Construction dynamics of a lava channel

We use a kinematic GPS and laser range finder survey of a 200 m-long section of the Muliwai a Pele lava channel (Mauna Ulu, Kilauea) to examine the construction processes and flow dynamics responsible for the channel–levee structure. The levees comprise three packages. The basal package comprises an 80–150 m wide ′a′a flow in which a ∼2 m deep and ∼11 m wide channel became centred. This is capped by a second package of thin (<45 cm thick) sheets of pahoehoe extending no more than 50 m from the channel. The upper-most package comprises localised ′a′a overflows. The channel itself contains two blockages located 130 m apart and composed of levee chunks veneered with overflow lava. The channel was emplaced over 50 h, spanning 30 May–2 June, 1974, with the flow front arriving at our section (4.4 km from the vent) 8 h after the eruption began. The basal ′a′a flow thickness yields effusion rates of 35 m³ s⁻¹ for the opening phase, with the initial flow advancing across the mapped section at ∼10 m/min. Short-lived overflows of fluid pahoehoe then built the levee cap, increasing the apparent channel depth to 4.8 m. There were at least six pulses at 90–420 m³ s⁻¹, causing overflow of limited extent lasting no more than 5 min. Brim-full flow conditions were thus extremely short-lived. During a dominant period of below-bank flow, flow depth was ∼2 m with an effusion rate of ∼35 m³ s⁻¹, consistent with the mean output rate (obtained from the total flow bulk volume) of 23–54 m³ s⁻¹. During pulses, levee chunks were plucked and floated down channel to form blockages. In a final low effusion rate phase, lava ponded behind the lower blockage to form a syn-channel pond that fed ′a′a overflow. After the end of the eruption the roofed-over pond continued to drain through the lower blockage, causing the roof to founder. Drainage emplaced inflated flows on the channel floor below the lower blockage for a further ∼10 h. The complex processes involved in levee–channel construction of this short-lived case show that care must be taken when using channel dimensions to infer flow dynamics. In our case, the full channel depth is not exposed. Instead the channel floor morphology reflects late stage pond filling and drainage rather than true channel-contained flow. Components of the compound levee relate to different flow regimes operating at different times during the eruption and associated with different effusion rates, flow dynamics and time scales. For example, although high effusion rate, brim-full flow was maintained for a small fraction of the channel lifetime, it emplaced a pile of pahoehoe overflow units that account for 60% of the total levee height. We show how time-varying volume flux is an important parameter in controlling channel construction dynamics. Because the complex history of lava delivery to a channel system is recorded by the final channel morphology, time-varying flow dynamics can be determined from the channel morphology. Developing methods for quantifying detailed flux histories for effusive events from the evidence in outcrop is therefore highly valuable. We here achieve this by using high-resolution spatial data for a channel system at Kilauea. This study not only indicates those physical and dynamic characteristics that are typical for basaltic lava flows on Hawaiian volcanoes, but also a methodology that can be widely applied to effusive basaltic eruptions.     

The morphology and evolution of the Stromboli 2002–2003 lava flow field: an example of a basaltic flow field emplaced on a steep slope

The use of a hand-held thermal camera during the 2002–2003 Stromboli effusive eruption proved essential in tracking the development of flow field structures and in measuring related eruption parameters, such as the number of active vents and flow lengths. The steep underlying slope on which the flow field was emplaced resulted in a characteristic flow field morphology. This comprised a proximal shield, where flow stacking and inflation caused piling up of lava on the relatively flat ground of the vent zone, that fed a medial–distal lava flow field. This zone was characterized by the formation of lava tubes and tumuli forming a complex network of tumuli and flows linked by tubes. Most of the flow field was emplaced on extremely steep slopes and this had two effects. It caused flows to slide, as well as flow, and flow fronts to fail frequently, persistent flow front crumbling resulted in the production of an extensive debris field. Channel-fed flows were also characterized by development of excavated debris levees in this zone (Calvari et al. 2005). Collapse of lava flow fronts and inflation of the upper proximal lava shield made volume calculation very difficult. Comparison of the final field volume with that expecta by integrating the lava effusion rates through time suggests a loss of ~70% erupted lava by flow front crumbling and accumulation as debris flows below sea level. Derived relationships between effusion rate, flow length, and number of active vents showed systematic and correlated variations with time where spreading of volume between numerous flows caused an otherwise good correlation between effusion rate, flow length to break down. Observations collected during this eruption are useful in helping to understand lava flow processes on steep slopes, as well as in interpreting old lava–debris sequences found in other steep-sided volcanoes subject to effusive activity.     

