dynamics of obsidian £ows inferred from microstructures: insights

Dynamics of obsidian £ows inferred from microstructures:insights from microlite preferred orientations

Jonathan Castro a;�, Michael Manga b, Katharine Cashman c

a Department of Geology, 52 W. Lorain St. Oberlin College, Oberlin, OH 44074, USAb Earth and Planetary Science, U.C. Berkeley, 307 McCone Hall, Berkeley, CA 94720-4767, USA

c 1272 Geological Sciences, University of Oregon, Eugene, OR, 94703-1272, USA

Received 5 December 2001; received in revised form 14 February 2002; accepted 21 February 2002

Abstract

The flow of obsidian lava leads to crystal alignments that reflect both the accumulated strain and the type of flowacross the surface. Microlite preferred orientations are used to investigate the emplacement dynamics, strain history,and structural evolution of Obsidian Dome, eastern California. Measurements of three-dimensional microlite trendand plunge in samples from the near-vent region, distal flow, and flow front show: (1) the flow directions along thedome margins, (2) the deformation style (e.g., pure versus simple shear) at the dome margins, and (3) the variation instrain as a function of position within the flow. Microlites form well-developed lineations in the plane of flow bandingin all samples. Stereographic projections indicate that lineations trend normal to the western flow front and plungeshallowly away from the margin. The radial flow pattern indicated by measurements made along the western marginsuggests that extrusion was from a roughly elliptical vent. These results highlight a strong correlation betweenmicrolite trend and the bulk flow direction inferred from the geometry of the flow. Along most of the easternperiphery, lineations trend parallel to the margin and likely reflect the local flow direction as influenced bycompression against the thickening flow crust, marginal talus piles, and topography. Orientation distributions implythat radial spreading accompanied by flattening was the dominant mechanism for flow emplacement. Comparisons ofmeasured orientation distributions with theoretical predictions suggest that microlite fabrics in flow front and near-vent samples developed in a pure shear flow. Microstructures in a sample from near the distal flow base records acomponent of simple shear. Variance in microlite trend provides a measure of the amount of strain acquired duringflow. Standard deviation in trend decreases from the near-vent region to the flow margins, reflecting progressivealignment of microlites during transport. Pure shear strain inferred from orientation distributions increases fromapproximately 0.3 near the vent to about 1.1 at the flow front. The difference between these strains (0.8) is an estimateof the strain associated with flow emplacement. Such strain is similar in size to that estimated from mesoscopicstructures (V1) and for horizontal spreading of a fluid whose volume is equal to that of Obsidian Dome (V1.6). Thissuggests that flow on the surface was sufficient to produce observed microlite preferred orientations in the flow front.These techniques can be applied to interpret older dissected lavas where erosion has erased much of the original flowfront or where larger-scale structures indicative of flow directions are poorly preserved. ß 2002 Elsevier Science B.V.All rights reserved.

0012-821X / 02 / $ ^ see front matter ß 2002 Elsevier Science B.V. All rights reserved.PII: S 0 0 1 2 - 8 2 1 X ( 0 2 ) 0 0 5 5 9 - 9

* Corresponding author. Tel. : +1-440-775-6701; Fax: +1-440-775-8038. E-mail address: [email protected] (J. Castro).

EPSL 6186 26-4-02 Cyaan Magenta Geel Zwart

Earth and Planetary Science Letters 199 (2002) 211^226

www.elsevier.com/locate/epsl

Keywords: obsidian; microlite; emplacement; strain; kinematics

1. Introduction

Flow structures in solidi¢ed lavas record straindeveloped during motion and thus contain impor-tant kinematic information about £ow emplace-ment processes. The emplacement history of alava £ow is de¢ned by its displacement directions,£ow type (i.e., strain regime), and ¢nite strain.These characteristics vary both temporally andspatially, thus an accurate assessment of lava£ow kinematics requires detailed structural anal-ysis at several scales. For example, AMS studies[1] and analyses of crystal preferred orientation[2,3] can be used to determine paleo£ow direc-tions and paleoslopes. Crystal preferred orienta-tions [2,3], folds [4^6] and deformed enclaves [7^9]have been used as rheologic and kinematic indi-cators to infer the £uid dynamics of silicic domeand £ow emplacement. These techniques are par-ticularly important for the analysis of £ows thatwere not observed when active, and thus lack ob-

servational constraints on their emplacement dy-namics and histories.

Silicic lavas (dacites and rhyolites) have beenthe focus of a range of kinematic and dynamicstudies [5,9,10]. For example, the geometry oflarge-scale folds and £ow banding patterns inrhyolite £ows suggests that £ow in the vent re-gions of silicic domes is dominated by coaxialspreading (i.e., pure shear £ow), whereas the dis-tal portions may involve signi¢cant componentsof non-coaxial £ow (i.e., simple shear £ow) [5].Two-dimensional (2D) strain analysis of deformedma¢c enclaves in a dacite £ow [9] corroboratethese inferences and have further shown that therelative proportion of simple shear increases to-wards the base of the £ow while the contributionof pure shear diminishes at depth. Analogue ex-periments [11] are consistent with the results ofthese ¢eld-based studies and such models provideadditional insight into the relations between ¢nitestrain axes (e.g., the strain ellipse) and displace-

Fig. 1. Location map [23] and aerial photo of Obsidian Dome. North is towards the top of the aerial photo. Scale bar in the ae-rial photo is about 0.5 km long.


J. Castro et al. / Earth and Planetary Science Letters 199 (2002) 211^226212

ment directions of lava £ows. While the ¢ndingsof both ¢eld-based and experimental studies haveshed valuable light on the emplacement mecha-nisms of silicic lava £ows, making it possible todocument the dynamics and history of a lava £owthrough relatively straight forward techniques ofmodern structural analysis, Ventura [9] notes that‘‘3D strain analyses of lava £ows (e.g., basalts andrhyolites) are required to assess the general valid-ity’’ of mechanical models of lava £ow emplace-ment. The purpose of this paper is to examine 3Dorientation distributions of acicular microlites inrhyolitic obsidian (OBS) from Obsidian Dome,CA, USA, as a means of deciphering the emplace-ment history of this young, well-preserved lavadome (Fig. 1).

