Bull Volcanol (2003) 65:294-305 DOIl 0.1007/s00445-002-026 l-y

E l i z a b e t h M c C l e l l a n d 9 P a t r i c k S. E r w i n

Was a dacite dome implicated in the 9,500 B.P. collapse of Mt Ruapehu? A palaeomagnetic investigation Received: 10 June 2000 / Accepted: 18 November 2002 / Published online: 25 January 2003 9 Springer-Verlag 2003

A b s t r a c t Constraining the process by which volcanoes

become unstable is difficult. Several models have been proposed to explain the driving forces which cause volcanic edifices to catastrophically collapse. These include models for destabilisation of volcanic flanks by wedging due to dyke intrusion and the weakening of mechanical properties by pressurisation of pore fluids. It is not known which, if any, of the models are relevant to particular sector collapse events, Recent developments in the palaeomagnetic estimation of emplacement temperatures of volcaniclastic rocks have shown that even relatively low emplacement temperatures can be recorded by votcaniclastics with high fidelity. We have carried out a palaeomagnetic study of emplacement temperatures to investigate the role of igneous activity in the initiation of the 9,500 B.P. Murimotu sector collapse of Mt Ruapehu, New Zealand. This debris avalanche deposit has three facies which are stratigraphically superimposed, and the lowermost facies contains three lithological assemblages representing different segments of the edifice which were transported with little internal mixing within the flow. We have determined that some of the dacite-bearing assemblage 1, facies I was hot (-350 ~ during transport and emplacement, whereas none of the other lithological assemblages of facies contained hot material. Our interpretation is that a dacite dome was active on the ancient Ruapehu edifice immediately prior to the Murimotu sector collapse. The partially cooled carapace of the dome and material shed from this part was incorporated into the Editorial responsibility: D. Dingwell E. McClelland (~-) Department of Earth Sciences, Oxford University, Parks Road, Oxford, OXI 3PR, UK e-mail: [email protected] Tel.: +44- 1865-272014 Fax: +44-1865-272072 P. S. Erwin Division of Transport, Environment and Taxation, Floor 4/19 Great Minster House, Department of the Environment, Transport and the Regions, 76 Marsham Street, London, SWIP 4DR, UK

avalanche deposit, along with cold lavas and volcaniclastics. We have not found evidence for incorporation o f material at or close to magmatic temperatures, at least in the sampled locations. Our palaeomagnetic work allows us to develop a comprehensive, new palaeomagnetic classification of volcaniclastics. Keywords Sector collapse 9 Mount Ruapehu 9 Volcano instability 9 Emplacement temperature 9 Palaeomagnetism

Introduction Recent advances in the understanding of the evolution of volcanoes has underlined the tact that degradation processes are just as important as growth processes in the lifetime of a volcano. The collapse and eruption of Mt St Helens in May 1980 alerted volcanologists to the possibility that volcanic slopes can fail catastrophically, producing huge avalanches of dry debris. In the Mt St Helens eruption, the slope failure initiated a major eruption when the high-level magma chamber was exposed. Recent work on Hawaii (Moore et al. 1989), Tenerife (Palacios 1994) and other volcanoes has shown that largescale sector collapse of volcanic edifices is a process of wide significance. Very large landslides are possible; the extent of submarine debris-flow deposits around oceanic volcanoes demonstrates that a significant proportion of the total erupted volcanic material is commonly redistributed in this fashion. Large-scale collapse is now recognised as a near-ubiquitous behaviour of large stratovolcanoes on land (e.g. Mt Etna, Sicily; Mt Rainier, Cascades), and of large oceanic shield volcanoes (e.g. Hawaii; Tenerife, Canary Islands; McGuire 1996). Instability of large stratovolcanoes arises, in part, from the inter-layering of mechanically weak and strong materials (pyroclastics and lavas) which may also be further weakened by hydrothermal alteration. The substrate on which the volcano is built is often not flat but tilted by tectonic processes, adding to instability. Insta-

