Broad accommodation of rift-related extension recorded...

ARTICLESPUBLISHED ONLINE: 26 SEPTEMBER 2010 | DOI: 10.1038/NGEO966

Broad accommodation of rift-related extensionrecorded by dyke intrusion in Saudi ArabiaJohn S. Pallister1*, Wendy A. McCausland1, Sigurjón Jónsson2, Zhong Lu1, Hani M. Zahran3,Salah El Hadidy3, Abdallah Aburukbah3, Ian C. F. Stewart3, Paul R. Lundgren4, Randal A. White1

and Mohammed R. H. Moufti5

The extensive harrat lava province of Arabia formed during the past 30 million years in response to Red Sea rifting and mantleupwelling. The area was regarded as seismically quiet, but between April and June 2009 a swarm of more than 30,000earthquakes struck one of the lava fields in the province, Harrat Lunayyir, northwest Saudi Arabia. Concerned that largerdamaging earthquakes might occur, the Saudi Arabian government evacuated 40,000 people from the region. Here we usegeologic, geodetic and seismic data to show that the earthquake swarm resulted from magmatic dyke intrusion. We documenta surface fault rupture that is 8 km long with 91 cm of offset. Surface deformation is best modelled by the shallow intrusionof a north-west trending dyke that is about 10 km long. Seismic waves generated during the earthquakes exhibit overlappingvery low- and high-frequency components. We interpret the low frequencies to represent intrusion of magma and the highfrequencies to represent fracturing of the crystalline basement rocks. Rather than extension being accommodated entirely bythe central Red Sea rift axis, we suggest that the broad deformation observed in Harrat Lunayyir indicates that rift margins canremain as active sites of extension throughout rifting. Our analyses allowed us to forecast the likelihood of a future eruption orlarge earthquake in the region and informed the decisions made by the Saudi Arabian government to return the evacuees.

Northwestern Saudi Arabia (excluding the Gulf of Aqabaregion) has long been considered a seismically quiet region,with few felt earthquakes during the past millennium1,2.

Similarly, the region has had few historical volcanic eruptions: abasaltic eruption near Al Madinah in ad 1256 (refs 3,4), anecdotalaccounts of possible 10th century volcanic activity at HarratLunayyir (Al-Shaqah) (Harrat Al-Shaqah is the local name for thenorthern part of Harrat Lunayyir. As the name Lunayyir is betterknown in the international literature and refers to the entire basaltfield, it is used in this paper) and possible eruptions between the13th century bc and ∼600 ad at harrats Rahat, Khaybar, Rahaand Uwayrid5,6 (Fig. 1). A low incidence of instrumentally detectedseismicity in the region during the 20th century is probably aresult of the lack of permanent seismic installations in Saudi Arabiabefore 1984. With expansion of the Saudi national seismic networkover the past 25 years, it is now recognized that the region issubject to occasional earthquake swarms. Such swarms have mostevents<M4, and several have been spatially associatedwithArabia’sQuaternary volcanic fields. The swarms are generally attributed totectonic adjustments within the crystalline rocks of the Arabianshield as they respond to tectonic stresses related to Red Sea rifting,to strike-slip motion along the Gulf of Aqaba–Dead Sea transformfault, and to asthenospheric underflow andmagmatic intrusions7,8.

2009 seismic crisis and response at Harrat LunayyirWhen an unusually energetic swarm of more than 30,000earthquakes took place in April–June 2009 beneath HarratLunayyir, it generated an immediate response from the SaudiGeological Survey (SGS), which installed a permanent telemetered

1Volcano Disaster Assistance Program, U.S. Geological Survey, Cascades Volcano Observatory, 1300 SE Cardinal Court, Vancouver, Washington 98683,USA, 2King Abdullah University of Science and Technology (KAUST), Thuwal 23955, Saudi Arabia, 3Saudi Geological Survey, National Center forEarthquakes and Volcanoes, P.O. Box 54141, Jeddah 21514, Saudi Arabia, 4Jet Propulsion Laboratory, M/S 300-233, 4800 Oak Grove Drive, Pasadena,California 91109, USA, 5Faculty of Earth Sciences, King Abdul Aziz University, P.O. Box 80206, Jeddah 21589, Saudi Arabia. *e-mail: [email protected].

network of seven broadband seismometers within the harrat area(Fig. 1, Supplementary Fig. S1). During the peak activity on19 May, 19 earthquakes of M4.0 or greater struck the region,including an M5.4 event (M5.4 is based on the SGS catalogueof events from local seismic stations; reported by USGS NationalEarthquake Information Center as M5.7) at 17:35 utc, whichcaused minor damage to structures in the town of Al Ays (40 kmsoutheast of the epicentre). Concurrently, a northwest-trending8-km-long surface rupture propagated across the northern partof the volcanic field.

A key question facing scientists and public officials was whetherthe earthquakes were purely tectonic or might they foreshadow avolcanic eruption? As the swarm progressed, our monitoring dataindicated that a magmatic intrusion was continuing and a volcaniceruption was possible. Three principal lines of evidence favoured amagmatic source: (1) locations of the earthquakes beneath an areaof morphologically young vents, (2) surface deformation indicatingintrusion of a shallow dyke, and (3) occurrence of earthquaketypes characteristic of those accompanying volcanic intrusionsand eruptions elsewhere.

