S44A-01
PERFORMANCE OF ROBUST SOURCE ESTIMATORS FOR LAST YEAR's LARGE EARTHQUAKES
We report on the performance of real-time estimators of source characteristics for seven [+] large earthquakes in the window November 2006 - September [December ?] 2007. The estimators include the improved mantle magnitude algorithm, the estimation of source moment using the body-wave algorithms Mwp in both the time and frequency domains, the slowness parameter THETA introduced by Newman and Okal [1998], and the high-frequency P-wave duration estimator Tau-1/3 [Reymond et al., 2006]. The earthquakes considered are the two events in the Kuriles (06319 and 07013), in Taiwan (06360), in the Moluccas (07021), in the Solomons (07091), Peru (07227) and Santa Cruz (07245). Among them, the Solomons and Peru events scale predictably, whereas the Santa Cruz event features a trend towards slowness, reminiscent of the 1980 source in the same region, and the 2006 Kuril event has a late source. By contrast, both the normal faulting 2007 Kuril source, the Moluccas and the Taiwan events feature high values of THETA, the former reminiscent of the outboard 1977 Sumbawa earthquake, and characteristic of intraplate earthquakes. These new data points generally uphold the deficiency of Mwp for sources exhibiting slowness or even a trend towards it. An attempt to process the 2006 Aysen, Chile earthquake yielded poor results, suggesting that these various algorithms fail below 10**26 dyn*cm.
S44A-02
The Challenge of Tsunami Warning in the Near Field
"If the ground shakes, get away from the ocean." That oft-repeated advice is soon ignored where earthquakes are common. Add the proviso "if shaking is so severe that you cannot stand," and people ignore slow earthquakes, if they are even aware of them. Tragedies such as the West Java tsunami of 2006 can only be avoided with a local warning system, but by the nature of the warning problem such systems cannot be as reliable and foolproof as those for tele\-tsunamis. Where local warning systems are well developed, such as Japan, the West Coast of North America, and Hawaii, initial warning messages are routinely issued within five minutes of an earthquake. Such rapid messages are necessarily based solely on seismic data, usually just epicentral location and earthquake magnitude. Since magnitude is not directly correlated with tsunami size, there is inevitably a high false-alarm rate. Direct measurement of the tsunami at sea would avoid false alarms, but the the density of sensors required is almost prohibitive. The first positive wave of a tsunami generated by a subduction earthquake will typically reach the adjacent shoreline in 35--45 minutes. If coastal residents are to be provided with twenty minutes of warning, the tsunami must reach the nearest sensor in no more than ten minutes (the additional time being consumed by the finite duration of the earthquake and the need to see several minutes of data on a gauge before a dangerous wave can be confirmed). The ten-minute restriction means that deep-ocean sensors can be no more than 20\thinspace km from the trench axis and must be spaced no more than 120\thinspace km apart along strike. The 1000-km-long Cascadia Subduction Zone thus requires a minimum of eight sensors. At present, it has four {\sc dart}s, but all are too far off shore to record the tsunami quickly enough for effective local warning. Hawaii faces an even more severe challenge: the first positive wave of a tsunami from a basal-slip earthquake will reach the adjacent shore in as little as six minutes. Until an inexpensive and robust technique can be devised for direct measurement of a tsunami very close to its source, local warning will have to depend on shore-based measurements. Dense GPS networks are promising, but only for earthquakes larger than about magnitude 7.6. For smaller earthquakes, coseismic deformation is too small to measure rapidly on land, even though a damaging local tsunami may have been generated. We are left with seismometers. MWP has been used successfully at close range in Hawaii, but for large or slow earthquakes it will saturate. To render the current rapid tsunami warnings more reliable, the performance of other P-wave-based estimates (such as mBC) in the near field must be explored. A rapid near-field measure of the energy- to-moment ratio, Θ, would be especially valuable.