Viscous Newtonian laminar flow in a rectangular channel: application to Etna lava flows

 We introduce a 3D model for near-vent channelized lava flows. We assume the lava to be an isothermal Newtonian liquid flowing in a rectangular channel down a constant slope. The flow velocity is calculated with an analytical steady-state solution of the Navier-Stokes equation. The surface velocity and the flow rate are calculated as functions of the flow thickness for different flow widths, and the results are compared with those of a 2D model. For typical Etna lava flow parameters, the influence of levees on the flow dynamics is significant when the flow width is less than 25 m. The model predicts the volume flow rate corresponding to the surface velocity, taking into account that both depend on flow thickness. The effusion rate is a critical parameter to evaluate lava flow hazard. We propose a model to calculate the effusion rate given the lava flow width, the topograhic slope, the lava density, the surface flow velocity, and either the lava viscosity or the flow thickness. Received: 20 January 1998 / Accepted: 8 January 1999     

Simulations of the 2004 lava flow at Etna volcano using the magflow cellular automata model

Since the mechanical properties of lava change over time, lava flows represent a challenge for physically based modeling. This change is ruled by a temperature field which needs to be modeled. MAGFLOW Cellular Automata (CA) model was developed for physically based simulations of lava flows in near real-time. We introduced an algorithm based on the Monte Carlo approach to solve the anisotropic problem. As transition rule of CA, a steady-state solution of Navier-Stokes equations was adopted in the case of isothermal laminar pressure-driven Bingham fluid. For the cooling mechanism, we consider only the radiative heat loss from the surface of the flow and the change of the temperature due to mixture of lavas between cells with different temperatures. The model was applied to reproduce a real lava flow that occurred during the 2004–2005 Etna eruption. The simulations were computed using three different empirical relationships between viscosity and temperature.     

A laboratory model of surface crust formation and disruption on lava flows through non-uniform channels

Crust formation on basaltic lava flows dictates conditions of both flow cooling and emplacement. For this reason, flow histories are dramatically different depending on whether lava is transported through enclosed lava tubes or through open channels. Recent analog experiments in straight uniform channels (Griffiths et al. J Fluid Mech 496:33–62, 2003) have demonstrated that tube flow, dictated by a stationary surface crust, can be distinguished from a mobile crust regime, where a central solid crust is separated from channel walls by crust-free shear zones, by a simple dimensionless parameter ϑ, such that ϑ<25 produces tube flow and ϑ>25 describes the mobile crust regime. ϑ combines a previously determined parameter ψ, which describes the balance between the formation rate of surface solid and the shear strain that disrupts the solid crust, with the effects of thermal convection (described by the Rayleigh number Ra).Here we explore ways in which ϑ can be used to describe the behavior of basaltic lava channels. To do this we have extended the experimental approach to examine the effects of channel irregularities (expansions, contractions, sinuosity, and bottom roughness) on crust formation and disruption. We find that such changes affect local flow behavior and can thus change channel values of ϑ. For example, gradual widening of a channel results in a decrease in flow velocity that causes a decrease in ϑ and may allow a down-flow transition from the mobile crust to the tube regime. In contrast, narrowing of the channel causes an increase in flow velocity (increasing ϑ), thus inhibiting tube formation.We also quantify the fraction of surface covered by crust in the mobile crust regime. In shallow channels, variations in crust width (d _c) with channel width (W) are predicted to follow d _c∼W ^5/3. Analysis of channelized lava flows in Hawaii shows crustal coverage consistent with this theoretical result along gradually widening or narrowing channel reaches. An additional control on crustal coverage in both laboratory and basaltic flows is disruption of surface crust because of flow acceleration through constrictions, around bends, and over breaks in slope. Crustal breakage increases local rates of cooling and may cause local blockage of the channel, if crusts rotate and jam in narrow channel reaches. Together these observations illustrate the importance of both flow conditions and channel geometry on surface crust development and thus, by extension, on rates and mechanisms of flow cooling. Moreover, we note that this type of analysis could be easily extended through combined use of FLIR and LiDAR imaging to measure crustal coverage and channel geometry directly.Editorial responsibility: A. Harris     

