Integration of micro-gravity and geodetic data to constrain shallow system mass changes at Krafla Volcano, N Iceland

New and previously published micro-gravity data are combined with InSAR data, precise levelling and GPS measurements to produce a model for the processes operating at Krafla volcano, 20 years after its most recent eruption. The data have been divided into two periods: from 1990 to 1995 and from 1996 to 2003 and show that the rate of deflation at Krafla is decaying exponentially. The net micro-gravity change at the centre of the caldera is shown, using the measured free air gradient, to be −85 μGal for the first and −100 μGal for the second period. After consideration of the effects of water extraction by the geothermal power station within the caldera, the net gravity decreases are −73±17 μGal for the first and −65±17 μGal for the second period. These decreases are interpreted in terms of magma drainage. Following a Mogi point source model, we calculate the mass decrease to be ∼2×10¹⁰ kg/year reflecting a drainage rate of ∼0.23 m³/s, similar to the ∼0.13 m³/s drainage rate previously found at Askja volcano, N. Iceland. Based on the evidence for deeper magma reservoirs and the similarity between the two volcanic systems, we suggest a pressure-link between Askja and Krafla at deeper levels (at the lower crust or the crust-mantle boundary). After the Krafla fires, co-rifting pressure decrease of a deep source at Krafla stimulated the subsequent inflow of magma, eventually affecting conditions along the plate boundary in N. Iceland, as far away as Askja. We anticipate that the pressure of the deeper reservoir at Krafla will reach a critical value and eventually magma will rise from there to the shallow magma chamber, possibly initiating a new rifting episode. We have demonstrated that by examining micro-gravity and geodetic data, our knowledge of active volcanic systems can be significantly improved.Editorial responsibility: A. Harris     

Net gravity decrease at Askja volcano,Iceland: constraints on processes responsible for continuous caldera deflation, 1988–2003

Askja caldera in northeast Iceland has been in a state of unrest for decades. Ground-deformation surveys show that the rate of deformation, i.e., deflation, is much higher then observed at any other dormant volcano in Iceland. This work presents the results from microgravity and deformation studies at Askja from 1988 to 2003. The deflation reaches a maximum of −0.46 m in the centre of the caldera, relative to a station outside the caldera, during the study period. The source of deformation is inferred to be at ∼3 km depth and a recent study infers a second deeper source at ∼16 km depth. The deflation is consistent with a subsurface volume change of −0.018 km³. We find a net microgravity decrease of 115 μGal in the centre of the caldera relative to the same station. This corresponds to a subsurface mass decrease of 1.6×10¹¹ kg between 1988 and 2003 based on the use of a point source model. A combination of magma drainage and cooling and contraction of the shallow magma reservoir at 3 km depth is our favoured model, consistent with the integrated observations. We suggest that extensional tectonic forces generate space in the ductile part of the crust to accommodate ongoing magma drainage from the shallow magma chamber.     

Multiple volcano deformation sources in a post-rifting period: 1989–2005 behaviour of Krafla,Iceland constrained by levelling,tilt and GPS observations

The Krafla rifting episode, which occurred in North Iceland in 1975–1984, was followed by inflation of a shallow magma chamber until 1989. At that time, gradual subsidence began above the magma chamber and has continued to the present at a declining rate. Pressure decrease in a shallow magma chamber is not the only source of deformation at Krafla, as other deformation processes are driven by exploitation of two geothermal fields, together with plate spreading. In addition, deep-seated magma accumulation appears to take place, with its centre ∼ 10 km north of the Krafla caldera. The relative strength of these sources has varied with time. New results from a levelling survey and GPS measurements in 2005 allow an updated view on the deformation field. Deformation rates spanning 2000–2005 are the lowest recorded in the 30-year history of geodetic studies at the volcano. The inferred rate of 2000–2005 subsidence related to processes in the shallow magma chamber is less than 0.3 cm/yr whereas it was ∼ 5 cm/yr in 1989–1992. Currently, the highest rate of subsidence takes place in the Leirbotnar area, within the Krafla caldera, and appears to be a result of geothermal exploitation.     

