Reviews of Recent Progress and Proposals on Chilean
Transcription
Reviews of Recent Progress and Proposals on Chilean
SATREPS Chile Tsunami Project Publication Series Volume 6 Reviews of Recent Progress and Proposals on Chilean Tsunami Warning System SATREPS Chile Tsunami Research Project on Enhancement of Technology to Develop Tsunami‐Resilient Community March 2016 Reviews of Recent Progress and Proposals on Chilean Tsunami Warning System Contents Chapter 1. Introduction 1.1 Review of Pre-SATREPS and Current Status of Chilean Tsunami Warning System Patricio CATALAN (UTFSM) 1.2 Improvement of Japanese Tsunami Warning System based on Lessons from the 2011 Great East Japan Earthquake Tsunami Tomoaki OZAKI (JMA) Chapter 2. Proposal on Tsunami Warning System Based on Seismic Observation 2.1 Developing Tsunami Database Patricio CATALAN (UTFSM) 2.2 Rapid Magnitude Determination of Megathrust Earthquake Akio KATSUMATA (MRI, JMA), Shigeki AOKI (JMA), Yasuhiro YOSHIDA (MC, JMA), Hiroshi UENO (JMA), Kazuki MIYAOKA (MRI, JMA), and Takashi YOKOTA (former JMA) 2.3 Application of Tsunami Forecast Chart Yutaka Hayashi (MRI, JMA) 2.4 Instrumental Modified Mercalli Intensity Akio KATSUMATA, Yutaka HAYASHI, Kazuki MIYAOKA, Hiroaki TSUSHIMA (MRI, JMA), and Toshitaka BABA (Tokushima Univ.) Chapter 3. Proposal on Tsunami Warning System Based on Tsunami Observation 3.1 Review on Measurement Technologies for Deep-Ocean Tsunami Toshitaka Baba (Tokushima Univ.) 3.3 Methodology of Optimizing Tsunami Observation Array Using Tsunami Inversion Hiroaki TSUSHIMA (MRI, JMA), Joaquín MEZA (UTFSM), César NÚÑEZ (SHOA), Patricio CATALÁN (UTFSM), Cecilia ZELAYA (SHOA), Tomohiro TAKAGAWA (PARI), Toshitaka BABA (Tokushima Univ.), and Yutaka HAYASHI (MRI, JMA) 3.4 Real-Time Inundation Estimation Using GPU Computing Tomohiro TAKAGAWA (PARI) Chapter 4. Proposal on Communication Protocol between Organizations Related to Tsunami Warning 4.1 Methods to Determine Thresholds to Issue Precaution Evacuation based on Modified Mercalli Intensity Yutaka HAYASHI, Akio KATSUMATA, Kazuki MIYAOKA, Hiroaki TSUSHIMA (MRI, JMA), Patricio CATALAN, Jose BAQUEDANO (UTFSM), ZELAYA Cecilia (SHOA), Victor ORELLANA (ONEMI), and Toshitaka BABA (Tokushima Univ.). Chapter 1. Introduction 1.1 Review of Pre-SATREPS and Current Status of Chilean Tsunami Warning System Patricio CATALAN (UTFSM) Over the last few years, Chile has been affected by a unique sequence of large earthquakes, which starting from the Mw8.8 Maule Earthquake, followed by the Mw8.2 Pisagua Earthquake, and the recent Mw8.4 Illapel Earthquake. Adding to this, the relative contribution of these events to existing gaps of seismic release, make it possible to occurrence of large magnitude earthquakes (Mw8.5< ) a strong probability within the next few years, in areas such as northern and central Chile. In addition, all these events created tsunamis, which varying degrees of destructiveness and damage. On the other hand, the first of these events occurred after a relatively long period of relative calmness where only moderate earthquake occurred in the period comprising 1960-2010. As a consequence, while the country had systems that considered the possibility of tsunamigenic earthquakes in their procedures and protocols, these systems had not been put to test. All of that changed on February 27th, 2010, when the Maule earthquake and tsunami not only brought damage and destruction, but also highlighted that the country was ill prepared to face a challenging situation like this. Several errors occurred, both in terms of hazard and warning level assessment, timely response, communications, and others. Chile has a unique setting on the issue of tsunami warning systems, when compared with other countries. Here, the seismic and tsunami assessment, and warning issuing are fragmented in three different institutions, each with limited and clearly defined responsibilities, resources and goals. Within the context of a Tsunami Warning System, The National Seismological Center (CSN) is responsible of providing a quick assessment of low level earthquake characteristics, such as hypocentral location and magnitude. This information is passed on to the National Hydrographic and Oceanographic Service (SHOA), which within its National Tsunami Warning System division (SNAM), should assess the tsunami hazard. Once the assessment is completed, the information is passed on to the Office of National Emergency (ONEMI), which evaluates the situation and is the only responsible of issuing evacuation alerts. As of early 2010, the system was not prepared to work on a 24/7 basis at CSN, and communication procedures were feeble. Regarding SNAM, the tsunami assessment relied on historical records and expert criteria, which may have been distributed among people with varying degrees of expertise. On the other hand, communication systems to the population were not well established, and relied on systems that failed during the emergency itself. Adding to that, the overall tsunami awareness on the Chilean population was low, with only people on certain coastal areas being specifically educated by either oral tradition and/or ad hoc training programs carried out by SHOA and ONEMI. However, the rest of the country was largely removed from the tsunami issue. It is of note that despite of the deficiencies, these local education programs helped to reduce the death toll since