Despite indicators of possible eruptive activity, and aside froma potential localized CO2 hazard, there was little immediatevolcanic threat to the population because of the remoteness ofthe area. A relatively low volcanic hazard was also indicated bythe characteristic Strombolian to Hawaiian style of past eruptionsat the harrat, which produced relatively slow-moving lava flowsand modest amounts of ash. Although explosive hydrovolcaniceruptions are documented at other harrat fields9,10, the risk ofsuch events at Lunayyir is considered low because of the lack of a

NATURE GEOSCIENCE | ADVANCE ONLINE PUBLICATION | www.nature.com/naturegeoscience 1© 2010 Macmillan Publishers Limited. All rights reserved.

http://www.nature.com/doifinder/10.1038/ngeo966

mailto:[email protected]

http://www.nature.com/naturegeoscience

ARTICLES NATURE GEOSCIENCE DOI: 10.1038/NGEO966

32° N

36° E 38° E 40° E 42° E 44° E 46° E 48° E 50° E 52° E

30° N

28° N

26° N

24° N

22° N

20° N

18° N

400 km

16° N

14° N

Arabian shield

Arabian Sea

Red Sea

HarratLunayyir

HarratKhaybar

HarratRahat

HarratRaha

HarratUwayrid

Yanbu

Al Ays

Riyad

MakkahJeddah

Madinah

Dammam

Saudi Arabia

APM

DSF

Gulf of Aden

Figure 1 | Index map, showing the 180,000 km2 harrat lava fields (black) within Saudi Arabia. Rectangle at Harrat Lunayyir indicates study area.Double lines indicate spreading axes in the southern Red Sea and Gulf of Aden (and the axis of crustal extension in the northern Red Sea), dotted vectorsshow inferred asthenospheric flow directions (NW along the Red Sea axis and N along the Makkah-Madinah-Nafud volcanic line of Camp and Roobol37).‘DSF’ refers to the Dead Sea Fault. Solid vector labelled ‘APM’ refers to 28 mm yr−1 of absolute plate motion of Arabia41. Two-headed vector showsorientation of crustal extension in the Harrat Lunayyir area inferred from this study. Dash–dot line indicates national border of Saudi Arabia.Scale bar=400 km.

significant hydrologic basin in the vent area. The greater hazard isthat of additional damaging earthquakes.

The pattern of seismicity is best described as a volcanicearthquake swarm characterized by: several tens of thousandsof small earthquakes (most <M3 and occurring between 24April and the end of July 2009), periods of increased seismicactivity that wax and wane, earthquakes of varying magnitude,and lack of a tectonic mainshock-aftershock sequence (Figs 2and 3). Real-time seismic amplitude (RSAM; ref. 11) values(Fig. 3) show that the energy release increased dramaticallyon 29 April and peaked on 19 May with the largest event(M5.4, off-scale line in Fig. 3). RSAM values then decreaseddramatically on 20 May, and thereafter retained a relativelyconstant level, except for a few increases in event rate, whichlasted less than 24 h and were energetically smaller by anorder of magnitude.

Other characteristics of this episode that are common to volcanicswarms include: a high rate of occurrence of small events comparedto large events (b-value, 1.2), shallow event locations clusteredbeneath the lava field at depths of 5–10 km, and a mixture ofevent types12. The swarm includes high-frequency earthquakes(13–28Hz), very low-frequency earthquakes (VLF, <1Hz), andhigh-frequency tremor with embedded spasmodic bursts of highfrequency earthquakes (13–28Hz) (Fig. 2).

Unusual type of volcanic earthquakesSeveral features of the earthquakes in this swarm are unusual.First is the retention of very high frequencies (>20Hz) atlarge distances (10 to >100 km), which indicates low seismicattenuation and fracturing of well-consolidated rock. Second is thecoincidence of VLF earthquakes with some of the high-frequencyearthquakes (Fig. 2b and c) over a range of magnitudes. Mixed

2 NATURE GEOSCIENCE | ADVANCE ONLINE PUBLICATION | www.nature.com/naturegeoscience

© 2010 Macmillan Publishers Limited. All rights reserved.



NATURE GEOSCIENCE DOI: 10.1038/NGEO966 ARTICLES

(Min)

5 1510 20 25 30

02:00

05:00

08:00

11:00

14:00

17:00

20:00

23:00

05/19

05/20

Tim

e (U

TC

)

VLF

M4.8,VLF

M3.5,VLFVLF

VLF

VLF

VLF VLF VLF

VLFVLF

VLF

M4.2,VLF

M3.9,VLF

M4.9,VLF M5.4,VLF

0

16

8

Pow

er (

× 10

6 )

1

0

¬1

Time

10

20

30

40

01:51:30 01:51:40 01:51:50 01:52:00

Time

01:51:30 01:51:40 01:51:50 01:52:00

Time

01:51:30 01:51:40 01:51:50 01:52:00

Freq

uenc

y (H

z)

Frequency (Hz)1010.1

b c

Time

0

4

¬4

¬8

8

Frequency (Hz)

17:48 17:50 17:52 17:54 17:56

Time17:48 17:50 17:52 17:54 17:56

15

5

10

Pow

er (

×10

6 )Fr

eque

ncy

(Hz)

Cou

nts

(×10

4)

a

LNYS

LNYS

Cou

nts

(×10

5 )C

ount

s (×

105 )

2

0

¬2

LNYS

LNYS

LNYS

Unfiltered

Low pass filtered (<2 Hz)

