National usgs marine Geohazards Workshop Menlo Park March 1, 2011 Notes by Amy Draut Tom Brocher, welcoming remarks

Yakutat terrane/microplate collision drives most of active faults that are not directly related to Aleutian trench. Pavlis et al., 2004 proposed it’s an oceanic plateau colliding with North America. Seismicity including 1964 earthquake associated with Yakutat collision. Flat-slab subduction in that area. Subducting Yakutat block has depth of ~100 km over 850 km distance. Subduction angle about 7 degrees. Fuis et al. (2008), Ferris et al. 2003; Christeson et al., Geology, 2010 got marine seismic data (STEEP survey in 2008) building on earlier USGS work (1975 surveys discussed by Bruns and Schwab, 1983). Yakutat crust is 20 km thick, with abrupt change in Moho depth across Transition Fault, with 7 km/s seismic velocity similar to oceanic Kerguelen or Ontong-Java Plateaus. It acts as a doorstop coming in to Alaska, causing significant uplift of St. Elias orogen as thickened crust tries to subduct. From hazards perspective, strikeslip motion works its way around the block as it resists subduction. Yakataga accretionary prism (Worthington et al., Tectonics, 2010), with deep decollement and deformation front out on the slope, implies potential tsunami source since it appears to be at seismogenic depth. Three large folds in deformation front, at updip end of one big Yakutat megathrust. Gulick et al., Geology, 2007. Difference in strike of motion between Yakutat block and Pacific, but no reason for a lot of motion along it – does it just take a long time between events? Some structures peel off the Fairweather Fault in the southeast, too, that are active faults.

1987-1992 had earthquake swarm out in middle of Gulf of Alaska, some >M7. Several branches of basement-involved faults. Slope magnetic anomalies are interesting too; Maus et al., 2007. May be due to block rotation as Pacific and Yakutat interact with possible coupling; does not have historic seismicity.

Fairweather fault system: Elliott et al., JGR 2010. 1958 event. A lot of the fault is under ice, hard to image. Has many interesting splays, active structures.

What are active geohazards? Pamplona zone acting as megathrust? Transition fault unsure, Gulf of Alaska shear zone, and definite concerns about Fairweather and inboard Denali Faults. Bruhn et al., 2004. Two large events in 1899, approximately M8+, recorded at gold mining camp near Hubbard Glacier. A 6 m tsunami happened, then glacial lake let go and caused more flooding, also 14 m coseismic uplift; not sure where this fault actually was, though. Seismic lines don’t show any obvious recent rupture, which is a problem. Another M8 event probably on Pamplona megathrust. Elliott and Freymuller GPS work shows a locked zone there accumulating strain now, post-1964.

Potential for Yakutat megathrust to be origin of first 1899 event; second one is an enigma. Prince William Sound earthquake likely nucleated on Yakutat megathrust and then dynamically triggered rupture on Aleutian megathrust. Shennan, QSR, 2009. May have synchronous uplift in both areas, too. See also Moore et al., Science, 2007, on Japan – do we have a similar megasplay fault here?

What causes segment boundaries, and how repeatable are they? How should we use their existence as an assessment tool?

Secondary effects: all hazards are local. Deaths caused by the earthquake-generated tsunamis rather than earthquake itself, as in 1964, and submarine landslides as the cause. Of 122 deaths in 1964, 106 were tsunami-related. Also, e.g., Latoya Bay, 1958 earthquake causd rockslide that generated runup of 200 m +. Fjords of coastal Alaska are ideal environment for producing submarine landslide. Recent glacial meltbacks have worsened those risks compared with fifty years ago. Submarine slides can be huge, 30 m thick by 7 km long, in one seismic example. Plafker et al., 1969 on runup from 1964 tsunami. Runup correlates with deaths. In Seward, 0.2 km3 slide volume (Resurrection Bay) from 1964; Haeussler et al., 2007; Suleimani et al., PAG paper showing Seward was rebuilt right in the old tsunami zone. Surveyor Slide (on Surveyor fan), a short, stumpy, apparently viscous slide 500 m thick, very large.