many people were capable of self evacuating. No casualties occurred over a long stretch of coast with coastal dwellings, and of the 124 tsunami victims, many concentrated at specific locations where evacuation was hampered by geography and local conditions, such as at Isla Orrego, near Constitución. Nevertheless, the significant conclusion was that something had to be done, quickly and efficiently, to improve preparedness and response. As a result, several initiatives where put in place, with a wide range of scopes and intervening actors. Among these, it can be mentioned Improvement of Tsunami Monitoring capability Improvement of Protocols, communications and required response times Improvement of the Tsunami Hazard Assessment 1.1.1 Improvement of Tsunami Monitoring capability As of 2010, Chile had a network comprising 18 sea level monitoring stations and 1 DART buoy. The sea level monitoring stations were located at main ports, and its main goal was to measure tides. As such, little redundancy was provided in terms of communications, and many stations only reported once hourly. Moreover, most of the stations were pressure sensors, which had a range of deployment depths but seldomly were placed at the sea bottom, hence they were prone to be clipped by the water level falling below the instrument position during a tsunami. As any tsunami warning based on tide gages, the system was only capable of detecting a tsunami once it reached the coast, thereby reducing the response time to a minimum. Moreover, latitudinal coverage of the country was sparse, and no stations existed between San Antonio and Talcahuano (33o 50´S and 36o44´S ). Significantly, this was the area affected by the tsunami, and the Talcahuano tide gage was destroyed shortly after the arrival of a second wave, when a wall collapsed over the sensor. The San Antonio gage was under repairs, and the other sensor within the main tsunami path, in the Juan Fernandez Archipelago, was destroyed within a minute of tsunami arrival. On the other hand, the only tsunami DART buoy was located off the coast of Pisagua, in northern Chile. Consequently, its main goal was to assess the occurrence of a transoceanic tsunami emanating from Chile but not providing early warning to Chile. The tsunami arrived to it about on hour after the event had caused massive destruction in Central Chile. After this situation, the network has been improved significantly. By 2015, it now consisted of 42 stations on nearly all relevant coastal dwellings along the coast. Moreover, significant changes have been made regarding robustness, including the use of two sensors transmitting and reporting in nearly real time, one under water and one radar sea surface tracker. Communications have been strengthened, using multiple systems such satellite (GOES or BGAN), cellular (GPRS), along with solar panels for electrical supply. The locations of the sensors are shown on Fig. 1, where it can be seen that a relatively uniform latitudinal coverage is now achieved. These sensors operated without issue on the tsunamis of April 1st, 2014 and September 16th, 2015. In addition, shortly after the 2010 event, SHOA had two DART tsunami buoys, located offshore the northern part of the country. Again, these sensors were not enough to provide a timely warning for the latest events. However, as recent as September 2015, SHOA has now deployed two additional DART buoys, of a newer generation, which allows deployment in shallower water, hence reducing the time for tsunami waves to arrive to them. In addition, they are located offshore Caldera and Constitución (see Fig 1 for reference), hence its aim is more in line for a national warning whilst still providing transpacific monitoring. It is of note, however, that despite these improvements, the system is still a monitoring system for evaluation once the tsunami has reached the coast, and no predictive ability is expected from this network. Its main aim, within the context of a TWS, is to provide information regarding the reduction in the tsunami threat, and warning and evacuation cancellation procedures. 1.1.2 Improvement of the Tsunami Hazard Assessment コメントの追加 [y1]: 1.1.3 Improvement of Protocols, communications and required As mentioned before, the system as of now, still relies on an expert assessment and historical data to provide the technical hazard evaluation that is passed on to ONEMI. However, in the intervening period, some significant advances have been reached that are aimed to significantly improving these capabilities. response times These have been reached within the context of the present SATREPS project, and through funding from the Chilean Government through the National Science and Technology Council (CONICYT) and its grant FONDEF D11I1119, Design and implementation of a tsunami database for early warning based on high performance computing. The project was led by UTFSM, with active collaboration of PUC and SHOA, and support and advising from JMA, Japan. The goal of the project was to implement the early stages of a Decision Support System, based on the implementation of tsunami catalog of modeled, hypothetic tsunamis, that would be queried upon an emergency. In turn, these tsunamis considered propagation to the coast up to depths of 100 and 200 m, after which they are shoaled to the coast using Green’s linear approximation. Based on the maximum amplitude at the coast, a