LNYS

LNYS

10

20

30

40

100.1 1

Figure 2 |Digital helicorder records for 19 May 2009 from seismic station LNYS (located∼15 km SE of the fault rupture). a, Most earthquakes arehigh-frequency impulsive volcano–tectonic (VT) events with dominant frequencies of 13–28 Hz and wide magnitude range; some high-frequencyearthquakes have accompanying very low-frequency (VLF) earthquakes (orange boxes) or accompanying high-frequency tremor. Red lines indicate clippedamplitudes. Representative signals (yellow boxes) detailed below. b, VLF associated with an M3.5 earthquake. VLF signal clearly visible in the unfilteredseismogram (top), has P- and S-wave phases (second from top) and clear separation between the high and very low frequencies as seen in thespectrogram (second from bottom) and power spectrum (bottom). c, VT earthquakes within high-frequency and high amplitude background signal, whichwe call tremor. Tremor occurred in bands that lasted from tens of minutes to about 5 h during periods of increased seismic activity (high-frequencyspasmodic bursts).

frequency earthquakes are common at volcanoes13,14; however,they have previously been observed mainly as mixtures of onlyhigh and low-frequency energy, recordings of mixed high andvery low frequencies are unusual, although another example has

recently been documented at Augustine volcano, Alaska15. TheLunayyir mixed frequency earthquakes contain little energy inthe intermediate band (1–10Hz), they are also clearly recordedover short and long distances (10–140 km), both the HF and





4/4 4/14 4/24 5/4 5/14 5/24 6/3 6/13 6/23 7/3 7/13

UMJS 10-min RSAM

LNYS 10-min RSAM

No data

VLF (<0.5 Hz)LF (1-10 Hz)HF (>8 Hz)

VLFs detectible at both UMJS and LNYS

5/19M5.4

No data

100

1

10,000

10-m

in R

SAM

100

1

10,000

10-m

in R

SAM

5

10

15

0

Dai

ly V

LP c

ount

20% of InSAR 75% of InSAR Negligible InSARdeformation deformation deformation deformation

~5% of InSAR

Figure 3 | Ten-minute real-time seismic amplitude (RSAM) plots showing peak seismic energy release for the vertical component of broadband seismicstations UMJS and LNYS (45 km WSW and 29 km SE from the epicentre of the M5.4 earthquake). RSAM values generated from filtered seismic records:high frequency (>8 Hz,red), low frequency (1–8 Hz, green), and very low frequency (<1 Hz,blue). InSAR-detected deformation periods indicated in shadesof yellow: 75% of the measurable deformation occurred within the darker yellow box, 20% within the lighter yellow box. Negligible deformation precededand approximately 5% followed these time windows. Lower graph shows daily counts of VLF earthquakes recorded at both UMJS and LNYS stations. Amajority of VLFs occurred during the deformation period constrained by InSAR, suggesting they are related to fracture opening and dyke intrusion.

VLF components have P- and S-wave phases (SupplementaryFigs S2, S3) and they have different waveforms for different events.Both low and very low frequency events are typically attributedto movement of fluids (magma, water or gases) and strombolianexplosions16–18; however no gas emissions or explosions werereported at Lunayyir. Although several explanations formixedVT+VLF events are possible (for example, slow slip along lubricatedfaults, superposed tensile opening and shear slip, bubble-inducedresonance, surface breaks and tensile opening), we favour aninterpretation involving coincident brittle rock fracture and acombined tensile and shear response of the crust in response topulses of shallow magmatic intrusion. This hypothesis is supportedby the fact that the VLF events are only concurrent with high-frequency events and during the main period of deformation, andtherefore dyke intrusion (Fig. 3).We note that a similar explanationinvolving pulses of magma injection is proposed to explain the VLPcomponents of mixed frequency events at Augustine volcano15, andwe suggest that as the use of broad band networks continues toexpand, suchmixedVT+VLP eventswill be seen at other volcanoes.

High frequencies at volcanoes are commonly attenuated bypath effects within the poorly consolidated and complexly layereddeposits that make up most Quaternary volcanic terrains. Incontrast, the thin basaltic deposits of Lunayyir directly overliemetamorphic and plutonic rocks of the Precambrian ArabianShield. Consequently, we attribute the unusually high-frequencycomponents of the Lunayyir earthquakes to brittle fracture of thesecrystalline rocks and to low attenuation of high-frequency signals asthey transit the dense basement rocks.

Geologic observations and seismic momentDuring the week preceding 19 May 2009 when maximumearthquake magnitudes exceeded M4, local observers reported

that a 3-km-long northwest-trending surface rupture appearedin the north-central area of the harrat. The rupture lengtheneddramatically to 8 km along a strike of 330◦–340◦ during the 19May M5.4 earthquake (Fig. 4a). We measured the offset of thePrecambrian basement surface near the southern end of the surfacerupture on 16 June (Fig. 4b). This measurement gave an absolutefault motion of 91 cm along a 035◦-trending, 63◦-plunging vector,which resulted in 45 cm of tensional opening at the ground surface.In areas where the fault transected basin sediments, the ruptureformed an open trench, with local vertical offsets of more than ametre and exposed depths of several metres.

Using relationships between seismic moment and momentmagnitude19,20 and the amount of graben-bounding faulting atLunayyir, as estimated from InSAR data (see below), we calculatea moment magnitude of Mw6.0. In contrast, the cumulativeinstrumentally determined magnitude for located earthquakes (ofM1 and greater) between the bracketing dates (8 May and 1 July2009) for the InSAR data is Mw5.6, indicating that the majorityof the moment release (75%) for the intrusion was aseismic, arelationship also documented in other rift-dyking events21,22. Inaddition, and as has been done for other volcanic swarms13, wecan use the dimensions of the Lunayyir dyke intrusion fromInSAR modelling (Fig. 5) as a proxy for faulting in the momentcalculation. In this case, the resulting moment magnitude isMw6.4,which would indicate that as much as 93% of the overall momentrelease could be aseismic.

