LETTERS PUBLISHED ONLINE: 27 JULY 2015 | DOI: 10.1038/NGEO2491 Triggered earthquakes suppressed by an evolving stress shadow from a propagating dyke Robert G. Green * , Tim Greenfield and Robert S. White Large earthquakes can generate small changes in static stress: increases that trigger aftershock swarms, or reductions that create a region of reduced seismicity—a stress shadow 1,2 . However, seismic waves from large earthquakes also cause transient dynamic stresses that may trigger seismicity 3,4 . This makes it difficult to separate the relative influence of static and dynamic stress changes on aftershocks. Dyke intrusions do not generate dynamic stresses, so provide an unambiguous test of the stress shadow hypothesis. Here we use GPS and seismic data to reconstruct the intrusion of an igneous dyke that is 46 km long and 5 m wide beneath Bárðarbunga Volcano, central Iceland, in August 2014. We find that during dyke emplacement, bursts of seismicity at a distance of 5 to 15 km were first triggered and then abruptly switched off as the dyke tip propagated away from the volcano. We calculate the evolving static stress changes during dyke propagation and show that the stressing rate controls both the triggering and then suppression of earthquake rates in three separate areas adjacent to the dyke. Our results imply that static stress changes help control earthquake clustering. Similar small static stress changes may be important for triggering seismicity near geothermal areas, regions being hydrofractured and deflating oil and gas fields. It is widely reported that regions of abundant aftershocks (or advanced main shock recurrence) following large earthquakes correlate spatially with the small static stress increases produced by permanent fault displacements 1,4,5 . Dynamic stress radiation patterns have also been invoked to explain aftershock clustering 3 , but dynamic triggering cannot impart a stress shadow that would reduce seismicity in response to stress decrease at a given location 6 . Many studies have suggested that in regions where the static stress is decreased, aftershocks are rare or seismicity rates are reduced as a consequence of the negative stress shadow caused by the fault rupture 1,6–9 . However, convincing observations of the stress shadowing effect have been notoriously hard to demonstrate, and some have argued that there is a lack of evidence that they exist at all 3,10,11 . The challenge has been to demonstrate convincing and significant earthquake rate drops that correlate unambiguously with a stress decrease, because a high preceding seismicity rate is required. This task relies on the correct determination of static stresses. Unfortunately, geometrical irregularities in large faults result in a complex stress field that is difficult to resolve close to an active fault, thereby hampering the detection of a sharp seismicity decline within a strong stress shadow near the source. Alternative metrics of rate counting have also suggested the absence of rate drops following large earthquakes 10 . The existence of static stress shadows has therefore remained a contentious question. Here we provide clear evidence of the stress shadow effect, with three separate regions showing unambiguous seismicity rate decreases in response to negative static stress perturbations. On 16 August 2014 volcanic unrest began at the subglacial central volcano Bárðarbunga in Iceland, with a surge of intense seismicity in the caldera (Fig. 1). Rapidly migrating earthquakes delineated a propagating dyke, which moved first southeast radially away from the volcano, then turned a sharp corner and propagated to the northeast. Our well-constrained locations of over 30,000 earth- quakes, mostly near the leading edge of the dyke, track its varying rate of segmented lateral growth 12 (see Supplementary Movie 1). Over a 10-day period, the dyke propagated 46 km at an average depth of 7 km below sea level, before an effusive fissure eruption broke out at Holuhraun, 5 km north of the Vatnajökull ice cap. As the intrusion propagated, several regions adjacent to the dyke lit up abruptly with bursts of increased seismicity. Earthquakes on the northeast flank of Bárðarbunga (region 1, Fig. 1) started simultaneously with the initial southeasterly dyke tip migration, and earthquake swarms at Kistufell (region 2) and Kverkfjöll geothermal field (region 3) began soon after. These regions have all been historically seismically active 13 at low background levels, orders of magnitude smaller than the swarm levels. Regions 2 and 3 exhibited low but measurable rates in the months preceding the unrest. Earthquake rates increased 50-fold in region 2 and 100- fold in region 3 as the opening dyke caused stress increases. All three regions of seismicity subsequently terminated abruptly, each at different times. Modelling of the evolving coulomb stress perturbation from a combination of the opening dyke and a deflating source beneath Bárðarbunga Caldera shows that the onset of seismicity in these regions is triggered when the stressing rate begins to increase. The subsequent shut-off of seismicity in each region correlates well with the time when the stressing rate from the propagating dyke became negative at that locality. As the dyke tip advanced past the glacier edge elevated seismicity was triggered in the Askja region further north (Fig. 1) by the increasing stress. This seismicity did not subsequently shut off because the coulomb stresses remained positive ahead of the final dyke tip location. Coulomb stress calculations rely on knowing the trigger fault geometry, the rake and the coefficient of friction, as well as correct determination of the subsurface deformation from the intrusion. We generate a time-dependent deformation model of the Bárðarbunga deflation and dyke opening by integrating the earthquake locations, which constrain the daily geometry of active dyke segments, with the amount of opening determined from surface displacements at Global Positioning System (GPS) stations adjacent to the dyke 12 . The total dyke length is 46 km and opening occurs between 2 and 8 km below sea level, with variable opening seen in each of the segments (Supplementary Movie 2). The largest opening of 5.1 m occurred on the segment north of the Vatnajökull ice-cap margin. The total volume of the dyke was 0.55 km 3 , similar to that deduced in ref. 12. Bullard Laboratories, Department of Earth Sciences, University of Cambridge, Madingley Road, Cambridge CB3 0EZ, UK. *e-mail: rgcg3@cam.ac.uk NATURE GEOSCIENCE | VOL 8 | AUGUST 2015 | www.nature.com/naturegeoscience 629 © 2015 Macmillan Publishers Limited. All rights reserved