Subaqueous Explosive Eruption and Welding of Pyroclastic Deposits

Peter Kokelaar and Cathy Busby

Silicic tuffs infilling an ancient submarine , at Mineral King in California, show without substantial interaction of the pyro- microscopic fabrics indicative of welding of glass shards and at temperatures clasts and ambient water. >500"C. The occurrence indicatesthat subaqueous explosive eruption and emplacement of pyroclasticmaterialscan occurwithoutsubstantialadmixture of the ambientwater,which Setting and Origin of the would cause chilling. lntracalderaprogressiveaggradation of pumiceand ash from a thick, Vandever Mountain Tuff fast-moving occurred during a short-lived explosive eruption of -26 cubic kilometers of in water 2150 meters deep. The thickness, high velocity, and Triassic and lower Jurassic metavolcanic abundant fine material of the erupted gas-solids mixture prevented substantial incorpo- and metasedimentary rocks of the Mineral ration of ambient water into the flow. Stripping of pyroclasts from upper surfaces of King area (Fig. 1) record early stages of subaqueous pyroclasticflows in general, bothabove the vent and alonganyflow path, may subduction-related magmatism that contin- be the main process giving rise to buoyant-convectivesubaqueous eruption columns and ued throughout Mesozoic time along the attendant fallout deposits. western edge of North America. The strata now occur as a screen between Cretaceous ~lutonsof the Sierra Nevada batholith and dip vertically to form an east-facing homo- Many volcanoes erupt beneath water, but beneath a depositional boundary layer of a cline locallv com~licated bv folds and subaqueous eruptions cannot be observed pyroclastic flow (4) or immediately follow- faults. ~lthdu~hthi rocks have been meta- directly, and their products are difficult to ing emplacement as a result of loading- morphosed at upper greenschist facies, recover and studv. Evidence of submarine compaction (5). Welding of rhyolitic glass- many primary textures and structures are explosive activity comes from sightings of es apparently requires temperatures >550" well preserved and, for brevity, we refer to surface manifestations, such as emergent to 600°C (6).> ,, although" if water diffuses into the rocks as volcanic and sedimentary jets, tsunamis, and water discoloration; ap- hot glass the glass viscosity will be reduced rocks. Present-day stratigraphic thicknesses pearances of extensive rafts of floating pum- and welding may occur at slightly lower (Table 1) are probably about half the orig- ice without subaerial sources; seismic and temperatures (7). inal value (10). acoustic-signal (T-phase) records; and also The possibility of welding of glass frag- Strata at Mineral King record several by inference from the geologic record. Sub- ments under water has long been controver- episodes of volcanic and tectonic activity marine explosive eruptions constitute a sig- sial (7-9), partly because of the supposition nificant hazard and have caused loss of life that cold ambient water becomes thoroughly (1). Volcanic explosivity beneath water admixed with the fragments during flow or must be hydrostatically suppressed (10 m subaaueous eru~tion(9). Theoretical con- \, water = 1 bar), but little is known about sideration (7) suggests that in the case of possible interactions of erupting magma and passage of a subaerial pyroclastic flow into ambient water and the depth and pressure water, formation and expansion of steam at controls on the nature and limits of sub- the flow margins may prevent ingress of the aqueous explosivity. Recent studies along ambient water that would cause cooling. the Izu-Bonin volcanic arc and associated The theory may not be applicable, however, back-arc rifts, south of Japan, indicate that as it assumes that a pyroclastic flow moves as 19" submarine explosivity can occur at depths a more or less coherent plug, whereas many h California of >500 m to perhaps >1800 m (2, 3). move as turbulent or laminar articulate Recognition of hot-state emplacement streams (4). Additionally, there are no of subaqueous pyroclastic deposits is funda- clearlv documented cases of welding- in tuffs Pendant mental to understanding submarine explo- at sites where their upper surfaces unequiv- sive , because the evidence of ocally lay below water at the time of depo- heat retention constrains possible eruption, sition (8). transport, and depositional processes. The In this article we present evidence, from most easily recognized feature indicative of the upper Triassic-lower Jurassic Vandever high emplacement temperature in pyroclas- Mountain tuff at Mineral King in Califor- MINERAL KING tic deposits is welding. Welding is hot-state nia, for welding of ash-flow tuff in a deep viscous deformation of glassy magmatic par- marine environment. Through facies anal- ticles (pumice clasts and shards). It can vsis of the enclosing" strata. we show that occur during emplacement of a tuff in or the site of deposition remained in deep water for a considerable time span before Mesozoic gr&itoids P. Kokelaar is in the Department of Earth Sciences, and after the tuff was em~laced.We infer [Sierra Nevada batholith] University of Liverpool, Liverpool L69 3BX, United that the eruption was subaqueous, and we Kingdom. C. Busby is in the Department of Geological Fig. 1. Location of the Mineral King metavol- Sciences, University of California, Santa Barbara, CA propose a model that accounts for subaque- canic and metasedimentary rocks in the Sierra 93106. ous ash-flow eruption and emplacement Nevada batholith, California.

