Abstract RED SEED stands for Risk Evaluation, Detection and Simulation during Effusive Eruption Disasters, and combines stakeholders from the remote sensing, modelling and response communities with experience in tracking volcanic effusive events. The group first met during a three day-long workshop held in Clermont Ferrand (France) between 28 and 30 May 2013. During each day, presentations were given reviewing the state of the art in terms of (a) volcano hot spot detection and parameterization, (b) operational satellite-based hot spot detection systems, (c) lava flow modelling and (d) response protocols during effusive crises. At the end of each presentation set, the four groups retreated to discuss and report on requirements for a truly integrated and operational response that satisfactorily combines remote sensors, modellers and responders during an effusive crisis. The results of collating the final reports, and follow-up discussions that have been on-going since the workshop, are given here. We can reduce our discussions to four main findings. (1) Hot spot detection tools are operational and capable of providing effusive eruption onset notice within 15 min. (2) Spectral radiance metrics can also be provided with high degrees of confidence. However, if we are to achieve a truly global system, more local receiving stations need to be installed with hot spot detection and data processing modules running on-site and in real time. (3) Models are operational, but need real-time input of reliable time-averaged discharge rate data and regular updates of digital elevation models if they are to be effective; the latter can be provided by the radar/photogrammetry community. (4) Information needs to be provided in an agreed and standard format following an ensemble approach and using models that have been validated and recognized as trustworthy by the responding authorities. All of this requires a sophisticated and centralized data collection, distribution and reporting hub that is based on a philosophy of joint ownership and mutual trust. While the next chapter carries out an exercise to explore the viability of the last point, the detailed recommendations behind these findings are detailed here.

Abstract George Patrick Leonard Walker (1926–2005) was a key figure in the development of volcanic geology and volcanology through his exceptional and broad-reaching contributions. The early part of his career was dominated by an innovative long-term study of the volcanic geology of eastern Iceland and he continued to contribute to the volcanic geology of many areas worldwide. The second part of his career was pivotal in turning volcanology from its previous descriptive style into a modern quantitative science through his quantitative studies of pyroclastic deposits, lava flows, hypabyssal intrusions and volcanic processes. His success rested on his flair for meticulous observation, insistence on measurement, and keen intuition to generate major advances in understanding. He had the ability to merge systematic and comprehensive data sets with novel conceptual models to yield fundamentally new insights. His work was characterized by its extreme originality and broad scope, and forms the underpinning of much of our modern understanding of how volcanoes erupt. He was an inspiration to many colleagues and students. He influenced volcanology not only by research, but also by his genius as a teacher.

Abstract Lava-rise structures form by endogenous growth where lava is injected under, and lifts up, its surface crust. They are especially common in pahoehoe where it flows on shallow gradients (.48). Examples are described from the Mauna Loa (Hawaii) 1859 and Xitle (Mexico) 1.5 ka basalt flow-fields, and range from ten to hundreds of metres wide by up to 13 m high. The lobes that host the lava-rise structures were initially thin (≥1 m) and it is estimated that at least 80% of their total volume was emplaced by subcrustal injection. They are bounded by lava-rise edges that are highly distinctive margins with steep banded and striated surfaces (BSS) upon which lava inflation was accommodated. The BSS are emergent surfaces formed where newly injected hot lava cooled and solidified against the air. The surface textures indicate the direction of motion of the lava crust and edges relative to the point of emergence. Some BSS are steep and inwardly dipping reverse faults. Others are divergent-facing pairs of curved surfaces, the walls of lava inflation clefts. Many such clefts commonly transect the main BSS and subdivide them into stacks of lava wedges. A steep foliation defined by the plane of flattening of vesicles also occurs and bulges out into the wedges, attributed to a general expansion of the lava-rise. ‘Lava-rise sutures’ occur between contiguous lava-rises or lava lobes that inflated synchronously against each other. They are common where cross-sections through pahoehoe lavas are seen, such as in flood basalt provinces. In cross-section these features have the aspect of facing stacks of lava wedges, previously interpreted in the Xitle lava and elsewhere to be spiracles (steam-escape structures).