Data needs: high resolution seismic reflection in coastal settings, to connect onshore and offshore faults, and for mass transport deposit identification. Crustal-scale seismics to understand segment boundaries, bathymetry as shallow as we can get for coastal ciites. Marine seismometer long-term deployment. Coastal/marine paleoseismology badly needed; should partner with GeoPrisms, EarthSCope, IODP.

Comment from Jody Bourgeois: Can you identify fast vs. slow slides? Tsunami-causing slide vs. non-tsunamigenic?

Sean: We judge slide “viscosity” from the deposit’s geometry, by analog with physical experiments [didn’t directly answer Jody’s question].

Steve Kirby: Deep water data availability?

Sean: Because of Law of the Sea work, have data along much of the margin plus what NOAA got. Will get multibeam this summer in deeper part of the fan, NOAA also has some that they have not seen, but that is only in patches. Some individual fjords are well mapped, others not at all.

Roland von Huene: The Sonne collected other data there, 15 years ago.

Ginger Barth: Other Law of the Sea work is planned but coverage still not clear. One more big NOAA cruise will happen in Gulf of Alaska.

Peter Haeussler: There’s a huge need for multibeam bathymetry along this margin. There’s no good multibeam over Queen Charlotte-Fairweather Fault, and not enough data to tie well from deepwater Law of Sea work onto very shallow work, going across the shelf. Effect of urbanization on submarine landslide effects. Seward is built on alluvial fan, whose stream was diverted through a tunnel to south edge of town; evidence for 40 m new sediment deposited there since 1964. In the future, only dumping sediment in small areas, not in the wide areas of the past. This will change effects of tsunami runup, and focuses effects of future landslides into smaller places.

Subducting plate geology constraining three earthquake ruptures, Kodiak to Unimak. 1964: two earthquakes, Kodiak segment and Yakutat terrane segment rupture. 1938 west of that, and enigmatic 1946 rupture west of that. Bilek (2010) on role of subduction erosion on seismicity. Kodiak-Bowie seamount hotspot ridge likely causing subduction erosion in Kodiak area. Lines up with main asperity offshore Kodiak. Patton-Murray seamounts also line up with asperity on margin off western Kodiak area. Kodiak Seamount in multibeam is 3 km high, base buried in trench floor sediment, 22 km wide. In Costa Rica, subducting seamounts are 2.5-3 km high that produce asperties with M5-7 earthquakes. Juan-Fernandez Ridge in central Chile also is 3.5 km high, imaged tomographically beneath the margin. So this Kodiak seamount produces relief enough to be comparable with other places that are asperities and barriers in the subduction zone. Chirikof Island, Trinity Island locked zone was barrier at southern end of 1964 rupture. Multibeam images show frontal accretionary prism, major faults, transverse escarpments (some with as much as 3 km of relief), detail of shelf. Using new algorithms on 30-year old data (John Miller processing in Denver), new interpretations.

Great earthquakes that sourced tsunamis along Alaska trench: 1938, ruptured right over subducted Zodiac Fan. Eocene fan sediment, 10 m.y. worth of continental debris from a single gateway. Fan head would have been by one of those hotspot ridges (fan head now subducted). Emile Okal plotted epicenters. Fan sediments, with high-contrast carbonate upper horizon, contrast with very pelagic sediment drilled elsewhere, and fan location may have dictated where great earthquakes occurred.

Divide Alaska Trench from Aleutian Trench where mid-slope terrace develops, near continental/oceanic crust transition.

Accreted frontal prism sediment accumulated in past 3 My as trench sediment became abundant. Before that, it was a more erosional margin. Segmentation seems to be controlled a lot by subducting lower-plate relief; we do have barrier between 1938 and 1964 rupture that is now a locked zone (Trinity Islands structure). Zodiac Fan location may have helped to limit rupture propagation. Lower plate relief was an asperity and barrier during 1964 earthquake rupture along the Kodiak margin.

Dave Scholl: Aleutians are one of the world’s most dangerous subduction zone.

There, we have very little data but lots of concerns. Have explosive volcanism, edifice collapse (can be offshore), slope failures (but don’t know where they are), large near- and far-field tsunamis, and great and giant earthquakes. USGS first work up there in 1950s on volcanic hazards (Okmok Volcano). Arc-building subduction zone concept was the big scientific payoff. Ancient collapse on Koska Volcano likely generated a significant tsunami in Bering Sea (Coombs et al., EPSL, 2007). Infrastructure includes fishing industries, lots at risk right at sea level.