196 SCIENCE VOL. 257 10 JULY 1992 characterized by submarine caldera-forming (14), and the rocks are inferred to have but not above it, demonstrating that sub- eruptions of rhyolitic pumiceous ash-flow formed in an arc graben-depression trace- sidence occurred on the bounding normal tuff (1l), small-volume eruptions of ande- able for 1000 km from the Sierra Nevada to faults during the eruption, (iii) absence of site (12), and growth of submarine sedi- Sonora, Mexico (15). The Vandever internal flow-unit boundaries or other forms mentary fans and aprons (13). These epi- Mountain tuff is the fill and outflow deposit of tuff stratification. and restriction of sub- sodes alternated with times of quiescence of one of the formed by a subma- aqueous suspension:fallout deposits to the characterized by deposition of prograda- rine explosive eruption of . top, showing that the entire section of tional clastic sequences and carbonate in Evidence for the intracaldera origin of massive tuff accumulated rapidly and con- nearshore to shelf environments and fine- the Vandever Mountain tuff includes: (i) tinuously, (iv) systematic mineralogic vari- grained sediments in relatively deep water abutment of the thick massive tuff against ation stratigraphically upward through the (14). Rates of tectonic subsidence were steep (nondegraded) normal-fault scarps on massive tuff (Table I), indicating that erup- high, averaging -180 m per million years either side, (ii) offset of strata below the tuff tion involved progressive draw-down of a

Table 1. Summary stratigraphy and sedimentological evidence for marine environment of the Vandever Mountain tuff (12, 14, 31). Section is broken into four map units.

Depositionalenvironment and Protolith Sedimentary structures and textures* Basal contacts and thickness* processest

IV. Slate Black shale and siltstone, commonly Shales massive or planar-laminated, Stawed deep-marine basin (23), Thin beds of black shale are pyritiferous,with minor thin-bedded non-bioturbated;sandstone beds show somewhat anoxic, which only interstratifiedwith uppermost several volcanic-lithic sandstone, rare thin Bouma turbidite sequences T,., or T,., occasionally received minor turbidity meters of Vandever Mountain tuff; beds of tuff, and very rare thick beds of and local basal scour; tuffs planar- current incursions and subaqueous minimum stratigraphic thickness tuff breccia. laminated or massive; tuff breccias fallout of ash, and very rare -150 m. massive to normally graded and volcaniclastic debris flows. indistinctly stratified. 111. Vandever Mountain tuff 4. Polymictic matrix-supportedbreccias: Largely disorganized; only minor localized Subaqueous debris-flowdeposition Interstratifiedwith uppermost 5 to 10 m up to 25% volcanic and sedimentary grading or stratification. inferred from: (i) lack of winnowed, of the tuff and overlies it in a horizon rock fragments in a black matrix of very clast-supportedconglomeratesor up to 15 m thick along northern 1.5 fine grained siliceous material other evidence of reworkingtypical of km of the caldera; locally shows representingremobilizedash (it is subaerially emplaced debris-flow loading-induced interpenetrationwith identical to finest tuffs in subaqueous deposits (32) and (ii) loading underlying tuff, on a scale of meters fallout successions at Mineral King); structures at contact with underlying to tens of meters. blocks 1 to 100 m long enclosed. tuffs, which indicate deposition upon water-saturatedash. 3. Distinctly bedded facies: medium- to Medium beds commonly graded, show Medium-beddedtuffs deposited from Gradational contact with indistinctly thin-bedded lapilli tuffs and scour-and-filland rip-up clasts at their turbidity currents; thin-beddedtuffs bedded facies indicated by local coarse-grained tuffs alternate with bases and small-scaleslumps; thin deposited from subaqueous fallout occurrence of thin layers (relatively thin-beddedmedium to very fine beds very well sorted and planar- and dilute sediment-gravityflows indistinct, discontinuous,or grained tuffs. laminated; crystal-rich tuffs are (turbidflows). disrupted) below the contact; 0 to 5 normal-graded. m thick. 