Dynamics of a confined lava flow on Kilauea volcano,Hawaii

This paper presents a new method of analysing lava flow deposits which allows the velocity, discharge rate and rheological properties of channelled moving lavas to be calculated. The theory is applied to a lava flow which was erupted on Kilauea in July 1974. This flow came from a line of fissures on the edge of the caldera and was confined to a pre-existing gully within 50 m of leaving the vent. The lava drained onto the floor of the caldera when the activity stopped, but left wall and floor deposits which showed that the lava banked up as it flowed around each of the bends. Field surveys established the radius of curvature of each bend and the associated lava levels, and these data, together with related field and laboratory measurements, are used to study the rheology of the lava. The results show the flow to have been fast moving but still laminar, with a mean velocity of just over 8 m s^–1; the lava had a low or negligible yield strength and viscosities in the range 85–140 Pa s. An extension of the basic method is considered, and the possibility of supercritical flow discussed.     

基于热流变运动学模型的长白山熔岩流数值模拟

热流变运动学模型是熔岩流复杂流动过程模拟的一种方法,其通过热量系统和地形的计算来模拟熔岩流的流动速度、温度变化和流动距离等.本文以气象站期碱流岩的野外调查数据为依据,应用该模型对其进行模拟研究.通过对模拟结果的分析发现,速度变化趋势与地形坡度变化趋势一致,但随着流动过程中黏度的增大,地形影响作用逐渐减小.在热量系统中,辐射热和微晶结晶热分别为熔岩流流动中最多的热损失和热增加.气象站期碱流岩模拟计算的最终流动距离为5.17 km,这与实测约5.4 km的长度相近.模拟计算结果与观测研究的认识和地质资料相吻合,说明了模拟方法的可行性和结果的可信性.此项工作的开展可为熔岩流的地质研究提供借鉴,为火山区熔岩流灾害区划提供新的思路和方法.     

Subaqueous, basaltic lava dome and carapace breccia on King George Island, South Shetland Islands, Antarctica

 On King George Island during latest Oligocene/earliest Miocene time, submarine eruptions resulted in the emplacement of a small (ca. 500 m estimated original diameter) basalt lava dome at Low Head. The dome contains a central mass of columnar rock enveloped by fractured basalt and basalt breccia. The breccia is crystalline and is a joint-block deposit (lithic orthobreccia) interpreted as an unusually thick dome carapace breccia cogenetic with the columnar rock. It was formed in situ by a combination of intense dilation, fracturing and shattering caused by natural hydrofracturing during initial dome effusion and subsequent endogenous emplacement of further basalt melt, now preserved as the columnar rock. Muddy matrix with dispersed hyaloclastite and microfossils fills fractures and diffuse patches in part of the fractured basalt and breccia lithofacies. The sparse glass-rich clasts formed by cooling-contraction granulation during interaction between chilled basalt crust and surrounding water. Together with muddy sediment, they were injected into the dome by hydrofracturing, local steam fluidisation and likely explosive bulk interaction. The basalt lava was highly crystallised and degassed prior to extrusion. Together with a low effusion temperature and rapid convective heat loss in a submarine setting, these properties significantly affected the magma rheology (increased the viscosity and shear strength) and influenced the final dome-like form of the extrusion. Conversely, high heat retention was favoured by the degassed state of the magma (minimal undercooling), a thick breccia carapace and viscous shear heating, which helped to sustain magmatic (eruption) temperatures and enhanced the mobility of the flow. Received: 1 August 1996 / Accepted: 15 September 1997     