Here we use 3D orientation distributions of mi-crolites in OBS to address three fundamentalcharacteristics of £ow emplacement: the £ow di-rection, £ow regime, and ¢nite strain. First, weshow from stereographic analysis that microlitepreferred orientation fabrics indicate the localand bulk £ow directions along the periphery ofthe dome. Microlite lineations de¢ne a stretchingpattern in the lava that we compare to strain pat-terns produced in analogue experiments [11].Next, we interpret measured orientation distribu-tions in a theoretical framework in order to elu-cidate the probable £ow type (e.g., pure vs. simpleshear). In particular, we show that most of ourmeasured orientation distributions resemble thoseof dominantly pure shear £ow, as predicted bynumerical studies [12^14]. We then analyze thevariance in microlite trend to estimate the strainacquired during £ow. Because ‘‘quantitative esti-mates of the strain accumulated by the magmaduring £ow and data on the partition of deforma-tion are lacking’’ [9], these strain estimates areuseful for determining the relative timing of mi-crolite alignment and the formation of larger-scale£ow structures such as surface folds. Strain esti-mates may also be relevant for understandinghow strain is partitioned on scales ranging frommicroscopic to macroscopic.

Obsidian lavas are well suited for studying £owdynamics because they contain small amounts ofmicrolites, acicular crystals typically less than 10Wm in length (Fig. 2). The characteristically low

volume fractions (0.01^0.1) insure that hydrody-namic interactions between microlites are negli-gible [15,16] and consequently, microlites serveas excellent £ow markers. Moreover, their pris-matic shape and large aspect ratio make thembetter £ow type and strain indicators than moreequant crystals, or those of tabular, cubic, andtransitional shapes [14]. Theoretical studies con-cerning the orientation distribution of rod-shapedcrystals in dilute systems provide a frameworkwith which to compare measurements of crystalorientation [17^19]. For example, Manga [13]showed that 3D microlite orientation distributionsin OBS from Glass Mountain, CA, USA, wheninterpreted within a theoretical context, can beused to constrain £ow properties such as sheartype, strain, and rheology. Obsidian, with itslow crystal contents, is therefore a good matchfor the conditions described in theoretical andcomputational studies. Techniques developedhere may be used to infer strain histories ofboth older (and more poorly preserved) OBS£ows, and samples from volcanic conduits.

2. Background

Obsidian Dome was extruded about 550 yr BP,during the most recent series of rhyolitic eruptionsin eastern California [20]. E¡usive activity was

Fig. 2. Photomicrograph of rod-shaped pyroxene microlitesin OBS. Scale bar is 10 Wm long.


J. Castro et al. / Earth and Planetary Science Letters 199 (2002) 211^226 213

preceded by a series of sub-Plinian eruptions fromthe Obsidian Dome, Glass Creek, and Deadmanvents [20]. This £ow is perhaps the most exten-sively studied OBS £ow on Earth, with abundantdata on the petrologic and textural developmentof the £ow made available through the InyoDomes Research Drilling Project [21^25]. Owingto its youth and the arid climate, surface expo-sures of Obsidian Dome contain a nearly pristinemicroscopic record of textural and structural evo-lution.

The lavas that compose Obsidian Dome aretexturally and structurally heterogeneous. Fink[10] noted that chemically homogeneous lavas ofrhyolitic OBS £ows generally exhibit three textur-al types: coarsely vesicular pumice (CVP), ¢nelyvesicular pumice (FVP), and obsidian (OBS).These lavas di¡er primarily in their vesicularitiesand microcrystallinities, and boundaries betweentypes may be sharp or gradational. These lavasare all present in varying proportions within Ob-sidian Dome. CVP has numerous (s 50%) largevesicles (diameters 1 mm), while FVP has bothlower vesicularity (6 40%) and smaller vesicles(diameter6 1 mm). Obsidian is, by de¢nition,vesicle poor. The three lava types have consistentcontact relations throughout £ows and occur in acharacteristic stratigraphic order (Fig. 3).

Microlites are found in the tephra, indicatingthat they formed prior to the onset of explosiveactivity (Fig. 2) [25]. Microlites in dome lavasconsist of dominantly Ca-poor pyroxene, oxides,and plagioclase feldspar. Flow banding, consistingof alternating microlite-rich and microlite-poorlayers, ranging in thickness from about 0.01^1 mm, is ubiquitous in all samples examined. Pre-vious work on the textures and volatile contentsof lavas from the £ow, conduit, and dike ofObsidian Dome suggests that crystallization be-gan with pre-eruptive growth of phenocrysts andmicrophenocrysts, followed by pre- to syn-erup-tive degassing-induced microlite nucleation andgrowth, and terminating in static crystallizationand devitri¢cation of the slower cooling interiorof the dome and vent [25^27].

Work by Miller [20], Fink [28] and the InyoDomes Research Drilling Project [21] con¢rmedthe presence of a feeder dike beneath the InyoDomes. Sampson [23] noted that the geometryof surface folds may qualitatively re£ect themean transport direction, and, in the case of Ob-sidian Dome, re£ect the initial eruptive style.Based on the elongate north^south morphologyof Obsidian Dome and the geometry of surfacefolds, Sampson [23] suggested that lava eruptedinitially from an elongate vent and eventually be-came localized in a pipe-like conduit during thewaning stages of eruption.

3. Methods

OBS samples were collected from 13 locationsalong the margin of Obsidian Dome (Fig. 4). Atthe £ow front, OBS occurs approximately 10 mbelow the £ow surface, where the lava is exposedin laterally-continuous £ow banded outcrops ap-proximately 10^20 m thick. While the base of theOBS layer is obscured by talus, samples were col-lected from positions that correspond to £owdepths of approximately 15 m. All £ow front sam-ples were oriented by noting the strike and dip of£ow banding, geographic north, and the up-direc-tion on the sample. Samples of near-vent and dis-tal lava were obtained from the Inyo drilling pro-gram (courtesy of T. Vogel), which produced

Fig. 3. Textural variations in Obsidian Dome as revealed by(a) £ow front exposures and (b, c) research drilling. From[22]. See text for abbreviations.