2. Indistinctly bedded facies: layers (2 to Sorting moderate to poor; layers laterally Flow unsteadiness during (relatively Gradational contact with underlying 30 cm) of tuff breccia, lapilli tuff, and discontinuous on a scale of meters due low-rate)progressive aggradation or massive facies; 0 to 30 m thick. tuff, with extremely gradational upper to rapid variation in the degree of laminar-shear-induced segregation of and lower contacts. sorting. finer from coarser material along differentiallyflowing layers in a pyroclastic debris flow. 1. Massive facies: rhyolitic welded Monotonouslymassive nonsorted lapilli Monotonous character and absence of Sharp concordant contact with pumiceous ash-flow tuff; up to -50% tuff lacking sedimentary or volcanic stratificationwith systematic underlying sedimentary rocks; latter pumice lapilli,2 to 20% volcanic and interbeds; moderate to intense welding mineralogicvariation indicate are interbedded with small-volume sedimentary lithic lapilli,and 20% fabrics preserved in strain shadows extremely rapid progressive pyroclastic-flowdeposits similar in phenocrysts (normalizedto 0% lithics); around intracaldera megablocks; aggradation during lateral flow; appearance to the Vandever K-feldspar to plagioclase ratio and welding-compaction possibly increased simultaneous shedding of slide Mountain tuff, suggesting conformity; quartz content decrease upward away from megablocks, as latter may megablocks and brecciasfrom underlying unit interpreted as caldera through the deposit: northernmost 2 km have caused some chilling. northerncaldera wall. floor also forms caldera walls and encloses megablocks (up to at least substrate to outflow tuffs and is 100 m long) and mesobrecciasof source of slide megablocks: 200 to underlying breccia-sandstoneat many 480 m thick. stratigraphic levels. 11. Breccia-sandstone Lithic tuffaceous sandstones,tuffaceous Deposits coarsen northward; coarse Apron of debris shed from the scarp of a Presence of similar siltstones several sandstone, polymictic (volcanic and nonstratified,nongraded breccias pass normal fault that acted as a conduit for tens of meters above and below sedimentary) lithic breccias,rhyolitic southward into stratified graded eruptions precursoral to Vandever contact with underlying unit indicative lapilli tuffs, and tuffs. breccias;medium- to thick-bedded, Mountain tuff eruption (11); breccias of concordant gradational contact: up medium- to coarse-grained sandstones are deposits of debris flows and to 200 m thick. with Bouma turbidite sequences T,., high-densityturbidity currents; and erosive loaded bases pass sandstones deposited by turbidity southward into thin- to medium-bedded currents; volcaniclastic deposits from graded planar-laminatedsandstones pyroclastic or debris flows, turbidites, with nonerosive bases; massive and subaqueous fallout. rhyolitic tuff breccias and lapilli tuffs pass southward into thin-beddedlapilli tuffs and planar-laminatedtuffs. I. Calcareous siltstone-limestone Calcareous thin-beddedsiltstones, very Tabular laterally continuous beds with Settling of ash and silt-to-mud-grade Approximately 200 m thick. fine grained sandstones and tuffs, and nonerosive bases: planar-laminated. terrigenous detritus through the water mudstones. column onto a limey substrate, and/or extremely dilute (turbid) flow. *Wave-generatedsedimentary structures are absent throughout. tDeposition was entirely below wave base. +Present-daythickness.

SCIENCE VOL. 257 10 JULY 1992 197 Fig. 2. (A) Back-scatteredelectron image (mozaic) of welded Vandever

Mountain tuff [Q,. quartz. crystals (black); F, feldspar crystals; L, lithic fragments]. Pumice clasts and shards pseudomorphicaliy replaced by lithic fragments and crystals. F~eldof vlew is 5.5 by 6 mm. (6)Higher crypto- to microcrystalline aggregates of quartz and feldspar, with vari- magnification view in plane-polarized light, showing viscous attenuation able small amounts of sericite, appear white. Arrows indicate sites of and pinching and molding of pumice clasts between and around quartz pronounced viscous pinching and molding of pumice clasts between crystals. Field of view is 2 by 2.5 mm. compositionally zoned magma body, and ing progressive aggradation. Horizons of tion of pumice clasts and shards attendant (v) presence in and on top of the tuff of dispersed mesobreccia (clasts <1 m) record with their molding and marked differential megablocks, commonly >I00 m long, de- periodic entrainment and transport of lithic compaction around the rigid crystals and rived from the vicinity of the fault-scarp fragments far into the caldera on or just lithic fraements. margins. If the caldera width in the dip above an aggradation surface. If the ring- ~tteniionhas recently been drawn to direction is the same as the strike ioutcro~) fracture vent was <1 km wide, the 7-km occurrences of flattening of clasts due to width, the intracaldera tuff volume wouid minimum diameter of the caldera requires syn- or postdiagenetic burial-compaction be -13 km3, and the original volume of that the pyroclastic material flowed laterally (18), in which early replacement of volcan- (nonvesiculated) magma erupted can be at least several kilometers (17). ic glass largely by clay minerals (19) was estimated to have been -26 km3. Although followed by collapse of the resultant aggre- a vent for