Internal differentiation of Kutsugata lava flow from Rishiri Volcano,Japan: Processes and timescales of segregation structures' formation

Internal differentiation processes in a solidifying lava flow were investigated for the Kutsugata lava flow from Rishiri Volcano in northern Japan. In a representative 6-m thick lava flow that was investigated in detail in this study, segregation products darker than the host lavas manifested mainly in the form of pipes (vesicle cylinders) and layers (vesicle sheets), occurring around 0.5–2.3 m and 2.0–4.0 m above the base, respectively. Both the cylinders and sheets are significantly richer in incompatible elements such as TiO₂ and K₂O than the host lavas, which suggest that these products essentially represent residual melt produced during solidification of the lava flow. Field observation and the geochemical features of the lavas suggest that the vesicle cylinders grew upward from near the base of the flow by continuous feeding of residual melt from the neighboring host lavas to the heads of the cylinders. On the other hand, the vesicle sheets were produced in situ in the solidifying lava flow as fracture veins caused by horizontal compression. The vesicle cylinders have a remarkably higher MgO content (up to 8 wt.%) than the host lava (< 6 wt.%), whereas the vesicle sheets display MgO depletion (as low as 3.5 wt.%). The relatively high MgO content of the vesicle cylinders cannot be explained solely by the mechanical mixing of olivine phenocrysts with the residual melt. It is suggested that the vesicle cylinders were produced by the extraction of olivine-bearing interstitial melt from an augite-plagioclase network in the host lava, whereas the vesicle sheets were formed by the migration of the residual melt from a crystal network consisting of plagioclase, augite, and olivine in the host lava into platy fractures. We infer that this selective crystal fractionation for forming the vesicle cylinders resulted from processes in which abundant vesicles rejected from the upward-migrating floor solidification front prevented olivine crystals from being incorporated into the crystal network in the host lava. The vesicle cylinders are considered to have formed in ∼ 1 day after the lava flow came to rest, while relatively large vesicle sheets (> 1 cm thick) appeared much later (after ∼ 9 days). The formation of these segregation products was essentially complete within 20 days after the lava emplacement.     

Distribution of thermal areas on an active lava flow field: Landsat observations of Kilauea,Hawaii, July 1991

A Landsat Thematic Mapper (TM) image acquired on 23 July 1991 recorded widespread activity associated with the Episode 48 of the Pu'u 'O'o-Kupaianaha eruption of Kilauea Volcano, Hawaii. The scene contains a very large number (>3500) of thermally elevated near infrared (0.8–2.35 m) pixels (each 900 m²), which enable the spatial distribution of volcanic activity to be identified. This activity includes a lava lake within Pu'u 'O'o cone, an active lava tube system (7.9 km in length) with skylights between the Kupaianaha lava shield and several ocean entry points, and extensive active surface flows (total area of 1.3 km²) within a much larger area of cooling flows (total16 km²). The production of an average flux density map from the TM data of the flow field, wherein the average flux density is defined in units of Wm^-2, allows for the chronology of emplacement of active and cooling flows to be determined. The flux density map reveals that there were at least three breakouts (>5000 Wm^-2) feeding active flows, but on the day that the data were collected the TM recorded a waning phase of surface activity in this area, based on the relatively large amount of intermediate power-emitting (cooling) flows compared to high power-emitting (active) flows. The production of a comparable flux density map for future eruptions would aid in the assessment of volcanic hazards if the data were available in near-real time.     