four-level hazard categorization is achieved. Multiple filters are set up to ensure a conservative assessment. The system is a complex development, and included the design of not only the database, but also the design and implementation of the hardware and software architecture, using open source software which has been integrated allowing for flexibility and ease of expansion. In this way, the system not only manages the emergency assessment, but also the modeling of the tsunami catalog, the administration and issuing of evaluation bulletins, in protocol-compatible format. The system also includes the flexibility of using different modeling modules. As of now, COMCOT is being used as main tsunami model engine, but a GPU based software was developed and implemented, with the expectation to be run in near real time (see section of 3.4 of this document). The system is now under testing in SHOA, and early assessment has shown a good predicting capability (see section 2.1). Moreover, the developing team has recently applied to an extension of the funding, to turn the early database into a complete Decision Support System, including integrated monitoring and graphical user interfaces. 1.2 Improvement of Japanese Tsunami Warning System based on Lessons from the 2011 Great East Japan Earthquake Tsunami Tomoaki OZAKI (JMA) 1.2.1. JMA tsunami warnings for the 2011 Great East Japan Earthquake The Japan Meteorological Agency (JMA) issued a tsunami warning on March 11, 2011, for the 2011 Great East Japan Earthquake at 14:49 based on a promptly estimated magnitude of 7.9. It was about three minutes after the earthquake occurred, and a warning of "Major Tsunami", which is the highest category of Japan's tsunami warning, was issued for Iwate, Miyagi and Fukushima prefectures. Predicted tsunami heights were 3m, 6m and 3m respectively. After the initial warning, the JMA upgraded tsunami height predictions to 10m or more for Miyagi prefecture and 6m for Iwate and Fukushima prefectures at 15:14, and to 10m or more for Iwate and Fukushima prefectures at 15:30. The upgrade was mainly based on offshore observations with GPS buoys that had been deployed 10-20km off the Tohoku coast by the Ministry of Land, Infrastructure, Transport and Tourism. On the other hand, cable-type sea-bottom pressure sensors deployed more offshore than GPS buoys by the Earthquake Research Institute (ERI) of the University of Tokyo also detected large tsunami arrival earlier than GPS buoys. However, these data could not be used for warning upgrade because the method for utilizing them in tsunami warnings was still under development. The above mentioned observation data and tsunami warning issuance timings are summarized in Fig. 1.2.1. As for the magnitude, JMA magnitude (Mj) (Katsunata, 2004) was used for the initial tsunami warning. Mj is based on the maximum displacement amplitude of strongmotion seismograms and has the advantage of quick availability. However, due to dependence on the short period (-10 sec.) wave component, Mj is prone to underestimation for gigantic or tsunami earthquakes. The JMA, therefore, calculate moment magnitude (Mw) operationally within 15 minutes of an earthquake using domestic broadband seismic waveform data, and the initial warning for the 2011 Great East Japan Earthquake would have been updated using Mw. However Mw could not be obtained because of scale out records at most of the broadband stations in Japan, and Mw8.8 was calculated from unsaturated overseas broadband records about 55 minutes after the earthquake, i.e., around 15:40. This value was not utilized for warning updates because the tsunami height prediction based on Mw8.8 was almost as high as the already updated warning at 15:30. 1.2.2. Lessons learned and improvements Considering the serious damage caused by the 2011 Great East Japan Earthquake, JMA improved its tsunami warnings in March 2013 based on the lessons learned from this devastating event. The major points of tsunami warning improvement are as follows. 1) Measures to avoid magnitude underestimation for huge earthquakes in an initial warning Although an initial warning is required to be issued as early as possible, currently there is no established method to estimate reliable huge magnitude value of much larger than 8.0 within several minutes. To deal with this matter, the JMA introduced tools with which operators can evaluate the probability of Mj underestimation before the initial tsunami warning issuance. If the possibility of Mj saturation is detected with the tools, the JMA will issue a tsunami warning using the maximum possible magnitude around the region close to the epicenter. The tools utilize various methods such as analyzing long periods of seismic waves (Katsumata et al., 2013) and measuring the length of a strong ground shaking area. In addition, the JMA introduced qualitative expression, e.g., "huge", in tsunami height prediction for such a warning that is based on the maximum possible magnitude, aiming at conveying impending danger at an early stage when large uncertainty of magnitude estimation is remained. Criteria and levels in numerical expression of tsunami height prediction were also revised (Table 1.2.1). 