Deflation during caldera unrest: constraints on subsurface processes and hazard prediction from gravity–height data

We have evaluated published gravity-height ((g/(h) data on Campi Flegrei, Kilauea, Askja and Krafla, in order to discriminate between subsurface processes during caldera subsidence. With respect to end member gravity-height correlations, such as the free air gradient (FAG) and the Bouguer corrected free air (BCFAG), (g/(h gradients must be interpreted in terms of subsurface mass redistribution, density changes or some combination of these. (g/(h gradients during subsidence plot (1) along or below the BCFAG, (2) between the BCFAG and the FAG or (3) along or above the FAG. We have evaluated each of these three regions in terms of subsurface processes during volcano subsidence. We have interpreted (g/(h gradients as possible indicators of precursors of volcanic activity and propose that gravity-height surveys may help to detect precursors of caldera collapse caused by magma drainage. In this context, the 1875 eruption of Askja in Iceland has been re-interpreted in terms of the beginning of the eruptive episode being induced by roof collapse of an evacuating magma chamber. Based on other examples of recent volcanic roof collapses, we evaluate the contribution of gravity-height surveys in assessing volcanic risks during caldera subsidence. Caldera-forming eruptions are environmentally and economically the most devastating volcanic events. Inflation is usually considered to be an important precursor to activity. Here, we show that deflation may be associated with the trigger mechanism for caldera-forming explosive eruptions.     

Analogue models of caldera collapse in strike-slip tectonic regimes

Regional-scale faulting, particularly in strike-slip tectonic regimes, is a relatively poorly constrained factor in the formation of caldera volcanoes. To examine interactions between structures associated with regional-tectonic strike-slip deformation and volcano-tectonic caldera subsidence, we made scaled analogue models. Tabular (sill-like) inclusions of creamed honey in a sand/gypsum mix replicated shallow-level granitic magma chambers in the brittle upper crust. Lateral motion of a base plate sited below half the sand/gypsum pack allowed simulation of regional strike-slip deformation. Our experiments modelled: (1) strike-slip deformation of a homogeneous brittle medium; (2) strike-slip deformation of a brittle medium containing a passive magma reservoir; (3) caldera collapse into sill-like magma reservoirs without regional strike-slip deformation; and (4) caldera collapse into sill-like magma reservoirs after regional strike-slip deformation. Our results show that whilst the magma chamber shape principally influences the development and geometry of volcano-tectonic collapse structures, regional-tectonic strike-slip faults (Riedel shears and Y-shears) may affect a caldera’s structural evolution in two main ways. Firstly, regional strike-slip faults above the magma chamber may form a pre-collapse structural grain that is exploited and reactivated during subsidence. Our experiments show that such faults may preferentially reactivate where tangential to the collapse area and coincident with the chamber margins. In this case, volcano-tectonic extension in the caldera periphery tends to localise on regional-tectonic faults that lie just outside the chamber margins. In addition, volcano-tectonic reverse faults may link with and reactivate pre-collapse regional-tectonic faults that lie just inside the chamber margins. Secondly, where regional-tectonic strike-slip faults define corners in the magma chamber margin, they may halt the propagation of volcano-tectonic reverse faults. The experiments also highlight the potential difficulties in assessing the relative contributions of volcano-tectonic and regional-tectonic subsidence processes to the final caldera structure seen in the field. Disruption of the pre-collapse surface by regional-tectonic faulting was preserved during coherent volcano-tectonic subsidence to produce a caldera floor of differentially-subsided fault blocks. Without definitive evidence for syn-eruptive growth faulting, thickness changes in caldera fill across such regional-tectonic fault blocks in nature could be mistaken as evidence for piecemeal volcano-tectonic collapse.     