Principal concern is great (8.0) and giant (8.5) earthquakes causing tsunamis. 1957 Mw8.7 earthquake caused Hilo, HI tsunami (from Atka).

Mission of USGS is very relevant to: must forecast probable occurrence, consequence of Aleutian-originated tsunamis. Need to be sector-specific because, as in Chile 2010, quakes can move along sectors between ridges or other barriers. What physical factors determine lengthy megathrust rupturing, and high seaward slip? What is geometry of strain-releasing faults beneath forearcs? Where and how is strain stored and released, both energetically and slowly, during earthquakes?

Importance of fault geometry: splay faults, subduction channels. Splay faults are very dangerous. Need geometry and lateral continuity data.

1946 Unimak Pass tsunami had critically important slow rupture component, contributed to large near-field and trans-oceanic tsunami. Normal faults strike parallel to magnetic anoalies seaward of Aleutian trnehc, could generate destructive near-field tsunamis.

Most critical needs: need paleoseismicity, paleotsunami studies offshore and onshore. Data priorities: coastal paleo studies. Reprocess MCS archived seismic data using new technology. Multibeam map of Aleutian forearc and trench in more detail badly needed, comparable to what’s available for Costa Rica and Chile, where you can see very detailed seamounts and slides (partner with NOAA, UNOLS, industry). We don’t have anything like this for Aleutians.

Most at risk: Dutch Harbor, coastal SW Alaska, western Canada, US; Hawaiian Islands. Also threatened from Kamchatka, not just Aleutian tsunamis. Bering Sea fishing community: largest US fish, crab catches, $6 billion industry. We know nothing about tsunamis coming from back side -- Beringian margin collapses?

Essential cooperative agencies: NOAA for multibeam, marine and coastal science support, tsunami centers. NSF for GeoPrisms, teaming with marine and onshore NSF scientists, and UNOLS for operational support. University of Alaska, Fairbanks: Tsunami and seismicity group (Alaska Eq Information Center)>

If funds available, in short term should focus on coastal paleo record, especially Fox Islands sector – unknown how risky, and not believed to have released much of the strain stored there. Longer term: coastal paleoseismicity record for all rupture sectors.

Summary of things we need to know: Splay and reverse faults, where located/length/history? Physical reasons favoring rupture continuation/termination and areas of high slip? What controls rupturing of slow or tsunami earthquakes? Need paleohistory of M8+ earthquakes, where did they rupture, where were large tsunamis launched?

Comment from David Oglesby, UC Riverside: numerical models could show how splay faults might propagate rupture.

Dave Scholl: wants physical reasons why we should model it one way vs. another way. Need better physical information that would feed into future models.

Emile Okal: caution against uniform character of barriers/termination features. The next earthquake might decide to jump across them. E.g., 1604 earthquake and tsunami in southern Peru stopped at Nazca Ridge, but in 1868 another major earthquake jumped across Nazca Ridge. As bad as the Alaska 1938 event was, some evidence that in 1788 there were one or two earthquakes that did not behave the same way. On large and small scales, the concept of segments and fragmentation is somewhat arbitrary and we can’t necessarily predict how different blocks will rupture. In Mexico 1932, and Solomon Islands 2007, rupture even jumped a plate boundary. Rupture can also continue into accretionary prism, sometimes coseismically, sometimes as an aftershock a while later (as in Kurils, 1963, 1975). Nothing behaves very uniformly, there’s much spatial and temporal diversity in how these concepts should be applied. We shouldn’t be constrained too much thinking that the next event would replicate earlier ones.

Dave Scholl: Yes. In Fox Islands scenario, they allow it to break over into the Alaska subduction system too, creating a larger 8++ earthquake. Record is very complex and should include clustering possibility.

Brian Atwater: In 20^th century, most of the Alaska/Aleutian subduction zone broke at some time (except Fox Islands). What do we know about probability of repetition?

Emile Okal: Yes and no, because magnitudes were quite different in different parts of the subduction system. Some areas where not much strain was released could be ripe for rupture. What about Feb. 4^th, 1965 event? What was extent of that tsunami, which directed itself toward big area of the Pacific that would have been oblique to where people were likely to observe?

