Lava Level Research Articles

Development, structure, and behavior of a perched lava channel at Kīlauea Volcano, Hawaiʻi, during 2007

Channelized lava flows are commonly produced during the early stages of basaltic eruptions. These channels usually maintain their morphology until the eruption ends or discharge is diverted. In some instances, narrower channels can roof over, developing into lava tubes. We report here on a channelized flow erupted at Kīlauea volcano in 2007 that evolved into a “perched lava channel” composed of a string of interconnected, elongate lava pools, forming a lava channel/lava pool hybrid. The lava channel, which had a time-averaged discharge rate of ~3–9 m3/s, initially fed a series of flow branches that exhibited cooling-limited and volume-limited controls on flow length, sometimes with each process controlling a different morphological aspect of a single flow branch. The perched lava channel grew vertically primarily by overplating of the channel levees from frequent overflows, forming a compound flow field. This vertical growth only occurred when the distal end of the channel was blocked. When levee failure at the distal end of the channel caused the lava level in the channel to drop below the levee rim, no vertical growth occurred. Seeps of spiny lava and slabby pāhoehoe were common, erupting from uplift scarps on the channel levees, apparently fed by sills from denser, relatively crystal-rich material filling the bottom of the channel. We infer that lava in the channel was stratified in vesicularity and velocity, with foamy, vesicular, faster-moving lava at the top of the lava stream and denser, relatively outgassed, slower-moving lava filling the bottom of the channel. The channel levees were unstable, failing on several occasions, perhaps triggered by the levee seeps. The appearance of seeps, therefore, is one way of assessing the collapse potential of similar perched lava structures.

Onset and evolution of Kīlauea's 2018 flank eruption and summit collapse from continuous gravity

Prior to the 2018 lower East Rift Zone (ERZ) eruption and summit collapse of Kīlauea Volcano, Hawai‘i, continuous gravimeters operated on the vent rims of ongoing eruptions at both the summit and Pu‘u ‘Ō‘ō. These instruments captured the onset of the 2018 lower ERZ eruption and the effects of lava withdrawal from both locales, providing constraints on the timing and style of activity and the physical properties of the lava lakes at both locations. At the summit, combining gravity, lava level, and a three-dimensional model of the vent indicates that the upper ∼200 m of the lava lake had a density of about 1700 kgm−3, slightly greater than estimates from 2011–2015 and possibly indicating a gradual densification over time. At Pu‘u ‘Ō‘ō, gravity and vent geometry were used to model both the density and the rate of crater collapse, which was unknown owing to a lack of visual observations. Results suggest the withdrawal of at least 11×106 m3 of lava over the course of two hours, and a material density of 1800–1900 kgm−3. In addition, gravity data at Pu‘u ‘Ō‘ō captured a transient decrease and increase about an hour prior to crater collapse and that was probably related to a small, short-lived fissure eruption on the west flank of the cone and possibly to dike intrusion beneath Pu‘u ‘Ō‘ō. The fissure was the first event in the subsequent cascade that ultimately led to the extrusion of over 1 km3 of lava from lower ERZ vents, collapse of the summit caldera floor by more than 500 m, and the destruction of over 700 homes and other structures. These results emphasize the importance of continuous gravity in operational monitoring of active volcanoes.

A review of controls on lava lake level: insights from Halema‘uma‘u Crater, Kīlauea Volcano