Abstract Walker (1973 ; Phil. Trans. R. Soc. Lond ., 274 , 107) argued that, for a limited set of compositions and flow types, effusion rate ( E ) was the principal influence on flow length, sparking a series of studies into the volume and cooling limits on flow extension. We here review these works, as well as the role of heat loss in controlling flow length. We also explore the applicability of Walker's idea to a larger compositional and morphological range. Heat loss plays a fundamental role in determining flow core cooling rates, thereby influencing cooling-limited flow length. Field measurements allow classification of four flow types with respect to heat loss. In this classification as we move from poorly insulated to well insulated regimes, decreased heat losses increase the length that a flow can extend for a given E , composition, morphology, or amount of cooling: (1) immature tube-contained, basalt - thin tube roofs provide minimal insulation, allowing cooling rates of c . 10 −2 ° C s −1 so that at low E , these flows extend only a few hundred metres; (2) poorly crusted, basalt - open channels with hot surface crusts also exhibit cooling rates of c . 10 −2 ° C s −1 so such flows extend a few kilometres at E < 1m 3 s −1 ; (3) heavily crusted, da-cite - heat losses are reduced when thick crusts form, reducing core cooling rates to c . 10 −4 ° C s −1 so these flows can potentially extend several kilometres even at low E and despite very high viscosities (10 9 −10 10 Pa I); (4) master tube-contained, basalt - thick tube roofs insulate flow, reducing heat losses and cooling rates to c . 10 −3 ° C s −1 . These cooling rates mean that at low BE, tube-contained flows can extend tens to hundreds of kilometres. Basically, if composition, insulation, and morphology are held constant flow length will increase with effusion rate.

Abstract Surtsey volcano of the south coast of Iceland (7 November 1963 to 5 June 1967), best known for its ‘Surtseyan’ eruptions, also featured prolonged episodes of effusive activity producing two partly overlapping pahoehoe lava shields. Here this effusive activity is examined to assess the overall volcanic architecture of the Surtsey lava shields and the mechanisms involved in their construction. The lava cone and apron are the principal structures of the shields. The cone is constructed during periods of relatively high magma discharge and vigorous fountain activity, producing surface flows of shelly pahoehoe and fast moving (up to 20 m s -1 ) slabby pahoehoe to aa sheet flows. The apron is formed during periods of passive lava effusion when the level of the lava stands well below the crater rims and lava is discharged through preferred internal pathways (i.e. lava tubes) to active flow fronts where it breaks out as inflating sheet lobes. The volcanic structure and architecture of monogenetic pahoehoe lava shields elsewhere in Iceland is essentially identical to the shields at Surtsey and it is proposed that they were constructed in a similar manner. A key conclusion is that the proportional size of the lava cone is a function of the average effusion rate. This relationship implies that shields with large cones were produced by relatively short eruptions (years to decades?) characterized by elevated, but fluctuating, magma discharge. Conversely, shields with small cones were produced by prolonged eruptions (decades to centuries?) typified by low and steady magma discharge.

Abstract The olivine alkali basalt flow field at Surtsey volcano consists of two pahoehoe lava shields largely made up of inflated sheet lobes and intercalated surface breakouts. Each sheet lobe exhibits a three-fold structural division into upper crust, core and lower crust, where the core corresponds to the liquid part of an active pahoehoe flow lobe sealed by the continually growing crusts. Segregations are common in Surtsey lavas and are confined to the core of individual lobes. Field relations and volume considerations indicate that segregation is initiated by generation of volatile-rich melt at or near the lower crust to core boundary via in-situ crystallization. Once buoyant, the segregated melt rose through the core during last stages of flow emplacement and accumulated at the base of the upper crust. The segregated melt is preserved as vesicular, coarse-grained material, of FeTi basalt composition, within vesicle cylinders, horizontal vesicle sheets and as floor-fill in megavesicles. Our results indicate that the segregated melt evolved from the host lava by 50–60% fractional crystallization of olivine, plagioclase and clinopyroxene. The eruption temperature of the host lavas was c . 1160 °C, whereas the segregated melt formed at c . 1130 °C and crystallized down to c . 750 °C. The oxygen fugacity during crystallization varied from approximately FMQ-buffer to values two order of magnitude lower at the time of final solidification. The reducing conditions were favourable for crystallization of olivine, which forms a near-continuous compositional trend from Fo 85 to Fo 13 . Feldspar compositions range from An 79 Or 0 to An 2 Or 57 and clinopyroxene compositions range from diopside-rich augites to aegerine–augites. Other mineral phases are magnetite, ilmenite and apatite, along with trace amounts of aenigmatite and nepheline. Vesicle cylinders in the lava core also contain residual phonolite glass of variable peralkalinity. The segregations have a strikingly similar whole-rock composition to the FeTi basalts from Katla and other volcanoes in South Iceland. These basalts are of relatively evolved composition with high melt densities. Volatile induced liquid transfer, as we propose for the formation of the segregations, may play an important role during magma differentiation and may explain the abundance of FeTi basalts in Iceland. Finally, closed-system fractional crystallization in Surtsey lavas produced melts of composition very close to the agpaitic lujavrites from the Illimaussaq intrusion in South Greenland.

