Lower crustal earthquakes (12–25 km depth) have been detected since August 2005 in the Askja volcanic system along the north Iceland rift, in the normally ductile part of the crust. The earthquakes occur in three clusters, which have stable dimensions and locations through time and are interpreted as positions of repeated melt supply from the mantle to the lower crust. Seismic velocity Vp/Vs ratios are consistent with the presence of partial melt in the lower crust at Askja. The spatial separation of the clusters shows that there are multiple positions of melt injection within this one magmatic segment and all three positions are currently active. This pattern of melt supply is more like that observed on fast spreading ridges than slow spreading ridges and is probably a consequence of the increased melt production beneath Iceland compared to the rest of the Mid-Atlantic Ridge. However, the relative number of earthquakes in each cluster shows that two thirds of the melt is supplied to the central volcano Askja (i.e., segment center). During the last major rifting episode shallow lateral melt migration occurred from the magma chamber beneath the volcano. Therefore on long time scales melt supply is probably greater at the segment center, with melt redistribution in the upper crust, even though there are multiple points of lower crustal injection along the segment.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Mantle flow models suggest that beneath slow-spreading ridges, melt is focused below the Moho and preferentially injected at spreading segment centers, from where it migrates at shallow levels towards the segment ends [Whitehead et al., 1984; Schouten et al., 1985; Magde and Sparks, 1997]. This is supported by crustal thickness variations [e.g., Lin and Phipps Morgan, 1992]. Seismic tomography has been used to infer positions of melt supply [Toomey et al., 1990; Magde et al., 2000; Dunn et al., 2000], but this method is sensitive to long-term average rather than instantaneous melt supply locations. The petrology of the lower crust and mantle exposed at oceanic core complexes has also been used to infer melt supply [Dick et al., 2008] but since the crust records all successive processes to which it has been subjected, it is difficult to determine timings for the injection positions observed. We report earthquakes recorded in the lower crust at the Askja volcanic rift system, interpreted as caused by melt movement, delineating patterns of ongoing melt supply from the mantle along a segment of the Mid-Atlantic Ridge.
 The mantle plume beneath Iceland causes increased melt production and consequently over-thickened crust, raising the Mid-Atlantic Ridge above sea level. As it crosses Iceland the rift splits into several en-echelon volcanic systems (inset Figure 1a), comprising central volcanoes intersected by fissure swarms [Sæmundsson, 1979]. These volcanic systems are analogous to the spreading segments of normal mid-ocean ridges. The last major rifting phase at Askja was in 1874–1875 with volcanic and tectonic activity at the central volcano and along the fissure swarm, creating the youngest of the nested calderas [Sigurdsson and Sparks, 1978]. The crust beneath Askja is ∼30 km thick, thinning away from the plume [Darbyshire et al., 2000]. Persistent upper-crustal (< 10 km) seismicity in the Askja region has been observed for more than 30 years and is attributed to tectonic faulting caused by plate spreading at 20 mm/a full rate [Einarsson, 1991; Soosalu et al., 2010].
 Data for this study come from successive local seismic networks deployed in the Askja area using Güralp 6TD broadband seismometers: a 3 week, 5 station network in August 2005; 2 month long summer campaigns of over 20 stations in 2006 and 2007 [Soosalu et al., 2010]; a 5 station trial network over winter 2007/08 and at least 15 stations deployed all year since summer 2008 (Figure 1a). Data from seven nearby national seismic network (SIL) stations run by the Icelandic Meteorological Office have also been used.
 The existence and characteristics of the lower crustal earthquakes at Askja were first reported by Soosalu et al. . Given that these earthquakes occur in a part of the crust that is normally aseismic and behaves in a ductile manner, we believe that they are caused by rapid melt movement generating sufficiently high strain rates to produce brittle failure. In this paper we present additional earthquakes and Vp/Vs ratios and focus on the larger scale interpretations that can be made about melt supply at spreading segments as a consequence of this discovery.
 The majority of the earthquakes occur in the upper crust (Figure 1b), with a sharp lower cut-off, representing the thermally controlled brittle-ductile transition at 6–8 km depth: for details see Soosalu et al. . Below the brittle-ductile transition the crust is sufficiently hot and ductile for tectonic stresses to be relieved before they cause earthquakes. However at Askja we have recorded earthquakes well within the ductile lower crust, mostly at depths of 12–25 km (Figure 1) and ML < 1.5.