Dave Scholl: One unusual thing is that the sediment is very thick here compared to many other island arcs, very different from Tonga, Marianas, Izu-Bonin. This place likes to store up its energy and break in long ruptures, not typical of island arcs because it has so much sediment that favors continuation of rupturing. That makes it more dangerous than some other arcs.

Tsunami from M9 earthquake in Puerto Rico trench could significantly damage US east coast. Earthquakes and tsunamis impact 4 million US citizens, visitors, US sphere of influence in Puerto Rico and Virgin Islands. Has model of predicted tsunami runup from M8.4 from PR trench in San Juan, Puerto Rico. Airport flooded, e.g. There have been 91 reported tsunamis in Caribbean, with casualties exceeding US west coast, Hawaii, Alaska combined. Length of plate boundary in Antilles arc is similar to the length of Cascadia margin. Also two very large strikeslip faults, including the one that ruptured in Haiti (2010). Thrust belt, Muertos Trough, on back side. Intra-arc earthquakes and tsunamis have occurred, including 3 large tsunamis in 1867, 1918, and 1946. Earthquake swarms from 2005 to 2009 in San Juan. Is a bigger one coming? Is there a potential mega-earthquake and mega-tsunami from PR Trench? Uri argues against that.

PR trench is deepest place in Atlantic, lowest gravity anomaly on earth, has a tilted carbonate platform. There may be a slab tear with asthenospheric flow through a gap, or seamount or ridge subduction. West of Puerto Rico, subduction is blocked by the tear, then resumes. Analyzed historical earthquakes; subduction zone earthquakes only in a 415-km-long segment. Largest historical “trench” earthquakes are not known elsewhere. May 2, 1787, event was previously suggested to be M8-8.25 event; but attenuation relationship for PR and Virgin Islands decays more slowly with distance than for the rest of the islands. Not sure why intensity does not die off as easily there; feel more shaking for smaller earthquakes. Uri thinks it was more likely about a 6.1.

GPS measured continuously for 3-5 years, and fit model to GPS data; model suggests no strain accumulation toward a future mega-earthquake. Depths of earthquake swarms (non-aftershocks), 2003-2009: lots of spread, but going to quite deep depths, >100 km. Probably influenced by the slab tear process, not just shallow swarms and not contributing to strain accumulation in shallow subduction zone. Most earthquakes have oblique focal mechanisms, oblique slip. Bunce Fault and absence of closer strikeslip faults to PR suggests very oblique slip, too, and static stress models show that strike slip motion on faults close to the trench is probably inhibiting motion on faults close to the arc.

Evidence for tsunami overwash on island (Anegada): breached sand ridges, ridges are 3 m high. Fields of cobbles and boulders up to 0.8 km from nearest shore. Sand and shell layer with mud cap, extended 1.5 km south of ridge (Atwater et al., in press; Steve Watt et al., in press). That layer was dated at after A.D. 1650 but was before the island was settled in 1800. From 1755 M9.0 Lisbon earthquake, possibly? Model shows that the deep trench would have diverted waves from Lisbon away from much of PR, but could have inundated Anegada. Modeled “orphan” tsunami in Europe from a hypothetical M9 event originating at Puerto Rico trench, but have not found any evidence of that in Europe.

Bathymetry, gravity, tilt of carbonate platform argue for trench collapse, slab retreat. No evidence for large historical events except in Hispanola. No known subduction earthquakes in Virgin Islands and Lesser Antilles.

Septentrional Fault – if ruptured, would affect Dominican Republic. Prentice et al. 2003 proposed 800-1200 yr recurrence interval. Last activity 1040-1230 AD, with ~4 m lateral slip. Slip accumulation rate there 12.3 mm/yr. So why hasn’t it ruptured yet? Alternative explanation: 2 large earthquakes in 1562 and 1842, Uri thinks actually 300 yr recurrence interval with 3-4 m slip, and fault is now in middle of its seismic cycle.

Is the Muertos Trough thrust belt a subduction zone or retro-wedge? Looks like accretionary prism but not much is known about it. Backarc thrusts? Can’t identify any historical earthquakes from there confidently.

Source of 1918 tsunami was previously thought to be an M7.2 earthquake. Now thought to be landslides. Seafloor mapping shows 150-m thick, 10 km3 slide area.