295 bility of large, oceanic shield volcanoes may arise from ment temperature, ranging from cold country rock to rifting associated with frequent dyke emplacement, per- material heated to high temperatures in close proximity to haps associated with erosional effects on the island dykes or magma chambers. Avalanches where the margins due to large, rapid changes in sea level produced collapse was generated by weakening due to overpresby glaciations (McGuire et al. 1997). However, the sured pore fluids should demonstrate a considerably driving forces which initiate such collapses are still poorly smaller range in emplacement temperatures, with the understood, and more information is required before there maximum temperature reflecting the equilibrium poreis any possibility of medium- or long-term prediction of fluid temperature. A combination of the above cases would result in a mixture of some very hot clasts, a catastrophic collapse. Elsworth and Voight (1996) have published phe- concentration of intermediate emplacement temperatures nomenological models for the destabilisation of volcanic representing the pore-fluid temperature, and a range of flanks by the combination of wedging due to dyke temperatures down to ambient. Completely cold deposits intrusion and the weakening of mechanical properties by can be recognised by the lack of magnetic overprint. The thermal remanent magnetisation of a deposit pressurisation of pore fluids. Collapse deposits exposed on Gran Canaria appear to have originated in moderately provides a method of estimating the emplacement temheated rocks (perhaps between 150-250 ~ with abun- perature of a volcaniclastic deposit. Aramaki and Akidant, pressurised pore waters which produced clastic moto (1957) used thermoremanent magnetisation (TRM) dykes (Day 1996). The effects of sector collapse were to qualitatively discriminate between deposits from hot directly observed during the eruption of Mt St Helens in nures ardentes and cold mudflows. The technique is 1980. The slopes of the volcano became oversteepened by described below. This palaeomagnetic technique has beeh the intrusion of a high-level magma chamber, and finally extended to derive semi-quantitative estimates of tempercollapsed in a massive landslide which unroofed the ature by Hoblitt and Kellogg (1979), and refined further magma chamber, triggering a pyroclastic blast and the by Kent et al. (1981) and McClelland and Druitt (1989). subsequent eruption sequence (Voight et al. 1981, 1983). The technique has been used in many subsequent studies Very little juvenile material was incorporated into the of pyroclastics, for example, Tamura et al. (1991) and resulting avalanche deposit, but direct temperature mea- Bardot et al. (1996). Clement et al. (1993) used the surements indicated that much of this deposit was hot technique to estimate the emplacement temperature of (97 ~ and effectively dry, indicating little pore-fluid Nevado de Colimia volcanic debris avalanche deposit, involvement. Sector collapses may have often been demonstrating that the technique is not limited to only pyroclastic rocks. associated with volcanic activity of some kind. In this paper we present the results of a palaeomagThere have been a number of possible mechanisms proposed for the initiation of volcanic debris avalanches. netic emplacement temperature study of the thermal For example, Siebert (1984) defined two main classes of structure of an avalanche deposit on Ruapheu. These eruptions which produce debris avalanches: the Bezymi- results allow us to better constrain the mechanisms anny type which has a magmatic component, and the leading to the collapse of the ancient Ruapehu edifice. Bandai type which is solely phreatic in nature. Other mechanisms which are either non-volcanic or indirectly volcanic may also cause an edifice to become unstable; Geological background these include response to active deformation or to longterm oversteepening, progressive weakening, peripheral The Murimuto Formation is an extensive volcaniclastic erosion or overloading. Field discrimination of deposits unit lying on the lower north-western flanks of Mount generated by these types of collapse is difficult and often Ruapehu, North Island, New Zealand. The formation was impossible. Ui (1983) has discussed the difficulty in originally interpreted as the product of lahars but has identifying hot, dry volcaniclastic debris flows. In this subsequently been interpreted by Hackett and Houghton paper, we use the palaeomagnetic emplacement temper- (1989) and Palmer and Neall (1989) as a debris avalanche ature technique which provides the most reliable means of deposit from the collapse of a sector of an ancient Ruapehu edifice. The age of the Murimoto Formation is differentiation between hot and cold deposits. The thermal structure of ancient avalanche deposits bracketed by the overlying Papakai andesitic tephra (older has hardly been addressed by previous researchers, and is than 5,430_+60 a B.C.) and the underlying tephra of the consequently very poorly constrained. However, infor- Taurewa Formation (ca. 10,500 a B.P.; Donoghue et al. mation on the variation in temperatures in such deposits 1997). The source area for the avalanche deposits is on should allow important constraints to be placed on the the northern flanks of Ruapehu, and the avalanche formation mechanisms. Determination of the range of travelled up to 17 km from source. Hackett and Houghton emplacement temperatures in a deposit should allow (1989) suggest that the source material is from the >120 ka differentiation between heating generated in proximity to Te Herenga Formation, produced in the earliest phase of magma and circulation of warm-to-hot pore fluids. For cone building on Ruapehu. Palmer and Neall (1989) example, debris flows where the major cause of slope suggest that the collapse was initiated by the intrusion of failure was mechanical oversteepening due to magma dykes into a hydrothermally weakened edifice. intrusion are likely to have a wide variation in emplace-

296

Fig. 1 Sketch map of the Murimotu Formation, with sample localities. Ticks on the side of the figure indicate grid numbers from the topographic map. Locality grid references: RPU9, 10 and 11, S 19/266228; RPU 12, S 19/268256; RPU 13, T20/237226 A classic feature of debris avalanche deposits is the presence of "debris avalanche blocks". These blocks will have reasonably internally homogeneous composition, and can be cracked or brecciated with an internal matrix made of crushed fragments of the source material. The blocks can be in sharp contact with blocks and matrix of a different composition. Original source stratigraphy can be preserved within the avalanche. This feature has been interpreted as the result of laminar plug flow (Voight et al. 1981) rather than turbulent flow. Palmer and Neall (1989) formally defined the Murimotu Formation on the northwestern flank of Mount Ruapehu. They interpret the deposit as the product of a debris avalanche from the Ruapehu edifice and recognise three distinct facies (see Fig. 1). Facies 1: This facies contains >50% debris avalanche blocks forming a clast-supported structure in an intrablock matrix. These originally coherent blocks have been completely shattered during transport and consist of lithic fragments which are commonly clast-supported within fine-grained matrix. The blocks are 1.5-36 m in length. They consist of one of three distinct lithologic associations: (1) hornblende dacite in a cream-yellow sandy-muddy matrix; (2) dark grey aphanitic andesite in a matrix of sandy gravel; and (3) plagioclase andesite in a dark grey matrix of gravely sands. The base of the facies 1 unit is exposed only at one location, giving a local thickness of 12 m;

the variation and range of thickness is therefore unknown. Facies 2: This facies contains <30% debris avalanche blocks in heterogeneous, yellow-brown to grey interblock matrix, mostly sandy mud. The maximum ctast size within this facies is 1.5 m. The clasts range from sub-angular to sub-rounded. The facies is well exposed and varies from 2 to 3.5 m in thickness. Facies 3: This facies contains 10-50% clasts of 2 crn or greater, with a maximum clast size of 70 cm. It is hetero-lithologic, containing rocks from all the lithological classifications within facies 1, some of which are hydrothermally altered. This deposit ranges in thickness from 0.7 to 2.1 m. Within this classification, Palmer and Neall (1989) interpret facies 1 and 2 as a primary collapse deposit of an ancient Ruapehu edifice. The lithological associations within facies 1 are thought to represent a preserved stratigraphy. Facies 3 represents laharic deposits associated with the collapse, resulting from at least three lahars. Palmer and Neall (1989) use their observations to infer that the Murimotu collapse was not triggered by an eruption, but by the gravitational collapse of a hydrothermally weakened body of rock on the north-west flank of the ancient Ruapehu edifice following the intrusion of dykes. By contrast, Hackett and Houghton (1989) suggest the intrusion of a shallow dacite cryptodome destabilised the edifice.