Caldera resurgence during magma replenishment and rejuvenation at Valles and Lake City calderas

A key question in volcanology is the driving mechanisms of resurgence at active, recently active, and ancient calderas. Valles caldera in New Mexico and Lake City caldera in Colorado are well-studied resurgent structures which provide three crucial clues for understanding the resurgence process. (1) Within the limits of ⁴⁰Ar/³⁹Ar dating techniques, resurgence and hydrothermal alteration at both calderas occurred very quickly after the caldera-forming eruptions (tens of thousands of years or less). (2) Immediately before and during resurgence, dacite magma was intruded and/or erupted into each system; this magma is chemically distinct from rhyolite magma which was resident in each system. (3) At least 1?km of structural uplift occurred along regional and subsidence faults which were closely associated with shallow intrusions or lava domes of dacite magma. These observations demonstrate that resurgence at these two volcanoes is temporally linked to caldera subsidence, with the upward migration of dacite magma as the driver of resurgence. Recharge of dacite magma occurs as a response to loss of lithostatic load during the caldera-forming eruption. Flow of dacite into the shallow magmatic system is facilitated by regional fault systems which provide pathways for magma ascent. Once the dacite enters the system, it is able to heat, remobilize, and mingle with residual crystal-rich rhyolite remaining in the shallow magma chamber. Dacite and remobilized rhyolite rise buoyantly to form laccoliths by lifting the chamber roof and producing surface resurgent uplift. The resurgent deformation caused by magma ascent fractures the chamber roof, increasing its structural permeability and allowing both rhyolite and dacite magmas to intrude and/or erupt together. This sequence of events also promotes the development of magmatic–hydrothermal systems and ore deposits. Injection of dacite magma into the shallow rhyolite magma chamber provides a source of heat and magmatic volatiles, while resurgent deformation and fracturing increase the permeability of the system. These changes allow magmatic volatiles to rise and meteoric fluids to percolate downward, favouring the development of hydrothermal convection cells which are driven by hot magma. The end result is a vigorous hydrothermal system which is driven by magma recharge.     

Lower-crustal earthquakes caused by magma movement beneath Askja volcano on the north Iceland rift

The lower crust of magmatically active rifts is usually too hot and ductile to allow seismicity. The Icelandic mid-Atlantic rift is characterized by high heat flow, abundant magmatism generating up to 25–30 km thick crust, and seismicity within the upper 8 km of the crust. In a 20-seismometer survey in July-August 2006 within the northern rift zone around the Askja volcano we recorded ~1700 upper-crustal earthquakes cutting off at 7–8 km depth, marking the brittle-ductile boundary. Unexpectedly, we discovered 100 small-magnitude (M_L <1.5) earthquakes, occurring in swarms mostly at 14–26 km depth within the otherwise aseismic lower crust, and beneath the completely aseismic middle crust. A repeat survey during July-August 2007 yielded more than twice as many lower-crustal events. Geodetic and gravimetric data indicate melt drainage from crustal magma chambers beneath Askja. We interpret the microearthquakes to be caused by melt moving through the crust from the magma source feeding Askja. They represent bursts of magma motion opening dykes over distances of a few meters, facilitated by the extensional setting of the active rift zone.     

1983–2003 decaying rate of deflation at Askja caldera: Pressure decrease in an extensive magma plumbing system at a spreading plate boundary

New deformation data from the Askja volcano, Iceland, show that the volcano's caldera has been deflating continuously for over 20 years, and confirm that the rate of subsidence is slowing down. The decay in subsidence rate can be fitted with a function of the form e ^{− t / τ} , where τ is 39 years. Reanalysis of GPS data from 1993–1998 show that these data can be fitted with a model calling for two Mogi point sources, one shallow, and another one much deeper (16.2 km depth). Pressure decrease occurs in both sources. The deeper source is responsible for observed horizontal contraction towards Askja at distances that cannot be explained by the shallower source. Plate spreading of 19 mm/year distributed evenly over about 100-km-wide zone is also favoured by the data.     

Body force equivalent calculations for subsidence of the Po valley

Summary Body force equivalents can be computed for subsidence of the Po delta in the twentieth century. Their preferred location suggests that not all of the subsidence is caused by pumping of fluids or other processes occurring at shallow depths. Some of the deformation may be due to crustal response to a north-south compressive stress.     