The changing morphology of an open lava channel on Mt. Etna

An open channel lava flow on Mt. Etna (Sicily) was observed during May 30–31, 2001. Data collected using a forward looking infrared (FLIR) thermal camera and a Minolta-Land Cyclops 300 thermal infrared thermometer showed that the bulk volume flux of lava flowing in the channel varied greatly over time. Cyclic changes in the channel's volumetric flow rate occurred over several hours, with cycle durations of 113–190 min, and discharges peaking at 0.7 m³ s⁻¹ and waning to 0.1 m³ s⁻¹. Each cycle was characterized by a relatively short, high-volume flux phase during which a pulse of lava, with a well-defined flow front, would propagate down-channel, followed by a period of waning flow during which volume flux lowered. Pulses involved lava moving at relatively high velocities (up to 0.29 m s⁻¹) and were related to some change in the flow conditions occurring up-channel, possibly at the vent. They implied either a change in the dense rock effusion rate at the source vent and/or cyclic-variation in the vesicle content of the lava changing its bulk volume flux. Pulses would generally overspill the channel to emplace pāhoehoe overflows. During periods of waning flow, velocities fell to 0.05 m s^–1. Blockages forming during such phases caused lava to back up. Occasionally backup resulted in overflows of slow moving ‘a‘ā that would advance a few tens of meters down the levee flank. Compound levees were thus a symptom of unsteady flow, where overflow levees were emplaced as relatively fast moving pāhoehoe sheets during pulses, and as slow-moving ‘a‘ā units during backup. Small, localized fluctuations in channel volume flux also occurred on timescales of minutes. Volumes of lava backed up behind blockages that formed at constrictions in the channel. Blockage collapse and/or enhanced flow under/around the blockage would then feed short-lived, wave-like, down-channel surges. Real fluctuations in channel volume flux, due to pulses and surges, can lead to significant errors in effusion rate calculations. Editorial responsibility: A. Woods     

Deformation patterns in a high-viscosity lava flow inferred from the crystal preferred orientation and imbrication structures: an example from Salina (Aeolian Islands,southern Tyrrhenian Sea,Italy)

Shape-preferred orientation and imbrication structures of crystals have been measured on samples representative of the base, centre and top of a highly viscous lava flow on Salina (Aeolian Islands, southern Tyrrhenian Sea). The data allow zones with different deformation patterns to be identified. In the base and top of the flow, deformation leads to the development of discrete preferred orientation and imbrication of the elongate crystals. The sense of shear is right-lateral at the base and left-lateral at the top of the flow. Shear strain can be estimated by the analysis of crystal preferred orientation. Deformation increases from the flow centre to the outer, more viscous boundary layers. Random orientation of crystals in the inner zone supports the presence of plug flow in a pseudoplastic lava. The textural features of the studied lava may be related to different mechanisms (i.e. lateral expansion). We conclude that the observed crystal alignments and imbrication structures may be related to a plug flow moving between two non-deforming walls. The walls are represented by the solidified, broken upper and basal crust of the flow. The low shear strain values calculated in the outer margins of the flow are indicative of the last deformation event. Crystal preferred orientation and imbrication structures may be related to the occurrence of velocity gradients existing between the inner zone of the flow and its solidus or near-solidus outer margins.     