 Over 1000 lower crustal earthquakes have been located using HYPOINVERSE-2000 [Klein, 2002], of which 796 are well constrained (horizontal and vertical errors < 1.0 km, azimuthal gap between stations < 180°, RMS time misfits < 0.2 s). They occur in three clusters in space (Animation S1 of the auxiliary material). The largest cluster, containing 65% of the earthquakes is beneath the northeast of the main Askja caldera, with its long axis extending along the fissure swarm (Figure 1a). This cluster looks like a continuous pipe when viewed in the across-rift profile (Figure 1d), but the along-rift profile (Figure 1c) and Animation S1 show that it is split into several smaller sub-clusters. The two remaining clusters are separated from the first by up to 10 km, one northeast and the other east of Askja. All three clusters of earthquakes can be active on the same day.
3. Time Distribution
 The lower crustal seismicity mostly occurs in swarms of fewer than 10 events lasting a few minutes, usually lacking a clear ‘mainshock’. Earthquakes within each swarm are located closely in space. They often follow in such quick succession that their codas interfere with other events in the swarm. On a day to week scale, the activity is episodic, some weeks having more than 100 events and others only a few. On longer time scales (weeks to months) the activity is continuous, with some lower crustal seismicity every week during the entire period we have been recording with a dense network.
 The lower crustal seismicity at Askja was first identified and thought to have started during 2006 [Soosalu et al., 2010] but re-examination of the data has found over 20 events during three weeks in August 2005. It is not possible to determine how long these events were occurring prior to 2005 because of a lack of seismometers close to Askja.
4. Vp/Vs Ratio
 The ratio of compressional to shear wave velocity, Vp/Vs, is sensitive to both temperature and the presence of partial melt. Wadati plots [Wadati, 1933] were used to determine Vp/Vs ratios, averaged over the area containing the ray paths between the event and the seismometers, from individual earthquakes (Figure 2a). Vp/Vs was found for all earthquakes with more than 10 stations with both P- and S-arrival time picks. The modal Vp/Vs ratio is 2.26 and the distribution is asymmetric, with a larger proportion of earthquakes having lower Vp/Vs than the peak (Figure 2b).
 We interpret the three clusters of earthquakes as zones of melt supply through the lower crust. The stable positions, shapes and dimensions of the clusters and the continuous occurrence of earthquakes for at least four years indicate that melt repeatedly follows the same pathways, probably through a network of cracks or dikes. Fresh pulses of magma that re-open cracks used by previous intrusions could generate each earthquake swarm.
 Earthquakes only occur when magma is moving, causing high strain rates. The sharp gaps in seismicity at the tops and bottoms of the clusters and sub-clusters are therefore locations where magma is stalling and are interpreted as sills. Our data provide support for a model of the lower and mid-crust built by in situ crystallization of multiple stacked sills [Kelemen et al., 1997]. This is not surprising given that two thirds of the melt supplied to mid-ocean ridges solidifies within the crust [White et al., 2008]. Lower crustal sills have been identified in the Oman ophiolite [e.g., Boudier and Nicolas, 1995]; imaged seismically at the Juan de Fuca Ridge [Canales et al., 2009] and the North Atlantic rifted margin [White et al., 2008]; and proposed on geochemical grounds for other volcanic systems of the north Iceland rift [Maclennan et al., 2001]. The bottom of the clusters below Askja and Vaðalda extend almost to the base of the crust calculated by Darbyshire et al.  (Figure 1d), but the bottom of the Kollóttadyngja cluster is much shallower (Figure 1c). It is possible that this difference in maximum depth could be explained by topography on the Moho, although it is more likely that the mantle melt supply recently ceased at Kollóttadyngja and that magma temporarily trapped in a 20 km deep sill is now moving to a shallower level.
 Lower crustal seismicity has only rarely been recorded in Iceland [Jakobsdóttir et al., 2008; Hjaltadóttir et al., 2009; Soosalu et al., 2010] and in all other cases is associated with discrete intrusion episodes lasting only days to months. In contrast, the Askja earthquakes are persistent and ongoing, representing a long-term feature. Although this is the first observation of such activity, it is possible that similar earthquakes are occurring elsewhere in Iceland, but are undetected thus far.