Gravity changes and passive SO2 degassing at the Masaya caldera complex, Nicaragua

An understanding of the mechanisms responsible for persistent volcanism can be acquired through the integration of geophysical and geochemical data sets. By interpreting data on micro-gravity, ground deformation and SO₂ flux collected at Masaya Volcano since 1993, it is now clear that the characteristically cyclical nature of the activity is not driven by intrusion of additional magma into the system. Rather, it may be due in large part to the blocking and accumulation of gas by restrictions in the volcano substructure. The history of crater collapse and formation of caverns beneath the crater floor would greatly facilitate the trapping and storage of gas in a zone immediately beneath San Pedro and the other craters. Another mechanism that may explain the observed gravity and gas flux variations is the convective overturn of shallow, pre-existing, degassed, cooled, dense magma that is replaced periodically by lower density, hot, gas-rich magma from depth. Buoyant gas-rich magma rises from depth and is emplaced near the surface, resulting in the formation and fluctuation of a low-density gas-rich layer centred beneath Nindirí and Santiago craters. As this magma vigorously degasses, it must cool, increase in density and eventually sink. Five stages of activity have been identified at Masaya since 1853 and the most recent data suggest that the system may have been entering another period of reduced degassing in 2000. This type of analysis has important implications for hazard mitigation because periods of intense degassing are associated with poor agricultural yields and reduced quality of life. A better understanding of persistent cyclically active volcanoes will allow for more effective planning of urban development and agricultural land use.     

Magma convection and mixing dynamics as a source of Ultra-Long-Period oscillations

Many volcanic eruptions are shortly preceded by injection of new magma into a pre-existing, shallow (<10 km) magma chamber, causing convection and mixing between the incoming and resident magmas. These processes may trigger dyke propagation and further magma rise, inducing long-term (days to months) volcano deformation, seismic swarms, gravity anomalies, and changes in the composition of volcanic plumes and fumaroles, eventually culminating in an eruption. Although new magma injection into shallow magma chambers can lead to hazardous event, such injection is still not systematically detected and recognized. Here, we present the results of numerical simulations of magma convection and mixing in geometrically complex magmatic systems, and describe the multiparametric dynamics associated with buoyant magma injection. Our results reveal unexpected pressure trends and pressure oscillations in the Ultra-Long-Period (ULP) range of minutes, related to the generation of discrete plumes of rising magma. Very long pressure oscillation wavelengths translate into comparably ULP ground displacements with amplitudes of order 10⁻⁴–10⁻² m. Thus, new magma injection into magma chambers beneath volcanoes can be revealed by ULP ground displacement measured at the surface.     

Magma mixing and magma plumbing systems in island arcs

Petrographic features of mixed rocks in island arcs, especially those originating by the mixing of magmas with a large compositional and temperature difference, such as basalt and dacite, suggest that the whole mixing process from their first contact to the final cooling (= eruption) has occurred continuously and in a relatively short time period. This period is probably less than several months, considerably shorter than the whole volcanic history. There may also be a prolonged quiescent interval, lasting longer than several days, between the magmas' contact and the mechanical mixing. This interval will allow the basic magma to cool and produce a semi-solidified boundary which is later disrupted by flow movements to produce basic inclusions.Mixing of magmas of contrasting chemical composition need not be the inevitable consequence of the contact of the magmas. It is, however, made more probable by forced convection caused by motive force such as the injection of a basic magma into an acidic magma chamber. A short interval between their initial contact and the final eruption requires that the acid magma chamber has a small volume, of the same order or less than that the introduced basic magma.The volcanic activity of Myoko volcano, central Japan, of the last 100,000 years shows alternate eruptions of hybrid andesite by mixing of basaltic and dacitic magmas, and non-mixed basalt to basaltic andesite. There was a repose period of 20,000 to 30,000 years between eruptions. The acidic chamber, eventually producing the mixed andesite activity, is formed during the repose period by the « in situ » solidification of the original basic magma against its wall. The volume of the chamber is very small, probably about 10^–2 km³. Basaltic magma with constant chemical composition is supplied to the shallow chamber from another deep seated basaltic chamber. The volume of the shallow magma chamber may be critical to the characteristics of volcanic activity and its products.     