nearly continuous core from the dome, conduit,and dike of the complex (Fig. 4). Conduit coresamples (obtained from drill core RDO-2B) arelargely microcrystalline, owing to the e¡ects ofdevitri¢cation. However, one sample collectedfrom a depth of approximately 10 m is glassy

and microlites are easily discerned from the ma-trix glass. This sample is hereafter referred to asthe ‘near-vent’ sample. Although this core samplewas not oriented (samples rotated when extractedand core orientations were not well constrained),it contains an important record of relative micro-structural development in the near-vent region. Inaddition, a sample from the distal £ow core(RDO-2A) was analyzed from a depth of 47 m.Unlike the other holes, RDO-2A was drilled ver-tically and consequently, orientations of samplesfrom this core are constrained in terms of theirup-directions and orientation with respect to ahorizontal plane. This core sample, hereafter re-ferred to as the ‘distal £ow’ sample, comes from abasal OBS layer with prominent horizontal folia-tions [22]. Thin sections were made in the plane offoliation. Samples from shallower £ow depths incore RDO-2A contain a high concentration ofangular vesicles, many of which disrupt microlitealignments. Such later-stage textures overprintearly £ow-induced fabric and thus preclude anaccurate analysis of strain history. Hence, we con-centrate on the deeper sample in the distal £owcore. With the exception of a 12 cm thick glassymargin, the dike is entirely crystalline (drill holeRDO-3A) [29] and consequently microlites aredi⁄cult to recognize. We do not include analysesof microlites in the dike because the main focus ofthis paper is to address the dynamics of subaerial£ow.

3.1. Measurement techniques

3D microlite orientation was determined ondoubly polished thin sections (30 Wm thick)made with the strike of £ow bands parallel tothe long dimension of the slide and the dip nor-mal to the plane of the slide. Orientation wasmeasured following techniques described in [13]whereby digital reconstructions of microlite size,shape, and orientations are made from a series ofhigh magni¢cation (500U) digital photomicro-graphs (Fig. 5). Images were collected from suc-cessive 2 Wm focal planes within a thin section.The outlines and focal points (i.e., the point thatis most in focus) of each microlite in each photo-micrograph were digitized in a computer-auto-

Fig. 4. Map showing locations of samples collected from theperiphery, distal £ow core RDO-2A, and near-vent (sampleC-1; upper map). Lower diagram provides a perspective viewof the near-vent (2B) and distal £ow core (2A) locations.Modi¢ed from [27].



mated drawing program. Each digitized photomi-crograph was shifted relative to the next in theseries by 2 Wm intervals to produce a stackedarray of images which were geometrically consis-tent with the di¡erent focusing planes in the thinsection. Next, the series of stacked images was

viewed in plan and lines were drawn to connectthe focal point markers on each end of the digi-tized microlite, thus producing a stick representa-tion of the length of each rod-shaped microlite inthe thin section (Fig. 5). Microlite length, width,and orientation were calculated from these digitalreconstructions.

Due to the tedious nature of digital reconstruc-tions, microlite orientation distributions for nineof the 14 samples were determined from directmeasurements on the petrographic microscope.This technique involves measuring apparent mi-crolite length and the di¡erence in vertical posi-tion between microlite tips with an optical micro-meter built into the petrographic scope and thefocusing knob, respectively (Fig. 6). 3D lengthwas calculated using the Pythagorean theorum giv-en the measured values for apparent length and

Fig. 5. Photomicrograph of OBS from the near-vent region(top) and a digital reconstruction of this OBS (bottom; sam-ple C-1). Scale bar is approximately 5 Wm long and the areain the reconstruction equals that in the photomicrograph.Measurement techniques are described in text and [13].

Fig. 6. (Top) Schematic showing how trend angle is deter-mined from comparison of microlite long axis with micro-scope stage goniometer. (Bottom) 3D length (L) and plunge(a) are determined from measurements of microlite apparentlength (A) and vertical di¡erence in microlite tips (H). Focalplanes 1 and 2 are the depths within the thin section wheremicrolite ends are completely in focus.



height di¡erence of each microlite. The plungeof each microlite was calculated trigonometri-cally once the true length (hypotenuse) and ver-tical height di¡erence was constrained. Micro-lite trend was measured with the goiniometer inthe microscope stage. These measurements weremade at 1000U and carry a precision of aboutþ 0.5 Wm.

4. Results

Microlite orientation data are plotted in Fig. 7together with the measured strike and dip of £owbanding at sample locations. Also shown are theaxial traces of buckle folds formed in response tocooling and shortening of the uppermost 10 m ofthe £ow [4]. Surface folds appear to be absent on

Fig. 7. Microlite orientations in £ow front, near-vent (C-1), and distal £ow core (RDO-2A) samples. Stereonets show trend andplunge of microlites. The number of microlites measured in each sample is given in parentheses. Red areas signify highly alignedmicrolites whereas blue demarcates more scattered microlite orientations. Great circles indicate the strike and dip of microlitic£ow banding, and the orientation of £ow banding. The precise orientation of the near-vent sample is unknown, however, £owbanding in the distal £ow core sample is horizontal and thus microlite orientations are constrained in terms of their plunge angleand relative degree of alignment, but not their trend.



the northwestern £ow lobe. Structural relationsalong the £ow front are indicated by the attitudesof £ow banding. Flow banding strikes at highangles to the margin at all but three sample loca-tions and is commonly oriented normal to the£ow front. Flow banding in two samples fromthe northernmost £ow margin and one from theeastern margin strike parallel to the £ow front.Flow banding dips shallowly (6 30‡) in mostsamples. As the orientation of surface folds isroughly parallel to the £ow front, the orientationof £ow banding in most samples is not compatiblewith the trend of large surface folds.

Microlite trend and plunge data for £ow front,distal, and near-vent samples are plotted on lowerhemisphere stereographic projections in Fig. 7.The strike and dip of microlite £ow banding areplotted as great circles. The data are contouredaccording to procedures described in [30,31],with only data of statistical signi¢cance shown,and variability in orientation frequency depictedas a range of colors. Red corresponds to the high-est orientation frequency (s 10%) and dark bluethe lowest (0^1%). The orientation frequency var-iation between these colors is linear. Red areastherefore represent microlite populations that areapproximately 10 times more aligned than micro-lites in the dark blue areas. The number of micro-lites measured, along with size measurements forseveral £ow front samples and the near-vent sam-ple are given in Table 1.