Abstract The eruption of El Jorullo (1759–1774) in Guanajuato, Mexico, generated substantial (100–300 m high) pyroclastic cones, an extensive ash blanket and a flow field of thick lavas. The cones have the aspect of scoria cones that result from Strombolian eruptions, but the ash blankets consist predominantly of sub millimetre-sized particles (comprising ≥80 wt% beyond 1 km from the vent). This combination of cones, fine deposit grain size, and moderate dispersal area has previously been attributed to ‘violent Strombolian’ eruptions. The ash blanket at El Jorullo comprises c . 40% of the erupted volume and contains hundreds of strictly parallel laminae, evidence for deposition by fallout from a great number of explosions such as those observed during the eruption of nearby Parícutin volcano (1943–1952). The high degree of fragmentation could have resulted from hydromagmatic activity, but the deposit mostly lacks evidence for significant involvement of external water. We consider that the predominantly fine grain size was probably produced by a combination of a high yield strength and viscosity of the erupting magmas, possibly high juvenile water content, and recycling and milling of pyroclasts within the vent. The lava flow field at El Jorullo constitutes c . 40% of the erupted volume. Eight major flows vary from 10 to 50 m thick. Flow thickness and yield strength (calculated from flow profiles) increased with time from c . 30 000 Pa for the earliest flows to c . 200 000 Pa for the latest.

Abstract During the climactic Plinian phase of the 1886 basaltic eruption of Tarawera, New Zealand, vents along the 17 km fissure erupted explosively with a wide range of dispersal. The 8 km long segment of the fissure which cuts across Mt Tarawera contains approximately 50 vents and includes the sources of both the weakest and most intense activity of the 5 h eruption. We seek to explain (1) what allowed the intensity to reach Plinian values that are rarely achieved by basaltic magma, and (2) what caused adjacent vents to erupt with very different dispersals and intensities despite identical magma composition. All juvenile clasts studied from this eruption have relatively high vesicle number densities ( c . 10 6 cm -3 ) and exceptionally high microlite crystallinities (60–90% of the groundmass), unlike the typical products of weaker Hawaiian and Strombolian basaltic explosions. Textural analysis of juvenile pyroclasts suggests that all the erupted magma experienced the same decompression history through to fragmentation. The Tarawera magma experienced a sudden, large, decompression producing nucleation of bubbles and microlites. The high microlite content was the primary means by which the magma’s viscosity increased, which kept the bubble population well coupled to the magma and allowed it to fragment explosively in a manner analogous to that postulated for silicic Plinian eruptions. The main differences at different sites along the Mt Tarawera fissure segment are in the amount and grain size of the wall rock lithic component of the deposits. We suggest that conduit/vent erosion and incorporation of significant volumes of cold wall rock into theeruptive jet prevented some vents from achieving Plinian intensity. Bubble size analysis suggests that coalescence ledto open-system degassing, ending the Plinian phase.

Abstract Vulcanian eruptions are frequent, small-scale, short-lived explosive volcanic eruptions, which are thought to be produced by impulsive sources. The experiments presented here, produced by injections of mixtures of water, alcohol or salt and solid particles into fresh water, created a wide variety of turbulent flows from steady and impulsive sources. We focus on the experimental flows analogous to Vulcanian events – unsteady, finite-volume releases of buoyancy (thermals) and momentum (puffs), and short releases driven by both momentum and buoyancy. Dimensional analysis, based on two controlling source parameters, total injected momentum ( M ) and total injected buoyancy ( B ), identified a universal scaling relationship for the propagation of the flows; the non-dimensional, time-varying velocity term ( ut 1/2 ), where u is flow front vertical velocity and t is time from flow onset, varies with the time-varying, non-dimensional ratio of source parameters ( M/Bt ), such that ut 1 / 2 / B 1 / 4 = 14; k(M/Bt) 1 / 2 . The quantitative relationship successfully describes experiments and several Vulcanian eruptions for a wide range of initial conditions. The utility of the relationship is demonstrated by estimating total mass erupted and vent mass flux as a function of time, two parameters important to hazards assessment, for the well-documented 7 August 1997 Vulcanian eruption at Soufrie`re Hills volcano, Montserrat. Results compare favour-ably to independent estimates of total mass erupted (based in part on deposit studies) and eruption duration (based on seismic analysis) with the advantage that our approach requires only the determination of the eruption cloud front velocity from conventional video analysis.