 Typical Vp/Vs ratios elsewhere in the Icelandic crust measured on seismic refraction profiles are mostly in the range 1.75–1.79 [Brandsdóttir and Menke, 2008, and references therein]. Due to the high geothermal gradient in Icelandic crust, the Vp/Vs ratio should increase slowly with depth. Using the best-fit gradient determined by Allen et al.  from broadband waveform inversions of local earthquakes in Iceland, Vp/Vs ratios should be in the range 1.83–1.89 at the depths of the Askja lower crustal earthquakes. Laboratory values of Vp/Vs for gabbros at suitable pressures for the lower crust are similar, at 1.85 [Christensen, 1996]. The high modal Vp/Vs ratio of 2.26 determined here therefore far exceeds that expected for normal Icelandic lower crust (Figure 2b). Since a significant proportion of the earthquakes have Vp/Vs ratios close to expected values in the lower crust, it is likely that the abnormally high values are caused by localized volumes with high Vp/Vs ratios, which are not sampled by all averaged source-receiver paths. Possible causes of increased Vp/Vs ratio at these depths are partial melt and/or high temperatures, but in order to create >20% increase in Vp/Vs ratio in the lower crust at least some melt must be present [Hammond and Humphreys, 2000]. Only a few percent of distributed melt is required to explain the high Vp/Vs values.
 The melt supply zones delineated by the Askja lower crustal earthquakes are ∼10 km wide, a similar size to velocity anomalies imaged with tomography by Magde et al.  on the Mid-Atlantic Ridge, thought to be high-temperature, low strength pathways left by repeated injection episodes of magma through the ductile lower crust. Zones of focused vertical transport through the shallow mantle exposed in the Kane Megamullion on the Mid-Atlantic Ridge are also 10 km wide [Dick et al., 2008]. The unique aspect of the melt injection reported here is that it has been recorded in real time. Since all three clusters have been active at the same time, melt must be simultaneously injected at three separate positions within a spreading segment. The first time this has been observed.
 There are some similarities with the ‘central injection zone’ often modeled for slow spreading ridges. The relative number of earthquakes in each cluster shows that the majority of melt is supplied beneath the central volcano, the ‘center’ of this magmatic segment. Some of the melt supplied in this location could eventually reach the geodetically modeled shallow magma chamber [e.g., Pagli et al., 2006] beneath the caldera, which is thought to have been the source of lateral dikes propagating up to 70 km during the last major rifting episode [Sigurdsson and Sparks, 1978].
 However, the multiple melt supply here is more complex than the single ‘central injection zone model’. It is unlikely that the melt in the two smaller clusters would ever feed the shallow magma chamber at Askja, instead it will freeze within sills and build the lower crust or might sometimes travel direct to the surface from depth. It is also possible that there are undetected injection zones, as this seismic array does not cover the entire length of the segment. Magma replenishment zones have been interpreted from tomography at intervals of 10–20 km along the East Pacific Rise [Toomey et al., 1990; Dunn et al., 2000], therefore at Askja, melt supply seems to resemble patterns at fast spreading ridges more than slow spreading ridges. This could be a result of the relatively high quantities of melt produced beneath Iceland by the mantle plume.
Figure 3 shows a cartoon summary for melt supply from the mantle to the crust of the Askja rift segment. Melt is focused in the mantle into 10 km wide regions and is injected into the lower crust in small pulses. It repeatedly follows previously used pathways and re-opens small cracks, generating the earthquakes we observe. This process is simultaneously occurring at multiple locations within the magmatic segment, not just at the segment center. The melt works its way shallower, stalling in a series of sills and crystallizing some material, the last remnants freezing in the mid-crust. Through time, magma supply positions may change, leaving behind a series of frozen stacked sills that build the lower crust. Despite multiple locations of melt supply, the majority of melt enters the lower crust at the center of the magmatic segment and time-averaged supply may be greater here then elsewhere along the rift. Although not currently occurring, sometimes melt injected at the Askja cluster makes it all the way to the upper crustal magma chamber, from which surface eruptions and long distance lateral dike injection may occur, building the shallow crust.
 Seismometers were borrowed from NERC GEF SEIS-UK, loans 842 & 857. We thank all those who helped with fieldwork and B. Brandsdóttir for assistance with logistics. Constructive reviews from Margaret Boettcher and an anonymous reviewer improved this manuscript. GMT software was used to draw figures. Dept. Earth Sciences contribution ESC.1980. The Editor thanks Margaret Boettcher and an anonymous reviewer.