the Vandever Mountain tuff has Evidence for Welding and "eates. In such cases. the flattened clasts not been positively identified, the southern Submarine Environment normally do not show internally a viscous- fault margin shows features [for example, attenuated microscopic vesicular structure dikes and phyllic alteration (1 I)] suggestive A variety of microscopic textural evidence as do welding-compacted pumice clasts, of a vent there, and there is no indication indicates that the Vandever Mountain tuff and individual shards are commonly not of any central vent. is welded (Fig. 2). Abundant magmatic gas preserved. The resultant streaky-textured The systematic mineralogic variation bubbles (vesicles) in undeformed pumice rocks are equivocal in respect of discrimi- with height [Table 1 (11, 16)] indicates clasts commonly form an internal close- nating welding from syn- or postdiagenetic that the massive, poorly sorted intracaldera packed microtubular fabric, but this fabric compaction. Among the best preserved tuff accumulated from the base upward by tends to collapse in welding, and in longi- welding fabrics in ancient rocks are those in steadily progressive aggradation from a per- tudinal section the bubbles form a sinuous which the glass of pumice clasts or shards sistent ~vroclasticstreamrather than from a and locally pinched fibrous texture. Shards has been pseudomorphically replaced by more cGotic accumulation by upwelling or formed by fragmentation of highly vesicular cxyptocrystalline to microcrystalline quartz- intrusion (17). The occurrence of mega- magma typically are polycuspate, but weld- feldspar aggregates as a result of devitrifica- blocks (1 m to tens of meters long) in ing results in flattening of the cusp forms tion [(20);as shown in Fig. 21. It is highly discrete horizons in the tuff, and alignment and development of shard interpenetration unlikely that glass can be replaced by quartz- of the blocks and horizons parallel to paleo- and interleaving as the porosity is reduced. feldspar aggregates (devitrified) and then horizontal, indicates that debris shed from Both of these characteristic welding tex- deformed during diagenesis or metamor- the caldera-fault scarps was emplaced along tures are evident in Fig. 2, which shows phism to produce a viscous-attenuated mi- successive tuff surfaces formed by and dur- pronounced viscous flattening and attenua- croscopic vesicular structure. Also unlikely

198 SCIENCE VOL. 257 10 JULY 1992 is the ~ossibilitvthat these textures can be (Table 1, units I, 11, and IV), show any commonly and largely water vapor). This produced if the' glass is first altered to clay wave-generated sedimentary structures that behavior characterizes the gas-thrust region mineral assemblages that then collapse and would indicate a shallow-marine environ- of an (24), and in sub- are then replaced by quartz and feldspar. ment. By analogy with modem marine sedi- aerial eruptions air must be incorporated The samples shown in Fig. 2 come from a ments, this observation suggests that the and heated here to reduce the bulk density zone of minimal tectonic deformation (strain Vandever Mountain tuff was deposited below of the column so as to permit its further shadow) adjacent and due to a megablock wave base, in water depths certainly greater wholesale ascent by development of buoy- two-thirds of the wav, uo. from the base of the than -60 m and possibly greater than -150 ancy [the "convective region" (24)]. The massive facies of the Vandever Mountain tuff to 200 m. The deeoer deoth is likelv if the mixing of ambient air into the eruption (Table I), toward the north end of the caldera sediments accumulated in a domain facing column, which effectively behaves as a where the tuff is -250 m thick. Although open ocean and subject to long-period waves fluid, is induced by turbulence and by shear mesoscopic fabrics of flattened large pumice (22). Although there is no direct evidence for and acceleration of the column-air contacts clasts occur widely (I I), microscopic welding such conditions at the time of the Vandever (Kelvin-Helmholtz and Rayleigh-Taylor in- textures are not preserved in the pervasively Mountain tuff eruption, the Mineral King stabilities). cleaved tuff that occurs outside the megablock stratigraphic section as a whole indicates that Under water, because of the increased strain shadows. In intracaldera tuffs in gener- such a setting was generally prevalent (14). pressure, volatile expansion is suppressed al, welding is typically less intense around the Additional evidence is that black pyritous relative to subaerial eruptions, and the sup- margins of megablocks (21), owing to the shales and siltstones like those immediately pression increases with water