Principles of palaeomagnetic emplacement temperature estimation Rocks almost ubiquitously contain grains of iron oxide or sulphide which retain a permanent record of the direction of the Earth's magnetic field imprinted at the time of rock formation or some later, significant geological event. In this paper, we use the fact that magnetism can be reset by heating to gain information about the temperature to which our samples have been heated. Above the Curie temperature (To), the atomic-scale magnetic alignment which gives rise to permanent magnetisation breaks down. For a permanent magnetic record of an ancient Earth field direction to be preserved, there must be a large energy barrier resisting any change of magnetisation direction, and this energy barrier is partially controlled by grain size and is reduced for higher and smaller grain sizes. A particular magnetic grain will retain a stable magnetisation below its blocking temperature (Tb), which is a function of grain mineralogy, size and shape; above the blocking temperature the magnetism is free to rotate, as the kinetic energy exceeds the energy barrier to change. Rocks are generally heterogeneous and thus contain a range of grain sizes and shapes of magnetic particles. This range of size and shape of the magnetic grains gives rise to a blocking temperature spectrum, with individual grains having blocking temperatures ranging from the Curie temperature down to the ambient temper-

297 ature. This property means the remanent magnetisation of a rock potentially carries a record of several directions of magnetisation acquired at different temperatures - in other words, a thermal history. For example, consider some lava which is erupted above the Tc of any crystals it contains, and then cools and solidifies. After initial deposition and cooling to ambient temperature, the rock will have its entire blocking temperature spectrum magnetised parallel to the prevailing magnetic field. The resulting magnetisation is termed a thermoremanent magnetisation (TRM). If part of this rock is subsequently incorporated into a pyroclastic flow and heated to a temperature between ambient and Tc, grains with blocking temperatures less than the reheated temperature will become unblocked and will no longer preserve their magnetisation. When the clast is deposited in a new orientation with respect to the prevailing magnetic field and again cools, the unblocked part of the blocking temperature spectrum will block (remagnetise) in alignment with this new field direction. This newly acquired remanence is termed a partial thermomanent magnetisation (pTRM). The rock will now carry two components of magnetisation: a younger component oriented parallel to the magnetic field at the time of the flow, carried by low blocking temperature grains, and an older, primary component, oriented in a random direction due to the turbulent transport in the pyroclastic flow, carried by those grains with higher blocking temperatures. It is possible for a clast to be reworked several times; as long as each subsequent reworking takes place at a lower temperature than the preceding one, then the clast potentially holds a record of the temperature of each reworking. Hoblitt and Kellogg (1979) introduced a four-fold palaeomagnetic classification for volcaniclastic rocks.

and then measuring the remaining remanence. During the heating, any magnetisation carried by grains with Tb values less than the set temperature is unblocked; on cooling in zero external magnetic field, the magnetisation of these grains then blocks into random orientations yielding a net zero magnetisation. Progressive measurement of the sample's magnetic intensity and direction, after heating to incrementally higher temperatures, yields a dataset to which best-fit vectors may be fitted; from which the separate components can be resolved. The intersection temperature of these vector components is the emplacement temperature of the clast. The blocking and unblocking process is a probabilistic one; at a particular temperature, some grains which unblocked in the natural reheating event will not be unblocked in the laboratory demagnetisation. This can cause a blurring of -10 ~ in the breakpoint between components. However, the method is to a greater extent limited by the practical size of temperature increments used in demagnetisation and with chemical overprinting associated with heating (McClelland Brown 1982; McClelland and Druitt 1989). Furthermore, the kinetic nature of remagnetisation means that the same distribution of grain sizes can be remagnetised either by heating in the Earth's magnetic field to an elevated temperature for a short time (a partial thermoremanent magnetisation, pTRM), or by exposure to the Earth's field at ambient temperature for a long time (a viscous remanent magnetisation, VRM).

-

Ten orientated block samples were taken from each of the three units in facies 1 and each of facies 2 and 3, in all cases from recent, welt-exposed sections. Sampling locations are plotted on Fig. 1. The blocks were oriented by gluing a rigid plastic plate onto the chosen surface, then marking strike and dip onto the plate. The strike was measured using a magnetic compass. Block sampling was chosen in preference to drilling in situ to minimise the possibility of orientation error due to block movement. Once in the laboratory, each block was set in plaster and one-inch cores cut from them; where the clast was large enough, two or more cores were taken.

-

-

-

Type I: all clasts are emplaced above their maximum blocking temperature at the time they come to rest. Clasts carry one component parallel to the geomagnetic field. When remanence directions are measured, they will be closely grouped around the geomagnetic field direction. The maximum blocking temperature recorded is a minimum estimate of the deposit's emplacement temperature. Type II: all clasts are emplaced at the ambient temperature. All remanence is acquired prior to deposition and, owing to rotations of clasts during transport, their measured remanence directions will be randomly oriented. Type III: all clasts are deposited below their maximum blocking temperature but above ambient temperature. Clasts will contain two pTVRM components, as described in the example above. Type IV: the deposit contains clasts with different temperature histories.

Emplacement temperature estimates are determined by

progressive thermal demagnetisation. This procedure is carried out by heating a sample in a laboratory furnace to a given temperature, then cooling it in zero magnetic field

Sampling and measurement Sampling

Palaeomagnetic measurements The total natural remanent magnetisation was measured for each core, using either a Molspin spinner magnetometer where remanence was strong or a CCL 2-axis cryogenic magnetometer if the magnetisation was weak. Thermal demagnetisation was carried out on all samples using furnaces with residual fields <10 nT. Vector structure remanence was analysed using IAPD software of Torsvik, Briden & Smethurst which incorporates the

298 Fig. 2 Palaeomagnetic data

from sites RPU9 (facies 1, assemblage 1), RPUI0 (facies l, assemblage 2), and RUP! 1 (facies 1, assemblage 3): Z plots of representative samples. Open symbols represent the projection of the remanence vector onto the vertical plane; closed symbols are projected onto the horizontal plane. Temperatures are indicated next to chosen data points and are in degrees Celsius. Components A, B and C are indicated on each plot

W, UP

W, UP

RPU9-5A

50 mAre"1

100 mAre"1

250 ~k

1200C ~' 5S0r~F~'~"- 390C

150C~"x'--~ NRM

W, UP

il~ RPU11-9A

1001 B ,550C

NRM

LINEFIND routines of Kent et al. (1983). Bulk susceptibility was measured at each step in order to assess the degree of thermal alteration. Where a second sample from a block was available, this was then analysed with a targeted, stepwise thermal demagnetisation using smaller step intervals over an identified temperature range.