Modelling gravity variations consistent with ground deformation in the Campi Flegrei caldera (Italy)

In recent years (1970–72 and 1982–84) two inflation episodes took place in the Campi Flegrei caldera (Italy), characterized by significant ground uplift and gravity variations. An elastic half-space model with vertical density stratification is employed to compute the displacement field and the gravity variations produced by the deformation of buried layers, following the inflation of a spherically symmetric deformation source. Contributions to gravity variations are produced by dilation/contraction of the medium, by the displacements of density interfaces (the free surface and subsurface layers) and of source boundaries and, possibly, by new mass input from remote distances into the source volume. Three cases were examined in detail: In case I, the magma chamber is identified as the deformation source and volume and pressure increase in the magma chamber is due to input of new magma from remote distances; in case II deformation is due to magma differentiation within the magma chamber (deformation source with constant mass); in case III the geothermal system is identified as the deformation source and a pressure increase, possibly driven by the exsolution of high temperature and high pressure volatiles in the magma chamber, is assumed to play a dominant role. From the comparison between measured and computed gravity residuals (free-air-corrected gravity variations) we can assess that, in case I, an inflation source with constant density would predict gravity residuals compatible with observations, whereas an expansion at constant mass (case II) would predict gravity residuals much lower than observed. The resolving power of gravity data however prevents accurate assessment of the density of the emplaced material. In case III, the pervasive density increase of the geothermal fluids induced by pressure increase is assumed to be the main source of gravity variations. The average porosity value required for this model to match both the ground deformation and the gravity residuals is found to be ˜10%, a value which is compatible with measured porosity values at Campi Flegrei in deep wells. The subsidence phases following both inflation episodes and the gravity residuals during subsidence lead us to consider case III as more plausible, even if a suitable combination of cases I and III cannot be discarded.     

An experimental study of structures arising in convective roll flows of magma melts in peripheral chambers of volcanoes

The morphology and internal convective structure of continuous roll flows in magma melts were experimentally studied with regard to peripheral (shallow) magma chambers and downgoing circular gravity flows in conduits. We found properties of continuous convective mixing in a magma melt in a peripheral chamber due to convective roll flows. We investigated the mechanism that is responsible for cumulus generation at the bottom of a peripheral chamber when homogeneous and heterogeneous melts are emplaced. We conducted an experimental study of how the contact surface is formed at the cumulus-melt boundary. We made a qualitative study of the mechanism that affects the composition of magma melts over time in a peripheral and a mantle (deeper) chamber owing to crystallization, as well as the special mechanism that is responsible for cumulus generation in the peripheral chamber.     

Petrology of lavas from the 2004–2005 flank eruption of Mt. Etna,Italy: inferences on the dynamics of magma in the shallow plumbing system

Following the 2001 and 2002–2003 flank eruptions, activity resumed at Mt. Etna on 7 September 2004 and lasted for about 6 months. This paper presents new petrographic, major and trace element, and Sr–Nd isotope data from sequential samples collected during the entire 2004–2005 eruption. The progressive change of lava composition allowed defining three phases that correspond to different processes controlling magma dynamics inside the central volcano conduits. The compositional variability of products erupted up to 24 September is well reproduced by a fractional crystallization model that involves magma already stored at shallow depth since the 2002–2003 eruption. The progressive mixing of this magma with a distinct new one rising within the central conduits is clearly revealed by the composition of the products erupted from 24 September to 15 October. After 15 October, the contribution from the new magma gradually becomes predominant, and the efficiency of the mixing process ensures the emission of homogeneous products up to the end of the eruption. Our results give insights into the complex conditions of magma storage and evolution in the shallow plumbing system of Mt. Etna during a flank eruption. Furthermore, they confirm that the 2004–2005 activity at Etna was triggered by regional movements of the eastern flank of the volcano. They caused the opening of a complex fracture zone extending ESE which drained a magma stored at shallow depth since the 2002–2003 eruption. This process favored the ascent of a different magma in the central conduits, which began to be erupted on 24 September without any significant change in eruptive style, deformation, and seismicity until the end of eruption.     