The height of the lava column is a fundamental measure of open-vent volcanic activity, but little continuous long-term data exist to understand this parameter. The recent (2008–2018) lava lake activity at the summit of Kīlauea Volcano provides an opportunity to track and understand the processes that control lava level over timescales ranging from minutes to years. We review recently published data as well as analyze the long-term record of lava level at Kīlauea’s summit during 2009–2017. Longer timescale fluctuations, over days to months, have a strong positive linear correlation with ground deformation, suggesting they reflect pressure changes in the summit magma reservoir. Shorter timescale fluctuations, over minutes to hours, have an inverse relationship with spattering (i.e., outgassing) intensity at the lake surface—increased outgassing lowers the lake level and vice versa. Our analysis of the long-term lava level data thus confirms that lava level at Halema‘uma‘u is controlled by these two types of processes: (1) those related to magma reservoir pressure changes (such as magma supply rate) and (2) shallow outgassing fluctuations (such as gas pistoning). Frequency filtering can isolate pressure- and outgassing-driven components of lava level change. Time series analysis indicates that there was no large, persistent periodicity in the lava level; a minor fortnightly oscillation might be related to Earth tides but was not continuously present. The correlation between lava level and deformation of the summit indicates that the lava lake acts as a reliable “piezometer”; tracking lava level over time may thus provide an indication of flank eruptive potential. We show that long-term interdisciplinary monitoring is necessary to discriminate the processes that control lava level.

Radar Altimetry as a Robust Tool for Monitoring the Active Lava Lake at Erebus Volcano, Antarctica

AbstractThe level of lava within a volcanic conduit reflects the overpressure within a connected magma reservoir. Continuous monitoring of lava level can therefore provide critical insights into volcanic processes and aid hazard assessment. However, accurate measurements of lava level are not easy to make, partly owing to the often dense fumes that hinder optical techniques. Here we present the first radar instrument designed for the purpose of monitoring lava level and report on its successful operation at Erebus volcano, Antarctica. We describe the hardware and data‐processing steps followed to extract a time series of lava lake level, demonstrating that we can readily resolve ∼1 m cyclic variations in lake level that have previously been recognized at Erebus volcano. The performance of the radar (continuous, automated data collection in temperatures of around −30 °C) indicates the suitability of this approach for sustained automated measurements at Erebus and other volcanoes with lava lakes.

Continuous Gravity and Tilt Reveal Anomalous Pressure and Density Changes Associated With Gas Pistoning Within the Summit Lava Lake of Kīlauea Volcano, Hawai‘i

AbstractGas piston events within the summit eruptive vent of Kīlauea Volcano, Hawai‘i, are characterized by increases in lava level and by decreases in seismic energy release, spattering, and degassing. During 2010–2011, gas piston events were especially well manifested, with lava level rises of tens of meters over the course of several hours, followed by a sudden drop to preevent levels. The changes in lava level were accompanied by directly proportional changes in gravity, but ground deformation determined from tilt was anticorrelative. The small magnitude of the gravity changes, compared to the large changes in volume within the vent during gas pistons, suggests that pistoning involves the accumulation of a very low‐density (100–200 kg/m3) foam at the top of the lava column. Co‐event ground tilt indicates that rise in lava level is paradoxically associated with deflation (the opposite is usually true), which can be modeled as an increase in the gas content of the magma column between the source reservoir and the surface. Gas pistoning behavior is therefore associated with not only accumulation of a shallow magmatic foam but also more bubbles within the feeder conduit, probably due to the overall decrease in gas emissions from the lava lake during piston events.

Forecasting the Eruption of an Open‐Vent Volcano Using Resonant Infrasound Tones

AbstractOpen‐vent volcanic systems with active degassing are particularly effective at producing infrasound that exhibits resonant tones controlled by the geometry of the volcano's crater. Changes in the infrasound character can thus provide constraints on a crater's lava level, which may vary dynamically in the lead‐up to an eruption. Here we show that the increasing frequency content and damping characteristics of the resonant infrasound at Volcán Villarrica (Chile) relate to lava lake position in its crater/conduit preceding its 2015 eruption. We model the acoustic response of Villarrica's crater to determine that the lake began to rise on 27 February and reached the flared upper part of Villarrica's crater before oscillating during the two days prior to the 3 March paroxysm and 1.5 km‐high lava fountain. This study demonstrates the utility of remote infrasound monitoring for future eruptions of Villarrica and other analogous open‐vent volcanoes.