Rock magnetic measurements In order to obtain a representative sample from each block, three to five chips of approximately 1 cm 3 were taken from each block. These were then crushed together to obtain coarse powder. Low field temperature-dependent susceptibility measurements were made on samples from each block, using an Agico CS-2 furnace and KYL-2 kappa bridge. Samples were heated to 700 ~ in air, and cooled back to 40 ~ in a measurement cycle. Based on the thermomagnetic data, three clasts were chosen from each site for further rock magnetic work. Chips of approximately 1 g were taken for 15 of the samples, trying in all cases to sample representative volumes of rock. For each of 15 samples, IRM, acquisition, hysteresis and remanence coercivity determinations were made by means of a Petersen variable field translation balance (VFTB). Coercivity spectra and quantitative values of He, Her, M~ and Mr~ were calculated to complement qualitative interpretation.

Palaeomagnetic and rock magnetic results Palaeomagnetic data The samples have relatively strong, stable magnetisations with a wide range of natural remanent magnetisation

W,!UP mArn'l t 5000

/

s

100

W, UP RPU10-8B

1000 mAm-1

RPU10-1A NRM

B/

RPUg-7A

sosc ~, .M._ 46SCar B

1 NRM

RPU11-10B W, UP

~'k

S

1000 mAm-1

500C

mAmTt 200

:00

NRM

B/ooc

:5ooo 1000

(NRM) intensities in the range 0.1-26 A m -l. Most samples in this study were fully demagnetised after treatment to between 550 and 600 ~

Facies 1: assemblage 1 : R P U 9

NRM intensities are weaker at this site than at other sites sampled; intensities range from 66 to 780 , n A m -I. Thermal demagnetisation of these samples reveals consistent behaviour among the samples. The remanence decreases in intensity throughout the series of thermal treatments (Fig. 2); this is termed a distributed blocking temperature spectrum. The remanence direction changes during this thermal demagnetisation process. Vector analysis of the change in remanence direction during demagnetisation reveals that up to three components of remanence are present. Figure 2 shows a vector plot for sample 9-5A; we refer to these components as A, B and C to represent the lowest-, intermediate- and highesttemperature components respectively. Component A is seen in five clasts; it is removed at 100-150 ~ and is poorly resolved. Component B is seen in all clasts, and is removed up to -318 to -405 ~ Component C is also seen in all clasts, and it persists up to the maximum blocking temperature. The time-averaged geomagnetic field at this location points upwards with an inclination of - 5 9 ~ If a pTRM or VRM is acquired in the deposit, then it will lie parallel to this direction. We compare the grouped remanence directions of each component from each site with this geomagnetic field direction to see if they are statistically identical. The lowest-temperature component is not well resolved. In order to determine its direction, we use the great circle analysis technique of McFadden and McE1hinny (1988). Great circles fitted to the low-temperature

299 Table 1 Directional data and statistics from each component from each site. Balics indicate the key data. At the 95% confidence level, distributions with 3R 2/N<7.81 are random. Asterisk (*) indicates too few samples for reliable statistics to be calculated Site

Component

Dec.

Inc.

N

1295

K

R

3RZ/N

9 9 9 9 9 10 10 11 11

A (VRM) B (all) B (group I) B (group 2) C A (VRM) B A (VRM) B

321.6 349.3 337.7 153.0 191.1 * 215.5 272.4 172,6

-2.6 -66.3 -21.2 -20.6 -15.6 * 11.0 -28.2 --47.3

5 10 5 5 10

* 90.6 38.1 77.8 108.9

1,3 4,98 2.01 1.17

2.81 4.2 3.01 2.3

2.367 10.584 5.436 1.58

10 5 10

43.4 84.1 81.6

2.4 1.8 1.32

5.62 2.77 3.16

9.475 4.603 2.995

11

C

130.7

22.9

5

70.2

2.15

3.14

5.915

12 12 12 12 13 13

A (VRM) B C D A (VRM) B

009.1 337.8 031.2 145.0 009.1 273.0

60.7 20.0 2.2 -22.4 -78.1 -36.0

5 10 3 2 4 8

67.3 84.7 * * 78.9 55.8

2.3 1.29

3.23 3.03

6.259 2.754

2.3 1.9

2.72 4

5.548 6.000

Comment Great circle analysis pole All samples Samples without VRM Samples with VRM Only two point lines

data intersect at (D=321.61, I=-2.6); this intersection point is poorly defined with a large error, but it gives us an estimate of the direction of component A which is subparallel to the geomagnetic field. We interpret A as a viscous remanent magnetisation acquired since deposition. In our analysis of component B, we looked at the grouping of all 10 directions. These yield a random distribution (see Table !). However, we identify two distinct groupings of clasts. Five samples (group 1) do not show a separate A component (e.g. sample 9-7A, Fig. 2), but the B component extends downwards into the temperature range where we would expect to see a VRM (i.e. from 20 to 150 ~ When we analyse component B from these samples from group 1 alone, we find that component B is significantly grouped at the 95% confidence level, and the direction is sub-parallel to the geomagnetic field (Table 1). Where component A is present (group 2; e.g. sample 9-5A, Fig. 2), the distribution of component B is random (Table 1). Component C is present in all samples and is randomly distributed (Table 1).

In some clasts there is a hint of the presence of very: low-temperature component A (e.g. sample 10-1A, Fig. 2) but the discrete blocking temperature spectrum means that VRM would be carried by only a very small fraction of the total remanence.

Facies 1: assemblage 3: R P U l l NRM intensities vary from 0.38 to 5 A m -I at this site. All samples are completely demagnetised by 500-600 ~ Samples with NRM intensities exceeding 2 A m -I have a more discrete blocking temperature spectrum: the lessstrongly magnetised samples have a well-distributed blocking temperature spectrum. Eight clasts from this site have a single-component remanence B (Fig. 2; sample l l-10B), and two clasts have two components of remanence, B and C (Fig. 2; sample 11-9A). Some of the single-component samples show curved demagnetisation trajectories, which may suggest movement during initial cooling. Again, some clasts show evidence of the presence of a small VRM component A. Components A, B and C are all randomly grouped (Table 1).