Microlites are well oriented in each of the £owfront samples ; the red clusters correspond to pre-ferred microlite alignments. Microlites with thehighest orientation frequency vary by no more

than 15‡ in trend and 10‡ in plunge in all samples.Less well-de¢ned secondary lineations appear asyellow and green ¢elds (e.g., in OD-3fp, OD-9fp,and OD-11fp). Microlite lineations do not plungesigni¢cantly out of the plane of £ow banding, asevidenced by the correspondence of preferredalignments with great circles. Indeed, microliteplunge in most samples does not exceed 30‡,and thus, these orientations are compatible withthe shallowly dipping character of £ow banding.Microstructures are therefore de¢ned by well-de-veloped lineations oriented within the plane of£ow banding, similar to 1-s fabrics commonlyfound in metamorphic tectonites [32]. Most line-ations trend parallel to the strike of £ow banding.These fabrics are identi¢ed by two red clusters oneither end of the great circle representing £owbanding. In other samples, lineations are inclinedto the strike direction. Lineation rake in thesesamples varies from approximately 10‡ (e.g., inOD-11-98) to 90‡ (e.g., in OD-4fp, OD-12fp,and OD-13-00).

When microlite lineation trend is compared tothe local geometry of the £ow front, two patternsemerge. First, lineations trend at high angles tothe £ow front along the western margin; this pat-tern mimics the orientation distribution shown bythe strike of £ow banding. Trend varies fromsouth^west in the southernmost sample (OD-13-98) to east^west in the western £ow (samples OD-4,9,11-98) to north^west in the northern £ow(OD-3fp and OD-4fp). While most linear fabricsare horizontal, three lineations plunge shallowlyaway from the £ow front. Second, of the ¢vesamples analyzed along the eastern margin of

Table 1Microlite properties in £ow front and near-vent samples measured from digital reconstructions

Sample No. Lengtha Width Aspect ratio Number density Volume fraction(Wm) (Wm) (no./cm3)

OD-4-98 127 8.7 (5.1) 2.1 4.2 1.4U108 0.020OD-7-98 402 6.3 (3.4) 1.5 4.1 2.8U108 0.014OD-9-98 486 5.9 (3.7) 1.9 3.1 3.6U108 0.028OD-11-98 381 6.9 (3.6) 1.7 3.9 4.7U108 0.032OD-13-98 182 9.1 (7.2) 1.7 5.4 2.3U108 0.028C-1b 389 12.1 (4.8) 2.3 5.4 4.6U108 0.090

Length, width, and aspect ratio are mean values.a Values in parentheses are standard deviations of mean length.b Near-vent sample.



the dome, four exhibit £ow front parallel lineationattitudes. All of these lineations are £at-lying toshallowly plunging. A single sample (OD-9fp)along the eastern margin shows a £ow front nor-mal trend. In summary, we ¢nd two preferredmicrolite orientation fabrics, one in which linea-tions trend parallel to the £ow front and one inwhich microlites trend approximately normal tothe £ow front.

Although the sample from the near-vent regionwas not geographically oriented, orientation dataare depicted on a stereonet to show the relativedegree of alignment (Fig. 7). In this sample, mi-crolites in the highest orientation density classtrend over a range of approximately 25‡ andplunge less than 10‡. The geographic orientationof the distal £ow sample is partially constrained inthat the drill hole was vertical and thus £owbanding attitudes (vertical, diagonal, horizontal)could be determined. The distal £ow sample ex-hibits well-developed horizontal foliations. Micro-lite orientations are therefore well-constrained interms of their plunge and relative degree of align-ment within the plane of £ow banding, but theirtrend directions relative to geographic north andoverall £ow geometry are unknown. Microlites inthis sample are well aligned and form two semi-circular regions on opposite edges of the stereonet(Fig. 7). These orientation clusters demarcate well-developed sub-horizontal lineations lying withinthe plane of £ow banding.

5. Discussion

Measured orientation distributions are the inte-grated product of bulk and local £ow kinematicsand re£ect the stretching pattern in various partsof the £ow. Here we discuss the relations betweenmicrolite preferred orientations and the bulk £owdirection inferred from the geometry of the £ow(i.e., dome shape and £ow front orientation).Next, we compare the measured fabrics to stretchtrajectories predicted by recent analogue models[11]. These comparisons may provide constraintson the £ow regime (e.g., pure and simple shear)operating in di¡erent parts of the £ow. In thediscussion that follows, we show that measured

changes in the degree of microlite alignment be-tween a near-vent sample and the £ow front pro-vide a means of quantifying the amount of strain(extension) associated with emplacement. Thesestrain values are useful for developing kinematicmodels [9], understanding larger-scale structuresin the £ow, and potentially for quantifying strainrates when estimates of emplacement time can beindependently assessed. Finally, we discuss appli-cations of these data to analysis of strain featuresof varying scales and criteria necessary for usingmicrolite fabrics to interpret the emplacement his-tories of prehistoric £ows.

5.1. Microlite orientation distribution^emplacement mechanism

Microlite preferred orientations develop during£ow in subsurface regions and during £ow em-placement. Microlite long axes become alignedin the local extension direction [13], which in thecase of a simple shear £ow, coincides with thetransport direction. In the case of a pure shear£ow, the extension direction may be either paral-lel or normal to the direction of lava transport[11]. At Obsidian Dome, the orientation of largesurface fold axes, which develop normal to mean£ow direction [10,33], the underlying topography,and the general shape of the dome indicate thatthe bulk £ow direction is normal to the £ow frontat each location along the margin. The ellipticaldome shape suggests that lava spread in a radialfashion away from a roughly elliptical vent[23,34]. The pattern of microlite lineations alongthe southwestern, western, and northwestern £owmargins records this radial £ow, as lineationstrend at high angles to the £ow front and thetrend angle varies sympathetically with the orien-tation of the £ow front. Sample OD-9fp, locatedon the eastern margin, may also record the bulk£ow direction at that location, as it too contains a£ow front normal lineation.

Along the eastern £ow front, margin-parallelpreferred alignments suggest that while the bulk£ow direction was approximately normal to the£ow front, the local extension direction was par-allel to the £ow front at several locations. Theserelations suggest that margin-normal shortening



may have played an important role in lineationdevelopment along the eastern part of the £ow.Compression of the £ow against a topographicbarrier or talus apron may have in£uenced thelocal extension direction and thus could havecaused earlier-formed lineations with margin-nor-mal alignments to rotate to their present margin-parallel positions. Indeed, surface folds are welldeveloped on the eastern £ow lobe where topog-raphy is slightly elevated immediately outboard ofthe £ow periphery; such folds are poorly devel-oped, if not absent on the western portions of the£ow. Fink [10] documents a similar, althoughlarger-scale deformation pattern on the northwestand northeastern lobes of Little Glass Mountain,CA, where overturned margin-parallel folds devel-oped during £ow advance. The shortening thataccompanies fold formation would be accommo-dated by stretching parallel to the hinge lines ofthese folds, and consequently, microlites wouldalign parallel to fold hinge lines.