S44A-03 INVITED
Recent Progress of Tsunami Forecast System of JMA
1. Procedure for issuance of Tsunami Forecast Japan Meteorological Agency (JMA) initiated tsunami forecast service for local tsunamis in 1952 with empirical method. We upgraded the forecast system to one with numerical simulation technique and database method in 1999, which is called Quantitative Tsunami Forecast System. Tsunami generation and propagation simulations were conducted for 100,000 different assumed faults in advance and their results were stored onto a database. We retrieve the most appropriate result from the database immediately after a large earthquake occurs, and assemble tsunami forecast to be issued. This method is also used for the international service to provide the nations in the Northwestern Pacific area with the Northwest Pacific Tsunami Advisory. 2. Application of the Earthquake Early Warning to Tsunami Forecast The Earthquake Early Warning (EEW) is a new service to announce the expected strong ground motion caused by earthquakes before it arrives. The EEW technique can estimate the hypocenter and magnitude very quickly, around a few seconds after an earthquake occurrence at the earliest. JMA aimed at the tsunami forecast dissemination within 3 minutes formerly, but it has become possible to shorten it by applying this technique since October 2006. The first application for the tsunami forecast was for the M6.9 Noto peninsula earthquake on 25 March 2007, and it took 2 minutes to issue tsunami forecast after the quake. The second one was for the M6.8 Chuetsu-oki earthquake on 16 July 2007 and the tsunami forecast was issued in 1 minute. 3. Incorporation of the CMT Solution into Tsunami Forecast Numerical simulation for the Quantitative Tsunami Forecast System was conducted under the assumption that all mechanisms were pure reverse faults with dip angle of 45 degree. From the viewpoint of the safety, JMA applies such conservative scenario as reverse fault generates tsunami most efficiently, considering that we cannot obtain the actual mechanism of the rupture in a very short time while tsunami forecast must be issued in a few minutes. By adopting the recent progress in the seismic data analysis, however, we have come to obtain the CMT solution in about ten minutes using the STS2 seismic waveform record. JMA began to incorporate CMT solution into the tsunami forecast operation in July 2007. Tsunami forecast once issued depending on the hypocenter location and Mjma will be upgraded if the CMT solution shows that the fault mechanism is normal or reverse and also that the Mw is larger than Mjma by 0.5 or more. On the other hand, if it is shown that the mechanism is a strike-slip fault, we can downgrade or cancel the forecast more promptly than before.
S44A-04
Real-time Experimental Forecast of the Peruvian Tsunami of August 15, 2007
At 23:41 UTC on 15 August 2007, an offshore magnitude 8.0 earthquake severely damaged central Peru and generated a tsunami. Severe shaking by the earthquake, which killed over 650 people throughout the region, collapsed buildings, while tsunami flooding of 5 m was observed at Paradas, Peru. The only real-time tsunami data available came from a deep ocean tsunami detection buoy within 1 hour of tsunami generation. The real-time forecast of the 15 August 2007 Peruvian tsunami event was another successful test of NOAA's experimental tsunami forecast system. The accurate, efficient, and reliable forecasts were instrumental in the quick cancellation of the warning within 2.5 hours after the earthquake, while demonstrating the great potential of the experimental system in real-time operation. The clear real-time tsunami signals recorded by the DART network guaranteed high-quality input to the forecast system. During the event, the data assimilation and inversion provided a DART-constrained tsunami source within 2 hours of tsunami generation and a refined tsunami source within 4.5 hours, based on longer DART time series. The initial and refined results produced by the forecast system at 2 hours 9 minutes and 4 hours 45 minutes, respectively, after the earthquake provided guidance for warning and mitigation in the Pacific, especially along the U.S. coastline. The forecast indicated that tsunami flooding would not occur in any of the 14 U.S. coastal communities. Comparison of modeling results with the observations at deep-ocean buoys and coastal tide gages showed excellent agreement for up to 24 hours after tsunami generation in both wave height and period. Comparison of the maximum wave heights presents less than a 5% error when the background noise of observations is small.
S44A-05 INVITED
Pacific Tsunamis Tested NOAA Experimental Forecast System
Tsunami Warning Centers (TWCs) of National Oceanic and Atmospheric Administration (NOAA) are tasked with issuing tsunami warnings for the U.S. and other nations around the Pacific. NOAA's Pacific Marine Environmental Laboratory has developed the methodology that combines real-time deep-ocean measurements with tested and verified model estimates to produce real-time tsunami forecast for coastal communities. This methodology (also known as Short-term Inundation Forecast - SIFT) is currently being implemented at NOAA's TWCs. Tsunami forecast should provide site- and event-specific information about tsunamis before the first wave arrives at threatened community. The next generation tsunami forecast provides estimates of all critical tsunami parameters (amplitudes, inundation distances, current velocities etc.) based on direct tsunami observation and model predictions. Results of real-time experimental forecasts for five recent tsunamis will be presented to quantify the accuracy and robustness of the NOAA experimental tsunami forecast system. The five tsunamis presented include Tonga Island, May,2006, Kuril Island, November, 2006, Kuril Island, January,2007, Solomon Island, April, 2007, and Peru, August ,2007. http://nctr.pmel.noaa.gov/