Facies 1: assemblage 2: RPUIO Facies 2 : R P U 1 2 Clasts from assemblage 2 have the highest range of N R M intensities found in this study (between 26 and 1.2 A m-l). The magnetisation of all samples from this site is dominated by component B which is demagnetised by 600 ~ (Fig. 2). The directions of component B are poorly but significantly grouped with a downward (positive) inclination (Table 1). The seven clasts which have N R M intensities greater than 5 A m -1 show little demagnetisation during temperature treatments up to 400 ~ then 5 0 90% of the remanence is demagnetised in a narrow temperature interval from 500 to 600 ~ (Fig. 2, samples 10-1A, 10-8B); this is termed a discrete blocking temperature spectrum. The other four samples have more distributed blocking temperatures.

Clasts from this site show more complex magnetisation characteristics than those in facies 1. N R M intensities vary between 0.16 and 4 A m -l. Seven clasts have singlecomponent remanence, B (Fig. 3; sample 12-1). One clast has two-component remanence (B and C; sample 12-8), and two have three-component remanence, B, C and D (Fig. 3; sample 12-9). In the three-component sample, the low-temperature component B is removed by 290-320 ~ The intermediate component C is removed by 450-550 ~ whereas the high-temperature component D persists up to 580 ~ In the two-component sample, the low-temperature component B is removed by 380--400 ~ The statistical analysis of the distribution of these directions is

300

RPU12-1A

RPU12-9A

W, UP

W UP

NRM

,f

%~C5ooc s ~-----~ 290C

NRM

~/ mAm" 1 NRM

roAm-1

C

W, UP

RPUla-IB~

B 2()0 N

NRM

W, UP

RPU13-2AB~

450C

% S

1000

mAm-1 1000

150(

Table 2 Hysteresis ratios of selected samples

Sample

Bc~/Bc

MrflM~

RPU9-3 RPU9-8 RPU9- l0 RPU 10- I RPU 10-6 RPUI0-7 RPU11-1 RPU 11-5 RPUI 1-9 RPU 12-1 RPU 12-6 RPU 12-7 RPU13-5 RPU 13-7 RPU 13-9

3.00 2.74 4.47 1.68 1.95 1.76 2.50 2.10 12.6 6.66 2.43 6.48 4.30 1.93 2.96

0.157 0.136 0.119 0.456 0.431 0.441 0.170 0.251 0.084 0.088 0.192 0.177 0.142 0.264 0.094

500

s s60

mAre-1

Fig. 3 Palaeomagnetic data from sites RPUI2 (facies 2) and 13 (facies 3): Z plots of representative samples. Symbols are as in Fig. 2. Component D is indicated for sample RPUI2-9A shown in Table I. We have calculated only the statistics for component B because of the small number of components C and D. Component B is randomly distributed. Some clasts have a poorly resolved VRM component (A).

Facies 3." RUPI3 Here, NRM intensities vary between 0.3 and 3 A m -~. Remanence is completely demagnetised by 550-660 ~ in most samples. However, remanence in clast 13-2 is demagnetised by 220 ~ All clasts have dominant singlecomponent remanence B (Fig. 3; samples 13-1B, 13-2A) which is randomly grouped whereas some clasts have a poorly resolved, low-temperature VRM component A sub-parallel to the Earth's magnetic field (Table 1).

Rock magnetic data

High-temperature susceptibility This technique is used to identify Curie temperatures of magnetic minerals to aid mineral identification. Data were collected from all clasts. Figure 4 gives examples of hightemperature susceptibility data from each site. Clasts from RPU9 show little variation in high-temperature susceptibility behaviour; 80% of samples behave in a similar way to RPU9-1. Samples show little variation in susceptibility until 500 ~ then a reasonably sharp drop occurs at the Curie temperature of magnetite (Fe304). The other sites show more variable behaviour and most clasts contain one or more mineral phases with Tc values of 500-585 ~ Many ctasts also show a high-temperature "tail" reaching,

in some cases, over 700 ~ in all cases this tail contains no discernible transitions but follows an approximately. exponential decay. We can see from the demagnetisation data that the minerals represented by this tail do not carry significant remanence; only two clasts carry more than 5% of their remanence above 600 ~

IRM acquisition, remanence coercivity and hysteresis data This technique is used to identity the presence of magnetic alteration products which commonly have a high coercivity, e.g. hematite and goethite. Three clasts were analysed per site. The remanence of the majority of samples saturated before an applied field of 100 mT was reached, and all samples saturated by 200 mT. The remanence coercivity of the samples ranged from 20 to 70 mT. The hysteresis data show a great deal of variability, with a large range of B~/Bc ratio,~ (Table 2). The data from all sites are consistent with the presence of pseudo-single domain to single domain magnetite with small but variable amounts of higher coercivity material, probably hematite. However, it is clear from the demagnetisation data that this higher coercivity phase does not carry significant remanence.

Interpretation of palaeomagnetic data The palaeomagnetic data allow us to construct a picture of the thermal and movement history of the deposit at each sampling site. The interpretation of low-temperature components is traditionally thought to be difficult. The kinetic nature of remagnetisation means that the same distribution of grain sizes can be remagnetised either by heating in the Earth's magnetic field to an elevated temperature for a short time (a partial thermoremanent magnetisation, pTRM), or by exposure to the Earth's magnetic field at ambient temperature for a long time (a viscous remanent magnetisation, VRM). Pullaiah et al.

301 Fig. 4 High-temperature susceptibility plots of three representative samples from each site, The horizontal axis is temperature in degrees Celsius; the vertical axis is normalised bulk susceptibility. Heating curves are shown by bold lines, cooling by faint lines. Sites RPU9, 10 and 11 contain only minerals with Curie temperatures above 500 ~ Minerals with Curie temperatures of 350 ~ or less are seen only at sites RPU12 and 13

RPU~IO

RPU~3

RPU9-1 1.4

1,6 1,4 1

12~

0"250

, 0

200

400

L

0.8 0.6 0.4 O.