Focal mechanisms for selected earthquakes (M> 3.5) from theswarm were calculated using moment tensor inversion of P-wavepolarities, assuming a double couple source mechanism23. Use ofa double couple mechanism for these events is consistent withfull moment tensor inversions by others (for example, St LouisUniversity, http://www.eas.slu.edu/Earthquake_Center/MECH.




http://www.eas.slu.edu/Earthquake_Center/MECH.SA/REPORT



45 cm opening

78 cm down drop

91 c

m to

tal m

otio

n

035

° str

ike,

063

° plu

nge

45 cm opening

335° strike ~90° dip

a

b

Figure 4 | Photographs showing the surface rupture produced by the 19May 2009 earthquake. a, Rupture cutting recent sediments and basaltictephra in this small basin, located at the south end of the fracture system.Wide aperture of the rupture compared to offset of nearby basement rocksattributed to differential compaction of sediments and ravelling of rupturewalls. Trace of fault indicated by arrows; uppermost arrow is location of b.b, Offset of cemented colluvium immediately overlying Precambrianbasement rocks. Offset of acacia plant indicates absolute motion of 91 cmalong a N35E (035◦),63◦ plunging vector with tensional opening of 45 cm.Local trend of the fault plane is N25W (335o).

SA/REPORT), which show a negligible volumetric componentto the moment tensor and result in a similar focal mechanism.These solutions allow for direct comparison of focal mechanismsbetween the high-frequency content of the mixed frequency andthe high-frequency earthquakes, which we can then relate tothe field observations of the ground fracture. Using the inferredcompression and tension axes from the fault plane solutions, we

derive the average regional stresses for these events (Fig. 5a). Thereis considerable range in the resulting focal mechanism solutions;however, the average tensional axes are oriented ENE–WSW, ata high angle to the mapped ground rupture and consistent withtensional opening across the rupture.

InSAR data and geodetic modellingMultiple Envisat, ALOS, and TerrSAR-X satellite interferogramsbracket the timing of the seismic swarm. These InSAR dataconstrain the main episode of deformation to the period ofmaximum total seismic energy and maximum very low-frequencyseismic energy release, with 75% of the deformation taking placeduring 8–27 May 2009 and 20% between 27 May and 17 June(Fig. 3). InSAR analysis of TerrSAR-X interferograms acquired forthe period 17–28 June show only a small amount of deformation(estimated as about 5% of the total) and negligible deformationin July, although a few small earthquake swarms continued intoJuly. Interferograms spanning the May and June activity show abroad area (∼2,000 km2) of deformation, indicating both uplift andhorizontal extension that we model as the result of dyke intrusion.A combination of ascending and descending Envisat interferogramsindicates about 40 cm of uplift and well over 1m of east–westextension (Fig. 5 and Supplementary Fig. S4). The uplifted regionis transected by a northwest-trending graben, bounded on thesouthwest by the surface rupture and on the northeast by severaldiscontinuities in the interferograms. The graben is coincident withthe main cluster of epicentres from the earthquake swarm and withthe mapped ground rupture. A complex pattern of tightly spacedfringes in the graben indicate subsidence exceeding 50 cm. Thedeformation field is best modelled by intrusion of a ∼10 km-long,NW-trending (340◦) dyke, with a top at less than 2 km depth andvolume of about 0.13 km3 (Fig. 5d). Faulting on graben-boundingnormal faults is also required to attain an adequate fit of the InSARmodel. These results imply that during intrusion the dyke progres-sively migrated to shallow crustal levels and triggered a meter-scalefault slip on thewestern graben-bounding faults. Similar interactionbetween shallow dyke intrusions and faulting is reported in the Afarrift, Ethiopia22 and near Lake Natron, Tanzania21. Dyke intrusionand propagation has also been documented for the ad 1256Madinah eruption3. Although the InSAR and field observationsindicate dyke propagation, we are unable to clearly establishmigration of earthquake locations, probably because our networkconfiguration was not adequate for high precision locations untilrelatively late in the swarm (seeMethods section).

Volcanic eruption and earthquake magnitude forecastsInterpretation of these geologic and geophysical data during thecrisis not only confirmed the magmatic nature of the activity butalso enabled probabilistic forecasts of potential eruption hazardsand maximum earthquake magnitudes. On 19 June 2009, weforecast a moderate probability of a basaltic eruption and a lowprobability of additional large earthquakes of M5 or greater withinthe following two months. Our forecasts were based on maximummagnitudes of earthquakes that typically accompany eruptionsat other volcanoes24, on the proportion of eruptions followingmultiple intrusions at dozens of volcanoes studied by the USGSVolcano Disaster Assistance Team and on our initial models ofthe InSAR anomaly. Consistent with the findings of Parsons andThompson25 andwith the large fraction of aseismicmoment release,we interpret the paucity of large magnitude earthquakes during thisintrusive episode to be related to the role of magma overpressurein suppressing earthquake magnitude—in effect, dyke intrusionoccupies space and reduces deviatoric stress that would otherwiseresult in normal fault slip. Because of a decline in seismic activityand effective end of InSAR-detected deformation by August 2009,we conclude that the volcanic–tectonic crisis has ended. Subsequent







c

b

25° 12' N

25° 16' N

25° 20' N

25° 12' N

25° 16' N

25° 20' N

25° 12' N

25° 16' N

25° 20' N

37° 36' E 37° 40' E 37° 44' E 37° 48' E 37° 52' E 37° 56' E

37° 36' E 37° 40' E 37° 44' E 37° 48' E 37° 52' E 37° 56 E' 37° 36' E 37° 40' E 37° 44' E 37° 48' E 37° 52' E 37° 56' E

a

0

315 045

270

180

090

225 135

15

10

5

0

Dyke opening/slip (m)