S44A-06
Recent improvements in earthquake and tsunami monitoring in the Caribbean
Following the catastrophic loss of life from the December 26, 2004, Sumatra-Andaman Islands earthquake and tsunami, the U.S. Government appropriated funds to improve monitoring along a major portion of vulnerable coastal regions in the Caribbean Sea, the Gulf of Mexico, and the Atlantic Ocean. Partners in this project include the United States Geological Survey (USGS), the National Oceanic and Atmospheric Administration (NOAA), the Puerto Rico Seismic Network (PRSN), the Seismic Research Unit of the University of the West Indies, and other collaborating institutions in the Caribbean region. As part of this effort, the USGS is coordinating with Caribbean host nations to design and deploy nine new broadband and strong-motion seismic stations. The instrumentation consists of an STS-2 seismometer, an Episensor accelerometer, and a Q330 high resolution digitizer. Six stations are currently transmitting data to the USGS National Earthquake Information Center, where the data are redistributed to the NOAA's Tsunami Warning Centers, regional monitoring partners, and the IRIS Data Management Center. Operating stations include: Isla Barro Colorado, Panama; Gun Hill Barbados; Grenville, Grenada; Guantanamo Bay, Cuba; Sabaneta Dam, Dominican Republic; and Tegucigalpa, Honduras. Three additional stations in Barbuda, Grand Turks, and Jamaica will be completed during the fall of 2007. These nine stations are affiliates of the Global Seismographic Network (GSN) and complement existing GSN stations as well as regional stations. The new seismic stations improve azimuthal coverage, increase network density, and provide on-scale recording throughout the region. Complementary to this network, NOAA has placed Deep-ocean Assessment and Reporting of Tsunami (DART) stations at sites in regions with a history of generating destructive tsunamis. Recently, NOAA completed deployment of 7 DART stations off the coasts of Montauk Pt, NY; Charleston, SC; Miami, FL; San Juan, Puerto Rico; New Orleans, LA; and Bermuda as part of the U.S. tsunami warning system expansion. DART systems consist of an anchored seafloor pressure recorder (BPR) and a companion moored surface buoy for real-time communications. The new stations are a second-generation design (DART II) equipped with two- way satellite communications that allow NOAA's Tsunami Warning Centers to set stations in event mode in anticipation of possible tsunamis or retrieve the high-resolution (15-s intervals) data in one-hour blocks for detailed analysis. Combined with development of sophisticated wave propagation and site-specific inundation models, the DART data are being used to forecast wave heights for at-risk coastal communities. NOAA expects to deploy a total of 39 DART II buoy stations by 2008 (32 in the Pacific and 7 in the Atlantic, Caribbean and Gulf regions). The seismic and DART networks are two components in a comprehensive and fully-operational global observing system to detect and warn the public of earthquake and tsunami threats. NOAA and USGS are working together to make important strides in enhancing communication networks so residents and visitors can receive earthquake and tsunami watches and warnings around the clock.
S44A-07
Data Fusion Concepts for Tsunami Warning
The tsunami source region for subduction zone earthquakes is near the base of the acreationary wedge. In that region the material properties are weak, and are unable to store significant elastic energy. Therefore, the tsunamigenic rupture is nearly aseismic. Instead, the seismic energy primarily radiates from deeper within the fault; thus resolving or imaging this shallow rupture is extremely difficult, particularly when factoring in the real time requirements of tsunami warning. In the current U.S. Tsunami warning systems, the operational paradigm is to initially determine the most probable source parameters in a maximum likelihood sense, then establish alert level based, and wait for sea level measurements (DART or tide gauge) to validate or cancel the alert. This approach results in numerous false alarms, with some missed tsunamis based on the initial alert. In fact, the maximum likelihood approach works well for typical events; however, the tsunamigenic events are outliers; in the last 30 years, there have been about 130 events of size and location that would warrant a expanding warning. Of those events, 43 were associated with tsunamis of greater than 1 m; however, most of those were local, with only 13 events having significant far field tsunamis. So, even for large events, tsunamigenic earthquakes are statistical outliers. To address this issue, we have conceptualized a data fusion based approach to tsunami warning that will incorporate multiple data types simultaneously to provide better, rapid estimates of tsunami source, partially by identifying "statistical outliers" and partially by running multiple hypothesis of tsunami sources. For example, we can test to see the error bias of the preliminary magnitude estimate; that is, is the system more likely underestimating or overestimating the magnitude. In addition, incorporating non-seismic data, such as hydroactoustic (T-phase) and GPS data may improve the ability to image the shallow rupture. The approach also allows for the incorporation of evidence for slumping/landslides, including anomalous long period Rayleigh waves (e.g., Eksrom, 2006) and hydroacoustic signals. In summary, a fusion approach to tsunami warning should provide for lower false alarm rates while increasing the probability of detection of the events in time period prior to sea level measurements.