Preferred mineral alignments and stretchedbubble fabrics have often been cited to indicatetransport directions [2,35,36]. Our results indicatethat the stretching direction in a 3D £ow maysometimes be perpendicular to the £ow direction.

Further insight into the origin of lineations andthe emplacement mechanism of Obsidian Domecomes from comparing microlite lineation pat-terns to strain patterns produced in experimentalstudies using analogue £uids. Merle [11] showedthat viscous silicone putty spreading radially atthe end of a channel over a solid horizontalbase experiences two shear regimes: (1) domi-nantly simple shear at the base and (2) domi-nantly pure shear at the surface. The local pro-portions of pure and simple shear also vary as afunction of spreading time, or lateral positionwithin the £ow, with the near-vent £ow dominat-ed by pure shear where gravity collapse drivesmotion, and distal £ow experiencing increasingamounts of simple shear. Field-based measure-ments of mesoscopic and macroscopic structures[5,9] con¢rm these experimental ¢ndings. Merle[11] further demonstrates that pure and simpleshear regimes correspond to a concentric stretchtrajectory (extension normal to £ow) and a radialstretch trajectory (extension parallel to £ow), re-

spectively. Concentric stretching is accompaniedby radial horizontal shortening and a componentof vertical shortening due to spreading. Thus, thepure shear regime is a 3D deformation. Concen-tric stretching creates shallow £ow-normal stretch-ing fabrics and surface folds, while radial stretch-ing produces preferred mineral alignments andstretched bubbles parallel to the £ow direction.These stretching regimes are separated by a nar-row zone of zero preferred stretching [11]. Similarstrain patterns have been produced in other ex-perimental studies [37].

Stretching patterns documented by Merle [11]are governed by the local proportions of simpleand pure shear and are thus indicative of a £ow’semplacement mechanism. We see evidence of bothradial and concentric stretching patterns in micro-lite orientation distributions along the ObsidianDome margin. Fabrics exposed along the western£ow front re£ect radial stretching, possibly due tosimple shear kinematics, while most samplesalong the eastern margin may have developed inaccordance with concentric stretching in a domi-nantly pure shear £ow. The presence of bothtypes of stretching fabrics at approximately thesame level (V15 m) within the £ow may re£ectthe proximity of these samples to the sharp tran-sition from concentric to radial stretching com-puted by Merle [11]. According to Merle’s [11]computations, the change from concentric stretch-ing (top) to radial stretching (base) is instantane-ous, without an intermediate stretch orientation.Two samples exhibiting oblique lineations alongthe £ow front (OD-7-98, OD-13-98) are inconsis-tent with these results [11]. The oblique lineationsmay be due to local strain perturbations causedby irregularities in topography or £ow rheology.Sample OD-13-98 is indeed positioned on theedge of a phreatic pit where the southern £owlobe partially ¢lls and obscures this depression.Both samples (OD-7-98, OD-13-98) have weaklydeveloped secondary lineations that trend approx-imately normal to the £ow front (Fig. 7). Weinterpret the secondary microstructures in thesesamples to record part of an integrated strainpath consisting of both early and later £ow-in-duced fabrics.

In summary, the measured stretching patterns



show that, on the scale of the entire £ow, a het-erogeneous deformation ¢eld develops in theuppermost OBS layer. Interaction of the £owwith topography is probably important in modi-fying early formed fabrics. Comparison of fabricsdeveloped in natural lavas to those produced inrelatively simple analogue systems provides a use-ful means of interpreting the £ow emplacementmechanism. Such comparisons may be especiallyhelpful when surface structures such as folds arepoorly developed or di⁄cult to recognize due tobrecciation of the £ow carapace.

5.2. Strain history

Comparison of the variance in microlite orien-tation distributions of near-vent, distal, and £owfront samples provides an estimate of the strainacquired during £ow emplacement. We expectscatter in orientation distributions to decreasewith increasing deformation as ‘‘the degree of pre-ferred orientation along the direction of elonga-tion in the deforming material increases’’ [18].Glassy lavas are particularly well-suited for study-ing lava £ow dynamics because microlite concen-tration is within the dilute limit (volume fractionscI1; Table 1) and hydrodynamic interactionsamong crystals are negligible [13,15,16]. Despitethe low abundance of microlites in these lavas,many of the samples contain abundant micro-phenocrysts and phenocrysts, which disrupt mi-crolite orientations. Indeed, much of the scatterin orientation is due to local dispersion of micro-lite long axes around larger objects. To isolate thee¡ects of £ow on microlite orientation distribu-tions, we have selected for analysis a £ow frontsample (OD-13-00) that contains relatively minoramounts of microphenocrysts and phenocrysts. Inthis sample, microlite orientation fabrics are un-disturbed by £owage around bubbles, pheno-crysts, and microphenocrysts. We also includeanalysis of the distal £ow sample, as it too ap-pears devoid of microphenocrysts and bubblesand thus provides important strain informationfrom the £ow base.

A theoretical framework exists for describingthe behavior of rod-shaped particles in pure andsimple shear £ows [17,18]. The theoretical results

apply to systems that: (1) are su⁄ciently dilutethat mineral grains are widely spaced, (2) exhibitplane strain deformation resulting from £ow, and(3) have Newtonian £ow rheologies. Blanchard etal. [12] and Manga [13] used these formulations tocompute the resultant orientation distributions ofcrystal-bearing magmas £owing, respectively, in adike and on the surface. These studies show thatacicular crystals in a low Reynolds number £owwill develop distinct orientation distributions de-pending on the type of shear involved (e.g., purevs. simple shear £ows) and the amount of shear-ing [14]. Manga [13] characterizes the orientationdistribution of microlites in Medicine Lake (CA)OBS in terms of standard deviations of micro-lite trend (cP) and plunge (ca). He shows thatwhile both of these values approach zero withincreasing strain in a pure shear £ow (i.e., crystalsbecome aligned in the £ow plane and with thedirection of extension), they approach ¢nite con-stants in a simple shear £ow owing to periodicparticle rotations. Most notable are the di¡eren-ces in (ca) between pure and simple shear £ows.Speci¢cally, in a simple shear £ow, ca reaches aminimum value of V20‡ for rod-shaped crystalsof aspect ratio 10 and strains greater than 10. Incontrast, ca is zero at an equivalent strain in apure shear £ow. Distinct orientation distributionsare therefore attributable to the type of shearingand the amount of strain.