10

8

6

4

2

0

Depth (km

)

d

a

Distance along strike (km)

LNY4

LNY3

LNY6

LNY2

LNY1

1 2 3 40

Figure 5 | Images showing seismic and deformation data from 2009 activity at Harrat Lunayyir. Composite map views a–c and cross section model forthe northern part of Harrat Lunayyir. a, False-colour satellite image with epicentres for 1,750 M≥ 2 earthquakes (yellow dots) between 13 May and 6October 2009. Dark areas are lava flows, slightly lighter circular areas are cinder cones, lighter greys are Precambrian crystalline rocks. Bold red linesindicate mapped surface ruptures. Focal mechanism inset is for the 19 May M5.4 event; red dots are upward first motions, blue are downward. Rosediagram gives average orientations of stresses derived from focal mechanisms for earthquakes with M4.0 to 5.4. Dark blue segments are tensionaldirections and light blue segments indicate the average tensional direction (078◦). Red segments are compression directions. b, Interferogram showing thedeformation that occurred in May and June 2009. Each fringe of this unwrapped interferogram is equivalent to a line-of-sight displacement of 10 cm.c, Modelled deformation pattern for intrusion of a vertical dyke (thick black line) with strike of 340◦ and faulting on graben-bounding normal faults (thinlines). See Supplementary Fig. S4 for the corresponding ascending-orbit InSAR observations and modelled deformation. d, Estimated dyke opening andnormal-fault slip distribution, shown in model cross-section.

to our hazard assessment and once the integrity of buildings in AlAys was addressed by a structural engineering team, the evacueeswere allowed to return to their homes and daily lives. However, asour data indicate that magma has risen to shallow levels, a pathwayis prepared for subsequent intrusions. Twelve intrusions over fiveyears following the 2005 mega-dyke event in the Afar illustratehow episodic periods of dyke intrusion are to be expected in riftboundaries26,27. Consequently, there is a need to remain vigilant forsigns of renewed activity. The SGS is continuing tomonitor the vol-canic field andwill issue advisories in the case of additional unrest.

Tectonic implications of the 2009 Lunayyir intrusionIn addition to their use in local volcano and earthquake hazardassessment, data from the Lunayyir intrusion provide unusualinsights into active volcanism and tectonics related to the RedSea28. The Precambrian crust of the Arabian-Nubian Shield wastectonically thinned by normal and detachment faulting andintruded by dyke swarms during the early stages (30–20Myr) ofrifting, forming a low-elevation rift valley29–31. Later, at about14Myr, asymmetric uplift of the eastern margin of the rift valleybegan, and only much later (about 5Myr) did organized sea-floorspreading begin along the southern Red Sea axis28,30–34. Organized

sea-floor spreading has not been established in the northern RedSea at the latitude of Lunayyir—this region remains a flooded riftbasin that is underlain mainly by highly extended and intrudedcontinental crust28,33,34. Consequently, although the adjacent crustof Arabia is technically a passive margin, as previous work hasshown, such margins are hardly ‘passive’ during early or prolongedperiods of rifting. In such situations, these are sites of activeextension and periodic magmatism26,29,35.

Uplift and the widespread alkali olivine volcanism of Arabiais related by most workers to asthenospheric upwelling, althoughmodels for the cause and extent of this process vary29,31,36,37. Theresulting harrat lava fields of Saudi Arabia are extensive, coveringabout 180,000 km2 (ref. 36). The most comprehensive study todate of Arabia’s volcanic fields suggests that volcanism along theMakkah–Medinah–Nafud volcanic line, including Harrats Rahatand Khaybar (Fig. 1), resulted from northward flow of astheno-spheric mantle along a pre-existing flexure in the continental litho-sphere (the ‘West Arabian Swell’)37,38. This hypothesis is consistentwith seismic anisotropy data, which show north–south orientedfast directions for the mantle beneath central Arabia, attributed toshear-alignment of olivine lattices at the base of the lithosphere39.This north–south orientation is interpreted as the vector sum






Figure 6 |Oblique aerial photograph showing basalt cinder cones and lavain northern Harrat Lunayyir. Basaltic tephra from these vents mantlesnearby ridges and forms fans at the base of the Precambrian granite ridgein the background. Well preserved geomorphic forms are indicative of therelative youth of this part of the volcanic field.

of northeast (040◦,22mmyr−1) Arabian plate motion40,41 andnorthwest directed (330◦) channelized asthenospheric flow37.

Our InSAR-derived deformation pattern, field measurementsof fault geometry and motion, and focal plane solutions forearthquakes provide a new means to evaluate crustal stress in thenorthwestern Arabian plate. The Precambrian basement in theregion is cut by lineaments and suture zones with prominent ENEand WNW trends42 at high angles to the NW-trending Lunayyirfault rupture. Although earthquakes have been recorded along someof these basement structures elsewhere in Saudi Arabia38, cross-cutting orientations of NW-trending Red Sea-related dykes43,44and dominantly N–S and NW–SE vent alignments in the volcanicfields6,9,36,37, as well as the mantle flow patterns deduced fromseismic anisotropy45, indicate that the modern stress field in thecrust of NW Arabia reflects primarily mantle flow dynamics ratherthan Precambrian basement structures.