Abstract Systematic measurements of the spatial variations of pyroclast grain size distributions in fall deposits are used to infer the mass eruption rates, eruption cloud heights, gas eruption speeds and released magma volatile contents during 31 explosive eruptions in regions where such events are common. The compilation of eruption cloud heights in this way is an important input to statistical models predicting volcanic hazard due to pyroclast dispersal. The estimates of mass eruption rate are found to be generally quite reliable, with estimated errors of no more than a factor of 2, corresponding to implied errors of c. 20% in eruption cloud height and clast dispersal. The estimates of magma volatile content and eruption speed, however, are very much less reliable. This highlights the extreme difficulty, despite meticulous fieldwork, of obtaining enough measurements of pyroclast size variations very near to a vent to allow good estimates of this parameter to be made. This is unfortunate, because improved values for released magma gas contents would allow quantification of the likely stability of eruption columns against collapse to form pyroclastic density currents, which would add an extra dimension to hazard modelling.

Abstract Ground-penetrating radar (GPR) is used to image the sub-surface architecture of the 1993 pumice flow deposits at Lascar volcano, northeastern Chile. This non-invasive geophysical technique allowed the determination of the three-dimensional, stratigraphic and facies variations of a deposit that is morphologically pristine but problematic to view internally. The geometry, sedimentological characteristics and compositional nature of the deposit make it ideal for investigation by GPR, which has been successfully used to map deposit-scale variations as well as detailed outcrop-scale features. This work both compliments and extends the interpretation of flow emplacement dynamics postulated for these flows by previous studies. Deposit shape, erosional and non-erosional contacts, buried units and stacked lobes are all informative deposit characteristics that can be readily ascertained from sub-surface imaging. Without such techniques, we are commonly unable to visualize these important features. In a broader context, we present this work in an attempt to demonstrate the suitability of GPR as a tool for studying the emplacement dynamics of pyroclastic currents and to encourage the diversification of applied field techniques in volcanology.

Abstract The probability of inundation of vulnerable areas by pyroclastic flows and surges generated by lava dome collapse at Soufriere Hills Volcano, Montserrat, is assessed. The runouts of the flows are simulated by two coupled one-dimensional models, the first representing the valley-following block-and-ash flows (avalanches) and the second the surges orthogonal to them. An ensemble of simulations is developed that honours the observed runout–frequency relationship of the flows and the likelihood of flows originating from a particular part of the dome. The former relationship is based on the observed eruption history from 1995 to 1998 and the second from 1995 to 2003. The avalanche component of the models is parameterized using a random selection from a library of triads of friction coefficients that produce simulation fits to field-observed flow deposits at 1 km runout increments.

Abstract The central Taupo Volcanic Zone (TVZ) in New Zealand currently is the most frequently active and productive area of silicic (overwhelmingly rhyolitic) volcanism on Earth. From 1.6 Ma to present, 25 caldera-forming eruptions have occurred, largely represented by ignim-brites, of which 23 are of dacitic to rhyolitic composition ( c . 65-77% SiO 2 ). These eruptions together represent c . 6000 km 3 of magma, but this record is known to be incomplete. Other large (but as yet unquantified) eruptions are recorded in deep-sea cores and now-uplifted Pleistocene marine sediments, but correlations to mapped ignimbrite sheets are incomplete and volumes unknown. From 61 ka to present, a mostly complete sequence is known from exposed deposits and 68 eruptions, totalling 782 km 3 of magma, are catalogued here. In this time period, three eruptions (1.8 ka, 35 km 3 ; 26.5 ka, 530 km 3 ; 61 ka, 80 km 3 ) were rhyolitic caldera-forming events that represent 82% of the volume erupted. The remaining 65 eruptions (137 km 3 , magma) consist of three basaltic, eight dacitic to rhyodacitic and 54 rhyolitic (≥72% SiO 2 ) events ranging in volume from <0.01 to 17.5 km 3 . Average central TVZ eruption rates from large caldera-forming events alone are 3.8km 3 kyr −1 , while since 61 ka the rate is 12.8 km 3 kyr −1 . The former value is similar to those for intracontinental silicic volcanic provinces like Yellowstone (2.1 Ma to present, 3.0 km 3 kyr −1 ) and the Southern Rocky Mountain volcanic field (29.4-26.9 Ma, 5km 3 kyr −1 ), where there is no clear record of smaller events. Magnitude-frequency relationships for large caldera-forming events in the central TVZ are broadly similar to those in the Southern Rocky Mountain volcanic field. However, the TVZ record since 61 ka shows magnitude-frequency relationships that imply an order-of-magnitude greater frequency in this latest period of activity for events of a given size up to >100 km 3 . TVZ eruptions are not evenly distributed in time, and clustering of events caused by interactions with tectonic processes are common.