The models described above apply to £owsundergoing plane strain. In a natural £ow, thepure shear component may correspond to a 3Ddeformation, as was demonstrated by [11]. Strainestimates derived from comparing measured ori-entation distributions in OBS to those predictedby theory will therefore only yield approximatevalues. The studies nonetheless provide an impor-tant framework for inferring accumulated strainfrom orientation measurements, as models de-scribing orientation distributions resultant from3D £ows are lacking. Fig. 8 shows histogramsof microlite trend (P) and plunge (a) (inset) forthe near-vent, distal, and £ow front samples,from which cP and ca are calculated. These dis-tributions show that (1) microlites are betteraligned in the £ow front and distal £ow than inthe near-vent region, and (2) very good microlite



alignment within the plane of £ow banding is de-veloped in all samples, as evidenced by their nar-row (a) distributions. Analysis of variance indi-cates that cP decreases by about a factor of twobetween the vent (40.2‡) and £ow front (21.8‡),with an intermediate value (24.2‡) in the distal£ow sample. ca shows little variation in thenear-vent and £ow front regions and is small(V6.5‡) in both of these samples. Small ca valuessuggest dominantly pure shear kinematics in thenear-vent and £ow front [13]. In contrast, ca isabout twice as large (12.7‡) in the distal £ow sam-

Fig. 9. Numerically derived curves showing the standard de-viation of microlite trend (cP) versus extensional strain forrod-shaped microlites of aspect ratio 5 in simple (solid) andpure shear £ows (dashed). Intersections of curves with ex-trapolated standard deviations for a near-vent sample (hexa-gons), distal £ow sample (circles), and £ow front sampleOD-13-00 (squares) yield estimates of strain due to £owacross the surface. The di¡erence in pure shear strain be-tween the £ow front and near-vent region (V0.8) is an esti-mate of the strain associated with £ow emplacement.

Fig. 8. Histograms of microlite trend and plunge (insets) forthe near-vent sample (top), distal £ow (middle), and a £owfront sample OD-13-00 (bottom). Trend and plunge are mea-sured in degrees. The total number of microlites measured ineach sample (n) is given in addition to the standard deviationof microlite trend (s.d.). Microlites are better aligned in the£ow front as indicated by the narrower trend distribution.Microlites are well aligned within the plane of £ow bandingin all samples, however, the relatively broader (a) distribu-tion in the distal £ow may re£ect the in£uence of basal shearstress (see text).6



ple. We interpret the higher ca in this sample tore£ect the in£uence of basal shear stress and aninherited component of simple shear strain [13]due to its location near the £ow base.

Fig. 9 shows numerically determined curves ofcP versus extensional strain for microlites of as-pect ratio 5 in dilute pure and simple shear £ows.The calculations assume plane strain, an initiallyrandom orientation distribution, and that the sus-pension consists of non-interacting microlites (i.e.,cI1; valid for dilute suspensions). Given thestrong component of pure shear strain indicatedby map patterns (e.g., Section 5.1) and small ca,we plot cP for near-vent and £ow front sampleson the pure shear curve to estimate strain. Valuesfor the distal £ow sample are also shown for com-parison. Graphical relations indicate strains of ap-proximately 0.3 and 1.1 for near-vent and £owfront samples, respectively. By virtue of their po-sition, microlites in the £ow front traveled fartherthan microlites in the near-vent region, and thusaccumulated more strain than the near-vent sam-ple. Some of the di¡erence in strain values mayalso re£ect the e¡ects of vesiculation near the ventand bubble collapse in the £ow [34,38]. Indeed,the near-vent sample contains a few bubbles (vol-ume fraction of a few percent) ; assessing theamount of bubble collapse in the £ow is di⁄cultand controversial [34,39]. Disruption of previ-ously developed, vent-inherited microlite align-ments by bubble growth and deformation e¡ec-tively resets the orientation distribution in thevent and thus, the di¡erence in strain values(0.8) provides an estimate of extensional strainassociated with £ow emplacement.

It is likely that £ow in Obsidian Dome hadcomponents of both pure and simple shear, andconsequently the estimates of pure shear straininferred from orientation distributions serve as alower bound for the possible strain experiencedduring emplacement. For example, if £ow in thedistal portions of Obsidian Dome were character-ized by simple shear, a strain of about 3.2 (thevalue for the distal £ow sample on the simpleshear curve, Fig. 9) may be a meaningful estimateof the strain experienced during £ow on the sur-face. This assumes that microlites at an equivalentdepth (47 m) in the near-vent region are randomly

oriented. We are cautious about this estimate be-cause we have no constraints on microstructureswithin deep parts of the vent. Nonetheless, ourpure shear strain estimate is of similar magnitudeas the range (1^1.6) determined by Ventura [9] fora dacitic lava £ow of similar dimensions (radiusV1 km). Pure and simple shear strain estimatesin near-vent, distal, and £ow front samples brack-et the strain accrued during emplacement.

Strain estimates are useful for determining therelative timing of microlite alignment and the for-mation of larger-scale structures in OBS £ows.The approximate timing of preferred microliteorientation can be assessed by comparing strain in-ferred from orientation distributions (e.g., Fig. 9)with the total strain experienced during £ow onthe surface. We consider the emplacement of thedome as a homogeneous £uid spreading radiallyunder the in£uence of gravity. We assume £owfront samples experienced a large component ofpure shear strain as a result of lateral spreading ofthe £ow. The strain due to axial spreading of the£ow can be estimated using an average dome ra-dius of 750 m and a mean thickness of 50 m [34].Flow volume (assumed constant) based on thesedimensions is approximately 0.1 km3. The equiv-alent radius of a sphere of equal volume is 290 m.Horizontal spreading of this £uid sphere yields anextensional strain (i.e., pure shear) of about 1.6.This pure shear estimate is similar to the strain ofV1 inferred from Fig. 9, and therefore suggeststhat £ow on the surface was su⁄cient to producethe preferred alignments in the £ow front. Exten-sional strain inferred from orientation distribu-tions is also similar in magnitude to shorteningstrains recorded in mesoscopic folds (usually6 1) and large surface folds that develop during£ow advance. This similarity suggests that exten-sional and shortening strains are compatibleacross a range of scales and that the contrastingstyles of deformation (i.e., extension vs. shorten-ing) may have occurred contemporaneously.