The submarine eruption of the Bombarda volcano,Milos Island,Cyclades, Greece

The Milos volcanic field includes a well-exposed volcaniclastic succession which records a long history of submarine explosive volcanism. The Bombarda volcano, a rhyolitic monogenetic center, erupted ∼1.7 Ma at a depth <200 m below sea level. The aphyric products are represented by a volcaniclastic apron (up to 50 m thick) and a lava dome. The apron is composed of pale gray juvenile fragments and accessory lithic clasts ranging from ash to blocks. The juvenile clasts are highly vesicular to non-vesicular; the vesicles are dominantly tube vesicles. The volcaniclastic apron is made up of three fades: massive to normally graded pumice-lithic breccia, stratified pumice-lithic breccia, and laminated ash with pumice blocks. We interpret the apron beds to be the result of water-supported, volcaniclastic mass-How emplacement, derived directly from the collapse of a small-volume, subaqueous eruption column and from syn-eruptive, down-slope resedimentation of volcaniclastic debris. During this eruptive phase, the activity could have involved a complex combination of phreatomagmatic explosions and minor submarine effusion. The lava dome, emplaced later in the source area, is made up of flow-banded lava and separated from the apron by an obsidian carapace a few meters thick. The near-vertical orientation of the carapace suggests that the dome was intruded within the apron. Remobilization of pyroclastic debris could have been triggered by seismic activity and the lava dome emplacement. Published online: 30 January 2003 Editorial responsibility: J. McPhie     

Groundwater dynamics in a till hillslope: flow directions,gradients and delay

Knowledge of groundwater dynamics is important for the understanding of hydrological controls on chemical processes along the water flow pathways. To increase our knowledge of groundwater dynamics in areas with shallow groundwater, the groundwater dynamics along a hillslope were studied in a boreal catchment in Southern Sweden. The forested hillslope had a 1‐ to 2‐m deep layer of sandy till above bedrock. The groundwater flow direction and slope were calculated under the assumption that the flow followed the slope of the groundwater table, which was computed for different triangles, each defined by three groundwater wells. The flow direction showed considerable variations over time, with a maximum variation of 75°. During periods of high groundwater levels the flow was almost perpendicular to the stream, but as the groundwater level fell, the flow direction became gradually more parallel to the stream, directed in the downstream direction. These findings are of importance for the interpretation of results from hillslope transects, where the flow direction usually is assumed to be invariable and always in the direction of the hillslope. The variations in the groundwater flow direction may also cause an apparent dispersion for groundwater‐based transport. In contrast to findings in several other studies, the groundwater level was most responsive to rainfall and snowmelt in the upper part of the hillslope, while the lower parts of the slope reached their highest groundwater level up to 40 h after the upper parts. This can be explained by the topography with a wetter hollow area in the upper part. Copyright © 2011 John Wiley & Sons, Ltd.     

Numerical simulation of three‐dimensional flow hydraulics in a braided channel

Results are presented from a numerical simulation of three‐dimensional flow hydraulics around a mid‐channel bar carried out using the FLUENT/UNS computational fluid dynamics (CFD) software package. FLUENT/UNS solves the three‐dimensional Reynolds‐averaged form of the Navier–Stokes equations. Turbulence closure is achieved using a RNG k–ϵ model. Simulated flow velocities are compared with measured two‐dimensional velocities (downstream and cross‐stream) obtained using an electromagnetic current meter (ECM). The results of the simulation are qualitatively consistent with the flow structures observed in the field. Quantitative comparison of the simulated and measured velocity magnitudes indicates a strong positive correlation between the two (r=0·88) and a mean difference of 0·09 m s⁻¹. Deviations between simulated and measured velocities may be identified that are both random and systematic. The former may reflect a number of factors including subgrid‐scale natural spatial variability in flow velocities associated with local bed structures and measurement uncertainty resulting from problems of ECM orientation. Model mesh configuration, roughness parameterization and inlet boundary condition uncertainty may each contribute to systematic differences between simulated and measured flow velocities. These results illustrate the potential for using CFD software to simulate flow hydraulics in natural channels with complex configurations. They also highlight the need for detailed spatially distributed datasets of three‐dimensional flow variables to establish the accuracy and applicability of CFD software. Copyright © 1999 John Wiley & Sons, Ltd.     