Shallow and deep controls on lava lake surface motion at Kīlauea Volcano

Lava lakes provide a rare window into magmatic behavior, and lake surface motion has been used to infer deeper properties of the magmatic system. At Halema'uma'u Crater, at the summit of Kīlauea Volcano, multidisciplinary observations for the past several years indicate that lava lake surface motion can be broadly divided into two regimes: 1) stable and 2) unstable. Stable behavior is driven by lava upwelling from deeper in the lake (presumably directly from the conduit) and is an intrinsic process that drives lava lake surface motion most of the time. This stable behavior can be interrupted by periods of unstable flow (often reversals) driven by spattering – a shallowly-rooted process often extrinsically triggered by small rockfalls from the crater wall. The bursting bubbles at spatter sources create void spaces and a localized surface depression which draws and consumes surrounding surface crust. Spattering is therefore a location of lava downwelling, not upwelling. Stable (i.e. deep, upwelling-driven) and unstable (i.e. shallow, spattering-driven) behavior often alternate through time, have characteristic surface velocities, flow directions and surface temperature regimes, and also correspond to changes in spattering intensity, outgassing rates, lava level and seismic tremor. These results highlight that several processes, originating at different depths, can control the motion of the lava lake surface, and long-term interdisciplinary monitoring is required to separate these influences. These observations indicate that lake surface motion is not always a reliable proxy for deeper lake or magmatic processes. From these observations, we suggest that shallow outgassing (spattering), not lake convection, drives the variations in lake motion reported at Erta 'Ale lava lake.

Insights into shallow magmatic processes at Kīlauea Volcano, Hawaiʻi, from a multiyear continuous gravity time series

AbstractContinuous gravity data collected near the summit eruptive vent at Kīlauea Volcano, Hawaiʻi, during 2011–2015 show a strong correlation with summit‐area surface deformation and the level of the lava lake within the vent over periods of days to weeks, suggesting that changes in gravity reflect variations in volcanic activity. Joint analysis of gravity and lava level time series data indicates that over the entire time period studied, the average density of the lava within the upper tens to hundreds of meters of the summit eruptive vent remained low—approximately 1000–1500 kg/m3. The ratio of gravity change (adjusted for Earth tides and instrumental drift) to lava level change measured over 15 day windows rose gradually over the course of 2011–2015, probably reflecting either (1) a small increase in the density of lava within the eruptive vent or (2) an increase in the volume of lava within the vent due to gradual vent enlargement. Superimposed on the overall time series were transient spikes of mass change associated with inflation and deflation of Kīlauea's summit and coincident changes in lava level. The unexpectedly strong mass variations during these episodes suggest magma flux to and from the shallow magmatic system without commensurate deformation, perhaps indicating magma accumulation within, and withdrawal from, void space—a process that might not otherwise be apparent from lava level and deformation data alone. Continuous gravity data thus provide unique insights into magmatic processes, arguing for continued application of the method at other frequently active volcanoes.

Automated tracking of lava lake level using thermal images at Kīlauea Volcano, Hawai’i

Tracking the level of the lava lake in Halema‘uma‘u Crater, at the summit of Kīlauea Volcano, Hawai’i, is an essential part of monitoring the ongoing eruption and forecasting potentially hazardous changes in activity. We describe a simple automated image processing routine that analyzes continuously-acquired thermal images of the lava lake and measures lava level. The method uses three image segmentation approaches, based on edge detection, short-term change analysis, and composite temperature thresholding, to identify and track the lake margin in the images. These relative measurements from the images are periodically calibrated with laser rangefinder measurements to produce real-time estimates of lake elevation. Continuous, automated tracking of the lava level has been an important tool used by the U.S. Geological Survey’s Hawaiian Volcano Observatory since 2012 in real-time operational monitoring of the volcano and its hazard potential.