depth. How- cooling effect of the inclusions. Thus original overlying the tuff (>I50 m thick; Table 1) ever, it seems that development and phys- welding fabrics that developed away from the typically, although not exclusively, form in ical processes of gas-thrust and convective blocks, throughout the bulk of the Vandever deep-water environments (>200 m deep) regions may be broadly similar to subaerial ~ountaintuff, may have been more intense (23). To accommodate the 30-m-thick (prox- counterparts in many respects (25). The than those shown in Fig. 2. imal) outflow tuffs beneath wave base, the nature and extent of likely effects due to Evidence of subaerial dissection or fluvial eruption most likely vented in water 230 m interactions of magma and 'ambient water, sedimentation is absent from the entire 5.5- deeper than the minimum required by the both in promoting explosivity and in km-thick (originally -11 km) Triassic-lower lack of wave-generated sedimentary structures quenching the system, are largely un- Jurassic stratigraphic section at Mineral King (see above). known, although from the Vandever (Table 1). The Vandever Mountain tuff is Mountain tuff it would appear that the two underlain and overlain by marine sedimentary Dynamics of Subaqueous substances can remain substantially sepa- rocks with conformable transitional contacts. Explosive Eruptions rate. and characteristics of the uppermost part of Because of suppression of magmatic-vol- the Vandever Mountain tuff also support the In magmatic explosive eruptions, solid and atile expansion, the extent of the gas-thrust interpretation of deposition in a submarine fluid particles are propelled through and region of a subaqueous eruption column environment. Neither the Vandever Moun- above the vent by the expansion to ambient must be reduced relative to its subaerial tain tuff nor its enclosing sedimentary strata pressures of exsolved magmatic gases (most counterpart. Thus the facility for early mix- ing with ambient fluid will be reduced, and capability for development of column buoy- Fig. 3. Schematic dia- air-water interface ancy by magma-water mixing with conse- gram illustrating pro- quent heat exchange and expansion will cesses of a high-mass- also be diminished. With sufficient suppres- discharge subaqueous sion a column that might otherwise have explosive eruption. No develo~edbuovancv mav thus remain rela- fixed relative scales are tively dense and fhrm H laterally moving im~lied. high-concentration particulate flow driven by gravity and by pressure from the vent. However, hydrostatic suppression is not essential for development of vent-fed pyro- clastic flows, as these can form subaerially, for example, as a consequence of large-scale collapse of a magma chamber roof. Cashman and Fiske (25). ,. oresented a model for the evolution of a subaqueous convective eruption column that results from heating and buoyant ascent of water with admixed pyroclasts and produces cold fallout deposits. Our model is complemen- tary and is concerned with erupted material that fails to become incoroorated in a con- vective column and instead flows from the vent to generate a hot laterally moving particulate flow (Fig. 3). Model for the Vandever Mountain tuff. The welding- seen in the Vandever Moun- tain tuff requires that at least part of the deposit, and probably most of it, aggraded at temperatures >500°C. The hot-em-

SCIENCE VOL. 257 10 JULY 1992 placed material cannot have mixed signifi- of the laterally moving flow and will ascend of late-deposited material near the vent. cantly with ambient water. Field and pet- with buoyant fluid. We propose that this Distal and last-deposited fine fallout ash was rographic evidence indicates that the de- process, which we refer to as subaqueous remobilized in debris flows (polymictic ma- posit aggraded progressively from a particu- pyroclastic-flow stripping, is how warm- trix-supported breccias; Table 1). late flow during a sustained high-mass- water buoyant-convective columns arise. It discharge explosive eruption in a submarine is the subaqueous counterpart of the sub- Conclusions environment. aerial processes whereby convective partic- When a subaqueous explosive eruption ulate dis~ersions arise directlv from the We suggest that the controversy over the is sustained at a high mass discharge, only gas-thrust region of an eruption column possible occurrence of subaqueous welding the outermost parts of the gaseous particu- (24) as well as from laterally moving pyro- of tuff has arisen not because it is rare, but late flow (column or laterally directed flow) clastic flows (27), largely as a result of because of the poor preservation potential may be able to mix significantly with ambi- admixture of air with consequent