Abstract Many calderas are the culmination of long-lived volcanic systems. Field-based studies provide detailed descriptions and interpretations of the origins of individual examples, while analogue and mathematical modelling provide insights about caldera formation. Caldera morphology and structure yield information on subsidence mechanisms and geometry of the associated magma chamber, while studies of eruptive products address aspects of magma composition and eruption dynamics. Combining field data with analogue and numerical modelling leads us to propose a genetic classification of calderas based on the stress conditions that permit formation of ring faults and the pressure evolution in the magma chamber during a caldera-forming eruption. Two main end-members, referred to here as overpressure and underpressure calderas, develop from different initial conditions and generate different sequences of caldera-forming deposits. With overpressure calderas, stress conditions leading to the formation of ring faults are achieved prior to initiation of the eruption when an overpressurized sill-like magma chamber is loaded by mag-matic regional doming or subjected to regional extension. Caldera collapse is initiated near at the beginning of the eruption and the resulting eruptions are often very large. By comparison, underpressure calderas result from ring fault subsidence after significant decompression of the magma chamber following a pre-caldera eruptive episode.

Abstract Textbooks in volcanology commonly emphasize the correlations observed between the volcanic and tectonic systems on Earth, implying direct cause-effect relationships between these two systems. When attention is focused on volcanic activity on other planets, moons and even asteroids, however, the causality implied by the correlations established on Earth becomes less evident. Questioning the cause-effect relationship between Tectonic and Volcanic systems on Earth constitutes the motivation to explore a different approach than that used in the past 30 years. In this paper I undertake this task based on a conceptual definition of a volcanic system that can be used regardless of tectonic scenario. A mechanical analogy of the planetary system serves as guide to insert fundamental physical principles in the conceptual framework, resulting in a hydrostatic model of volcanism (HMV). The HMV is used to constrain the thickness of magma source regions that is most likely to sustain volcanic activity as a function of depth, and other aspects concerning the temporal evolution of a volcanic system. Those quantitative estimates are found to be in agreement with real observations made on Earth, leading to the conclusion that, by decoupling the tectonic and volcanic systems, it is possible to provide a better explanation of the origins of volcanism not only on Earth but on other planets of our solar system as well.

Abstract Large igneous provinces (LIPs) form in both oceanic and continental settings by the emplacement and eruption of voluminous magmas ranging from basalt to rhyolite in composition. Continental flood basalt provinces are the best studied LIPs and consist of crustal intrusive systems, extensive flood lavas and ignimbrites, and mafic volcaniclastic deposits in varying proportions. Intrusive rocks are inferred to represent the solidified remnants of a plumbing system that fed eruptions at the surface, as well as themselves representing substantial accumulations of magma in the subsurface. The vast majority of intrusive rock within the upper crust is in widespread sills, the emplacement of which may structurally isolate and dismember upper crustal strata from underlying basement, as well as spawning dyke assemblages of complex geometry. Interaction of dykes and shoaling sills with near-surface aquifers is implicated in development of mafic volcaniclastic deposits which, in better-studied provinces, comprise large vent complexes and substantial primary volcaniclastic deposits. Flood lavas generally postdate and overlie mafic volcaniclastic deposits, and are emplaced as pahoehoe flows at a grand scale (up to 10 4 km 2 ) from eruptions lasting years to decades. As with modern Hawaiian analogues, pahoehoe flood lavas have erupted from fissure vents that sometimes show evidence of high lava fountains at times during eruption. In contrast to basaltic provinces, in which volcaniclastic deposits are significant but not dominant, silicic LIPs are dominated by deposits of explosive volcanism, although they also contain variably significant contributions from widespread lavas. Few vent sites have been identified for silicic eruptive units in LIPs, but it has been recognized that some ignimbrites have also been erupted from fissure-like vents. Although silicic LIPs are an important, albeit less common, expression of LIP events along continental margins, the large volumes of easily erodible primary volcaniclastic deposits result in these provinces also having a significant sedimentary signature in the geologic record. The inter-relationships between flood basalt lavas and volcaniclastic deposits during LIP formation can provide important constraints on the relative timings between LIP magmatism, extension, kilometre-scale uplift and palaeoenvironmental changes.