Shallowly driven fluctuations in lava lake outgassing (gas pistoning), Kīlauea Volcano

Lava lakes provide ideal venues for directly observing and understanding the nature of outgassing in basaltic magmatic systems. Kīlauea Volcano's summit lava lake has persisted for several years, during which seismic and infrasonic tremor amplitudes have exhibited episodic behavior associated with a rise and fall of the lava surface (“gas pistoning”). Since 2010, the outgassing regime of the lake has been tied to the presence or absence of gas pistoning. During normal behavior (no gas pistoning), the lake is in a “spattering” regime, consisting of higher tremor amplitudes and gas emissions. In comparison, gas piston events are associated with an abrupt rise in lava level (up to 20 m), during which the lake enters a “non-spattering” regime with greatly decreased tremor and gas emissions. We study this episodic behavior using long-term multidisciplinary monitoring of the lake, including seismicity, infrasound, gas emission and geochemistry, and time-lapse camera observations. The non-spattering regime (i.e. rise phase of a gas piston cycle) reflects gas bubbles accumulating near the top of the lake, perhaps as a shallow foam, while spattering regimes represent more efficient decoupling of gas from the lake. We speculate that the gas pistoning might be controlled by time-varying porosity and/or permeability in the upper portions of the lava lake, which may modulate foam formation and collapse. Competing models for gas pistoning, such as deeply sourced gas slugs, or dynamic pressure balances, are not consistent with our observations. Unlike other lava lakes which have cyclic behavior that is thought to be controlled by deeply sourced processes, external to the lake itself, we show an example of lava lake fluctuations driven by cycles of activity at shallow depth and close to the lake's surface. These observations highlight the complex and unsteady nature of outgassing from basaltic magmatic systems.

Lava lake level as a gauge of magma reservoir pressure and eruptive hazard

Forecasting volcanic activity relies fundamentally on tracking magma pressure through the use of proxies, such as ground surface deformation and earthquake rates. Lava lakes at open-vent basaltic volcanoes provide a window into the uppermost magma system for gauging reservoir pressure changes more directly. At Kīlauea Volcano (Hawai‘i, USA) the surface height of the summit lava lake in Halema‘uma‘u Crater fluctuates with surface deformation over short (hours to days) and long (weeks to months) time scales. This correlation implies that the lake behaves as a simple piezometer of the subsurface magma reservoir. Changes in lava level and summit deformation scale with (and shortly precede) changes in eruption rate from Kīlauea’s East Rift Zone, indicating that summit lava level can be used for short-term forecasting of rift zone activity and associated hazards at Kīlauea.

Characterization of very-long-period seismicity accompanying summit activity at Kīlauea Volcano, Hawai'i: 2007–2013

Eruptive activity returned to the summit region of Kīlauea Volcano, Hawai'i with the formation of the “Overlook crater” within the Halema'uma'u Crater in March 2008. The new crater continued to grow through episodic collapse of the crater walls and as of late 2013 had grown into an approximately elliptical opening with dimensions of ~160×215m extending to a depth of ~200m. Occasional weak explosive events and a persistent gas plume continued to occur through 2013. Lava was first observed in the new crater in September 2008, and through 2009 the lava level remained deep in the crater and was only occasionally observed. Since early 2010 a lava lake with fluctuating level within the Overlook crater has been nearly continuously present, and has reached to within 22m of the Overlook crater rim. Volcanic activity at Kīlauea Volcano is episodic at all time scales and the characterization of very-long-period seismicity in the band 2–100s for the years 2007–2013 illuminates a portion of this broad spectrum of volcanic behavior. Three types of very-long-period events have been observed over this time and each is associated with distinct processes. Type 1 events are associated with vigorous degassing and occurred primarily between 2007 and 2009. Type 2 events are associated with rockfalls onto the lava lake and occurred primarily after early 2010. Both of these event types are induced by pressure and momentum changes at the top of the magma column that are transmitted downward to a source centroid ~1km below the northeast corner of the Halema'uma'u Crater where the energy couples to the solid Earth at a geometrical discontinuity in the underlying dike system. Type 3 events are not related to surficial phenomena but are associated with transients in mass transfer that occur within the dike system. Very-long-period tremor has also accompanied the return of eruptive activity, with increasing amplitude associated with hours- to months-long changes in gas emission rates and summit deformation.