expansion of diagnostic microstructures and the diffi- ent water, especially near the vent. In and by sedimentation of large and dense culty of unequivocal demonstration of the depths where the gas-thrust region of a high particles. Pyroclastic flows invariably pro- paleoenvironment of deposition. Moderate- mass discharge column fails to penetrate the duce some form of overlying buoyant dis- to-large-volume silicic explosive eruptions air-water interface, or only just does so, the persion, which produces fallout deposits mostly occur on continental crust, and eruption must involve a laterally moving (for example, co-ignimbrite ash). Because those in a submarine setting generally mark high-concentration particulate flow that of the small density difference between a regime of considerable crustal extension. originates directlv from the vent. Stream- erupted particles and water, slower upward Such regimes are characterized by high heat lines from the ve'nt will be like those in a velocities of buoyant fluid are required to flow with steep geothermal gradients, in- suppressed fountain, and the flow will tend entrain the particles in subaqueous columns tense magmatism, and, commonly, subse- to assume a bell-like shape (Fig. 3), which than in subaerial columns (25). quent pronounced crustal shortening. will be relatively tall in shallower water and For a given erupted volume, the relative Hence intense hydrothermal alteration [in- flatten out with increasing depth. Given amounts of stripped as opposed to flow- cluding mineralization (20, 29)], intrusion, that in near-vent regions the flow is thick emplaced material, and hence the propor- metamorphism, and deformation are all to and traveling fast (owing to the high mass tion of subaqueous fallout versus flow depos- be expected. All of these tend to result in discharge), and that it is a poorly sorted it, will be greater if the eruption is relatively obliteration of microstructures diagnostic of

"gas-solids disoersion rich in fine material. it lone-livedu so that the surface area of the welding. Furthermore, paleoenvironmental will be impossible for either immediate pyroclastic flow is large relative to the interpretation of tuffs is commonly difficult complete escape of magmatic gases or im- underlying volume of flowing material. in volcanic terranes where fossils may be mediate thorough admixture of the ambient Conversely, a short-lived high mass dis- rare and depositional environments difficult fluid. Because of this, regardless of whether charge eruption will produce a short-lived to compare with standard facies models due the flow is turbulent or laminar. deflation flow with a relatively low proportion of to extremely rapid and episodic deposition and wholesale change from a gaseous to an fallout material. For the Vandever Moun- (30)- aaueous dis~ersioncannot occur near to the tain tuff the limited amount of fallout tuff vent during a sustained high mass discharge (Table 1) could be taken to indicate that REFERENCESANDNOTES subaqueous eruption. In addition, explosive the amount of material originally suspended 1. R. Morimoto, Bull Volcanol. 22, 151 (1960), expansion (flashing to steam) of heated was small and hence that the particulate and J, Ossaka, Bull Earthquake Res. Inst. ambient water in the outer sheath (bound- stream was thick and of short duration 33, 221 (1953). ary layer) of the flow, where mixing with (perhaps hours or days!). However, an un- 2. S. Nagaoka etal., Rep. Hydrogr Res. Marit. Saf. Agency (Jpn.) 27, 145 (1991); R. S. F~ske,person- water does occur, should act to prevent known amount of suspended material is al communication. incorporation of large volumes of water likely to have drifted away. 3. J. Gill etal., Sclence248, 1214 (1990) deep into it. The exclusion of ambient We infer that the Vandever Mountain 4. M. J. Branney and P. Kokelaar, Bull. Volcanol. 54, water is similar to that inferred by Kokelaar tuff aggraded rapidly beneath a thick and in press. 5. R L.Smith, U.S. Geol. Sun/. Prof Pap. 354 (1960), (26) for much smaller scale sustained sub- persistent particulate flow, and, although J. R. Riehle, Geol. Soc. Am. Bull. 84, 2193 (1973). aaueous basaltic eru~tions.Exclusion of material esca~edas outflow to the south. 6. 1 Fr~edmanet al., J. Geophys. Res. 68, 6523 ambient water is only likely to develop most flooded into and was ponded within (1963); K. Yagl, Bull. Volcanol. 29, 559 (1966). 