2) More precise update of tsunami warning At the time of the 2011 Great East Japan Earthquake, JMA could not calculate Mw within 15 minutes as with JMA's normal operation because of scale out records in most of the domestic broadband seismometers. In addition, cable-type sea-bottom pressure sensors' data could not be applied to warning update because such procedure and relevant technique did not established at that time. To deal with these issues, the JMA deployed broadband strong motion seismometers to acquire full scale of broadband seismic wave data and to obtain Mw robustly. The JMA also started the operation of utilizing sea-bottom pressure sensors' data for warning update from March 2012. In addition, the JMA deployed three buoy-type sea-bottom pressure sensors along the Japan Trench, and the Japan Agency for Marine-Earth Science and Technology (JAMSTEC) deployed 20 of cable-type sea-bottom pressure sensors. These data started to be used for JMA's tsunami warning operation from March 2013. Currently, in estimating a coastal tsunami height from offshore data, the method of multiplying a coefficient defined for each location to an offshore tsunami height is used. In the near future, the method of applying inversion technique (Tsushima et al., 2009, 2011, 2012) is planned to be introduced to JMA's tsunami warning system. REFERENCES . Hoshiba, M., T. Ozaki, 2012, Earthquake Early Warning and Tsunami Warning of the Japan Meteorological Agency, and their performances for the 2011 off the Pacific coast of Tohoku Earthquake (Mw9.0). Kamigaichi, O (2009), Tsunami Forecasting and Warning, En-cyclopedia of Complexity and System Science, Springer., 9592-9618. Katsumata, A. (2004), Revision of the JMA displacement magnitude, Quart. J. Seism., 67, 1-11 (in Japanese with English abstract). Ozaki, T. (2011), Outline of the 2011 off the Pacific coast of Tohoku Earthquake (Mw9.0) - Tsunami warnings/advisories and observations -, Earth Planets Space, 63, 827-830. Ozaki, T. (2012), JMA's Tsunami Warning for the 2011 Great Tohoku Earthquake and Tsunami Warning Improvement Plan, Journal of Disaster Research, Vol.7 Special Edition. Katsumata, A., H. Ueno, S. Aoki, Y.Yoshida and S. Barrientos (2013) : Rapid magnitude determination from peak ampli-tude at local stations, Earth Planet Space. Tsushima, H., R. Hino, H. Fujimoto, Y. Tanioka and F. Imamura (2009), Near-field tsunami forecasting from cabled ocean bottom pressure data, J. Geophys. Res., Vol.114, B06309, doi:10.1029/2008JB005988. Tsushima, H., K. Hirata, Y. Hayashi, Y. Tanioka, K. Kimura, S. Sakai, M. Shinohara, T. Kanazawa, R. Hino and K. Maeda (2011), Near-field tsunami forecasting using offshore tsunami data from the 2011 off the Pacific coast of Tohoku Earthquake, Earth Planets Space, Vol.63, pp. 821-826, doi:10.5047/eps.2011.06.052. Tsushima, H., R. Hino, Y. Tanioka, F. Imamura and H. Fujimoto (2012), Tsunami waveform inversion incorporating permanent seafloor deformation and its application to tsunami forecasting, J. Geophys. Res., Vol.117, B03311, doi:10.1029/2011JB008877. Fig. 1.2.1 Coastal and offshore sea level observation data for the period of the initial and the following two upgraded tsunami warnings. Dotted lines in observation curves indicate data recovered afterward through site surveys. Table 1.2.1 Categories and criteria of tsunami warning and tsunami height prediction. Current (March 2013-) Previous Category Expression of Tsunami height Tsunami Major Tsunami Warning Tsunami Tsunami Advisory 10m or more, 8m, 6m, 4m, 3m Expression of Tsunami height Numerical Predicted height Qualitative Over 10m 10m 10m Huge 5m 5m to 10m 3m to 5m 2m, 1m 3m High 1m to 3m 0.5m 1m (Blank) 0.2m to 1m Chapter 2. Proposal on Tsunami Warning System Based on Seismic Observation 2.1 Developing Tsunami Database Patricio CATALAN (UTFSM) 2.1.1 SIPAT Overview One of the key developments of the last few years, and with a significant support of the SATREPS project, was the development of the so called tsunami-database, which in reality turned out to be the baseline of a Decision Support System, termed Integrated System of Tsunami Prediction and Warning, SIPAT (in Spanish). The system was originally intended to be a simple database of precomputed tsunami scenarios, but it evolved into a complete hardware and software system, which allows modeling, pre and post processing, storage, administration and query of a large tsunami catalog during the non-emergency state. This whole subsystem is called Creator, and is accessed through a web interface on the SHOA servers. Regarding the modeling, within the SATREPS project a study was carried out on which it could be seen that reasonable modeling predictions at depths between 100 and 200m could be achieved. This reduces the computing load, as linear implementations of the NSWE can be used. In consequence, a total of 962 forecast points is used, which include also DART bouys and tide gages. In case of the true forecast points, time series data are stored, which are then shoaled to the nearest coastal location (shortest path) by using Green’s Law. Then, the shoaled time series are processed, estimating several parameters as shown in Table 1. From these, the relevant hazard category level (0-3) is stored for query when in emergency. The levels are defined as 0: No Hazard. 