Continuous gravity measurements reveal a low-density lava lake at Kīlauea Volcano, Hawai‘i

On 5 March 2011, the lava lake within the summit eruptive vent at Kīlauea Volcano, Hawai‘i, began to drain as magma withdrew to feed a dike intrusion and fissure eruption on the volcanoʼs east rift zone. The draining was monitored by a variety of continuous geological and geophysical measurements, including deformation, thermal and visual imagery, and gravity. Over the first ∼14 hours of the draining, the ground near the eruptive vent subsided by about 0.15 m, gravity dropped by more than 100 μGal, and the lava lake retreated by over 120 m. We used GPS data to correct the gravity signal for the effects of subsurface mass loss and vertical deformation in order to isolate the change in gravity due to draining of the lava lake alone. Using a model of the eruptive vent geometry based on visual observations and the lava level over time determined from thermal camera data, we calculated the best-fit lava density to the observed gravity decrease — to our knowledge, the first geophysical determination of the density of a lava lake anywhere in the world. Our result, 950±300 kgm−3, suggests a lava density less than that of water and indicates that Kīlaueaʼs lava lake is gas-rich, which can explain why rockfalls that impact the lake trigger small explosions. Knowledge of such a fundamental material property as density is also critical to investigations of lava-lake convection and degassing and can inform calculations of pressure change in the subsurface magma plumbing system.

Time-lapse camera observations of gas piston activity at Pu‘u ‘Ō‘ō, Kīlauea volcano, Hawai‘i

Gas pistoning is a type of eruptive behavior described first at Kīlauea volcano and characterized by the (commonly) cyclic rise and fall of the lava surface within a volcanic vent or lava lake. Though recognized for decades, its cause continues to be debated, and determining why and when it occurs has important implications for understanding vesiculation and outgassing processes at basaltic volcanoes. Here, we describe gas piston activity that occurred at the Pu‘u ‘Ō‘ō cone, in Kīlauea’s east rift zone, during June 2006. Direct, detailed measurements of lava level, made from time-lapse camera images captured at close range, show that the gas pistons during the study period lasted from 2 to 60 min, had volumes ranging from 14 to 104 m3, displayed a slowing rise rate of the lava surface, and had an average gas release duration of 49 s. Our data are inconsistent with gas pistoning models that invoke gas slug rise or a dynamic pressure balance but are compatible with models which appeal to gas accumulation and loss near the top of the lava column, possibly through the generation and collapse of a foam layer.

Cyclic spattering, seismic tremor, and surface fluctuation within a perched lava channel, Kīlauea Volcano

In late 2007, a perched lava channel, built up to 45 m above the preexisting surface, developed during the ongoing eruption near Pu‘u ‘Ō‘ō cone on Kīlauea Volcano’s east rift zone. The lava channel was segmented into four pools extending over a total of 1.4 km. From late October to mid-December, a cyclic behavior, consisting of steady lava level rise terminated by vigorous spattering and an abrupt drop in lava level, was commonly observed in pool 1. We use geologic observations, video, time-lapse camera images, and seismicity to characterize and understand this cyclic behavior. Spattering episodes occurred at intervals of 40–100 min during peak activity and involved small (5–10-m-high) fountains limited to the margins of the pool. Most spattering episodes had fountains which migrated downchannel. Each spattering episode was associated with a rapid lava level drop of about 1 m, which was concurrent with a conspicuous cigar-shaped tremor burst with peak frequencies of 4–5 Hz. We interpret this cyclic behavior to be gas pistoning, and this is the first documented instance of gas pistoning in lava well away from the deeper conduit. Our observations and data indicate that the gas pistoning was driven by gas accumulation beneath the visco-elastic component of the surface crust, contrary to other studies which attribute similar behavior to the periodic rise of gas slugs. The gas piston events typically had a gas mass of about 2,500 kg (similar to the explosions at Stromboli), with gas accumulation and release rates of about 1.1 and 5.7 kg s−1, respectively. The time-averaged gas output rate of the gas pistoning events accounted for about 1–2% of the total gas output rate of the east rift zone eruption.