7. R S. J. S~arksetal. J. Volcanol. Geotherm. Res. effectivelv during a steadilv sustained or the activelv subsiding caldera. For the bulk 7 97 (1980) smoothly'waxing-and waning eruption that of the erupted material heat loss was mini- 8 R A F Cas and J V Wr~qhtBull Volcanol 53 allows development of a steadily flowing mal and welding occurred at temperatures 357 (1991). particulate stream. Exclusion of water may >500°C. In consideration of the apparently 9. J. Stix, Earth-SCI Rev. 31, 21 (1991) 10. An average of 50% shortening perpendicular to be less likely (not impossible) if the gas- steady and hot aggradation, combined with cleavage and bedding has been demonstrated thrust region of the eruption column rises likely water depths of r 150 to 200 m, it for similar strata of the same age in the Ritter roof substantially above the water surface such seems unlikely that the eruption column pendant of the Sierra Nevada, 130 km north of M~neralKing, see 0. T. Tob~schand R. S. Fiske, that the pyroclastic flow is largely of mate- ever rose substantially above the water sur- Geol. Soc Am. Bull. 87, 1411 (1976). rial that has fallen back into water, albeit face. 11. C J. Busby-Spera J Geophys. Res. 89, 8417 close to the vent. Toward the end of the eruption, as (1984). 12. , J. Volcanol. Geotherm. Res. 27, 43 Although most of the fast-moving par- explosivity diminished, any form of erup- (1986). ticulate flow cannot interact with ambient tive fountain must have become smaller, 13. , J. Sed. Petrol. 55, 376 (1985). water in near-vent regions, mixing and heat and the flow perhaps became unsteady. We 14. , in Tectonics and SedimentationAlong the exchange between the magma and water California Margin, S. Bachman and J. Crouch, interpret the onset of stratification in the Eds. (Society of Economic Paleontologists and are bound to occur in the outer sheath uppermost tuffs (indistinctly and distinctly Mineralogists, Los Angeles, 1984) pp. 135-1 56. (boundary layer) of the flow, possibly as a bedded facies; Table 1) as recording this 15. C. J. Busby-Spera, Geology 16, 1121 (1988); result of turbulence and also by shear and evolution (28). The bedded tuffs record etal. Geol. Soc.Am. Spec. Pap. 255, 325 (1990). acceleration of the contacts (Fig. 3). Pyro- fallout of flow-stripped material as well as 16. C J. Busby-Spera, thesis, Princeton Un~versity, clasts will be removed from the top margins remobilization of pyroclasts due to slumping Princeton, NJ (1983).

SCIENCE ' VOL. 257 10 JULY 1992 17. Cas and Wright (8) suggested that intracaldera 26. P. Kokelaar, Bull. Volcanol. 48, 275 (1986), figure designed to address the relation of LTP to tuffs do not result from pyroclastic flow and hence 4. A bell-shaped exclusion zone postulated to learning; we have produced mice with mu- are not true pyroclastic flow deposits. They do not occur above the vent was referred to in that case explain the processes they envision, but in this as a "cupola of steam," because the explosivity tations in individual enzymes likely to be situation pyroclastic emplacement without some involves bulk incorporation of water in the eruptive involved in the regulation of candidate form of lateral flow is clearly impossible, even in conduit, so that the gas of the erupted dispersion memory mechanisms, such as LTP. These near-vent locations. is dominated by (nonmagmatic) steam; see P. 18. M. J. Branney and R. S. J. Sparks, J. Geol. Soc. Kokelaar. J. Geol. Soc. London 140. 939 11 983). specific mutations were made with the use London 147, 919 (1990) 27. P. D. ~ohle~et a/., U.S. Geol. Sun/. ~rdfpap. of gene targeting (3). As a first step in this 19. For example, halloysite, smectites, sericite, mont- 1250 (1981), p. 489. program, we report studies on a strain of morillonite, kaolinite, and illite. 28. Compare with R. S. Fiske and T. Matsuda, Am. J. 20. R. L. Allen, Econ. Geol. 83, 1424 (1988). Sci. 262, 76 (1964) mutant mice that do not express the a 21. P. W. Lipman, Geol. Soc. Am. Bull. 87, 1397 29. H. Ohmotoand B.J. Skinner, Econ. Geol. Monogr. isoform of calcium-calmodulin-dependent (1976) 5, 1 (1983). protein kinase type I1 (a-CaMKII). This 22. L. Draper, Marine Geol. 5, 133 (1966); P. D. 30. P. Kokelaar, Geol. Soc. Am. Bull., in press. Komar, J. Sed. Petrol. 44, 169 (1974); B. Butnam 31. C. J. Busby-Spera and J. Saleeby, Geologic enzyme is neural-specific and is present eta/., J. Geophys. Res. 84, 1182 (1979). Guide to the Mineral King Area, Sequoia National presynaptically and appears abundantly ad- 23. See K. T. Pickering et a/., Deep Marine Environ- Park, California (Society of Economic Paleontolo- jacent to the postsynaptic membrane at ments (Unwin Hyman, London, 1989). gists and Mineralogists, Los Angeles, 1987). 24. R. S. J. Sparks, Bull. Volcanol. 48, 3 (1986). 