1: Tsunami amplitude at the coast is between 0.3 and 1 m 2: Tsunami amplitude at the coast is between 1 and 3 m. 3: Tsunami amplitude exceeds 3 m. in accordance with the SHOA ONEMI protocol. Regarding the modeling, Creador allows for multiple modeling alternatives, which are controlled by a Modeling Subsystem. As of now, only COMCOT and a GPU model are implemented, but it is flexible enough to incorporate other models if needed by SHOA. 2.1.2 Tsunami Catalog One of the key elements of the system is the earthquake catalog to be used in the analysis. A report by the National Seismological Center showed that by varying 9 parameters (hypocentral latitude, longitude and depth; aspect ratio relationship; dip, rake, strike angles; magnitude; shear modulus) could lead to a total number of scenarios in excess of 1.8 million. However, in many cases, the parameter space considered values not consistent with the fault geometry. Nevertheless, in order to optimize the database populating process, a decision was taken in order to ensure Latitudinal coverage Complete range of depths Only one shear modulus This led to a total of 7656 scenarios. However, given the time constraint imposed by the project, this was further reduced to focus on only the range of earthquake magnitudes where tsunami generation was less certain (Mw 7.5, 8.0, 8.5) and only one aspect ratio. This reduced the number of proposed scenarios to 1528. Table 2.1.1: Table output of the Extract, Transform and Load post processing algorithm 2.1.3 Tsunami Hazard Evaluation Procedure With the database set up, SIPAT can now provide an effective tool for estimating the tsunami hazard. To this end, it is assumed that seismic parameters are provided from several sources (CSN, USGS, GEOFON). Hypocentral location data and magnitud are introduced in accordance to prescribed formats, and with this information SIPAT begins its table look up procedure. Three filters are setup in order to estimate the hazard i) A spatial filter, by which catalog candidates are identified within an interrogation window centered on the epicenter, with a length of 2L and width of 2W. L and W are the source length and width, as estimated by scaling laws. The rationale is that given the uncertainty in the location of the epicenter relative to the rupture, and the closeness of the tsunamigenic zone to the coast, it is necessary to include all possible configurations. ii) A magnitude filter, by which the candidates are reduced to those with magnitudes within [00,5] Mw of the estimated magnitude. With this, it is expected that possible variations in tsunami hazard due to non uniformity of the slip is covered by the increased magnitude. iii) Once all candidate scenarios are identified, the hazard at each coastal location is estimated as the maximum hazard level among all candidates. Figure 2.1.1: Offline assessment of SIPAT performance, for the Sep 16, Mw 8.4 Illapel Earthquake. Colors correspond to the hazard level as predicted by the system. It is of note that this procedure is increasingly conservative, and could led to an overestimation of the hazard. However, at this stage, it is considered a reasonable compromise given all the epistemic uncertainties involved. In addition, the final evaluation does not relate to a single tsunami model, but rather to the hazard envelope of a set of candidates. Testing of the system offline has shown the benefits of the system, as well as testing of its accuracy. For instance, in Fig. 1, a sample output of the first hazard solution for the Mw 8.4 Illapel Earthquake and Tsunami. It is of note that the first seismic estimation yield Mw 7.9 (provided by the USGS). Nevertheless, despite the smaller magnitude, the conservative approach used leads to a very accurate assessment of the hazard, on which on the region of Coquimbo and Valparaiso the hazard level reach the Alarm state, that is, waves exceeding 3 m. This was indeed the real tsunami case, as measured by tsunami survey efforts. In addition, the benefits of having the system can be observed as well. The evaluation took less than 15 seconds, and is capable to provide a regional assessment, where different administrative regions having different hazard levels. This is clearly an improvement over the existing procedure. 2.2 Rapid Magnitude Determination of Megathrust Earthquake Akio KATSUMATA (MRI, JMA), Shigeki AOKI (JMA), Yasuhiro YOSHIDA (MC, JMA), Hiroshi UENO (JMA), Kazuki MIYAOKA (MRI, JMA), and Takashi YOKOTA (former JMA) 2.2.1 Introduction One of major problems in the tsunami warning for the 2011 off the Pacific coast of Tohoku Earthquake (Mw 9.0) was a lack of awareness of underestimation of the earthquake magnitude at the time soon after the occurrence. The first tsunami warning should be issued within a few minutes after the detection of the earthquake occurrence. Displacement magnitude (Katsumata, 2004) is used for the first tsunami warning. Seismic moment tensor is also estimated after the first tsunami warning to estimate size of the earthquake more accurately (Usui et al., 2010). At the time of the occurrence of the 2011 off the Pacific coast of Tohoku Earthquake (Mw 9.0), the magnitude was not estimated properly due to too short cutoff period (six seconds) of the filter for the displacement magnitude compared with the rupture duration (about three minutes [Yoshida et al., 2011a]). Seismic moment could not be estimated from the regional seismological network data due to over-range of broadband sensor (STS-1 and STS-2) outputs, and it took longer time to estimate it from global data. Magnitude determination is the key to the proper tsunami warning. Some method had been developed after the earthquake. Here those rapid magnitude estimation methods are described based on Katsumata et al. (2012). 