 1980b) plot in the field of a stable convective column. Transition to the generation of pyroclastic flows may have been the result of either large increases in magma discharge rate associated with caldera formation (Druitt and Sparks 1984) or significant reduction in the magmatic volatile content. Because we have no petrologic evidence of greatly reduced volatile content in the pyroclastic flow material, we favour the former interpretation for the evolution of the eruptive sequence.

VOLCANIC VOLATILE EMISSIONS

A direct estimate of the volcanic volatile mass released during the Minoan eruption can be made on the basis of petrologic studies of the ejected tephra. Glass inclusions in phenocrysts can be considered as representative samples of the pre-eruption magma composition, which contain original concentrations of volatile elements such as sulphur, chlorine, fluorine and magmatic water, quenched in the glass. On the other hand, matrix glass of the erupted tephra represents the degassed magma, and the difference in concentration of volatile species between glass inclusions and matrix glass is a measure of the yield to the atmosphere during degassing of the magma. Our analytical methods and the limitations of this approach have been discussed in Devine et al. (1984), but it is emphasized that this petrologic method can only give a minimum estimate of volcanic volatile yield.

The composition of glass inclusions in plagioclase phenocrysts and matrix glasses from the Minoan plinian tephra are given in Table 1. Glass inclusions have a systematically lower total of analysed oxides than the matrix glass, and we take this to reflect primarily the amount of water dissolved in the glass inclusions. Thus the pre-eruption water content of the Minoan magma was probably of the order 5.5 wt.%. When normalized to the same total, glass inclusions and matrix glass compositions are very close (compare second and third column in Table 1). The only oxide which falls significantly outside one standard deviation of the average is Na₂O, which is inherently difficult to determine in silicic glasses (Nielsen and Sigurdsson 1981). We therefore conclude that the inclusions are representative of the pre-eruption composition of the Minoan magma, and that their volatile content can be taken as a basis of estimating volatile yield to the atmosphere during the eruption.

On the basis of microprobe analysis of glass inclusions, Devine et al. (1984) reported a sulphur content of the Minoan rhyodacite magma of about 74 (± 12) ppm. And a sulphur content of the degassed matrix glass of about 53 (± 12) ppm. A new and more comprehensive analysis of sulphur in Minoan glass inclusions gives significantly different results, when the effects of valency-dependent peak wavelength shifts on the concentration of sulphur during electron bombardment are taken into account. The new results (see footnote in Table 1) indicate a range of sulphur concentrations in glass inclusions from about 44 to over 100 ppm, presumably reflecting real heterogeneities in the Minoan magma. The data average around 70 ppm, but for the sake of the sulphur degassing mass estimate, we have adopted 100 ppm as the upper limit. Matrix glass of the Minoan tephra has a sulphur concentration of 35 ± 14 ppm, i.e. the maximum fraction of sulphur released to the atmosphere during the Minoan eruption was about 65 ppm of the total erupted mass. The low initial sulphur content and low atmospheric yield are typical of silicic magmas in general, in contrast to basic and trachytic magmas (Palais and Sigurdsson 1989). When scaled to the total erupted mass of magma, these figures give an estimate of the sulphur output to the atmosphere from the Minoan eruption as 5.5 x 10⁹ kg, equivalent to a 1.7 x 10¹⁰ kg volcanic aerosol composed of H₂SO₄. The error in this value is ± 20% due to the microprobe analysis of sulphur; the additional error due to uncertainty in the estimate of total erupted magma is believed to be of the same order. The above mass yield may be high by about 10% due to the crystal content of the magma as, strictly speaking, the sulphur output from the eruption should be calculated only on the basis of the melt fraction mass, excluding the mass of erupted crystals. The crystal content of the Minoan magma is 9.9% (Sparks and Huang 1980).

The glass inclusion data also indicate that pre-eruption chlorine content of the Minoan magma was of the order 2600 ppm, or typical for silicic magmas (Palais and Sigurdsson 1989). As chlorine content of inclusions and matrix glasses are not significantly different (Table 1), we conclude that the magmatic chlorine yield to the atmosphere was negligible during the Minoan eruption. By far the major volatile transfer to the atmosphere during the eruption involved magmatic water. Our data on glass inclusions indicate pre-eruption H₂O content of 5.5 wt.%, corresponding to about 4.6 x 10¹² kg magmatic water output to the atmosphere.