The Lunayyir fault rupture (310◦–330◦) and InSAR-modelleddyke (340◦ strike, vertical dip) are sub-parallel to the Red Sea axis(∼330◦), aeromagnetically imaged Tertiary Red Sea rift dykes43 andalignment of volcanic vents in the Lunayyir field (320◦–340◦; ref. 6).Consequently, the recent Lunayyir deformation indicates thatcrustal stress has a principal tensional axis oriented perpendicularto the Lunayyir rupture, that is, striking approximately 070◦. Weconclude that in the northwestern coastal region of Saudi Arabiacrustal stress is controlled primarily by asthenospheric flow awayfrom the Red Sea rift axis, as opposed to channelized flow along theWest Arabian Swell, which plays amore significant role in the basaltfields farther to the east37. This result indicates a more distributedmode of extension that includes dyke intrusion in the rift margin,as opposed to accommodation entirely by axial spreading, as in

the southern Red Sea segment. Our focal solutions for the largerearthquakes in the Lunayyir swarm indicate a more northerlyoriented fault plane (∼348◦); however, there is considerable rangein these solutions, and we regard the overall pattern of deformationas themost reliable indicator that ENE–WSWdirected tension is thedominant direction of regional crustal stress.

The case for volcano monitoring in Saudi ArabiaAlthough the volcanic deposits at Lunayyir are not radiometricallydated, the report of a possible 10th century eruption5 and youthfulappearing geomorphic features of the deposits indicate recent ac-tivity. Young geomorphic features include pristine vent structures,unweathered lava surfaces, lava flows overlying Neolithic archeo-logical sites, and easily eroded tephra deposits still present atop thenearbywindswept Precambrian basementmountains (Fig. 6).

More broadly, Camp et al.3 noted 21 documented or inferrederuptions on the Arabian Peninsula within the past 1,500 years; thelatest in 1937 near Dhamar, North Yemen. In addition, subaerialeruptions with fatalities took place in 2007–2008 at Jabal at TairIsland, located along the southern Red Sea axis46. Basaltic lavas arethe most voluminous eruptive products of the harrats; however,more hazardous explosive eruptions produced voluminous tuffrings, pyroclastic flows and domes at two of the volcanic fields, andurban encroachment is increasing10,47. The 2009 intrusive episodeat Harrat Lunayyir, along with geomorphically young lava andtephra deposits (Fig. 6) are reminders that, although eruptionsare not frequent, the harrat fields remain active and potentiallyhazardous9,10. In response, the SGS is increasing its program ofvolcanicmonitoring, hazard assessment and communication.

MethodsSeismic methods. With the onset of the seismic swarm at the end of April,regional networks of broadband instruments were augmented with a localbroadband seismic network at Harrat Lunayyir. Instrument type, installationdate and locations are given in the Supplementary Table. Earthquake locationsand magnitudes were determined by SGS personnel using Nanometrics’ Atlassoftware and the regional velocity model. Errors in locations using the regionalnetwork are greater than 20 km, they reduce to ∼16 km for May 2–13, and afterthis date they improve to approximately 1 km with the completion of the networkat the harrat. The located event catalogue is complete over M3.5, but the seismicnetwork is capable of detecting and locating events greater than M1.5. RSAM(ref. 11) values were calculated as a proxy for energy release in 10-min windowsusing continuous data records for broadband stations LNYS and UMJS. RSAMvalues were determined in three frequency bands: high (>8Hz), low (1–8Hz), andvery low (<1Hz). The VLF component of the mixed frequency earthquakes wasdetermined using continuous data from two regional stations, UMJS and YNBS,and the first installed local station, LNYS. The continuous data were low passfiltered (zero-phase four-pole filter with a corner frequency of 1Hz) and visuallyinspected for VLF events. Only events that were visible on more than one stationwere counted as an event.

Geodetic methods. The ascending and descending interferograms were processedusing a standard 2-pass procedure with the topographic phase contributionremoved using the SRTM 3-arc-second digital elevation model. We thenunwrapped the phase within the interferograms and used a quadtree sub-samplingscheme to reduce the number of InSAR observations for the modelling. Theinteferograms shown in Fig. 5 are created from Envisat descending-orbit imagesdraped over a shaded-relief digital-elevation model.

Dislocation sources within an elastic halfspace were used to model openingof a dyke and slip on faults48. We estimated the dislocation parameters byminimizing the difference between the predicted surface displacements and theInSAR observations. This estimation was carried out in two separate steps. In thefirst step we assumed uniform opening/slip on each of the dislocation surfacesand used nonlinear optimization to search for the best dislocation parameters,that is, the location, strike, dip, dimensions, and opening of the dyke as well asthe dip, downward extension, and slip of the faults. The mapped surface faulttrace and discontinuities in the interferograms were used as constraints for thelocation, length, and strike of each fault segment. In the second step we usedthe optimum dyke dip, location, and strike, and the estimated fault dip to fixthe geometry of the model. We then solved for variable opening of the dyke andnon-uniform slip on the inward dipping fault planes using linear least-squaresminimization with positivity and smoothening constraints. The resulting modelis shown in Fig. 5d.





Received 5 February 2010; accepted 25 August 2010;published online 26 September 2010

References1. Ambraseys, N. N. & Melville, C. P. Evidence for intraplate earthquakes in

northwest Arabia. Bull. Seismol. Soc. Am. 79, 1279–1281 (1989).2. Ambraseys, N. N. Reassessment of earthquakes, 1900–1999, in the Eastern

Mediterranean and the Middle East. Geophys. J. Int. 145, 471–485 (2001).3. Camp, V. E., Hooper, P. R., Roobol, M. J. & White, D. L. The Madinah

eruption, Saudi Arabia: Magma mixing and simultaneous eruption of threebasaltic chemical types. Bull. Volcanol. 49, 489–508 (1987).