2.2.2 Magnitude Determination from Span of Strong Motion Area Large earthquakes cause strong motion in a wide area. Span of strong-motion area is related to earthquake magnitude. Seismic intensity distribution in Japan can be known in a few minutes after earthquake occurrence owing to a dense on-line network of seismic intensity meter in Japan. It is possible to estimate earthquake magnitude roughly from the area of strong shaking. Figure 1 shows the distributions of seismic intensity of the 2011 off the Pacific coast of Tohoku Earthquake (Mw 9.0) and the 2003 Off-Tokachi Earthquake (Mw 8.0). The span of 5-lower or the greater of the JMA seismic intensity scale of the 2011 off the Pacific coast of Tohoku Earthquake reached about 700 km. The span of the 2003 Off-Tokachi Earthquake was about 300 km, which was much less than that of the 2011 off the Pacific coast of Tohoku Earthquake. Fig. 1. Distributions of seismic intensity of the 2011 off the Pacific coast of Tohoku Earthquake and the 2003 Off-Tokachi Earthquake. The contours in maps denote slip distributions estimated by Yoshida (2005) and Yoshida et al. (2011a). 2.2.3 Source Area Estimation from Seismic Intensity Distribution The place of observed strong motion would be close to the seismic source. Source area can be estimated from the area of strong-motion. At each station, distance from the fault is estimated from the seismic intensity with the formula by Si and Midorikawa (1999). To estimate the source area, grid points are set on the plate boundary. At each grid points, number of stations where the distance between the station and the grid point is larger than the estimated fault distance is counted. If the grid point is on the source area, the number of inconsistent stations (where the distance between the station and the grid point is larger than the estimated fault distance) should not exceed some number. To exclude far offshore areas, the grid points are restricted to those at which more than ten seismic stations are expected to detect seismic intensities equal to 4.5 or the greater. Figure 2 shows the estimated source area of the 2011 off the Pacific coast of Tohoku Earthquake and the 2003 Off-Tokachi Earthquake from the number of the inconsistent stations. Fig. 2. Estimated source area of the 2011 off the Pacific coast of Tohoku Earthquake and the 2003 Off-Tokachi Earthquake from seismic intensity distribution [Ueno et al., 2013]. 2.2.4 Duration of Strong Motion The duration of the strong motion becomes also longer for larger earthquakes. Durations of strong motion were investigated for large earthquakes in and around Japan. Figure 3 shows distribution of strong-motion durations for the 2003 Off-Tokachi Earthquake. The colors denote strong-motion duration at stations. Data from K-NET [Kinoshita, 1998] and KiK-net [Aoi et al. 2000] is used for this analysis. The durations in the north of the epicenter are shorter than those in the other directions in the figure. The duration difference is considered due to directivity effect. Figure 4 shows azimuthal distribution of the duration, which shapes a sinusoidal curve. It is possible to estimate fault length and rupture direction from the distribution (Izutani and Hiraswa, 1987). The arrow in Figure 3 indicates estimated fault length and rupture direction. When the faulting is simple unilateral, this method is useful for rapid estimation of the fault parameters. Figure 5 shows relationship between moment magnitude and strong-motion duration of earthquakes which occurred in and around Japan. Good correlation is seen between strong-motion duration and earthquake magnitude. The duration of the earthquake in March, 2011 exceeded eighty seconds, which was the longest among those of the large earthquakes. It is difficult to estimate the accurate moment magnitude on the basis of the duration due to the large scatter. We can, however, judge possible magnitude underestimation by the duration. Fig. 3. Distribution of strong-motion duration of the 2003 Off-Tokachi Earthquake [Aoki et al., 2011]. The star indicates the epicenter of the earthquake, and the arrow indicates estimated fault length and rupture direction by the method of Izutani and Hirasawa (1987). Fig. 4. Azimuthal distribution of strong-motion duration of the 2003 Off-Tokachi Earthquake [Aoki et al., 2011]. Blue curve indicates the theoretical distribution of the estimated fault. Fig. 5. Relationship between moment magnitude (after the Global CMT Project) and strong-motion duration Dobs of earthquakes which occurred in and around Japan [Aoki, et al., 2011]. The red circle denotes that of the 2011 off the Pacific coast of Tohoku Earthquake. 