,

600

0,21 0

. 200

, 400

. 600

1 0.8

200

",,=

400

600

0.8 0.6 0.4 0.2

"5

oa

,.0

0.6

,

0.2 1 o

O

0

200

400

,

200

400

,

600

600

12

08 0.6 0.4 0.21 0 0

0.6 0,4

200

400

600

0

0

0.8

i 0.8

0.6

0.6

0.4

0.4

02

0.2

600

400

600

RPU12-7 t8

14 12 1 0.8 0.6 0.4 O2 0 -

0 200

600

RPU13-1

400

600

9

9

200

400

600

RPU13-5

RPU13-2

18 1.6 1.4 ~ 12 1 0.8 0.6 0.4 0,2 0

400

200

RPU12-6 14

400

200

02

RPU12-1

200

0

RPU11-7

16

12

0

0

RPU11-5

0.4

.~

0.2 0

RPU11-1 1.4

600

1,2

0.6 0.4

0

400

RPUI~9

1.4 1.2

1.1]

0"51 0

200

RPUI~2

RPUIO-1 2.~] ~

;~,

0

2.

1 0.8-

~.s I

^

~

06

O. 200

400

600

0.20

200

400

600

200

400

600

Temperature (~

(1975) calculated the kinetic relationship for magnetite and hematite, magnetic minerals which are commonly present in igneous rocks. The longest period of time that a rock can have been held in a constant magnetic field is the length of the present Bruhnes normal magnetic polarity chron (780 ka). The maximum laboratory temperature required to demagnetise this VRM is 163 ~ for magnetisation carried by magnetite. Hence, if one finds a remanence carried by magnetite, with a direction parallel to the Earth's magnetic field which demagnetises at -150 ~ it could either be a VRM acquired over 780 ka

or a pTRM acquired at 150 ~ over a short time period. The theoretical maximum temperature required to demagnetise a VRM, carried by magnetite and acquired over 10 ka (the age of the Murimotu Formation), is 135 ~ Components with higher unblocking temperatures cannot be explained by the presence of VRM. Many clasts carry poorly resolved components A approximately parallel to the geomagnetic field direction; this is removed in the first few demagnetisation steps (between 20 and 150 ~ The presence in some samples of this component, roughly parallel to the Earth's field

302 with unblocking temperatures up to 150 ~ may therefore indicate either a VRM acquired over the last 10,000 years, or that some of the clasts were emplaced at temperatures below - 1 5 0 ~ This imposes a lower limit in the resolution of this technique. In the following discussion of the data we therefore do not consider this component A any further.

behaviour at all in any sample from site RUP9. The rock magnetic data tell us that there is only one magnetic mineral species present at site RUP9 - namely, almost pure magnetite - as there is only one Curie temperature and this is very close to the Curie temperature of pure magnetite (580 ~ Since there is only one magnetic mineral present, the change in direction at 350 ~ must be due to a reheating event, as described above.

Facies 1: assemblage l: RPU9 Facies 1: assemblage 2 : R P U I 0 Hornblende dacite in a cream-yellow sandy-muddy matrix. We find two populations of clasts at this site. All the clasts carry both components B and C; component B is removed in a narrow temperature interval from -318 to -405 ~ In half the clasts (group 2) component B is randomly oriented; component B is statistically indistinguishable from the geomagnetic field in the remaining clasts (group 1). Component C is randomly oriented in all clasts. Our interpretation is that group 1 clasts where component B is sub-parallel to the geomagnetic field were hot (-350 ~ at the time of emplacement, whereas group 2 clasts were at less than -130 ~ at deposition. We are thus unable to distinguish a separate A component from the hot clasts, as A was acquired in these hot clasts in the same direction as the B component. Furthermore, the heat carried by the hot clasts was not sufficient to raise the equilibrium emplacement temperature of assemblage 1 above -130 ~ These group 2 ctasts with random B components will have been reworked at elevated temperature (-350 ~ prior to the collapse event. Is another explanation possible for these data? We have made the interpretation that some of the blocks were hot. on the basis of our observation of a uniformly directed magnetic direction recorded in these blocks, which persists up to about 350 ~ in each case and which we interpret as a thermal magnetic overprint acquired during cooling in the deposit, so that the blocks were at 350 ~ on deposition. This interpretation would be incorrect if the following were true (except that this is not the case). If a change in chemistry had occurred @ e r deposition, forming a new magnetic mineral with a Curie temperature (maximum possible temperature to which remanence can persist) of 350 ~ then this would give us the same remanence picture of a consistent remanent component up to 350 ~ as this new mineral would have been magnetised in the Earth's field when it formed. If this formation occurred after deposition, directions would be the same for all clasts. Possible minerals with the correct Curie temperature include titanomagnetites and hemoilmenites. The critical data which demonstrate that this scenario is not correct are the high-temperature susceptibility data presented in Fig. 4. If a mineral with a Curie temperature of about 350 ~ were present at site RUP9, then this would show up as a sharp decrease in susceptibility at this temperature. Evidence for a titanomagnetite phase is seen in sample RUPI2-7. However, we do not see this

Dark grey aphanitic andesite in a matrix of sandy gravel. Eight clasts from each facies carry a single component of magnetization. The other two clasts carry two components (B and C) and component B is removed in a narrow temperature interval from -250 to -380 ~ Component B is significantly well grouped but lies in a direction (D=215.5; I=+ll.0; a95=43.4) which is statistically. different from the geomagnetic field. Neither of the C components is close to the geomagnetic field direction. This result suggests that the deposit was emplaced at less than -130 ~ and contains two clasts which had been reworked at elevated temperature prior to the Murimotu collapse. The implication of the poor but non-random grouping of component B is not clear, since it is not parallel to the present Earth field and cannot have been acquired in situ. However, this grouping may be a consequence of the whole block having been rolled over, so it is now almost completely inverted by a non-turbulent transport mechanism of the debris avalanche which only produced partial internal randomisation of the avalanche block.

Facies l: assemblage 3: R P U l l Plagioclase andesite in a dark grey matrix of gravely sands. All of the clasts carry a single component of magnetisation (B) which is not systematically grouped. Therefore, these clasts were at less than -130 ~ at the time of emplacement.