COMPARISON WITH ACIDITY LAYERS IN ICE CORES

Initially, Hammer et al. (1980) and Hammer (1984) reported an acidity layer in Greenland Camp Century ice core at about 1390 ± 50 BC, which they correlated with the Minoan eruption of Santorini. The acidity peak value of 5 μequiv. H⁺/kg corresponds to 5.4 ± 0.6 μequiv. SO₄/kg (Hammer 1980). The layer represents an acid fall-out of 98 kg/km² in Greenland, and they estimated a total global acid fall-out (H₂SO₄+HX) of 1.25 x 10¹¹ kg from this eruption. We have pointed out previously (Sigurdsson et al. 1985) that the estimated total acid fall-out implied by this ice-core layer is about 30 times higher than the petrologic estimate of volcanic acid yield of Devine et al. (1984), calling into question the correlation of the Minoan eruption with this acidity peak.

More recently, Hammer et al. (1987) have abandoned their correlation of this acidity layer with the Minoan event, and instead propose that an acidity layer at depth of 1227.5 m in the Dye 3 south Greenland ice-core, with an age of 1645 ± 20 yr BC corresponds to the Santorini eruption. It is instructive to compare the acidity yield of the eruption which generated this acid fall-out in Greenland with the volcanic aerosol yield estimated by us for the Minoan eruption. Electric conductivity of the ice core indicates a H⁺ concentration in excess of 4 μeq/kg ice and a peak sulphate concentration of about 11 μeq SO₄^2-/kg ice (Fig. 2 and 3 of Hammer et al. 1987). Summing the entire 25 cm width of the acidity peak gives a mean SO₄^2- deposition of 5.63 μeq/kg, or a sulphate deposition of 1.24 meq/m², corresponding to H₂SO₄ deposition of 62 kg/km². Using Clausen and Hammer's (1988) scaling factors for derivation of global fall-out, the acidity layer corresponds to a total H₂SO₄ mass of 1.92 x 10¹¹ kg if scaled to a low northern latitude (2 to 17° N) nuclear fall-out model and 0.94 x 10¹¹ kg if scaled to a high northern latitude (75° N) fall-out. Santorini (36.4° N) would fall closer to the former.

The mass of volcanic aerosol fall-out that generated the 1645 BC acidity layer is thus between 6 and 12 times higher than the total volcanic aerosol yield of 1.7 x 10¹⁰ kg H₂SO₄ for the Minoan eruption, estimated by us on petrologic grounds. As indicated above, uncertainties in the total erupted mass of magma and in microprobe determination of sulphur could lead to a petrologic underestimate, but we have adopted the highest sulphur concentration observed in glass inclusions, in order to arrive at the maximum value. Even with such generous assumptions in the petrologic estimate, it is difficult to reconcile the petrologic data with the ice-core evidence. It is possible that re-calibration on the basis of data from the 1982 El Chichon eruption may lower ice-core volcanic acidity estimates by a factor of three (Paolo Laj, pers. comm.), and thus bring the ice-core and petrologic estimates closer. But on the basis of the present evidence, the 1645 BC ice-core acidity layer may be either (a) from a mid- to low-latitude eruption with a volcanic aerosol yield at least 6 to 12 times higher than indicated for the Minoan eruption by our data, or (b) from high-latitude eruptions, such as the 3430 ± 100 ybp Aniakchak caldera eruption in Alaska or the frequent fissure eruptions in Iceland which lead to sulphur deposition on the nearby Greenland ice cap.

A possible source of excess sulphur in the volcanic aerosol, in addition to the degassing of erupted magma, is sulphur derived from evaporation of sea water in contact with hot pyroclastic flows entering the ocean around Santorini. There are several lines of evidence that indicate extensive interaction between sea water and the Minoan magma at the time of the eruption. Thus the discovery of stromatolites on Santorini contemporaneous with the eruption may be taken as an indication of a caldera flooded by sea water in Minoan times (Eriksen et al. 1990). Second, features of volcanic deposits from phase 2 and particularly phase 3 of the Minoan eruption indicate extensive magma-water interaction (Sparks and Wilson 1990). Third, the vast majority of the ignimbrite flows generated during the Minoan eruption entered the ocean around Santorini.