2.2.5 Magnitude Estimation from P-wave Some magnitude determination methods from P-wave have been proposed. Yoshida (1995) proposed a magnitude determination method from P-wave displacement amplitude, Mp. Figure 6 shows the station magnitude Mp of the 2011 off the Pacific coast of Tohoku Earthquake. The horizontal axis of Figure 6 denotes time of magnitude determination from the origin time of the earthquake which corresponds to S-wave arrival time at the station. The magnitude grows before about four minutes. This reflects the extension of the rupture area. Tsuboi et al. (1995) developed a method of estimating moment magnitude, Mwp, from integrated P-wave. Figure 7 shows the magnitude Mwp of the 2011 off the Pacific coast of Tohoku Earthquake. Ogawara et al. (2004) showed a magnitude determination method, Mwliss, based on squared amplitude of broadband seismic wave. Figure 8 shows the magnitude Mwliss of the 2011 off the Pacific coast of Tohoku Earthquake. These magnitudes show similar variation along the time from the origin time, and reach about nine in five minutes. It is possible to determine the magnitude of this earthquake properly in five minutes with these methods. Dispersion of Mwp is smaller compared with other P-wave magnitudes. Fig. 6. Mp of the 2011 off the Pacific coast of Tohoku Earthquake (Mw 9.0) [Yoshida et al., 2011b]. The red circle denotes station magnitude calculated from seismic wave between P and S arrivals at each station. The blue small dot denotes that calculated from seismic wave contaminated with S-wave. The horizontal axis shows the time when the magnitude is estimated at each station measured from the origin time. Fig. 7. Mwp [Tsuboi et al., 1995] of the 2011 off the Pacific coast of Tohoku Earthquake [Yoshida et al., 2011b]. The symbols are the same as those in Figure 6. Fig. 8. Mwliss [Ogawara et al., 2004] of the 2011 off the Pacific coast of Tohoku Earthquake [Yoshida et al., 2011b]. The symbols are the same as those in Figure 6. 2.2.6 Magnitude Determination from Long-period Seismic Wave It takes a long time to complete a rupture of a large earthquake. Excitation of long-period seismic wave is one of features of large earthquakes. The cutoff period (6 s) for the displacement magnitude used for the first tsunami warning was too short for the 2011 off the Pacific coast of Tohoku Earthquake. Usage of long period components of seismic wave would help to estimate earthquake magnitude properly. Figure 9 shows seismic waves processed with filters of different frequency responses. The figure shows seismic waves of the 2011 off the Pacific coast of Tohoku Earthquake (the left) and the 2003 Off-Tokachi Earthquake (the right). The upper of the figure shows seismic waves with the same response for the displacement magnitude. The lower shows those processed with a filter of 200-1000 second pass-band. While amplitudes of short-period seismic wave are not so different between the two earthquakes, those of long-period seismic wave differ very much. This is one method to distinguish the difference of these magnitudes. Magnitude determination methods from various frequency ranges are developed. Peak displacements of seismic waves of 1, 2, 5, 10, 20, 50 and 100 second cutoff periods are used to determine magnitude. Phase type is not cared in this method, and peaks are possibly those of S-waves. Figure 10 shows estimated durations to determine magnitude only from P-wave and from S-wave. It is necessary to wait for the completion of fault rupture to get enough length of data. At a station close to source area, S-P time would be shorter than the rupture duration. It is considered that magnitude determination from S-wave peak is more rapid than that only from P-wave peaks when local seismic data are available. Figure 11 shows growth of the magnitude with time from the origin time on the horizontal axis for the earthquakes of the 2010 Maule Earthquake and the 2011 off the Pacific coast of Tohoku Earthquake. The averaged magnitudes calculated from closest ten stations are shown in the figure. The magnitudes from long-period seismic wave reach the final values within three minutes for the earthquake. Fig. 9. Filtered Seismic waves of the 2011 off the Pacific coast of Tohoku Earthquake (the left) and the 2003 Off-Tokachi Earthquake (the right) [Yoshida et al., 2011b]. The upper shows seismic waves for the displacement magnitude used for the tsunami warning, and the lower shows those processed with a filter of a 200-1000 second pass-pand. Fig. 10. Assumed time to estimate earthquake magnitude from P-wave (the green broken curve) and S-wave (the red solid curve) [Katsumata et al., 2011]. The curves indicate relationships between epicentral distance and sum of travel time and assumed rupture duration. Sixty seconds is assumed as the rupture duration here. Fig. 11. Growth of magnitude from seismic waves of various cutoff-periods [Katsumata et al., 2013]. The horizontal axis shows time from the origin time of the earthquake. The numerals indicate cutoff periods of the filters. 2.2.7 Estimation of Fault Plane with High-sample-rate GNSS Data High sample rate (1 sample/second) GNSS data is currently available. Those data can be used for rapid estimation of the seismic source parameters. very-long period seismic data. The GNSS data can be used as Here we used the time-series GNSS data to obtain CMT solutions with W-phase analysis (Kanamori and Rivera, 2008), and then a finite fault plane model is estimated so that calculated static co-seismic displacements would be consistent with the observed ones. (Fig. 12). Fig. 12. Estimation steps of fault plane with high-sampling-rate GNSS data. 2.2.8 Conclusions Several magnitude determination methods are being developed to estimate earthquake magnitude soon after occurrence of a large earthquake for tsunami warning. Magnitude estimation from span of strong-motion area, strong-motion duration, P-wave, and S-wave amplitudes were examined. Combination of these methods is expected to help us to issue a proper tsunami warning for the next