4. Moufti, M. R. H. The Geology of Harrat Al-Madinah Volcanic Field, HarratRahat, Saudi Arabia. 465, PhD thesis, Univ. Lancaster, England (1985).

5. Van Padang, N. M. Catalogue of the Active Volcanoes of the World, IncludingSolfatara Fields of Arabia and Indian Ocean, Vol. 16 1–64 (Int. Assoc. Volcanol.,1963).

6. Roobol, M. J. Geology, Structure and Seismicity of Harrat Lunayyir VolcanicField. Al-Eis region, northwestern Saudi Arabia. Saudi Geol. Surv. Rep. 14 (2009).

7. Al Amri, A. M. S. Recent seismic activity in the northern Red Sea. J. Geodyn.20, 243–253 (1995).

8. El Isa, Z. H. & Al Shanti, A. Seismicity and tectonics of the Red Sea and westernArabia. Geophys. J 97, 449–457 (1989).

9. Camp, V. E. & Roobol, M. J. The Arabian continental alkali basalt province:Part I. Evolution of Harrat Rahat, Kingdom of Saudi Arabia.Geol. Soc. Am. Bull.101, 71–95 (1989).

10. Camp, V. E., Roobol, M. J. & Hooper, P. R. The Arabian continental alkalibasalt province: Part II. Evolution of Harrats Khaybar, Ithnayn, and Kura,Kingdom of Saudi Arabia. Geol. Soc. Am. Bull. 103, 363–391 (1991).

11. Murray, T. L. & Endo, E. T. A real-time seismic-amplitude measurementsystem (RSAM). US Geol. Surv. Bull. 1966, 5–10 (1992).

12. Hill, D. P. et al. The 1989 earthquake swarm beneath Mammoth Mountain,California; an initial look at the 4 May through 30 September activity.Bull. Seismol. Soc. Am. 80, 325–339 (1990).

13. Lahr, J. C., Chouet, B. A., Stephens, C. D., Power, J. A. & Page, R. A. Earthquakeclassification, location, and error analysis in a volcanic environment:Implications for the magmatic system of the 1989–1990 eruptions at RedoubtVolcano. J. Volcanol. Geotherm. Res. 62, 137–151 (1994).

14. McNutt, S. R. inMonitoring and Mitigation of Volcano Hazards (eds Scarpa, R.& Tilling, R. I.) 99–196 (Springer, 1996).

15. Jacobs, K. M. & McNutt, in The 2006 Eruption of Augustine Volcano, Alaska(eds Power, J. A., Coombs, M. L. & Freymueller, J. T.) (US Geological SurveyProfessional Paper, (1769, in the press).

16. Ohminato, T., Chouet, B. A., Dawson, P. & Kedar, P. Waveform inversion ofvery long period impulsive signals associated with magmatic injection beneathKilauea Volcano, Hawaii. J. Geophys. Res. 103, 23839–23862 (1998).

17. Chouet, B. A. Volcano seismology. Pure Appl. Geophys. 160, 739–788 (2003).18. McNutt, S. R. Volcano Seismology. Annu. Rev. Earth Planet. Sci. 33,

461–491 (2005).19. Dowrick, D. W. & Rhodes, D. A. Relations between earthquake magnitude and

rupture dimensions. How regionally variable are they? Bull. Seismol. Soc. Am.94, 776–788 (2004).

20. Hanks, T. C. & Kanamori, H. A moment magnitude scale. J. Geophys. Res. 84,2348–2350 (1979).

21. Calais, E. et al. Strain accomodation by slow slip and dyking in a youthfulcontinental rift, East Africa. Nature 456, 783–787 (2008).

22. Wright, T. J. et al. Magma-maintained rift segmentation at continental rupturein the 2005 Afar dyking episode. Nature 442, 291–294 (2006).

23. Kanamori, H. & Given, J. W. Use of long-period surface waves for rapiddetermination of earthquake source parameters. Phys. Earth Planet. Inter. 27,8–31 (1981).

24. Zobin, V. M. Seismic hazard of volcanic activity. J. Volcanol. Geotherm. Res.112, 1–14 (2001).

25. Parsons, T. & Thompson, G. A. The role of magma overpressure insuppressing earthquakes and topography: Worldwide examples. Science 253,1399–1402 (1991).

26. Ebinger, C. et al. Length and timescales of rift faulting and magma intrusion:The Afar rifting cycle from 2005 to present. Annu. Rev. Earth Planet. Sci. 38,439–466 (2010).

27. Keir, D. et al. Evidence for focused magmatic accretion at segment centresfrom lateral dike injections captured beneath the Red Sea rift in Afar. Geology37, 59–62 (2009).

28. Girdler, R. & Styles, P. Two-stage Red Sea floor spreading. Nature 247,7–11 (1974).

29. Pallister, J. S. Magmatic history of Red Sea rifting. Geol. Soc. Am. Bull. 98,400–417 (1987).

30. Bohannon, R. G. Tectonic configuration of the western Arabian continentalmargin, southern Red Sea. Tectonics 5, 477–499 (1986).

31. Bohannon, R. G., Naeser, C. W., Schmidt, D. L. & Zimmermann, R. A.The timing of uplift, volcanics, and rifting peripheral to the Red Sea: A case forpassive rifting? J. Geophys. Res. 94, 1683–1701 (1989).