The heat content of these flows would have resulted in heating and evaporation of sea water, which may in turn have affected the volatile budget of the volcanic eruption plume and the resulting volcanic aerosol. Ignimbrites retain high heat during flow from source. Thus the 1980 Mount St. Helens pyroclastic flows were generally at 300^o to 400^o C within 5 to 7 km from the crater (Banks and Hoblitt 1981). The heat output from hot ignimbrites to evaporation of sea water can be estimated as 6.8 x 10⁵ J/kg, or total of 1.4 x 10¹⁹ J, assuming cooling from 400^o to 20 ^oC of half the ignimbrite mass (2.1 x 10¹³ kg; latent heat 3 x 10⁵ J/kg; specific heat 10³ J/kg/^o C). Assuming a fully efficient energy transfer and 2.6 x 10⁶ J/kg heat consumption in sea water evaporation, this heat source could evaporate 5.5 x 10¹² kg sea water. With a mean sulphate concentration of 28 moles/m³ or approx. 2700 ppm in sea water, evaporation of this mass could conceivably yield 1.5 x 10¹⁰ kg SO₄^2- as a contribution to the Minoan volcanic aerosol, or comparable to the 1.7 x 10¹⁰ kg sulphate mass estimated above on a petrologic basis. Sea water evaporation thus could be considered as one possible factor in accounting for the discrepancy between the petrologic and ice-core estimates, but such a process should also result in much greater enrichment of the HCl component in the volcanic aerosol (1 x 10¹¹ kg Cl on basis of the above factors). As chlorine concentrations are not enhanced above background in the 1645 BC ice-core acidity layer (Fig. 3 in Hammer et al. 1987), we conclude that contribution from sea water evaporation can be ruled out as a factor in explaining this discrepancy. The potential contribution from oceanic sulphur to the Greenland acidity peak can, however, be tested by sulphur isotope analysis, as the δ³⁴S of sea water has a value of about +20‰, whereas the value for magmatic sulphur is between zero and +4‰.

A non-correlation of the 1645 BC acidity peak with the Minoan eruption is further supported by radiocarbon dating of organic matter carbonized during the eruption, giving ¹⁴C calibrated ages corresponding to 1615 ± 17 yrs BC (Hammer et al. 1987), i.e. significantly younger than the Greenland acidity layer. Finally, the discovery of a major eruption of Aniakchak in the Aleutian arc (Alaska) that is virtually indistinguishable in age from the Minoan eruption raises further questions regarding the source eruption of the Greenland acidity peak. The caldera-forming Aniakchak eruption has a ¹⁴C age of 3430 ± 100 ybp and produced more than 50 km³ of volcanic deposits (Miller and Smith 1987). No petrologic studies have been carried out on the products of this important eruption and consequently its potential sulphur yield to the atmosphere is unknown but, because of its high-latitude source, the eruption would have led to the deposition of a major volcanic acidity layer on the Greenland ice sheet at about the time of the Minoan event.

In Fig. 6 we present the petrologic estimate of H₂SO₄ volcanic aerosol from the Minoan eruption and show for comparison the sulphuric acid mass implied by the Greenland acidity layer from 1645 BC. Also shown are mass estimates of sulphuric acid yield from a number of major eruptions for comparison.

POTENTIAL ATMOSPHERIC AND CLIMATE EFFECTS

Devine et al. (1984) have shown that estimates of volcanic sulphur aerosol mass correlate well with mean northern hemisphere temperature decrease following several historic eruptions, where the observed temperature decreases are related to the sulphur mass by a power function close to that of a cube root. Inclusion of additional volcanic events in this data base, shown in Fig. 7, has further strengthened the correlation of sulphur aerosol mass and volcano-induced temperature decline (Palais and Sigurdsson 1989). This relationship is consistent with a linear increase in optical depth of the atmosphere (and consequent surface temperature decrease) with the cube root of the change in mass and volume of the stratospheric sulphuric acid volcanic aerosol.

On the basis of our petrologic estimate of 5.5 x 10⁹ kg sulphur emission during the Minoan eruption and the temperature/sulphur dependence as revised by Palais and Sigurdsson (1989), the mean northern hemisphere surface temperature decline following the Minoan eruption is estimated as 0.5^o C. The Minoan H₂SO₄ volcanic aerosol estimate of 1.7 x 10¹⁰ kg is somewhat larger than those of the Krakatau 1883 and Agung 1963 eruptions in Indonesia (2.9 x 10⁹ and 2.8 x 10⁹ kg, respectively) and the Katmai 1912 eruption in Alaska (7.9 x 10⁹ kg; Palais and Sigurdsson 1989). We emphasize, however, that the mass estimates for the Krakatau and Agung events are less accurate than that for the Minoan eruption, as the Indonesian volcanic deposits have yet to be mapped in detail. Estimated northern hemisphere surface temperature decrease following these events is 0.2 to 0.3^o C (Rampino and Self 1982; Jones et al. 1981).

The climatic effects of the Minoan eruption have been correlated with a frost event recorded in tree rings in the western USA and dated by dendrochronology as 1626 ± 2 yrs BC (LaMarche and Hirschboeck 1984). Narrow tree rings have also been identified in Irish oak trees from the period 1626 to 1628 BC and attributed to the Santorini event (Baillie and Munro 1988). These frost events are within the limits of ¹⁴C dating of 1615 ± 17 yrs BC of Santorini vegetation carbonized by the eruption (Hammer et al. 1987).

It is evident that many notable frost ring events recorded in trees correlate with volcanic eruptions, and from the data of LaMarche and Hirschboeck (1984) it appears that approx. 60% of frost ring events correlate with major volcanic eruptions, whereas about 50% of major eruptions correlate with frost ring events. Notably absent from the frost ring record is e.g. any evidence of the major eruptions in Japan and Iceland in 1783, and the eruption of an unknown volcano in 1259, which form the largest known volcanic acidity peaks in Greenland ice cores (Langway et al. 1988).

CONCLUSIONS

New evidence on thickness and distribution of the tephra fall-out layer of the Minoan eruption leads to a revised total fall-out volume of 19.1 km³ dense-rock equivalent (4.2 x 10¹³ kg), partitioned between 4 x 10¹² kg plinian fall and 3.8 x 10¹³ kg co-ignimbrite ash. Adopting the proportions of 46/54 for co-ignimbrite ash/fines-depleted ignimbrite (Sparks and Huang 1980), the volume of the latter is estimated as 20 km³ DRE (4.2 x 10¹³ kg). Total magma output during the Minoan eruption is therefore 39 km³ DRE, or 8.4 x 10¹³ kg. This estimate is based on the widest isopachs that can be constructed for the Minoan fall-out, and therefore is likely to be an upper limit, with a maximum error of a factor of two. The peak mass eruption rate (intensity) of the plinian phase has been estimated as 2.5 x 10⁸ kg/s on the basis of isopleths of maximum lithics clasts (Bond and Sparks 1976; Carey and Sigurdsson 1989). The Minoan eruption would have lasted about four days with this mass eruption rate. This intensity corresponds to a 36 km high plinian column, thus leading to the injection of volcanic aerosol and tephra into the Earth's stratosphere. The combined evidence of magma volatile content and mass eruption rate give a basis for evaluating peak exit velocity (380-520 m/s) and vent radius (~160 m) during the maximum intensity of the plinian phase.

Petrologic evidence indicates that prior to the eruption the rhyodacite magma resided in a reservoir at 2.5 to 5 kbar, or approximately at 8 to 15 km depth in the crust (Carey and Sigurdsson in prep.). Mineralogic compositions are consistent with a temperature of 850° C, fO₂ 10^-13 and a water content of 3-4 wt.%. The magma was thus water-undersaturated in the reservoir, but reached saturation at approximately 1 kbar during ascent to the surface. This evidence of relatively low fugacity of oxygen in the magma (lower than the NiO-Ni buffer curve) and the absence of a free vapour-dominated gas phase in the magma reservoir speak against the presence of sulphur-rich phases in the magmatic system that could have contributed to the sulphuric acid aerosol. Firstly, the oxidation state of the magma is much lower (by an order of magnitude) than required for stabilization of a sulphur-rich phase such as an anhydrite (Carroll and Rutherford 1985). Secondly, accumulation of sulphur gases in a vapour-dominated gas phase in the magma reservoir is very unlikely, as the phase equilibria indicate a vapour-free system.

Sulphur yield to the atmosphere is estimated as 5.5 x 10⁹ kg S during the Minoan eruption, on the basis of electron microprobe analyses of non-degassed glass inclusions in phenocrysts and degassed matrix glasses, or equivalent to a 1.7 x 10¹⁰ kg volcanic aerosol composed of H₂SO₄. The low magmatic sulphur content and degassing are typically those observed for a large number of other rhyodacite and rhyolite explosive eruptions, reflecting the low sulphur solubility of Fe-poor silicic magmas (Palais and Sigurdsson 1989). Although chlorine is quite high in the magma, there is no evidence of significant chlorine-loss to the atmosphere during the eruption, as both matrix glasses and glass inclusions have comparable Cl concentrations.

Previous estimates of the volcanic aerosol mass from the Minoan eruption have been made from acidity layers in ice cores. These estimates range from 1.25 x 10¹¹ kg (H₂SO₄+HX) for the 1390 ± 50 yrs BC Greenland ice-core acidity layer (Hammer et al. 1980), to 1.92 x 10¹¹ kg H₂SO₄ for the 1645 ± 20 yrs BC acidity layer in a more recent study (Hammer et al. 1987). The difference by an order of magnitude (petrologic/ice-core factor 0.03) between the petrologic and ice-core estimates is very hard to reconcile, if this ice-core acidity layer is accepted as representing the Minoan eruption. Errors in the Minoan magma volume determination and sulphur content could increase the petrologic estimate by no more than a factor of two, leaving a large discrepancy with the ice core. This is curious, as petrologic estimates for other eruptions in general match the ice core evidence quite well; thus the acid volcanic aerosol masses for well-studied events, determined by these two methods (petrologic estimate of sulphur mass/ice-core total acids estimate) differ by only a factor of 0.9 for the Laki 1783 eruption, 0.5 for the 1815 Tambora eruption, 0.6 for the AD 934 Eldgja eruption, 0.4 for the Katmai 1912 eruption, 0.5 for the Coseguina 1835 eruption and 0.4 for the Santa Maria 1902 eruption (Palais and Sigurdsson 1989). The petrologic estimate tends to be slightly lower in general, probably due to the inclusion of acids other than H₂SO₄ in the ice-core layer, such as HCl and possibly HF.

The observed discrepancy between the petrologic sulphur yield and volcanic aerosol ice-core estimates may be related to analytical problems, such as scaling factors in deriving global acidity mass from ice cores. Alternatively the discrepancy may indicate that the 1645 ± 20 yrs BC ice-core acidity layer is not derived from deposition of the Minoan aerosol and represents an unknown eruption, or possibly the 3430 ± 100 ybp Aniakchak eruption in Alaska. The ice-core layer is somewhat older than the most recent ¹⁴C dating of 1615 ± 17yrs BC for the Minoan event, based on vegetation on Santorini carbonized by the eruption (Hammer et al. 1987). This ¹⁴C date correlates with possible climatic effects of the Minoan eruption, as shown by a western USA tree-ring dated frost event in 1626 ± 2yrs BC (LaMarche and Hirschboeck 1984) and narrow tree rings in Irish oak trees from 1626 to 1628 BC (Baillie and Munro 1988). Our results indicate that the Minoan event produced a stratospheric sulphuric acid aerosol (1.7 x 10¹⁰ kg H₂SO₄), which was significantly larger than the Krakatau 1883 (2.9 x 10⁹) and Agung 1963 (2.8 x 10⁹ kg) volcanic aerosols, resulting in about 0.5° C northern hemisphere surface temperature decrease. This is consistent with the occurrence of comparable frost-ring events in the tree-ring record for all three eruptions (LaMarche and Hirschboeck 1984).

------------------------------------------------

--------------------------------------------------