Hot fluid migration: an efficient source of ground deformation: application to the 1982–1985 crisis at Campi Flegrei-Italy

Some solutions of the forced heat advection problem in compressible media are worked out employing a perturbative approach and their implication for thermoelastic deformation are discussed. A sharp temperature front, which migrates at a speed in the order of Darcy flow rate, develops in the medium, giving rise to significant deformation via thermal expansion. A thermally induced pressure source accompanies the temperature front, which may be significant only in very high temperature cases. Results are applied to interpreting the uplift episode of 1982–1984 at Campi Flegrei (C.F.), near Naples, Italy. A mechanism is envisaged for uplift at Campi Flegrei in which a sudden connection is established between a deep, hot, high-pressure fluid reservoir and a shallow, relatively cold, low-pressure aquifer. The inclusion of fluid migration in the deformation model allows simple explanations of several geochemical and geophysical observations made during the bradyseismic crisis. It appears that the proposed mechanism may explain the large observed uplift, without requiring unreasonable pressure increase within the magma chamber. Furthermore, the deformation source may be allowed to be shallower than the magma chamber, as required by any reasonable deformation model at C.F.     

Surface deformation at the Krafla volcano,North Iceland, 1982–1992

Extensive measurements of ground deformation at the Krafla volcano, Iceland, have been made since the beginning in 1975 of a series of eruptions and intrusions into the fissure system that extends north and south of the volcano. I concentrate on measurements before and after the eruption of September 1984, the last event of this series when the largest volume of lava was erupted. The patterns of ground deformation associated with the 1984 eruption, determined by precision levelling, electronic distance measurements and lake level observations, were similar to earlier intrusions and eruptions, in that the surface of the volcano subsided and the fissure system widened as magma moved laterally from a shallow central reservoir into the fissure system. The shallow magma reservoir of Krafla continued to expand for about five years after the eruption, but a slow subsidence of the central area began in 1989. Besides the presence of an inflating and deflating shallow magma reservoir at a depth of 2.5 km beneath the Krafla caldera, another inflating magma reservoir may exist at much greater depth below Krafla. The accumulation of compressive strain by numerous rift intrusions and eruptions since 1975 along the flanks of the north-south Krafla fissure swarm is being released slowly and will probably be reflected in the results of deformation measurements near Krafla for the next several decades. The total horizontal extension of the Krafla rift system in 1975–1984 was about 9 m, equal to about 500 years of constant plate divergence. The extension is twice the accumulated divergence since previous rifting events and eruptions in 1724–1729     

The fissure swarm of the Askja volcanic system along the divergent plate boundary of N Iceland

Divergent plate boundaries, such as the one crossing Iceland, are characterized by a high density of subparallel volcanic fissures and tectonic fractures, collectively termed rift zones, or fissure swarms when extending from a specific volcano. Volcanic fissures and tectonic fractures in the fissure swarms are formed during rifting events, when magma intrudes fractures to form dikes and even feeds fissure eruptions. We mapped volcanic fissures and tectonic fractures in a part of the divergent plate boundary in northern Iceland. The study area is ~1,800 km², located within and north of the Askja central volcano. The style of fractures changes with distance from Askja. Close to Askja the swarm is dominated by eruptive fissures. The proportion of tectonic fractures gets larger with distance from Askja. This may indicate that magma pressure is generally higher in dikes close to Askja than farther away from it. Volcanic fissures and tectonic fractures are either oriented away from or concentric with the 3–4 identified calderas in Askja. The average azimuth of fissures and fractures in the area deviates significantly from the azimuth perpendicular to the direction of plate velocity. As this deviation decreases gradually northward, we suggest that the effect of the triple junction of the North American, Eurasian and the Hreppar microplate is a likely cause for this deviation. Shallow, tectonic earthquakes in the vicinity of Askja are often located in a relatively unfractured area between the fissure swarms of Askja and Kverkfjöll. These earthquakes are associated with strike-slip faulting according to fault plane solutions. We suggest that the latest magma intrusions into either the Askja or the Kverkfjöll fissure swarms rotated the maximum stress axis from being vertical to horizontal, causing the formation of strike-slip faults instead of the dilatational fractures related to the fissure swarms. The activity in different parts of the Askja fissure swarm is uneven in time and switches between subswarms, as shown by a fissure swarm that is exposed in an early Holocene lava NW of Herðubreið but disappears under a younger (3500–4500 BP) lava flow. We suggest that the location of inflation centres in Askja central volcano controls into which part of the Askja fissure swarm a dike propagates. The size and amount of fractures in the Kollóttadyngja lava shield decrease with increasing elevation. We suggest that this occurred as the depth to the propagating dike(s) was greater under central Kollóttadyngja than under its flanks, due to topography.     