Challenges: estimating intra arc earthquake and tsunami potential, landslide tsunami potential difficult in carbonate environment.

Question from Patrick Muffler: of the 91 tsunamis in historical record, did any affect US east coast?

Uri: no, not that we know of. Any tsunamis affecting east coast are thought to have been from landslides. 1755 Lisbon earthquake had devastating tsunami all the way to central Brazil. But US east coast was spared then because bathymetry near the source was such that it dispersed waves. It hit Newfoundland, Caribbean, Brazil, but not US east coast.

Steve Kirby: Convergence there is ~2 cm/yr. Maybe this margin just hasn’t been that seismically active in last few hundred years?

Uri: definitely possible, so we need to tackle problem from many angles, not just historical record.

Carolyn Ruppel: Uri has effectively accomplished in the Caribbean the kind of work that is being talked about in Aleutians. How? How did he get ship time, etc.?

Uri: Remember, this is a much smaller area than the Aleutians, and more accessible than the Aleutians. Have also been lucky because of opportunities on ships that were already in the area, and funding opportunities that opened up after the Sumatra earthquake, and excellent technical staff.

Steve Kirby: Uri’s also had very supportive management in the long-term for going to sea.

Emile Okal: The death toll may be skewed when comparing Caribbean to west coast, because Native populations may not have been counted well. Does Uri think there is no danger from M9 in Lesser Antilles?

Uri: Extended the models as far as Guadalupe. There is more work needed in that area (Lesser Antilles); but as of now, does not see reason to argue for that big an earthquake there.


Steve Kirby: Subduction-related tsunami sources outside the US that threaten Pacific shorelines.

Half Moon Bay, CA, on April 1, 1946, from Mw8.6 Unimak, Alaska earthquake 3600 km away.

Types of tsunamigenic faults in subduction zones: interplate thrust events on megathrust, seismic slip on splay faults, and off-trnech events under outer rise (OR) and outer trench slope (OTS). Also there are intraslab and strikeslip events, but those aren’t known to be tsunamigenic. Important to sort out which is which: USGS NEIC does this best. Typically only Mw >8.3-8.5 produce tsunami waves big enough to be damaging at regional and trans-oceanic scales. The 1946 earthquake affected Antarctica (4 m runup, 13,000 km away). Waves as big as 42 m on Unimak. West coast of US also significant, and Hawaii, Marquesas, New Zealand. Disaggregation of subduction sources that have the largest effects on distant coastlines: Alaska, NW Pacific (Kurile, Kamchatka), and Cascadia dominate the Pacific Basin for our most populous Pacific coastlines (work of Hong Kie Thio). Hawaii is the “catcher’s mitt” for all of those.

Modeled tsunami propagation from 1755 Lisbon earthquake (Barkan et al., Marine Geology, 2009). Complex compressional tectonics, huge tsunami waves in S. Europe, NW Africa, down to Brazil, and Newfoundland. This is largest known in Atlantic basin.

NOAA NGDC compilation of tsunami sources, in terms of loss of life. Western Pacific Izu-Bonin, Tonga-Kermadec, Vanuatu, Marianas are some of fastest converging places in the world but have not supported M8+ thrust earthquakes. To get M>8.5 events (21 included in this analysis), factors are: (1) incoming plate age, where slab dynamics include stronger coupling. Not in extensional island arcs, where Coulomb friction is lower. 18 of 21 fit this factor. (2) Trench sediment fill; cf. seafloor roughness. Subduction channels can smooth plate boundaries, enabling long-runout ruptures. 21 of 21 fit this factor. (3) Thick crust in upper plate, or low plate boundary dip in island arcs: rupture width, thicker crust can extend the depth limit to serpentinized mantle. 21 of 21 fit. (4) Forearc basins: continuous basin-centered gravity lows, with potential for multi-segment rupture; indicates high coupling. 20 of 21 fit this factor, but Nias Sumatra earthquake of 2005 is an exception.