Facies 2 : R P U I 2 Less than 30% debris avalanche blocks, heterogeneous yellow-brown to grey inter-block matrix. Seven clasts carry a single component of magnetisation (B). One clast carries two components of magnetisation (B and C). Two clasts carry three components of magnetisation (B, C and D). None of these components show any systematic alignment to the geomagnetic field (Table 1). This facies was therefore deposited at less than - t 3 0 ~ Furthermore, the presence of components C and D shows that some clasts have been reworked at elevated temperature prior to the collapse event in one or more reworking event.

303 Facies 3 : R P U 1 3 Hetero-lithologic, 10-15% clasts of 2 cm or greater. All clasts have a dominant single component of remanence B which is not systematically grouped. Therefore, these clasts were at less than - 1 3 0 ~ at the time of emplacement.

deposit. We do not see any evidence of the incorporation of material at, or close to magnetic temperatures into the Murimotu Formation. It follows from this interpretation that a sizeable fraction of a hot dacite dome was left high on the edifice at the time of collapse. It is likely that this would have generated pyroclastic flow activity and may be a source for the dacite seen in facies 3.

Overview of all sites

New classification

We propose the following scenario to explain this variability in temperatures. In this scenario, a dacite dome was active on the ancient edifice of Ruapehu immediately prior to the 9,500 a B.P. collapse event. This dome shed material during its growth, producing a series of proximal volcaniclastic deposits emplaced at moderate temperatures, high on the ancient edifice. These breccias then cooled to ambient temperatures, and were then later remobilised into the far-travelled avalanche and transported out onto the ring plain (our facies 1, assemblage 1, group 2). During the collapse of the ancient volcano, the still hot dome was shattered and incorporated with the more voluminous but cold volcaniclastics to form facies 1, assemblage l, group 1. The rocks surrounding the dome and its proximal deposits formed the remainder of facies 1 and facies 2. Facies 3 was formed in lahars associated with, and following the main collapse, perhaps reworking some of the collapsed material. The group 1 dacite clasts which were at elevated temperature when emplaced yield emplacement temperatures o f - 3 5 0 ~ Similarly, the emplacement temperature of the group 2 volcaniclastic breccias in their original proximal position was also - 3 5 0 ~ The similarity of these temperatures may indicate that similar processes governed the formation of these two fractions of the deposit. This temperature may reflect the surface temperature of an active dacite dome, suggesting that only the partially cooled carapace of the dome and the material shed from this part were incorporated into the avalanche

The approach of Hoblitt and Kellogg (1979) considers the deposit as a whole, and implicitly assumes that only deposits in which all clasts have the same thermal history can provide useful volcanological information. In the Hoblitt and Kellogg classification, the deposits of the Murimotu Formation are either type II or type IV deposits. Under this classification they would all be interpreted as being emplaced at ambient temperature, as we did not find a systematic low-temperature component parallel to the present Earth field in any sample at any site. By taking a clast-based approach, we have identified multi-component magnetic signatures and consistent temperature ranges over which components are isolated; for example, we have identified early emplacement temperatures for events which occurred prior to the formation of the Murimoto avalanche. Furthermore, this clast-based approach has allowed us to identify two populations of clasts in RPU9, one emplaced at elevated temperatures (group 1), the other at ambient temperature (group 2). Clearly, Hoblitt and Kellogg's classification conveys only part of the information obtainable from palaeomagnetic analysis of volcaniclastic rocks. Any classification imposes simplification on the data and therefore has to be used with care. However, we put forward a new, clastbased classification for this type of palaeomagnetic data (Table 3). Each clast is assigned a type on the criteria of Table 3, where Te is the emplacement temperature, HT is high temperature, MT is medium temperature, and LT is

Table 3 New palaeomagnetic classification of volcaniclastic rocks. TE emplacement temperature, 7", Cmie temperature, H T high temperature, L T low temperature, M T intermediate temperature Type

Description

Interpretation

x al

No resolvable components Clast carries one component orientated along ambient field during cooling Clast carries one component randomly orientated Clast carries two components: HT randomly orientated, LT orientated along ambient field during cooling Clast carries two components: HT randomly orientated, LT randomly orientated Clast carries three components: HT and MT randomly orientated, LT orientated along ambient field during cooling Clast carries three components: HT, MT and LT randomly orientated

Unstable and/or altered magnetic mineralogy Emplaced at or above the Curie temperature (TE>Tc)

a2 fll 32 71 72

al reworked at ambient temperature (TE=ambient temperature) al reworked at an intermediate temperature (Tc>TE>ambient temperature) /31 clast reworked and emplaced at ambient temperature. (TE=ambient temperature) /31 clast reworked as a "secondary" deposit (TEl>TE2>ambient temperature) ?1 clast reworked and emplaced at ambient temperature (TE=ambient temperature)

304 Table 4 Sites classified using new classification (see text for discussion; n . a . not applicable)

Deposit

Classification

aTE (~

fif E (~

yTE (~

Facies Facies Facies Facies Facies

(50% f12, 50% ill) ( 1 0 0 %a2) (80% a2, 20% f12) (70% a2, 10% /32, 20% y2) (100% a 1)

Ambient Ambient Ambient Ambient Ambient

320 n.a. 250 315 n.a.

n.a. n.a. n.a. -500 n.a.

1 LA 1 1 LA 2 1 LA 3 2 3

low temperature. This principle is extendable to larger numbers of components; for example, a ,51-type clast would show four components, the lowest-temperature one being aligned to the geomagnetic field at the time of cooling. The proportions of different clast types are reported to give a deposit description (e.g. (30%/31 70% a l ) . This new classification, despite being more complex than Hoblitt and Kellogg's scheme, allows complicated palaeomagnetic data to be described and classified relatively concisely without losing important information. The data in this paper are presented using this new classification in Table 4.