32. Coleman, R. G. & McGuire, A. V. Magma systems related to the Red Seaopening. Tectonophysics 150, 77–100 (1988).

33. Cochran, J. R. A model for the development of the Red Sea. Am. Assoc. Petrol.Geol. Bull. 67, 41–69 (1983).

34. Steckler, M. S. Uplift and extension at the Gulf of Suez—indications of inducedmantle convection. Nature 317, 135–139 (1985).

35. LeTourneau, P. M. & Olsen, P. E. The Great Rift Valleys of Pangeain Eastern North America, Tectonics, Structure, and Volcanism(Colombia Univ. Press, 2003).

36. Coleman, R. G., Gregory, R. T. & Brown, G. F. Cenozoic Volcanic Rocks ofSaudi Arabia.US Geol. Surv. Open-file Rep. 83–788 and Saudi Arabian DeputyMinist. Miner. Resour., Open File Rep. USGS-OF-03-93 (1983).

37. Camp, V. E. & Roobol, M. J. Upwelling asthenosphere beneath western Arabiaand its regional implications. J. Geophys. Res. 97, 15255–15271 (1992).

38. Al-Saud, M. M. Seismic characteristics and kinematic models of Makkah andcentral Red Sea regions. Arab. J. Geosci. 1, 49–61 (2008).

39. Hansen, S., Schwartz, S., Al-Amri, A. & Rogers, A. Combined platemotion and density-driven flow in the asthenosphere beneath Saudi Arabia:Evidence from shear-wave splitting and seismic anisotropy. Geology 34,869–872 (2006).

40. McClusky, S. et al. Global positioning system constraints on plate kinematicsand dynamics in the Eastern Mediterranean and Caucasus. J. Geophys. Res. 105,5695–5719 (2000).

41. McClusky, S., Reilinger, R., Mahmoud, S., Ben Sari, D. & Tealeb, A. GPSconstraints of Africa (Nubia) and Arabia plate motions. Geophys. J. Int. 155,126–138 (2003).

42. Johnson, P. R. Proterozoic Geology of Western Saudi Arabia, North-CentralSheet. Saudi Geol. Surv. Open-file Rep. SGS-OF-2005-5, 1:500,000 map (2005).

43. Johnson, P. R. Explanatory Notes to the Map of Proterozoic Geology ofWestern Saudi Arabia. Saudi Geol. Surv. Tech. Rep. SGS-TR-2006-4,1:500,000 map (2006).

44. Zahran, H. M., Stewart, I. C. F., Johnson, P. R. & Basahel, M. H.Aeromagnetic-Anomaly Maps of Central and Western Saudi Arabia. SaudiGeol. Surv. Open-file Rep. SGS-OF-2002-8, scale 1:2,000,000 (2002).

45. Hansen, S. E., Rodgers, A. J., Schwartz, S. Y. & Am-Amri, A. M. S. Imagingruptured lithosphere beneath the Red Sea and Arabian Peninsula. Earth Planet.Sci. Lett. 259, 256–265 (2007).

46. Smithsonian-Institution. Jebel at Tair. Bull. Global Volc. Pgm. 32 (10) (2007).47. Baker, P. E., Brosset, R., Gass, I. G. & Neary, C. R. Jebel al Abyad: A recent

alkalic volcanic complex in western Saudi Arabia. Lithos 6, 291–314 (1973).48. Okada, Y. Internal deformation due to shear and tensile faults in a half-space.

Bull. Seismol. Soc. Am. 82, 1018–1040 (1992).

AcknowledgementsThis work is the result of a joint effort of the Saudi Arabian Geological Survey (SGS), andthe US Geological Survey (USGS), conducted with the assistance of the US Consulate,Jeddah, Saudi Arabia. We thank many SGS colleagues (too numerous to list individuallyhere) who participated in field and laboratory work and contributed to the success of thecrisis response.We also acknowledge support to the Volcano Disaster Assistance Programprovided by USAID’s Office of Foreign Disaster Assistance and the USGS VolcanoHazards Program, and we thank SGS President Z. Nawab for facilitating our worktogether. Original Envisat radar raw data are copyrighted by the European Space Agency(ESA) and were provided by ESA. We also acknowledge the Nature Geoscience reviewersand editor Whitchurch, whose contributions substantially improved this paper.

Author contributionsH.M.Z. organized the overall crisis response and coordinated the results with governmentofficials. S.E.H. contributed earthquake focal solutions and seismic data. A.A. providedcomputer programming and seismic network data support. I.C.F.S. contributed to thegeophysics and field geology, seismic location interpretations, and overall communicationbetween the groups. W.A.M. analysed and interpreted all of the continuous seismic data.Z.L. processed and interpreted Envisat data, S.J. modelled and interpreted the radardata, P.R.L. processed and interpreted Terra-SAR-X data, J.S.P. conducted field work,contributed to volcanic hazards analysis and with assistance fromW.A.M. and S.J., wrotethe paper. R.A.W. summarized data on comparable seismic swarms and magnitudesfrom other volcanic fields for use in forecasting. M.R.H.M. contributed volcanologic dataand context based on his extensive field work on the harrat fields of Saudi Arabia.

Additional informationThe authors declare no competing financial interests. Supplementary informationaccompanies this paper on www.nature.com/naturegeoscience. Reprints and permissionsinformation is available online at http://npg.nature.com/reprintsandpermissions.Correspondence and requests for materials should be addressed to J.S.P.





http://npg.nature.com/reprintsandpermissions


Broad accommodation of rift-related extension recorded...

Documents

Transcript of Broad accommodation of rift-related extension recorded...