TSWG 2005 report ranked hazards for interplate-thrust earthquake tsunami sources in western Pacific. Highest ones were Kamchatka, southern Kuriles, Manila trench, a few others (because of sediment accumulation). Candidate stealth subduction segments that have not produced great plate boundary earthquakes in historical times but should be important research targets – those sectors with 3 or 4 of the above 4 contributing factors and generally low convergence rates (<20-30 mm/yr, so probably long return intervals). Geist et al. Sci. Am., 2005, and TSWG. These include those suspected to be slow or clogged: Makran, Yakutaga, several others.

Double seismic zone near-trench seismicity: Gamage et al., GJI, 2009. Upper zone shows normal faulting, lower zone shows reverse faulting deeper. There is complex stress transfer between treat OR/OTS flexural earthquakes and subduction IPT earthquakes and their aftershocks in Japan subduction zone. Japan trench gravity data shows exceptionally large outer rise gravity anomaly and bathymetric relief, indicating lots of flexural stress. Levitt and Sandwell (JGR, 1995) flexure model. Geist modeled tsunami from 1933 earthquake from Japan, Mw8.6. Modeling off-coast wave amplitudes within factor of 2 to 3 of those observed in Japan and Hawaii, encouraging that it was on the right track.

Other great OR/OTS earthquakes including 2009 Tonga/Samoa event had these features in common: Mesozoic incoming plate (thermally mature, thick, high stress for given plate curvature), focal mechanisms showing rupture planes crosscut seafloor spreading fabric at high angles (>30 degrees), normal fault scarps were long and parallel to the trench, earthquakes occurring where outer rise gravity anomalies were large and positive, and higher dip angles and larger ocean depths in epicentral areas.

Data in some of the most important areas could include “low hanging fruit”, especially in places like Fox Islands, Aleutians.

Summary: tsunamis from distant sources are infrequent but big threats to US shorelines. Alaska, NW Pacific, Cascadia, and Lisbon area pose biggest threat to populated US shorelines. Need marine geologic, geophysical data and coastal paleotsunami data for best insights into recurrence times and still many knowledge gaps: subduction zone kinematics, seismic/aseismic slip, processes that limit IPT rupture lengths, and physics of slow-rupture earthquakes.

Jody Bourgeois: she disputes “infrequent but very significant threats”. They don’t have to be very infrequent; every 20-30 years just from Kurile area, and US population has grown so much on west coast. They are very present in historical record. The 2004 Sumatra earthquake was the largest of its size in 40 years, but that’s a long time considering funding cycles and how fast human population has grown.

Emile Okal: Hawaii seems to have minimal danger from the part of the Aleutian arc where we might expect activity soon (looking at Hong Kie Thio’s map). What controls that physically? Are we sure that is not just a plot of earthquake history that has damaged Hawaii?

Steve Kirby: those are modeled, not historical, results based on bathymetry affecting simulated tsunami waves. Hong Kie Thio is not here so can’t be sure what bathymetric feature caused the “V” dip in central-western Aleutians in his lower plot.

Emile Okal: only one event in western Pacific is pure, typical subduction event which is 1933. Some of the comparisons of that with other events aren’t valid.

Sean Gulick: in those 4 categories of characteristics in tsunamigenic earthquakes, one is the Wells et al., forearc basin gravity low concept and another is the sediment/roughness argument. Need to think about these more broadly. Where do we roll over from velocity weakening to velocity strengthening rheology?

Steve Kirby: gravity indicates deepest parts of basins, corresponds to downdip limit of high slip. It’s closely related to basin geometry.

Dave Scholl: splay fault system is probably updip limit. Can it get out into frontal accretionary prism, slow rupture? Where do splay faults define forearc basins?

Sean Gulick: that transition of rheology of the overriding block should be apparent in gravity, and may extend beyond the basin.

Chris Goldfinger: his class on mega-thrust earthquakes did exercise using Wells et al. paper to see how well the fit matches the basins, and got only about 50/50 success, even for Nankai on which the model was based. Alaska doesn’t work. When you include Chile 1960, that really changes the statistics because it caused about 80% of the moment in the 20^th century. So Nias 2005 is probably not the only exception. Ray Wells isn’t here to join in this discussion.

Emile Okal: another exception should be that where the mold is fit by no earthquake because we just don’t know of any earthquake even though you’d think you should be able to predict one. E.g., in Makran we are waiting for a M9 earthquake where there’s extremely thick sediment. Also southern Chile.

After the afternoon coffee break: