Palaeomagnetic estimates of emplacement ... - Springer Link

Volc~~~~~ölogy

Bull Volcanol (1989) 51:16-27

© Springer-Verlag 1989

Palaeomagnetic estimates of emplacement temperatures of pyroclastic deposits on Santorini, Greece Elizabeth A M c C l e l l a n d ~ and Timothy H Druitt 2 i Department of Earth Sciences, University of Oxford, Oxford OX1 3PR, United Kingdom 2 Department of Geology, University College, PO Box 78, Cardiff CF1 1XL, United Kingdom

Abstract. Thermal remanent magnetism provides

a method of quantitatively determining the emplacement temperature of individual lithic clasts in a volcaniclastic rock. The technique is reviewed and applied to two types of Quaternary pyroclastic deposit on Santorini. Emplacement-temperature estimates for lithic clasts from two co-ignimbrite lithic breccias (Cape Riva and Middle Pumice eruptions) range from 250°C to >580°C, showing unambiguously that the breccias were emplaced hot. Good precision on temperature estimates (about +20°C) were obtained from the Cape Riva breccias. Lithics in a Plinian airfall deposit from the Middle Pumice eruption give less precise results because the primary magnetisation has been partly overprinted by chemical (and/or viscous) remanence, and some clasts may have rotated during compaction of the deposit. Temperatures from proximal airfall are consistent with welding of the deposit within 1.5 km from vent. Temperature estimates for lithic clasts further from vent scatter, but a falloff of temperature away from vent can be recognised if an average emplacement temperature for the whole deposit is identified at each location. The study highlights some difficulties in interpreting quantitative temperature estimates for prehistoric pyroclastic deposits.

Introduction

Clastic deposits around volcanoes are emplaced at a variety of temperatures. Pyroclastic flows come to rest at elevated, even near-magmatic temperatures, and commonly weld. Fallout tephra Offprint requests to: TH Druitt

cool by entrainment of air into eruption columns and during settling through the atmosphere, but can weld close to the source vent where accumulation rates are highest. Reworking of pyroclastic ejecta by water can generate clastic deposits that are essentially cold. Assessment of emplacement temperatures can be crucial in interpreting ancient volcaniclastic sediments. For example, a common problem is distinguishing between non-welded pyroclastic-flow deposits and deposits from mudflows derived by water-remobilization of poorly sorted volcanic debris. Field criteria favouring hot emplacement include the presence of charcoal, baked lithic clasts, or gas-escape pipes, but the evidence is sometimes ambiguous. Charcoal and baked lithics may be source-derived while gas-escape pipes occur in mudflows emplaced on top of hot lava or ejecta. Only gas-escape pipes rooted on charcoal provide a criterion for hot emplacement that is reasonably unambiguous. Quantitative knowledge of temperatures of pyroclastic deposits may prove valuable in modelling explosive eruptions. The thermal remanent magnetism (TRM) of a deposit provides a method of temperature estimation. Aramaki and Akimoto (1957) used TRM in a qualitative manner to distinguish deposits of nuée ardentes from those of mudflows. The nuée deposits contain lithic clasts which had been heated above their Curie Points and had a single, uniform component of magnetisation acquired during cooling; lithics in many mudflows had randomly oriented magnetisations because they had not been heated significantly during emplacement. A similar study of deposits from the Minoan eruption of Santorini was reported by Wright (1978). Subsequently, the method has been extended by Hoblitt and Kellogg (1979) to give quantitative estimates, and refined by Kent et al.

McClelland and Druitt: Emplacement temperatures of pyroclastic deposits

(1981) using more sophisticated data-analysis techniques. We applied the thermal demagnetisation method to two types of pyroclastic depos!t from Santorini. First, we studied two co-ignimbrite lithic breccias (Middle Pumice and Cape Riva eruptions). These breccias are coarse-grained, lithicrich deposits believed to be formed by proximal sedimentation of lithic blocks from pyroclastic flows (Druitt and Sparks 1982; Wright 1981; Walker 1985). This interpretation is based on field evidence where the breccias are well exposed, but where poorly exposed the breccias resemble some mudflow, flash flood, or rock-avalanche deposits. Our temperature estimates demonstrate unequivocally a bot, primary origin for the breccias. Second, we studied a Plinian airfall deposit (Middle Pumice eruption) which is welded within 1.5 km from its source vent. Our purpose was to investigare the potential of the TRM method in quantitative thermal studies of prehistoric airfall deposits by seeing if: (1) temperatures yielded by proximal, welded airfall were physically plausible, and (2) a falloff in temperature away from vent could be detected and quantified. Our data highlight some difficulties in interpreting quantitative temperature estimates, and suggest that the thermal histories of individual clasts in pyroclastic deposits may be complex. We begin by reviewing the concept of TRM and the principles of the thermal demagnetisation method.

Principles of palaeomagnetic temperature determination The remanent magnetisation of a rock has the capacity to record several directions of magnetisation acquired during its thermal history. The temperature at which a particular grain of a magnetic mineral acquires its magnetisation (the blocking temperature, Tb) depends essentially on grain size, shape, and mineralogy, and may range from the Curie Point (580°C and 680°C for pure magnetite and hematite respectively) to the ambient (room) temperature. Above Th, the available tl-lermal energy is sufficient to prevent the grain from retaining a permanent magnetisation; below Tb a magnetisation parallel to the ambient magnetic field is frozen in. In general Tb increases with increasing grain size. When a rock or magma containing a large number of magnetic grains is cooled from high temperature, magnetisation begins to be recorded by grains with the highest Tb s as the temperature falls through the Curie Point.

17

As cooling progresses, more and more grains freeze in a remanent magnetisation recording the ambient fiel& Once the rock or magma has cooled to ambient temperatures, it has acquired a thermal remanent magnetisation (TRM). Let us consider that a rock carrying a TRM is now moved into a different orientation with respect to the Earth's magnetic field, where it is heated. Here we are concerned with the situation in which a lithic clast is ripped from the walls of a volcanic vent or eroded from the ground surface and incorporated into a pyroclastic deposit. Three situations can be envisaged. 1. If the emplacement temperature of the clast in the deposit (Ternpl) is greater than the Curie temperature, all grains will be demagnetised and on re-cooling will acquire a uniformly oriented remanence in the new field direction. 2. If Templis equal to the ambient temperature, no demagnetisation will occur, and all grains will retain their original remanence, which will appear as a randomly oriented component in the deposit. 3. In the case of Templ being less than the Curie Point, grains with Tb > Templ will retain their original remanence, hut those with Tb < Templ will remagnetise. Thus, in general, lithic clasts in pyroclastic deposits have a two-component magnetisation" a randomly oriented high-Th remanence, and a uniformly oriented 1ow-Tb one. The emplacement temperature of a lithic clast can be determined by thermal demagnetisation. The sample is heated to an initially low temperatute and then cooled in the absence of a magnetic field. The remanence in grains with TbS less than the chosen temperature will be randomised by the heating and this random orientation will be frozen in on cooling. The remanent direction and intensity is then measured. An increment of the total remanence will be removed by this process, and the process is repeated at progressively higher temperatures until all the remanence is removed. At temperatures up to Templ the low-Tb component direction is removed; above this temperature the original high-Tb component is removed. This is clearly displayed if the data are plotted on a vector diagram where both the changes in direction and intensity are plotted. Figure 1 is a vector plot of two-component remanence in a lithic clast from the Cape Riva lithic breccia. The remanence vector is visualised as a line in 3-D space with length equal to the remanent intensity, one end anchored at the origin and pointing in the direction of the magnetisation. As the total remanence

18


CR48A

Up N

w , ' ~'o ' "~~~i!j124'

over this temperature range. Above 480 o C the original remanence is removed, which is SSW and upwards. There is some slight curvature of the path between the two straight line segments, and the laboratory estimate of the re-heating temperature lies between the highest temperature at which the vector point statistically lies on the overprint straight line, and the next temperature step above this (457-465 ° C in this example). See McClelland Brown (1981) for a more detailed explanation of this method. Estimation of Temp~ is subject to errors. The blocking and unblocking process is statistically controlled so some grains which unblocked in the natural re-heating event will not be unblocked in the laboratory experiment even though the laboratory temperature is equivalent to the natural temperature. This causes a blurring of the 'breakpoint' between the straight line segments of about 10 ° C. Measurement errors in direction and intensity are taken into account in a straight-line-fitting package by Kent, Briden and Mardia (1983), (the L I N E F I N D program). Errors in temperature measurement are about 5°C. However, the method is more often limited by the size of temperature steps it is feasible to undertake, and by problems with chemical overprinting associated with the heating (see McClelland Brown, 1982 and results section of this paper).

,E

S Down Fig. 1. Vector plot for thermal demagnetisation of a lithic clast with two-component remanence. Dots give the magnetisation vector of the clast, projected onto the horizontal plane, at different laboratory temperatures (given by the numbers beside the dots, in °C); open trian9les give the vector projected onto a vertical plane in an E / W direction at the same laboratory temperatures. Best fit lines to the data were calculated by the L I N E F I N D program of Kent et al. (1983), slight curvature around the break-point is attributable to a small amount of chemical overprinting of the primary remanence. N R M is natural remanent magnetisation at room temperature

vector changes during progressive thermal demagnetisation, the free end of the line traces out a path in space. Figure 1 shows the projection of this path onto the horizontal and a vertical plane. The points on the path represent the vector endpoint after thermal demagnetisation to the indicated temperature. In this example, the low-T» overprint direction is removed during thermal demagnetisation to about 450°C and a straight line in the direction of the vector being removed (NW and downwards) is traced out on the vector plot

Geological background Santorini (Greece) has been the site of at least 12 pyroclastic eruptions since the late Pliocene. Pyroclastic deposits are weil exposed in caldera-wall successions as thick as 200 m. We focus on coignimbrite lithic breccias from the Middle Pumice and Cape Riva eruptions, and a welded Plinian airfall deposit from the Middle Pumice eruption.

Sample Iocalities Fig. 2 a. Map of Sant0rini showing sample sites for the Middle Pumice airfall, Middle Pumice lithic breccias, and Cape Riva lithic breccias and ignimbrite. Vent sites for the initial airfall phases of each eruption and the dispersal axis of the Middle Pumice airfall are also shown. Abbreviations: CAN: Cape Ayios Nikolaos; CK: Cape Katothira; CA: Cape Athinios; CB: Cape Balos; CL: Cape Loumaravi; b thickness (uncompacted, in cm) and isopach contours of the Middie Pumice airfall, modified after Sparks and Wright (1979). Thicknesses marked by asterisks are unpublished data of R. Mellors

Isopach map,MiddlePumiceairfall

Approx. position of Cape Riva airfall vent

Therasia(~

/

Approx. positionof f " Middle Pumice airfalf

veet ~

Middle Pumice airfall / dispersal axis ~/

Localities

•

~

• Middle Pumice airfall • Middle Pumice

//

// /

//

/ ~ «

/

/I

l/I 1// / I

"-f t

i / ~~

"25o ~260

ù, , ~ Y ° , 2~8 /

(

Ther

14 ~

/

I =o

/'«

airffäll~b+ r

/I

,

;;~9o

~

o Cape RJva breccia (also airfall+ignimbrite)

0

km

b


The co-ignimbrite lithic breccias were described by Druitt and Sparks (1982). The Middle Pumice breccias directly overlie the airfall to a depth of up to 60 m. They are clast- to matrixsupported, massive to bedded, and contain lithic blocks as large as 2 m in diameter. Juvenile andesitic scoriae and rounded pumice lapilli are the subordinate component. The Cape Riva eruption (18000 B.P.; Pichler and Friedrich 1976) began with a Plinian-style phase from a vent in northern Therasia (Fig. 2a). A welded ignimbrite was emplaced when the Plinian column subsequently collapsed (Druitt 1985). The co-ignimbrite lithic breccias formed during the third eruptive phase, probably concomitant with caldera subsidence. The breccias are as thick as 25 m, and contain lithic blocks up to 2 m across and subordinate rhyodacitic pumice. Locally the lithic breccias grade vertically or laterally into non-welded ignimbrite (jointly termed the breccia unit in this paper). The breccia unit is overlain by welded ignimbrite from the final phase of the Cape Riva eruption. The Middle Pumice and Cape Riva lithic breccias are believed to have formed by proximal sedimentation from pyroclastic flows (Druitt and Sparks 1982) because: (1) they are very extensive, with original areas exceeding 60 km2; (2) locally they grade laterally and vertically into nonwelded ignimbrite; (3) they contain gas-escape structures. Orange oxidation zones which transgress lithological boundaries in the Middle Pumice breccias suggest emplacement while hot. T h e r m a l remanent magnetism can be used to test the interpretation that the breccias were hot, primary volcanic deposits. The Middle Pumice airfall deposit was described by Sparks and Wright (1979), who called it the Thera Tuff. It varies in thickness from 2 m to 6 m and mantles topography in the manner typical of fallout. Isopachs (Fig. 2b, modified after Sparks and Wright 1979) identify a source vent north of Cape Katothira, and a SSW-directed dispersal axis (Fig. 2a). Within about 1.5 km from the vent the airfall is welded; but beyond about 1.5 km it is a non-welded, inversely graded, Plinian pumice-fall deposit. The transition from welded to non-welded facies occurs rapidly in caldera wall exposures. The airfall is compositionally zoned; pumice is dacitic throughout (Sparks and Wright 1979; Druitt 1983) but becomes slightly more mafic (from 65% to 63% SiO2) upwards in the deposit. The well-exposed transition from welded to non-welded facies in the Middle Pumice airfall provides an opportunity to test t h e use of TRM in estimating quantitatively the em-

19

placement temperature of a prehistoric pyroclastic deposit.

Sampling and measurement Twenty-seven lithic clasts were collected at ten localities from the Middle Pumice airfall (Fig. 2a); lithologies include porphyritic and aphyric lavas, lithified tuffs, and a granitoid. Fifteen pumice blocks were collected from the airfall at nine localities. The Middle Pumice breccia was sampled at two localities with a total of five lithic clasts. Twenty-nine lava clasts collected from the Cape Riva airfall, welded ignimbrite, and breccia unit (lithic breccia and non-welded ignimbrite) at four localities include mafic, silicic, porphyritic and aphyric types. It is common for palaeomagnetic samples to be taken by drilling oriented cores in situ. This procedure was not followed on Santorini because of the generally unconsolidated nature of the deposits, the precipitous cliffs from which the samples were taken, and the lack of cooling water. Clasts were oriented using a levelling table with suncompass attachment. Once in the laboratory, one-inch-diameter cores were cut from the oriented clasts. Pairs of cores were cut from each of five clasts in order to obtain a measure of precision of Templ. The total magnetisation (NRM) of the samples was measured using a Digico spinner magnetometer and values of NRM were found to lie within the range of 100-20000 mAm -1. Samples were thermally demagnetised using furnaces with residual fields of less than 5nT. Vector structure of remanence was analysed using the line-fitting algorithm of Kent et al. (1983) (the L I N E F I N D program).

Results and interpretation

General features of the data The rocks analysed in this study have one or more stable components of magnetisation, the directions of which can be determined using the LINEFIND program; none display random behaviour.

Lithic clasts. Most lithic clasts have a two-component remanence which reflects a low-Tb component acquired while cooling in the deposit, and an earlier high-Tó magnetisation. Precision as good as about _+20°C on the estimate of Templ is obtainable from lithics clasts from the Cape Riva

20


breccias. Many Cape Riva lithics show almost ideal behaviour upon thermal demagnetisation. Examples of well-behaved, two-component remanences from the Cape Riva breccias are shown in Figs. 1 and 3a. Data on each vector plot define two clear lines in each projection, with a sharp break in direction between them. The estimate of Temp! is the temperature interval between the highest temperature at which the vector endpoint lies on the low-Tb line and the next temperature,

this being 465-495°C for sample CR48A (Fig. 1) and 521-528°C for CR66A (Fig. 3a). Fewer lithic clasts from the Middle Pumice ejecta have clear two-component remanences. Examples of Middle Pumice lithics with almost ideal behaviour are shown in Figs. 3b and 3c. Examples of more complex behaviour are discussed below. Pumice clasts. Almost all pumices (collected from Middle Pumice only) have single-component re-

Up CR66A

Up ~NRM I

W

ù,,

MP46B

Up w,

MP104A

381

w,

B

I~-.-----~_+.71"ooE

,

56O

600 ~ • = ~ . . . o _ , . . . ~-~,~~,~

s,

- N

looo

S '

~l~J

....t mAr~1 ~

mAn3 +

\52o '~521

I

,+\~

" (a)

iooc

"\ 400-

l

i

1òoo '

' N

~cl

"~~ ~ " ~ ' ~ ~

NRM

\

mA~ ~

'~NRM 2oo(

2000

E Down

Down

Up N

Dffwn

MP74B Up

ù

4OOO .

jl

i"""

MP32

400

/~-- \ 474

200£

Up W

s,

°'~

~ (d)

417

400C

B 570 274-.% 224=

I

201oo

'i\~+

I

ùoo

,

mAr~ I

E

S'

~

(e) mA~ 1

200C ~ooo

~7,

mA~'

• I/'"

MP69

~«.-~:~:~~ò--, 200C

/

t, ,30..~..--,\250 ~~

(f)

,tX~___~ NRM E Down

E

~'=NRM

Down Down

Fig. 3. Typical vector plots of thermal demagnetisation data; symbols as in Fig. 1 ; a two-component remanence of lithic clast from the Cape Riva breccia. Clast emplacement temperature, Temp~= 521-528 °C; b two-component remanence of lithic clast from the Middle Pumice airfall, 1.65 km from vent. TempJ= 522-560° C; e two-component remanence of lithic clast from the Middle Pumice airfall, 6.8 km from v e n t . Templ = 4 0 0 - 4 4 3 o C; d single-component remanence of pumice from the Middle Pumice airfall; e twocomponent remanence of lithic clast from the Middle Pumice airfall, with curvature between low- and high-T» lines indicating chemical overprinting. Templ = 274-320 ° C; f three-component remanence of lithic clast from the Middle Pumice breccia, with curvature between intermediate- and high- Tb lines. Templ = 250-350 ° C or 474-536 ° C


manence acquired after deposition (e. g., Fig. 3d). Pumices discharged at magmatic temperatures will cool rapidly during transport in the eruption cloud. We expect that magnetisation in pumice would only be acquired in a coherent fashion after it has come to rest in the deposit; the remanence should therefore be single-component with blocking temperatures less than Templ. Grains with Tó > Temo~ should have been magnetised while the pumice was tumbling in the air and the total magnetisation with blocking temperatures between T»

21

and the Curie Point should have been cancelled out by this randomisation process. In view of this argument, it is a puzzling feature of our data that all pumices have remanence which is carried in blocking intervals up to the Curie Point of magnetite (580 ° C). Zlotnicki et al. (1984) have observed this phenomena in pumice samples from Guadeloupe, and attributed it to chemical remanent magnetisation (CRM) with high blocking temperatures.

N

N

+

d

Middle Pumice pumices

~

N

]

J b

a 4-

N

\

~.i "\

MiddFe Pumice lithics, high TB • Airfall • Breccla

C

•° •

o

+

Cape Riva lithics, low TB1 / • Breccia / • Airfall / • Wetded ignimbrite

d

/

\ \

~

+

\ • Bre~~ia \,~irfa,,

.

~

/ /

/ /

e

l,'ig. 4. Groupings of remanence directions shown on equal-angle stereograms; open symbols are vectors projected onto upper hemisphere, solid symbols are vectors projected onto lower hemisphere. Asterisks give the mean axial geocentric dipole field direction for Santorini (termed ambient field in text). Stars give the data mean and the circles around them show the 95% confidence limits on the mean; a pumices from the Middle Pumice airfall; b lithics from the Middle Pumice airfall and breccia, low-Tb directions; c lithics from the Middle Pumice airfall and breccia, high-Th directions; d lithics from Cape Riva airfall, welded ignimbrite, and breccia unit (includes lithic breccia and non-welded ignimbrite), low-Tb directions; e lithics from Cape Riva airfall, welded ignimbrite, and breccia unit, high-Th directions (randomly grouped)

22


its from the Middle Pumice and Cape Riva eruptions show contrasting grouping behaviour. The high-Th components in lithic clasts from Cape Riva airfall, welded ignimbrite, and breccia unit are randomly distributed (Table 1; Fig. 4e), indicating that the high-Th components predate emplacement and were randomised by the deposition process. Lithics in the Cape Riva ejecta show virtually ideal TRM behaviour. In contrast, the grouping of high-Th components from Middle Pumice lithic clasts is significant at the 95% confidence level (Table 1; Fig. 4c), although with a95=32.5 ° and 3R2/n=15.1 the grouping is not as tight as for low-Tb components of these clasts. The non-randomness of high-T» components of Middle Pumice lithics indicates that some CRM overprinting has occurred in grains with T»s over Templ , biassing the distribution towards the ambient field direction.

Directional data. Stereograms in Fig. 4 show the distribution of directions of high- and low-T» components in lithics and pumices from the two eruptions. Calculated means and circles of 95% confidence (a95) a r e also shown in Fig. 4. Table 1 tabulates mean directions and component statistics. Directions of the single components in pumices from the Middle Pumice airfall (Fig. 4a) are well grouped and lie close to the direction of the Earth's magnetic field at the latitude of Santorini over the last hundred thousand years (hereafter termed the ambient field), indicating that the remanence was acquired in situ in the deposit. The angular e r r o r (a95) on the estimate of the mean direction is 9.4 °, and the mean direction is D = 6.9 °, I = 57.3 °. This is statistically indistinguishable from the ambient field direction for Santorini. Rayleigh (1919) derived a test for random grouping of directions. This test uses the vector resultant R of a set ot n vectors, and compares 3R2/n with the value of the Chi-squared distribution (Z2, three degrees of freedom). At a level of 95% confidence a group of directions is randomly distributed if 3R2/n is less than Z2=7.81. 3 R 2 / n for the Middle Pumice pumices is 40.32 (Table 1), hence the directions are statistically significantly grouped. Directions of low-Tó components in all lithic clasts (Table 1; Figs. 4b and 4d) are also significantly grouped about the ambient field (Table 1). The mean of low-T»-component directions for Middle Pumice lithics is statistically identical to that of pumices from the Middle Pumice airfall at the 95% confidence level. These features are to be expected if the low-Tó components of the lithic clasts were acquired while cooling in situ in the deposit. High-T» component directions in depos-

Evidence for chemical remanence Chemical remanent magnetisation (CRM) is defined as a new magnetisation which is formed by the growth or change in shape of a magnetic mineral. This can be due to the breakdown of a nonmagnetic parent to form a magnetic mineral, alteration of a pre-existing magnetic phase into another magnetic phase (e.g. exsolution of a titanium-rich magnetite into ilmenite and pure magnetite), or growth or shape change of existing magnetic grains. Commonly, at least one of these changes takes place during heating of any rock type and CRM is formed. If Tb (an intrinsic property due to grain size, shape and composition) of the new or modified grains is less than the tern-

Table 1. Statistics for magnetisation components Eruption

Deposit

Clast

Middle Pumice

airfall

pumice

airfall and breccia

lithic

Cape Riva

airfall, ignimbrite, breccia

Component

Declination

Inclination

a95

n

R

k

3R2/n

6.9

57.3

9.4

15

14.2

17.5

40.32

low-T»

352.1

52.4

12.3

41

31.6

4.2

72.06

lithic

high-T»

338.0

42.3

32.5

29

12.1

1.6

15.10

lithic

low-Tb

14.3

50.9

9.7

28

24.9

lithic

high- Tb

75.5

76.2

> 180

24

1.5

89 1.0

66.40 0.28


perature at which they were formed (generally less than Ternpl) , then they will be unblocked on formation, and will acquire a TRM in the ambient field direction on subsequent cooling below their blocking temperature. Their magnetisation is completely indistinguishable from the overprint TRM of the original (primary) magnetic grains. If, on the other hand, Tb is greater than the formation temperature (and possibly greater than Templ), the new or modified grains will become magnetised in the ambient field direction at the instant of formation. It is therefore possible that there will be two suites of grains with TbS greater than TempO,one primary group carrying the earlier randomly oriented remanence direction, the other secondary group carrying a remanence in the ambient field direction. The estimate of the high-Tb direction will therefore be biassed towards the overprint direction. One effect of the overlapping blocking temperature above Templis to cause curvature in the vector plots of thermal demagnetisation above Tenw~. If small, this curvature does not affect the estimate of emplacement temperature. Several lithic clasts from the Cape Riva breccia show small curvature above Tempb but this does not prevent precise estimation of Templ. Figure 1 shows that three data points between 465°C and 526°C lie between the low- and high-Th lines. The data point at 457 ° C does not lie exactly on the Iow-Th line, but the LINEFIND algorithm makes a rigorous analysis of which points lie with 95% confidence on a line defined by a group of points. This algorithm has always been used in this study to decide which is the first point to lie oft the loW-Tb line when estimating Templ. The amount of CRM overprint with blocking temperatures above T~mpl taust be very small since the mean high blocking temperature direction from Cape Riva lithics is completely random (a9» > 180°). The intensity of chemical remanence in Middle Pumice lithics is generally greater than in those from the Cape Riva ejecta. Vector plots from most Middle Pumice lithics show a small amount of curvature due to CRM. In the typical example in Fig. 3e, most data points lie statistically on either the high-T» or low-Tb best-fit lines, but the point at 320°C does not. Although a straight line is obtained on demagnetisation above 320 ° C, the direction of this line is probably slightly deflected towards the ambient field direction. Rotation of the primary high-T» directions towards the ambient field by CRM is the cause of the weak statistical grouping of high-Th component directions in Middle Pumice lithics.

23

Three-component remanences Four lithic samples from the Middle Pumice ejecta have three-component remanence. In one example (Fig. 3f), significant curvature occurs between the intermediate- and high-T» components and a sharp break separates intermediate- and IoW-Tb components. The cause of this behaviour is unclear. The 1oW-Tb component may have been acquired in situ, with the high- and intermediateTb components being part of an earlier two-component remanence. The estimate of Templ would therefore be 250-350°C. Alternatively, the direction of the intermediate-Tb component could record the ambient field in this example, giving an estimate of Templ of 474-536°C, with CRM overprinting causing the curvature above this temperature. The low-Tb component (blocking temperatures up to 200 ° C) may be a low-stability viscous overprint in a direction which is not geologically significant, or the existence of intermediate- and low-Tb components may indicate that this clast rotated by an unusually large amount during cooling and compaction of the deposit (Hoblitt et al. 1985). Further indication that rotation may have occurred is provided by 4 out of 15 pumice samples which have kinked or slightly curved lines. We note that estimation of Templdepends on the breakpoint between high- and low-Tb remanences in lithic clasts, and is not affected by slight rotation during post-emplacement slumping or sample collection.

Emplacement temperatures of the co-ignimbrite lithic breccias The first facet of the study concerned the Middle Pumice and Cape Riva co-ignimbrite lithic breccias. Special attention was paid to the Cape Riva breccias, since they were described in detail by Druitt and Sparks (1982). The purpose was to test the interpretation, based on field studies, that the breccias are primary deposits from hot pyroclastic flows. The Middle Pumice breccia was sampled near Cape Katothira, where it is 60 m thick, and at Cape Athinios, where it is 10 m (Fig. 2a). Five lithic clasts were thermally demagnetised. Four clasts have two magnetisation components; directions of low-Tb components in these clasts cluster close to the direction given by pumices and lowTb components of lithic in the Middle Pumice airfall. One lithic clast has a single component. Thermal demagnetisation data from three clasts show

24


strong curvature (e. g. Fig. 3f), probably due to the presence of a chemical remanence. Estimates of Templ range from 250-350°C to _>580°C if the single-component clast was deposited above the Curie temperature (Fig. 5). The dataset for the Cape Riva breccia unit is more extensive. A total of 25 lithic clasts, 3-45 cm in diameter, were collected from four locations (Fig. 2a). Two-component clasts yield values of Templ with low uncertainties (Fig. 5) because the lithics lack significant chemical remanence, as shown by the random distribution of high-Th component directions and scarcity of strong curvature in demagnetisation data. Determinations of Te~pl on pairs of cores from each of four clasts show that a precision of _+20°C is typical for the Cape Riva breccia unit. Low-Tb component directions statistically resemble those of three lithic clasts from the airfall and welded ignimbrite from earlier phases of the Cape Riva eruption. Five single-component clasts were probably emplaced above their Curie temperatures because there is no evidence for a strong chemical remanence in the Cape Riva breccia unit. Estimates of Templ range from 312°C to >580°C. No systematic correlations between temperature and clast diameter or temperature and stratigraphic level are evident. Further data would be required to assess if an apparent upwards increase of Temp~ at Cape Balos is a real feature of the breccia, which is lithologically uniform at this locality. At Cape Loumaravi, where the breccia grades upwards into 200

300 I

400 I

~Middle Pumice breccia

'

500 I I I

'

F- . . . . . . . . . . . .

-I

I I

I

I

C a p e Athinios

!

I

CapeBalos

~

I]

I

I I I

I

I

I NIl

~-4]

Ignirnbrite I

The aims of this reconnaisance study were to test if TRM temperature estimates from the welded facies of the Middle Pumice airfall were consistent with welding, and if a falloff in temperature could be detected away from vent in the nonwelded facies. The airfall was sampled at ten localities downwind from vent, from where welded to where not welded (Fig. 2a). Because localities do not lie exactly on the dispersal axis, we projected each locality on to the axis in order to calculate distance from vent. No locality lies more

Cape Katothira

Cape Riva breccia and ignimbrite

I I

Test study of the Middle Pumice welded airfafl

600 I

I----~

I

non-welded, pumiceous ignimbrite (see Fig. 7 of Druitt and Sparks 1982), the decrease in lithic content is not reflected in systematic changes of Temp» Overall, the temperature estimates are consistent with the lack of welded pumice in the Cape Riva breccia unit if, as argued below, the lowest recorded Temp~values most closely approximate an 'avërage' emplacement temperature for a pyroclastic deposit once neighbouring clasts have come to thermal equilibrium. The temperature estimates show conclusively that the Middle Pumice 'and Cape Riva lithic breccias were emplaced bot, and cannot have been reworked by water. They are consistent with the interpretation, based on field observations, that the breccias formed by proximal sedimentation of lithic clasts from pyroclastic flows.

tCapeLournaravi

I

I

Cape A y i o s Nikolaos I

I

200

I 300

I 400

I 500

Lithic clast emplacement temperature (°C)

I

600

Fig. 5. Estimates of emplacement temperatures of lithic clasts in the Middle Pumice lithic breccias and the Cape Riva breccia unit (lithic breccia and non-welded ignirnbrite). Arrows depict samples with single-component remanences which may indicate emplacement above the Curie Point. Square brackets link results from different cores from the same lithic clast, which provide a measure of the precision of the method


than 2 km oft the dispersal axis. At each locality we collected between two and four lithic blocks 5-20 cm in diameter for thermal demagnetisation. Emplacement-temperature estimates are plotted in Fig. 6 as a function of distance from vent. Repeat determinations on single clasts are included on Fig. 6. Lithics having three components of magnetisation give two possible values for Ten~pl shown as dotted bars. Estimates of Templ range from 224 ° C to > 580 ° C (the Curie Point of magnetite). Three lithic clasts from welded airfall (most proximal locality, Fig. 6) have single-component magnetisation, in the ambient field direction, consistent with emplacement at >580°C. Temperatures from the transition zone between partially welded and non-welded airfall are 520 ° to > 580 ° C, two clasts having single-component remanences. The welding temperature of pumice depends on non-volatile composition, water content, and load pressure. Data are sparse, but the experiments of Yagi (1966) suggest at least 585 °C is required to weld rhyolitic to dacitic pumice (30 kgcm -2 load). Pumice in the Middle Pumice airfall is dacitic. Therefore the TRM temperature estimates are broadly consistent with welding in the proximal airfall deposit. In contrast, lithics from the non-welded airfall facies give temperatures of 224°C to > 5 8 0 ° C , and considerable scatter is evident at most localities. No simple falloff of Templ is observed away from vent. One factor contributing to the scatter is the common chemical overprinting in lithic clasts in this deposit, which reduces the precision of Templ compared with the Cape Riva breccias. The

25

single components of some clasts in non-welded airfall might be due to particularly strong chemical overprinting of high-Th remanence. However, because significant scatter is present even in our most precise data (samples with little CRM), variability of Templ is probably a real feature of the deposit. Another factor is that lithic samples were collected from all levels in the deposit, which is slightly zoned compositionally, and was probably zoned thermally following eruption. It is therefore likely that some scatter in Templ is due to thermal zoning, although no systematic differences between clasts from the top and base of the deposit are evident in our dataset. A third, and possibly important, factor causing scatter in Templis actual variation in initial emplacement temperatures of lithic clasts. For example, clasts plucked from deep conduit walls might be hotter than those eroded from near the surface. Also, large clasts retain heat longer than small ones during fallout, and would be emplaced at higher temperatures (no correlation between Templ and clast size is evident in our data). A combination of experimental error, thermal zoning, and natural variation in initial emplacement temperature may explain the scatter of Templ values in Fig. 6. Although no systematic falloff of Templ is obvious in Fig. 6, two lines of argument suggest that the dashed line, the approximate locus of lowest TernpI values, may crudely represent a decrease away from the vent of some 'average' emplacement temperature of the deposit. First, consider the thermal histories of lithic clasts emplaced with a range of initial temperatures (T~nit). Following deposition, heat transfer will act so that: (1) clasts

Middle Pumice airfall 600

"{,[ i

5OO

" ~ 400

,

I

'X\! I ± I

E

~ 300

"~....... I .........

E E

I ~

"

1

{

...................

~ 200

some pumices blackened Ji

»

welded

.L

no blackening

i

B

not welded

IO0

Distance from vent projected onto dispersal axis (km)

Fig. 6. Variation of emplacement temperature of lithic clasts in the Middle Pumice airfall with distance form the source vent. Dotted bars are alternative temperature estimates where the clast has a three-component remanence. Arrows (number of arrow heads equals number of samples) depict clasts with a single-component magnetisation which may indicate emplacement above the Curie Point. Alternative interpretations of single-component remanences are discussed in the text. The dotted line shows how minimum emplacement temperature estimates fall off away from vent, and probably approximates the average temperature of the deposit once small-scale (cm to dm) temperature heterogeneities had been eradicated by conduction

26


at a given level in the deposit will heat up or cool down to some average temperature (denoted Taver to distinguish it from Temp] of a single clast) given by redistribution of heat between neighbouring hot lithics, cool lithics, and pumice; (2) the entire deposit (2-6 m compacted thickness) will cool down e n m a s s e by heat transfer to its surroundings. Because, under normal conditions, both processes are governed primarily by conduction (see Riehle 1973), we expect the time scale of process (2) to be longer than that for (1). An exception might be if the entire deposit was rapidly cooled immediately after eruption, perhaps by intense rainfall (Hoblitt et al. 1985). Under normal conditions, lithics with Tinit> Tawr would cool monotonically with time after emplacement, and all ferromagnetic grains with blocking temperatures less than Tinit will be remagnetised. In contrast, clasts with Tinit( Taver will initially heat up to Tawr before then cooling to ambient, so that remagnetisation of all grains with blocking temperatures less than Tav~r will occur in initially cool lithics. Lithic clasts in the airfall will therefore record T¢mp~ values ranging from Taver to the maximum T~nit, which may exceeded the Curie temperature. Although probably an oversimplification, this scenario illustrates that if a single 'average' emplacement temperature can be recognised as representative of the deposit at a particular distance from vent, it is most closely approximated by the lowest value of Templ. Based on present data, this value (dashed line, Fig. 6) decreases from _>580°C in the welded region to 200°C 8 km from vent, and shows a marked kink in rate of decline at - 2 km from vent. The second line of argument is based on thickness (uncompacted) variations of the airfall in cal-

dera-wa11 exposures (Fig. 7). Thickness is a measure of accumulation rate (if sedimentation occurs for approximately equal times at all locations), which is believed to exert a strong influence on the temperatures of airfall deposits (Sparks and Wright 1979). Rapid accumulation inhibits cooling of clasts during fallout and promotes fast burial and insulation. Thickness of Middle Pumice airfall falls oft rapidly to ~ 2 km, then is almost uniform to ~ 7 km. The kink in slope is due to the caldera wall trending SSE across the isopachs, then swinging into approximate parallelism with them (Fig. 2b). It is an artifact of location projection onto the dispersal axis. The striking feature of Fig. 7 is the close similarity in shape between the thickness (accumulation rate) curve and dashed curve of Fig. 6. This similarity lends support to the idea that the dashed curve is a crude measure of the average emplacement temperature of the airfall deposit.

Conclusions Our data confirms the conclusions of previous workers that thermal demagnetisation is a valuable tool for determining emplacement temperatures of prehistoric volcaniclastic deposits. Most lithic clasts in ejecta from the Middle Pumice and Cape Riva eruptions have two components of magnetisation, a high-Th component acquired prior to eruption and a low-Tb one acquired after eruption. Emplacement temperatures of single clasts can be determined precisely (_+20°C), but overprinting by chemical ( a n d / o r viscous) remanence, or clast movement during compaction of

Middle Pumice airfall 600

12

~

,,\

\

10

500

E

~8

400

"~'emBla..... cementtem~e--, *" ,~ttJre

~6

~o

300 ~

.........................

E

E

2OO

84

Q~

0~

0%0

2 welded

o

ö

~

not welded

~

~

~

~

~

Distance from vent projected onto dispersal axis (km)

100

Fig. 7. Curves comparing the falloff of average emplacement temperature (dotted line, Fig. 8) and thickness (caldera-wall data of Fig. 2b) of the Middle Pumice airfall away from source vent. Similarity in the shapes of the curves suggests that average accumulation rate (proportional to the thickness of the final deposit) was a major influence on the temperature of the deposit. Sharp inflexions in the curves are exagerated because distance from source vent was calculated by projection of ealdera wall locations onto the airfall dispersal axis


the deposit, can reduce the precision. Emplacement temperature estimates for lithic clasts in the Middle Pumice and Cape Riva co-ignimbrite lithic breccias range widely (250 ° to >580°C). However, they show unambiguously that the breccias were deposited hot, and that TRM can be used to recognise such deposits elsewhere. Study of the Middle Pumice airfall focusses attention on certain difficulties in extracting quantitative temperature information from a prehistoric airfall deposit. Such temperature information might be useful in the thermal modelling of eruptions. Temperatures from clasts in proximal, welded airfall seem physically reasonable, but interpretation of the scatter of temperatures obtained further from the vent, probably attributable to a combination of chemical overprinting, thermal zoning, and variable initial clast temperatures, is not straightforward. A possible falloff of temperature away from vent can only be recognised in the Middle Pumice airfall if some 'average' emplacement temperature is identified. In detail, the thermal histories of individual lithic clasts in volcaniclastic deposits may be quite complex, and further detailed studies are required. Acknowledgements. Field work for this study was financed by a grant from the Royal Society of London. We appreciate the permission of the Greek authorities to undertake field work on Santorini, and thank Rick Hoblitt, Steve Sparks, and Colin Wilson for their helpful reviews of the manuscript.

References Aramaki S, Akimoto S (1957) Temperature estimation of pyroclastic deposits by natural remanent magnetism. Am J Sci 255:619-627 Druitt TH (1983) Explosive volcanism on Santorini, Greece. Ph D thesis, University of Cambridge, pp 269 Druitt TH (1985) Vent evolution and lag breccia formation during the Cape Riva eruption of Santorini, Greece. J Geol 93:439-454 Druitt TH, Sparks RSJ (1982) A proximal ignimbrite breccia

27

facies on Santorini, Greece. J Volcanol Geotl~erm Res 13:147-171 Hoblitt RP, Kellogg KS (1979) Emplacement temperatures of unsorted and unstratified deposits of volcanic rock debris as determined by palaeomagnetic techniques. Geol Soc Am Bull 90:633-642 Hoblitt RP, Reynolds RL, Larson EE (1985) Suitability of nonwelded pyroclastic-flow deposits for studies of magnetic secular variation: A test based on deposits emplaced at Mount St. Helens, Washington, in 1980. Geology 13:242-245 Kent DV, Ninkovich D, Pescatore T, Sparks SRJ (1981) Palaeomagnetic determination of emplacement temperature of Vesuvius AD79 pyroclastic deposits. Nature 290:393-396 Kent JT, Briden JC, Mardia KV (1983) Linear and planar structure in ordered multivariate data as applied to progressive demagnetisation of palaeomagnetic remanence. Geophys J R Astr Soc 62:699-718 McClelland Brown E (1981) Palaeomagnetic estimates of temperatures reached in contract metamorphism. Geology 9:112-116 McClelland Brown E (1982) Discrimination of TRM and CRM by blocking temperature spectrum analysis. Phys Earth Planet Inter 30:405-414 Pichler H, Friedrich WL (1976) Radiocarbon dates of Santorini volcanies. Nature 262:373-374 Rayleigh Lord (1919) On a problem of vibrations, and of random flights in one, two and three dimensions. Phil Mag (6) 37:321-347 Riehle JR (1973) Calculated compaction profiles of rhyolitic ash-flow tuffs. Geol Soc Am Bull 84:2193-2216 Sparks RSJ, Wright JV (1979) Welded air-fall tuffs. Geol Soc Am Special Paper 180:155-166 Walker GPL (1985) Origin of coarse lithic breccias near ignimbrite source vents. J Volcanol Geotherm Res 25:157-171 Wright JV (1978) Remanent magnetism of poorly sorted deposits from the Minoan eruption of Santorini. Bull Volcanol 41:131-135 Wright JV (1981) The Rio Caliente ignimbrite: analysis of a compound intraplinian ignimbrite from a major Quaternary Mexican eruption. Bull Volcanol 44:189-212 Yagi K (1966) Experimental study on pumice and obsidian. Bull Volcanol 29:559-572 Zlotnicki J, Pozzi JP, Boudon G, Moreau MG (1984) A new method for the determination of the setting temperature of pyroclastic deposits (example of Guadeloupe: French West Indies). J Volcanol Geotherm Res 21:297-312

Received April 11, 1988/Accepted June 12, 1988