32. T. G. Gloppen and R. J. Steel, Soc. Econ. Pale- synapses that express LTP (4, 5). Calmod- 25. K. V. Cashman and R. S. Fiske, Science 253, 275 ontol. Mineral. Spec. Publ. 31, 49 (1981) ulin (CaM) loaded with calcium (Ca2+) (1991). These authors described a fallout deposit 33. Supported by a NATO collaborative research activates this enzyme and induces its auto- resting above a coeval hot-emplaced (-450°C) grant 910549 and NSF grant EAR9018606 (to phosphorylation. Once autophosphory- flow deposit, which we suggest is most likely to C.B.).We are grateful to G. Lloyd for assistance have formed from the material that would not be with production of the SEM image and to M. J. lated, the CaMKII holoenzyme no longer incorporated into the convective column above Branney, R. V. Fisher, S. Self, and R. S. J. Sparks requires Ca2+ or CaM for activity. This the vent. for reviews of a draft of the article. switch-like mechanism can maintain the enzyme in an active state beyond the dura- tion of the activating Ca2+ signal (6) and has been invoked in learning models (7). Pharmacological experiments have impli- cated this holoenzyme in the induction of LTP (8, 9). Although postsynaptic mecha- Deficient Hippocampal Long-Term nisms seem normal in the CAI hippocam- pal region of these mutant mice, we find Potentiation in a-Calcium- little or no LTP. We thus have a strain of mice that should be suitable for studying the behavior implications of normal synap- Calmodulin Kinase II Mutant Mice tic transmission but deficient LTP. In the accompanying paper we show that these Alcino J. Silva, Charles F. Stevens, Susumu Tonegawa, mutant mice are impaired in performing a Yanyan Wang spatial-learning task (I0). a-CaMKII mutant mice. In order to As a first step in a program to use genetically altered mice in the study of memory produce mice with a mutation in the a- mechanisms, mutant mice were produced that do not express the a-calcium-calmodulin- CaMKII locus, we constructed the plasmid dependent kinase II (a-CaMKII).The a-CaMKII is highly enrichedin postsynaptic densities p23 (Fig. 1A) which contains a 6.1 kilobase of hippocampus and neocortex and may be involved in the regulation of long-term po- (kb) mouse genomic a-CaMKII sequence tentiation (LTP). Such mutant mice exhibited mostly normal behaviors and presented no that is disrupted by insertion of a neomycin- obvious neuroanatomicaldefects. Whole cell recordings reveal that postsynaptic mech- resistance gene (neo) from the plasmid pgk- anisms, including N-methyl-D-aspartate(NMDA) receptor function, are intact. Despite neo (I I). The insertion is within the normal postsynaptic mechanisms, these mice are deficient in their ability to produce LTP a-CaMKII exon encoding most of the reg- and are therefore a suitable model for studying the relation between LTP and learning ulatory domain, and the inserted sequence processes. replaced a 130-bp mouse genomic sequence flanked by a pair of Sph I sites; the expres- sion product of the 130-bp sequence in- cludes the entire inhibitory domain and five Long-term potentiation (LTP) is an elec- observation that pharmacological agents amino acids in the amino end of the cal- trophysiological manifestation of a long- that block hippocampal glutamate receptors modulin-binding domain (12). We trans- lasting increase in the strength of synapses of the N-methyl-D-aspartate (NMDA) class fected the linealized (Fig. 1A) p23 plasmid that have been used appropriately (I). Al- and thus prevent the induction of LTP also into El4 embryonic stem (ES) cells (13) by though LTP has been studied as a mecha- impair spatial learning in rodents (2). The electroporation (13) and isolated 150 (neo) nism responsible for some types of learning problem with this evidence is that blocking colonies, of which two (E14-20 and E14- and memory, the actual evidence for this NMDA receptors disrupts synaptic function 84) were shown by Southern blotting anal- hypothesis is not extensive. The main sup- and thus potentially interferes with the in ysis to harbor a homologously integrated port for LTP as a memory mechanism is the vivo computational ability of hippocampal plasmid. circuits. Perhaps the failure of learning We injected the E14-20 ES cells into results not from the deficit in LTP but C57B116J blastocysts and transferred the A. J. Silva and S. Tonegawa are at the Howard Hughes Medical Institute, Center for Cancer Research and simply from some other incorrect operation blastocysts into pseudo-pregnant mothers. Department of Biology, Massachusetts Institute of of hippocampal circuits that lack NMDA Twelve male chimeric mice were born; they Technology, Cambridge, MA 02139. C. F. Stevens receptor function. were bred with BALBIc females. Hybridiza- and Y. Wang are at the Howard Hughes Medical Institute, Salk Institute, 10010 North Torrey Pines We have adopted a strategy for the study tion blot analysis (Southern) of tail DNA Road, La Jolla, CA 92037. of the mechanisms of mammalian memory, from the offspring of the chimeric males,

SCIENCE VOL. 257 10 JULY 1992