Lava tube shatter rings and their correlation with lava flux increases at Kīlauea Volcano, Hawai‘i

Shatter rings are circular to elliptical volcanic features, typically tens of meters in diameter, which form over active lava tubes. They are typified by an upraised rim of blocky rubble and a central depression. Prior to this study, shatter rings had not been observed forming, and, thus, were interpreted in many ways. This paper describes the process of formation for shatter rings observed at Kīlauea Volcano during November 2005–July 2006. During this period, tilt data, time-lapse images, and field observations showed that episodic tilt changes at the nearby Pu‘u ‘Ō‘ō cone, the shallow magmatic source reservoir, were directly related to fluctuations in the level of lava in the active lava tube, with periods of deflation at Pu‘u ‘Ō‘ō correlating with increases in the level of the lava stream surface. Increases in lava level are interpreted as increases in lava flux, and were coincident with lava breakouts from shatter rings constructed over the lava tube. The repetitive behavior of the lava flux changes, inferred from the nearly continuous tilt oscillations, suggests that shatter rings form from the repeated rise and fall of a portion of a lava tube roof. The locations of shatter rings along the active lava tube suggest that they form where there is an abrupt decrease in flow velocity through the tube, e.g., large increase in tube width, abrupt decrease in tube slope, and (or) sudden change in tube direction. To conserve volume, this necessitates an abrupt increase in lava stream depth and causes over-pressurization of the tube. More than a hundred shatter rings have been identified on volcanoes on Hawai‘i and Maui, and dozens have been reported from basaltic lava fields in Iceland, Australia, Italy, Samoa, and the mainland United States. A quick study of other basaltic lava fields worldwide, using freely available satellite imagery, suggests that they might be even more common than previously thought. If so, this confirms that episodic fluctuation in lava effusion rate is a relatively common process at basaltic volcanoes, and that the presence of shatter rings in prehistoric lava flow fields can be used as evidence that such fluctuations have occurred.

ON A CONSERVATIVE LAVA FLOW AUTOMATON

A simple lava flow automaton is discretized, three different variants are shown to exhibit a wide range of behavior, while some geometrical and shape characteristics are analysed. We respect, on an underlying square topology, equalization of lava height between nearest neighbors (whenever possible), and update the evolution in parallel. This model is a discretized version (as regards the lava level) of a very successful continuous height automaton, but with no local loss through cooling.

A model for the formation of lava tubes by roofing over a channel

The formation of lava tubes is a common phenomenon on some basaltic volcanoes, such as Etna. A model for tube formation by roofing of a channel is proposed and involves first describing lava as a Bingham liquid flowing down a slope. It is further assumed that lava flows in a channel with rectangular cross section: as a result of heat loss into the atmosphere, a crust is gradually formed on the upper surface of the flow and this crust eventually welds to the channel levées. We assume that a lava tube is formed when such a crust is sufficiently thick to resist the drag of the underlying flow and to sustain itself under its own weight. The minimum thickness of the crust satisfying such conditions depends on the tensile strength and shear strength of the crust itself. Assuming that the growth of the crust produces a downflow linear increase of the shear stress at the interface between flowing lava and the crust, the distance is evaluated between the eruption vent and the point where the tube is formed. The model predicts that if the flow rate is constant, the thickness of the flow increases as the crust fragments grow and weld to each other, and the velocity of the crust decreases to zero. Once the lava tube is formed, the initial flow rate can be achieved by a flow thickness smaller than the vertical size of the tube, with the same viscous dissipation: this may explain why under steady state conditions, the lava level inside a tube is frequently lower than the roof of the tube itself.