Conclusions Our results support the interpretation of facies 1 and 2 as debris avalanches, rather than lahar deposits. We also provide constraint on the mechanism of destabilisation which initiated the Murimotu collapse. Hackett and Houghton (1989) suggested that the intrusion of a shallow dacite cryptodome destabilised the edifice in a similar fashion to the 1980 eruption of Mount St Helens. Our evidence indicates that a dome, which broke the surface of the edifice, played some contributory part in the destabilisation process. I. Some of the dacite blocks from assemblage l of facies 1 (Palmer and Neall 1989) were hot (350 ~ at the time of emplacement whereas others were cold. The equilibrium temperature of the whole deposit was no more than 130 ~ This suggests that a dacite dome was active on Ruapehu immediately prior to the collapse which led to the deposition of the Murimotu Formation. This may indeed have contributed to the destabilisation of the ancient edifice. 2. The other two assemblages of facies 1 and the deposits of facies 2 and 3 were cold (or no hotter than - 1 3 0 ~ at the time of emplacement. 3. Some material from facies 1 and 2 shows evidence of one or more phase of reworking at intermediate temperatures prior to the Murimotu collapse. This indicates either pyroclastic or hot debris avalanche activity, which produced volcaniclastic material which has been incorporated into the formation. 4. Material from facies 3 was cold (or no hotter than 130 ~ at the time of emplacement, and shows no evidence of reworking at intermediate temperatures prior to the Murimotu collapse.

5. The process which led to the partial collapse of the ancient Ruapehu edifice did not involve a significant input of heat into the edifice as a whole.

Acknowledgements This work was carried out by E.M. under ROPA grant ROPA/96/98(GR3/R138) and by P.E. under a NERC research studentship, the receipt of which is gratefully acknowledged. Kate Hobson provided assistance in field sampling.

References Aramaki S, Akimoto S (1957) Temperature estimation of pyroclastic deposits by natural remanent magnetization. Am J Sci 255:619-627 Bardot L, Thomas R, McClelland E (1996) Emplacement temperatures of pyroclastic deposits on Santorini deduced from palaeomagnetic mechanisms: constraints on eruption mechanisms. In: Morris A, Tarling DH (eds~ 1996, palaeomagnetism and tectonics of the Mediterranean region. Geol Soc Spec Publ 105:354-357 Clement BM, Connor CB, Graper G (1993) Palaeomagnetic estimate of the emplacement temperature of the long-runout Nevado de Colima volcanic debris avalanche deposit, Mexico. ESPL 120:499-510 Day SJ (1996) Hydrothermal pore fluid pressure and the stability of porous, permeable volcanoes. In: McGuire WJ, Jones AP, Neuburg J (eds) Volcano instability oil the Earth and other planets. Geol Soc Lond Spec Publ 110:77-93 Donoghue SL, Neall VE, Palmer AS, Stewart RB (1997) The volcanic history of Ruapehu during the past 2 millennia based on the record of Tufa Trig tephras. Bull Volcanol 59:136-146 Elsworth D, Voight B (1996) Evaluation of volcano flank instability triggered by dyke intrusion. In: McGuire WJ, Jones AP, Neuburg J (eds) Volcano instability on the Earth and other planets. Geol Soc Lond Spec Publ 110:45-53 Hakett WR, Houghton BF (1989) A facies model for a Quaternary andesitic composite volcano: Ruapehu, New Zealand. Bull Volcanol 51:51-68 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 Kent DV, Ninkovich D, Pescatore T, Sparks SRJ (1981) Palaeomagnetic determination of emplacement temperature of Vesuvius AD 79 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 demagnetization of palaeomagnetic remanence. Geophys J R Astron Soc 62:699-718 McClelland Brown E (1982) Discrimination of TRM and CRM by blocking temperature spectrum analysis. Phys Earth Planet Interiors 30:405--414 McClelland EA, Druitt TH (1989) Palaeomagnetic estimates of emplacement temperatures of pyroclastic deposits on Santorini, Greece. Bull Volcanol 51:16-27 McFadden PL, McElhinny MW (1988). The combined analysis of remagnetization circles and direct observations in palaeomagnetism. Earth Planet Sci Lett 87:161-172

305 McGuire WJ (1996) Volcano instability: a review of contemporary themes. In: McGuire W J, Jones AP, Neuburg J (eds) Volcano instability on the Earth and other planets. Geol Soc Lond Spec Publ 110:1-23 McGuire WJ, Howarth RJ, Firth CR, Solow AR, Pullen AD, Saunders SJ, Stewart IS, Vita-Finzi C (1997) Correlation between rate of sea-level change and frequency of explosive volcanism in the Mediterranean. Nature 389:473-476 Moore JG, Clague D, Holcomb R, Lipman P, Normark WR (1989) Prodigious submarine landslides on the Hawaiian ridge. J Geophys Res 94:17465-17484 Palacious D (1994) The origin of certain wide valleys in the Canary Islands. Geomorphology 9:1-18 Palmer BA, Neall VE (1989) The Murimotu Formation - 9500 year old deposits of a debris avalanche and associated lahars, Mount Ruapehu, North Island, New Zealand. N Z J Geol Geophys 32:477-486 Pullaiah G, Irving E, Buchan KL, Dunlop DJ (1975) Magnetization changes caused by burial and uplift. Earth Planet Sci Lett 28:133-134

Siebert L (1984) Large volcanic debris avalanches: characteristics of source areas, deposits and associated eruptions. J Volcanol Geotherm Res 22:163-197 Tamura Y, Koyama M, Fiske RS (1991) Palaeomagnetic evidence for hot pyroclastic debris flow in the shallow submarine Shirahama Group (Upper Miocene-Pliocene) Japan. J Geophys Res 96 B13:2179-2178 Ui T (1983) Volcanic dry avalanche deposits - identification and comparison with non-volcanic debris stream deposits. In: Aramaki S, Kushiro I (eds) Arc volcanism. J Volcanol Geotherm Res 18:135-150 Voight B, Glicken H, Janda RJ, Douglas PM (1981) Catastrophic rockslide avalanche of May 18th 1980. In: Lipman PW, Mullineaux D (eds) The 1980 eruptions of Mount St. Helens, Washington. US Geol Surv Prof Pap 1250:347-378 Voight B, Janda ILl, Glicken H, Douglas PM (1983) The nature and mechanics of the Mount St. Helens rockslide-avalanche of 18th May 1980. Geotechnique 33:243-273