Strain determined from orientation measure-ments may also provide a check on estimates oferuption duration and £ow rate derived fromanalyses of £ow morphology [40]. Fink and Grif-¢ths [40] used observations of overall £ow mor-phology and results from laboratory experiments



to infer emplacement rates and durations for aseries of prehistoric rhyolite and dacite £ows, in-cluding Obsidian Dome. Their simulations sug-gest emplacement times of about 400 days andcorresponding £ow rates of 0.6^11 m3/s for theObsidian Dome eruption. A strain of 0.8 inferredfrom microlite orientations accrued over 400 daysimplies a strain rate of about 2.3U1038 s31. Theaverage volumetric £ow rate of 5 m3/s inferredfrom dome morphology [40] divided by the totaldome volume [40] yields an approximate strainrate of 3U1038 s31. The consistency of thesestrain rates suggests that estimates of £ow rate[40] for the Obsidian Dome eruption are good,as larger strains recorded in the dome would im-ply larger strain rates for a corresponding em-placement time of 400 days.

Our strain estimates can also be used to assessemplacement times of OBS eruptions if indepen-dent estimates of strain rates are available. Cur-rently there are no direct estimates of strain ratefor OBS eruptions, however, promising new mi-crostructural techniques are being developed todetermine shear rate and £ow type in OBS lavas[41].

6. Conclusions

3D microlite orientation distributions measuredin OBS are used to place constraints on the em-placement conditions and strain history of Obsi-dian Dome, CA. These measurements reveal themicroscopic strain pattern (i.e., stretching direc-tions) in the shallow parts of the £ow periphery,and provide a basis for estimating strain associ-ated with £ow emplacement. The results of thisstudy should be applicable to analyses of otherglassy lava domes provided they contain dilutequantities of acicular microlites. The salient ¢nd-ings of this study are summarized below.

1. Microlite lineations are well developed in allsamples. In many samples, lineations re£ect lo-cal stretching directions, which may di¡er frombroad-scale £ow patterns. Thus, paleo£ow di-rections based on orientation distributions of afew samples may be misleading, and we suggest

that several samples be analyzed along a £owperiphery to resolve the bulk £ow direction.Additionally, microstructures should be ana-lyzed in the context of large-scale £ow struc-tures (e.g., surface folds) to elucidate £ow em-placement dynamics and kinematics.

2. Our measurements of small ca, in both near-vent and £ow front samples suggest that £owsurface (upper 15 m) samples record predom-inantly pure shear £ow during emplacement.This ¢nding is consistent with results fromanalog models [11], ¢eld-based studies [5,9],and numerical simulations [14] and suggeststhat £ow emplacement was gravity-driven andcan be modeled as simple viscous spreading[9,11]. Analysis of a distal £ow sample fromnear the base suggests that simple shear maybe more important at depth.

3. Strains of approximately 1 are inferred fromthe orientation distributions of a near-ventand £ow front sample. This estimate re£ectspart of the bulk strain acquired during trans-port of lava from the vent to the £ow margin,and thus is a measure of strain associated withemplacement. This value coincides with strainmeasured in mesoscopic and large-scale surfacefolds [5,6]. It also matches the value for coaxialspreading of a viscous blob of equal volumeand thus suggests that £ow on the surface issu⁄cient to produce microlite alignments. Mi-crolite preferred orientations may thereforeyield valuable insights into paleo£ow directionsand paleoslope attitude of ancient and highlydissected lava £ows.

4. Estimates of £ow emplacement time can becalculated from strain and independent esti-mates of strain rate. To date, there are no pub-lished values for strain rates in OBS £ows. Thishighlights the need for continued study intothe kinematics and microstructural characterof glassy lavas.

Acknowledgements

This work was supported by NSF Grant EAR-9805305. Thomas Vogel generously provided thinsections of core samples. Helpful reviews from



Oliver Merle and an anonymous reviewer greatlyimproved this manuscript.[SK]

References

[1] E. Canon-Tapia, G.P.L. Walker, E. Herrero-Bervera, Theinternal structures of lava £ows-insights from AMS mea-surements I: Near vent aPa, J. Volcanol. Geotherm. Res.70 (1996) 21^36.

[2] D. Shelley, Determining paleo-£ow directions fromgroundmass fabrics in the Lyttelton radial dykes, NewZealand, J. Volcanol. Geotherm. Res. 25 (1985) 69^79.

[3] G. Ventura, R. DeRosa, E. Colletta, R. Mazzuoli, Defor-mation patterns in high viscosity lava £ow inferred fromcrystal preferred orientation and imbrication structures:an example from Salina (Aeolian Island, S. TyrrhenianSea, Italy), Bull. Volcanol. 57 (1996) 555^562.

[4] J.H. Fink, Surface folding and viscosity of rhyolite £ows,Geology 8 (1980) 250^254.

[5] J.V. Smith, H.C. Houston, Folds produced by gravityspreading of a banded rhyolite £ow, J. Volcanol. Geo-therm. Res. 63 (1994) 89^94.

[6] J. Castro, K.V. Cashman, Constraints on rheology ofobsidian lavas based on mesoscopic folds, J. Struct.Geol. 21 (1999) 807^819.

[7] Q. Williams, O.T. Tobish, Microgranitic enclave shapesand magmatic strain histories: constraints from drop de-formation theory, J. Geophys. Res. 99 (1994) 24359^24368.

[8] R. DeRosa, R. Mazzuoli, G. Ventura, Relationships be-tween deformation and mixing processes in lava £ows: acase study from Salina (Aeolian Islands, Tyrrhenian Sea),Bull. Volcanol. 58 (1996) 286^297.

[9] G. Ventura, The strain path and emplacement mechanismof lava £ows: an example from Salina (southern Tyrrhe-nian Sea, Italy), Earth Planet. Sci. Lett. 188 (2001) 229^240.

[10] J.H. Fink, Structure and emplacement of a rhyolitic ob-sidian £ow: Little Glass Mountain, Medicine Lake High-land, northern California, Geol. Soc. Am. Bull. 94 (1983)362^380.