Emplacement of subaerial pahoehoe lava sheet flows into water: 1990 Kūpaianaha flow of Kilauea volcano at Kaimū Bay, Hawai`i

Episode 48 of the ongoing eruption of Kilauea, Hawai`i, began in July 1986 and continuously extruded lava for the next 5.5 years from a low shield, Kūpaianaha. The flows in March 1990 headed for Kalapana and inundated the entire town under 15–25 m of lava by the end of August. As the flows advanced eastward, they entered into Kaimū Bay, replacing it with a plain of lava that extends 300 m beyond the original shoreline. The focus of our study is the period from August 1 to October 31, 1990, when the lava buried almost 406,820 m² of the 5-m deep bay. When lava encountered the sea, it flowed along the shoreline as a narrow primary lobe up to 400 m long and 100 m wide, which in turn inflated to a thickness of 5–6 m. The flow direction of the primary lobes was controlled by the submerged delta below the lavas and by damming up lavas fed at low extrusion rates. Breakout flows through circumferential and axial inflation cracks on the inflating primary lobes formed smaller secondary lobes, burying the lows between the primary lobes and hiding their original outlines. Inflated flow lobes eventually ruptured at proximal and/or distal ends as well as mid-points between the two ends, feeding new primary lobes which were emplaced along and on the shore side of the previously inflated lobes. The flow lobes mapped with the aid of aerial photographs were correlated with daily observations of the growing flow field, and 30 primary flow lobes were dated. Excluding the two repose periods that intervened while the bay was filled, enlargement of the flow field took place at a rate of 2,440–22,640 square meters per day in the bay. Lobe thickness was estimated to be up to 11 m on the basis of cross sections of selected lobes measured using optical measurement tools, measuring tape and hand level. The total flow-lobe volume added in the bay during August 1–October 31 was approximately 3.95 million m³, giving an average supply rate of 0.86 m³/s.     

Kinematic significance of mingling-rolling structures in lava flows: a case study from Porri Volcano (Salina, Southern Tyrrhenian Sea)

 A basaltic andesite lava flow from Porri Volcano (Salina, Southern Tyrrhenian Sea) is composed of two different magmas. Magma A (51 vol.% of crystals) has a dacitic glass composition, and magma B (18 vol.% of crystals), a basaltic glass composition. Magma B is hosted in A and consists of sub-spherical enclaves and boudin-like, banding and rolling structures (RS). Four types of RS have been recognized: σ–type;δ–type; complex σ-δ–types and transitional structures between sub-spherical enclaves and rolling structures. An analysis of the RS has been performed in order to reconstruct the flow kinematics and the mechanism of flow emplacement. Rolling structures have been selected in three sites located at different distances from the vent. In all sites most RS show the same sense of shear. Kinematic analysis of RS allows the degree of flow non-coaxiality to be determined. The non-coaxiality is expressed by the kinematic vorticity number Wk, a measure of the ratio Sr between pure shear strain rate and simple shear strain rate. The values of Wk calculated from the measured shapes of microscopic RS increase with increasing distance from the vent, from approximately 0.5 to 0.9. Results of the structural analysis reveal that the RS formed during the early–intermediate stage of flow emplacement. They represent originally sub-spherical enclaves deformed at low shear strain. At higher strain, RS deformed to give boudin-like and stretched banding structures. Results of the kinematic analysis suggest that high viscosity lava flows are heterogeneous non-ideal shear flows in which the degree of non-coaxiality increases with the distance from the vent. In the vent area, deformation is intermediate between simple shear and pure shear. Farther from the vent, deformation approaches ideal simple shear. Lateral extension processes occur only in the near-vent zone, where they develop in response to the lateral push of magma extruded from the vent. Lateral shortening processes develop in the distal zone and record the gravity-driven movement of the lava. The lava flow advanced by two main mechanisms, lateral translation and rolling motion. Lateral translation equals rolling near the vent, while rolling motion prevailed in the distal zones. Received: 6 November 1997 / Accepted: 29 November 1997     

Relationships between deformation and mixing processes in lava flows: a case study from Salina (Aeolian Islands,Tyrrhenian Sea)