Persistent activity and violent strombolian eruptions at Vesuvius between 1631 and 1944

During the period 1631–1944, Vesuvius was in persistent activity with alternating mild strombolian explosions, quiet effusive eruptions, and violent strombolian eruptions. The major difference between the predominant style of activity and the violent strombolian stages is the effusion rate. The lava effusion rate during major eruptions was in the range 20–100 m³/s, higher than during mild activity and quiet effusion (0.1–1 m³/s). The products erupted during the mild activity and major paroxysms have different degree of crystallization. Highly porphyritic lava flows are slowly erupted during years-long period of mild activity. This activity is fed by a magma accumulating at shallow depth within the volcanic edifice. Conversely, during the major paroxysms, a fast lava flow precedes the eruption of a volatile-rich, crystal-poor magma. We show that the more energetic eruptions are fed by episodic, multiple arrival of discrete batches of magma rising faster and not degassing during the ascent. The rapidly ascending magma pushes up the liquid residing in the shallow reservoir and eventually reaches the surface with its full complement of volatiles, producing kilometer-high lava fountains. Rapid drainage of the shallow reservoir occasionally caused small caldera collapses. The major eruptions act to unplug the upper part of the feeding system, erupting the cooling and crystallizing magma. This pattern of activity lasted for 313 y, but with a progressive decrease in the number of more energetic eruptions. As a consequence, a cooling plug blocked the volcano until it eventually prevented the eruption of new magma. The yearly probability of having at least one violent strombolian eruption has decreased from 0.12 to 0.10 from 1944 to 2007, but episodic seismic crises since 1979 may be indicative of new episodic intrusions of magma batches.     

Magma supply and storage at Kilauea volcano, Hawaii, 1956–1983

Shallow crustal magma reservoirs beneath the summit of Kilauea Volcano and within its rift zones are linked in such a way that the magma supply to each can be estimated from the rate of ground deformation at the volcano's summit. Our model builds on the well-documented pattern of summit inflation as magma accumulates in a shallow summit reservoir, followed by deflation as magma is discharged to the surface or into the rift zones. Magma supply to the summit reservoir is thus proportional to summit uplift, and supply to the rift zones is proportional to summit subsidence; the average proportionality constant is 0.33 × 10⁶ m³/γrad. This model yields minimum supply estimates because it does not account for magma which escapes detection by moving passively through the summit reservoir or directly into the rift zones.Calculations suggest that magma was supplied to Kilauea during July 1956– April 1983 at a minimum average rate of 7.2 × 10⁶ m3/month. Roughly 35% of the net supply was extruded; the rest remains stored within the volcano's east rift zone (55%) and southwest rift zone (10%). Periods of relatively rapid supply were associated with the large Kapoho eruption in 1960 and the sustained Mauna Ulu eruptions in 1969–1971 and 1972–1974. Bursts of harmonic tremor from the mantle beneath Kilauea were also unusually energetic during 1968–1975, suggesting a close link between Kilauea's deep magma supply region and shallow storage reservoirs. It remains unclear whether pulses in magma supply from depth give rise to corresponding increases in shallow supply, or if instead unloading of a delicately balanced magma transport system during large eruptions or intrusions triggers more rapid ascent from a relatively constant mantle source.