Abstract Over 30 years ago G. P. L. Walker wrote an article in Nature on excess spreading axes and spreading rate in Iceland. His statement that ‘the spreading rate is several times greater in part of Iceland than elsewhere on the Mid-Atlantic Ridge’ was immediately rejected and has rarely been discussed since. The problem of excess spreading in Iceland has appeared in various ways in the geological data obtained since that time. Here the drift (spreading half-rates) of several central volcanoes is studied to find out the behaviour of the spreading in Iceland. All the volcanoes seem to be drifting faster than the accepted spreading rate in the North Atlantic. The mean rate is 70% greater than on the ocean floor around Iceland. The spreading has been essentially symmetric about the rift zone axis. The enhanced spreading seems to evince a steady and long-lasting geological process because the behaviour of young and ancient volcanoes is similar.

Abstract The surface expressions of most Holocene rift-zone volcanic systems in Iceland are 40–150 km long a nd 5–20 km wide swarms of tension fractures, normal faults and basalt volcanoes each of which extends from a central volcano (a composite volcano or a caldera). Below the Holocene surface, the swarms are mainly composed of subvertical dykes and normal faults except within the central volcanoes where the main tectonic elements are inclined sheets. The inclined sheets are mostly 0.5 m thick, whereas the regional dykes are commonly 3–6 m thick and occasionallyas thick as 50–60 m. The two principal ways by which rift-zone volcanic systems become loaded are (1) the magmatic overpressure induced by dykes, and (2) the plate ‘pull’ associated with extension in the direction of the spreading vector (1058). This paper shows that both loading conditions give rise to mechanical interaction between volcanic systems in general, and their central volcanoes in particular. Here we show that the magmatic overpressure ofa regional dyke may reach tens of mega-pascals. We model the effects of simultaneous dyke injections, each dyke with an overpressure of 10 MPa, in the echelon systems on the Reykjanes Peninsula. The results indicate north-trending zonesof high shear stress between the nearby ends of the volcanic systems, favouring strike–slip faulting. Geometrically similar shear-stress zones develop between the volcanic systems on the peninsula when acted on by a plate pull of 5 MPa in a direction parallel with the spreading vector. The results agree with the observation that there are many north-trending strike–slip faults on the Reykjanes Peninsula. When the same plate pull is applied to a cluster of eight central volcanoes in Central Iceland, zones of high tensile stress develop between many of the volcanoes. These highly stresses zones encourage mechanical interaction between the volcanoes, such as simultaneous dyke emplacement and seismogenic faulting, as is supported by observational data.

Abstract The Geitafell Volcano, an extinct Tertiary central volcano, was active in the Icelandic rift zone between 5 and 6 Ma. Because of deep glacial erosion, the interior of the volcano is exposed from its flanks to the top of the extinct crustal magma chamber represented by several gabbro plutons. Here we present the results of a detailed study of the infrastructure of the Geitafell Volcano. The study includes measurements of the shape, size and fracture pattern of the extinct crustal magma chamber and its remarkably sharp contact with a dense swarm of inclined sheets for which the chamber acted as a source. A comparison of the attitude of 1087 joints within the basaltic magma chamber (the pluton) with the attitude of 48 sheets cutting the chamber indicates that the injection of late-formed sheets from the still-molten inner part of the crustal magma chamber was structurally controlled by the cooling joints in the solidified outer part of the chamber. Similarly, the main trends of 408 mineral veins of the fossile geothermal system, ENE–WSW, NW–SE and NNE–SSW, coincide with those of cooling joints, suggesting that geothermal fluids circulated through the heat source of the geothermal system via cooling joints. Our study of 592 sheets from the dense swarm surrounding the crustal magma chamber suggests that most are inclined sheets whereas some are radial dykes. Their thickness distribution follows a negative exponential law with an average arithmetic thickness of 0.64 m. Most inclined sheets dip from 40 to 70° with the general dip decreasing with increasing distance from the magma chamber. The sheet swarm is generally bowl-shaped and concave upwards. Reconstruction of the magma chamber indicates that at its maximum size it was about 8 km in diameter, 1–2 km thick and with a top at 1.5 km depth below the surface. Numerical models using this magma-chamber geometry as a basis can formally account for the intensity and geometry of the sheet swarm of the Geitafell Volcano.