 The massive unit of a lava flow from Porri volcano (Salina, Aeolian Islands) displays many unusual structures related to the physical interaction between two different magmas. The magma A represents approximately 80% of the exposed lava surface; it has a crystal content of 51 vol.% and a dacitic glass composition (SiO₂=63–64 wt.%). The magma B has a basaltic-andesite glass composition (SiO₂=54–55 wt.%) and a crystal content of approximately 18 vol.%. It occurs as pillow-like enclaves, banding, boudin-like and rolling structures which are hosted in magma A. Structural analysis suggests that banding and boudin-like structures are the result of the deformation of enclaves at different shear strain. The linear correlation between strain and stratigraphic height of the measured elements indicates a single mode of deformation. We deduce that the component B deformed according to a simple shear model. Glass analyses of the A–B boundary indicate that A and B liquids mix together at high shear strain, whereas only mingling occurs at low shear strain. This suggests that the amount of deformation (i.e. forced convection) plays an important role in the formation of hybrid magmas. High shear strain may induce stretching, shearing and rolling of fluids which promote both forced convection and dynamical diffusion processes. These processes allow mixing of magmas with large differences in their physical properties. Received: 15 July 1995 / Accepted: 30 May 1996     

Magma dynamics, crystallization, and chemical differentiation of the 1959 Kilauea Iki lava lake, Hawaii, revisited

Using constraints from an extensive database of geological and geochemical observations along with results from fluid mechanical studies of convection in magma chambers, we identify the main physical processes at work during the solidification of the 1959 Kilauea Iki lava lakes. In turn, we investigate their quantitative influence on the crystallization and chemical differentiation of the magma, and on the development of the internal structure of the lava lake. In contrast to previous studies, vigorous stirring in the magma, driven predominately by the descent of dense crystal-laden thermal plumes from the roof solidification front and the ascent of buoyant compositional plumes due to the in situ growth of olivine crystals at the floor, is predicted to have been an inevitable consequence of very strong cooling at the roof and floor. The flow is expected to have caused extensive but imperfect mixing over most of the cooling history of the magma, producing minor compositional stratification at the roof and thermal stratification at the floor. The efficient stirring of the large roof cooling is expected to have resulted in significant internal nucleation of olivine crystals, which ultimately settled to the floor. Additional forcing due to either crystal sedimentation or the ascent of gas bubbles is not expected to have increased significantly the amount of mixing. In addition to convection in the magma, circulation driven by the convection of buoyant interstitial melt in highly permeable crystal-melt mushes forming the roof and the floor of the lava lake is envisaged to have produced a net upward flow of evolved magma from the floor during solidification. In the floor zone, mush convection may have caused the formation of axisymmetric chimneys through which evolved magma drained from deep within the floor into the overlying magma and potentially the roof. We hypothesize that the highly evolved, pipe-like ‘vertical olivine-rich bodies’ (VORBs) [Bull. Volcanol. 43 (1980) 675] observed in the floor zone, of the lake are fossil chimneys. In the roof zone, buoyant residual liquid both produced at the roof solidification front and gained from the floor as a result of incomplete convective mixing is envisaged to have percolated or ‘leaked‘ into the overlying highly-permeable cumulate, displacing less buoyant interstitial melt downward. The results from Rayleigh fractionation-type models formulated using boundary conditions based on a quantitative understanding of the convection in the magma indicate that most of the incompatible element variation over the height of the lake can be explained as a consequence of a combination of crystal settling and the extensive but imperfect convective mixing of buoyant residual liquid released from the floor solidification front. The remaining chemical variation is understood in terms of the additional influences of mush convection in the roof and floor on the vertical distribution of incompatible elements. Although cooling was concentrated at the roof of the lake, the floor zone is found to be thicker than the roof zone, implying that it grew more quickly. The large growth rate of the floor is explained as a consequence of a combination of the substantial sedimentation of olivine crystals and more rapid in situ crystallization due to both a higher liquidus temperature and enhanced cooling resulting from imperfect thermal and chemical mixing.