On the variations of flow rate in non-explosive lava eruptions

A physical model is developed to determine factors which influence the dynamics of non-explosive lava eruptions. In the model, magma rises in a laminar regime from an overpressured chamber surrounded by elastic rock and erupts at the surface, where the accumulation of lava may alter the vent pressure. In a flow rate versus volume diagram, an eruption follows a path which is sensitive to relative changes in flow variables such as viscosity, conduit dimensions and thickness of lava over the vent. The path does not depend on the unknown elastic properties of the country rock or the chamber dimensions. With appropriate volcanological data, an observed flow rate versus volume path can be interpreted. The approach is applied to three well-documented eruptions. In each case the eruption rate was partially influenced by decreasing chamber pressure. The 1988–1990 eruption of Lonquimay (Chile) exhibits an early phase of conduit shrinkage which lasted about 100 days, probably due to magma solidification against country rock. The 1979 eruption of La Soufrière de Saint Vincent (West Indies) was initially affected by lava dome growth and later passed through a phase possibly caused by constriction of the conduit. The 1943–1952 eruption of Paricutín (Mexico) was affected by changes of lava level in the cinder cone and magma properties. In these three eruptions, the pre-eruptive chamber overpressures were similar (10–20 MPa) even though the erupted volumes differed by as much as two orders of magnitude.

Morphological, structural and textural variations in the 1988–1990 andesite lava of Lonquimay Volcano, Chile

AbstractThe 1988–1990 eruption of Lonquimay Volcano, Chile (38°S) formed a 10.2 km long andesite lava with a volume of 0.23 km3 over a period of 13 months. The lava extrusion rate decreased with time as chamber pressure and vent dimensions decreased. The velocity of the flow front decreased exponentially with distance from vent as a consequence of cooling and the increase of apparent viscosity at the flow front. The lava developed a central channel which decreased in width and depth with time. Three prominent lava levées were formed on each margin and resulted from abandonment as the channel decreased in width as a result of a rapid decrease of flow rate over the first 100 days of activity. A fourth major levée developed in February, during a brief period of flow rate increase down the main channel, but its walls were gradually exposed as the lava depth again decreased due to declining flow rate. The structure of lava levées depended on their age and longevity of the flow in the adjacent channel. Initial levées were formed in the first few days as the lava spread laterally and then retreated, leaving levées of massive lava. More mature rubble levées were formed during the next month by the lava pushing and then shearing aa and blocky breccia which formed on the cooling flow margin. Fragmentation and abrasion formed a characteristic zonation in the levées. A basal zone consists of very poorly sorted matrix-rich breccia with very rounded vesicular clasts and bimodal grain size distribution. The basal breccia zone strongly resembles block and ash flow deposits. This zone passes up into a zone of clast-supported clinker breccia which becomes increasingly matrix-poor and coarser with clasts becoming more angular upwards. The crest of the levée is composed of large (10–100 cm) angular to subangular blocks with no matrix. The zoned levées form after the active lava channel suddenly narrows. Lava depth initially increases and breccias are deposited on the channel margins and acquire the zoned structure by progressive shearing and accretion of clinkery aa breccia. The lava level then drops exposing the steep inner scarp of a levée. The most mature levée type formed in a long-lived channel over several months. The outer wall of the levée consists of zoned breccia, but the inner wall consists of a massive curving wall of strongly foliated lava with well-developed horizontal striations and ductile Reidel shears. The massive foliated facies is a consequence of prolonged flow which coats strongly sheared lava onto the inner levée wall. Scanning electron microscopy shows that the aa clinker clasts and foliated lava from the levée walls form at low melt fractions (⋚ 15%). In the last three months of the eruption the flow front ceased to advance but thickened as lava drained from proximal regions and intruded into the interior of the distal lava. The last stages of lava movement were characterized by updoming in the central channel. A lava surface feature, named here ‘Armadillo structure’, was formed by deformation of the cooler but still ductile lava crust. The deformation caused by underflow produced Reidel shears dipping upstream and doming of the lava due to rotation of the shear planes. The study demonstrates that lava morphology, structure and texture are strongly influenced by variations of effusion rate, local flow rate, channel topography and thermal maturity of the lava, which is reflected in downstream changes in viscosity.