[11] O. Merle, Internal strain within lava £ows from analoguemodelling, J. Volcanol. Geotherm. Res. 81 (1998) 189^206.

[12] J.P.L. Blanchard, P. Boyer, C. Gagny, Un nouveau criterede sens de mise en place dans une caisse ¢lonienne le‘pincement’ des mineraux aux epontes, Tectonophysics53 (1979) 1^25.

[13] M. Manga, Orientation distribution of microlites in obsi-dian, J. Volcanol. Geotherm. Res. 86 (1998) 107^115.

[14] G. Iezzi, G. Ventura, Crystal fabric evolution in lava£ows: results from numerical simulations, Earth PlanetSci. Lett. (in press).

[15] M. Rahnama, D.L. Koch, E.S.G. Shaqfeh, The e¡ect ofhydrodynamic interactions on the orientation distribution

in a ¢ber suspension subject to simple shear £ow, Phys.Fluid. 7 (1995) 487^506.

[16] C.A. Stover, D.L. Koch, C. Cohen, Observations of ¢breorientation in simple shear £ow of semi-dilute suspen-sions, J. Fluid. Mech. 238 (1992) 277^296.

[17] G.B. Je¡rey, The motion of ellipsoidal particles immersedin a viscous £uid, Proc. R. Soc. Lond. A 102 (1922) 161^179.

[18] N.C. Gay, Orientation of mineral lineation along the £owdirection in rocks a discussion, Tectonophysics 3 (1966)559^562.

[19] L.G. Leal, The slow motion of slender rod-like particles ina second-order £uid, J. Fluid. Mech. 69 (1975) 161^179.

[20] C.D. Miller, Holocene eruptions at the Inyo volcanicchain, California: Implications for possible eruptions inLong Valley Caldera, Geology 13 (1985) 14^17.

[21] J.C. Eichelberger, C.D. Miller, L.W. Younker, 1984 drill-ing results at Inyo Domes, California, EOS Trans. AGU66 (1985) 384.

[22] J.H. Fink, C.R. Manley, Origin of pumiceous and glassytextures in rhyolite £ows and domes, in: J.H. Fink (Ed.),The Emplacement of Silicic Domes and Lava Flows,Geol. Soc. Am. Spec. Pap. 212 (1987) pp. 77^89.

[23] D.E. Sampson, Textural heterogeneities and vent areastructures in the 600-year-old lavas of the Inyo volcanicchain, eastern California, Geol. Soc. Am. Spec. Pap. 212(1987) 89^101.

[24] T.A. Vogel, L.W. Younker, B.C. Schuraytz, Constraintson magma ascent, emplacement and eruption: Geochem-ical and mineralogical data from drill-core samples atObsidian Dome, Inyo Chain, California, Geology 15(1987) 405^408.

[25] S.E. Swanson, M.T. Naney, H.R. Westrich, J.C. Eichel-berger, Crystallization history of Obsidian Dome, InyoDomes, California, Bull. Volcanol. 51 (1989) 161^176.

[26] B.E. Taylor, Degassing of Obsidian Dome rhyolite, Inyovolcanic chain, California, in: H.P. Taylor, Jr., J.R.O’Neil, J.R. Kaplan (Eds.), Stable Isotope Geochemistry:A Tribute to Samuel Epstein, Geochem. Soc. (1991)pp. 339^353.

[27] H.R. Westrich, H.W. Stockman, J.C. Eichelberger, De-gassing of rhyolitic magma during ascent and emplace-ment, J. Geophys. Res. 93 (1988) 6503^6511.

[28] J.H. Fink, The geometry of silicic dikes beneath the InyoDomes, California, J. Geophys. Res. 90 (1985) 1127^1133.

[29] T.A. Vogel, J.C. Eichelberger, L.W. Younker, B.C. Schu-raytz, J.P. Horkowitz, H.W. Stockman, H.R. Westrich,Petrology and emplacement dynamics of intrusive andextrusive rhyolites of Obsidian Dome, Inyo Craters Vol-canic Chain, eastern California, J. Geophys. Res. 94(1989) 17937^17956.

[30] W.B. Kamb, Ice petrofabric observations from Blue Gla-cier, Washington, in relation to theory and experiment,J. Geophys. Res. 64 (1959) 1891^1909.

[31] W.B. Kamb, Theory of preferred orientation developedby crystallization under stress, J. Geol. 67 (1959) 153^170.



[32] G. Davis, Structural Geology of Rocks and Regions,John Wiley and Sons, New York (1984) 492 pp.

[33] J.H. Fink, R.C. Fletcher, Ropy pahoehoe: Surface fold-ing of a viscous £uid, J. Volcanol. Geotherm. Res. 4(1978) 151^170.

[34] J.C. Eichelberger, C.R. Carrigan, H.R. Westrich, R.H.Price, Non-explosive silicic volcanism, Nature 323 (1986)589^602.

[35] M.P. Coward, The analysis of £ow pro¢les in a basalticdike using strained vesicles, J. Geol. Soc. Lond. 137(1980) 605^615.

[36] M. Polacci, P. Papale, The evolution of lava £ows fromephemeral vents at Mount Etna: insights from vesicledistribution and morphological studies, J. Volcanol. Geo-therm. Res. 76 (1997) 1^17.

[37] S. Blake, Viscoplastic models of lava domes, in: J.H. Fink

(Ed.), Lava £ows and Domes: Emplacement Mechanismsand Hazard Implications, IAVCEI Proc. Volcanol. 2(1990) pp. 208^243.

[38] M.V. Stasiuk, J. Barclay, M.R. Carroll, C. Jaupart, J.C.Ratte, R.S.J. Sparks, S.R. Tait, Degassing during magmaascent in the Mule Creek vent (USA), Bull. Volcanol. 58(1996) 117^130.

[39] J.H. Fink, S.W. Anderson, C.R. Manley, Textural con-straints on e¡usive silicic volcanism: Beyond the perme-able foam model, J. Geophys. Res. 97 (1992) 9073^9083.

[40] J.H. Fink, R.W. Gri⁄ths, Morphology, eruption rates,and rheology of lava domes: Insights from laboratorymodels, J. Geophys. Res. 103 (1998) 527^545.

[41] A.C. Rust, M. Manga, K.V. Cashman, Determining £owtype and shear rate in magmas from bubble shapes andorientations, EOS Trans. AGU 82 (2001) 47.



dynamics of obsidian £ows inferred from microstructures: insights

Documents