great earthquake. Acknowledgements Data from National Research Institute for Earth Science and Disaster Prevention (K-NET, KiK-net, F-net), University of Chile, Geospatial Information Authority of Japan, and the Japan Meteorological Agency were used in this study. Hypocenter parameters of the unified seismic catalog of Japan and Global CMT Project are referred to. The unified seismic catalog of Japan is based on data from the National Research Institute for Earth Science and Disaster Prevention, Hokkaido University, Hirosaki University, Tohoku University, University of Tokyo, Nagoya University, Kyoto University, Kochi University, Kyushu University, Kagoshima University, the National Institute of Advanced Industrial Science and Technology, Tokyo metropolitan government, Shizuoka prefectural government, Kanagawa prefectural government, the City of Yokohama, the Japan Marine Science and Technology Center, and the Japan Meteorological Agency. We thank Y. Kaida for many valuable discussions about source area estimation from seismic intensity distribution. REFERENCES . Aoki, S., Y. Yoshida, and A. Katsumata (2011), “The characteristics of rupture propagation deduced from strong motion duration”, Programme and Abstract of The Seismological Society of Japan Fall Meeting, P2-42. Aoi, S., K. Obara, S. Hori, K. Kasahara, and Y. Okada, (2000), “New strong-motion observation network: KiK-net”, EOS. Trans. Am. Geophys. Union, Vol. 81, F863. Usui, Y., S. Aoki, N. Hayashimoto, T. Shimoyama, D. Nozaka, and T. Yoshida (2010), “Description of and advances in automatic CMT inversion analysis”, Quart. J. Seism., Vol 73, 169-184. Izutani, Y. and T. Hirasawa (1987), “Use of strong motion duration for rapid evaluation of fault parameters”, J. Phys. Earth, Vol. 35, 171-190. Kanamori, H. and L. Rivera (2008): Source Inversion of W phases: speeding up seismic tsunami warning, Geophys. J. Int., 175, 222-238. Katsumata, A. (2004), “Revision of the JMA displacement magnitude”, Quart. J. Seism., Vol. 67, 1-10. (in Japanese with English abstract) Katsumata, A., S. Aoki, Y. Yoshida and K. Kimura (2011), “Quick magnitude determination based on peak velocity and displacement”, Programme and Abstract of The Seismological Society of Japan Fall Meeting, P2-67. Katsumata, A., H. Ueno, S. Aoki, Y. Yoshida and S. Barrientos (2013): Rapid magnitude determination from peak amplitudes at local stations, Earth Planet Space, 65, 843-853. Kinoshita, S. (1998), “Kyoshin Net (K-NET)”, Seismological Research Letters, Vol. 69, 309-332. Ogawara, T., T. Furudate and M. Okada (2004), “Preliminary empirical formula to estimate moment magnitude of teleseismic event by using LISS data”, Technical Reports of the Matsushiro Seismological Observatory, JMA, Vol. 21, 75-81. Si, H. and S. Midorikawa (1999), “New attenuation relationship for peak ground acceleration and velocity considering effect of fault type and site condition”, J. Struct. Constr. Eng., Vol. 523, 63-70. (in Japanese with English abstract and figure captions) Tsuboi, S., K. Abe, K. Takano, and Y. Yamanaka (1995), “Rapid determination of Mw from broadband P waveforms”, Bull. Seism. Soc. Am., Vol. 85, 606-613. Ueno, H., A. Katsumata, Y. Kaida, and T. Yokota (2013): Rapid source area estimation based on seismic intensity distribution, Japan Geoscience Union Annual Meeting 2015,SSS33-P26. Yokota, T. and Y. Kaida (2011), “Estimation of magnitude using distribution of seismic intensity”, Programme and Abstract of The Seismological Society of Japan Fall Meeting, P2-22. Yoshida, Y (1995), “Magnitude determination from P-wave amplitude”, JMA Technical Note of Seismology and Volcanology, Vol. 71, 41-52. Yoshida, Y. (2005), “Rupture process of the 2003 Off-Tokachi Earthquake”, Technical Report of Japan Meteorological Agency,Vol. 126, 9-14. Yoshida, Y., H. Ueno, D. Muto and S. Aoki (2011a), “Source process of the 2011 off the Pacific coast of Tohoku Earthquake with the combination of teleseismic and strong motion data”, Earth Planets Space, Vol. 63, 565-569. Yoshida, Y., S. Aoki, A. Katsumata and T. Yokota (2011b), “Evaluation of rapid Mw determination method”, Programme and Abstract of The Seismological Society of Japan Fall Meeting, P2-69. 2.3 Application of Tsunami Forecast Chart Yutaka HAYASHI (MRI, JMA) 2.3.1 Background Tsunami database [Tatehata, 1997] is a tool to realize tsunami early warning based on seismic observation. And as described in Section 2.1, Chilean prototype tsunami database has developped as the key technique used for near-future near-field tsunami forecast system in Chile. It will enables us to divide coast lines of mainland Chile into forecast areas. However, from the viewpoint of reliability and robustness of Chilean tsunami forecast system, it also is preferable to develop a couple of independent methods as backups in preparation to earthquakes beyond the scenario set engaged in the database, the troubles of a system, and data transmission failure between organizations of tsunami warning operation. In this section, application potency of Tsunami Forecast Chart (hereafter TFC) is discussed as a backup of oncoming Chilean tsunami database system and also as a benchmark for determining thresholds to issue precaution evacuation based on Modified Mercalli Intensity (Section 4.1). 2.3.2 Outline of Tsunami Forecast Chart (TFC) Various kind of TFC had been used in Japan until starting quantitative tsunami forecast by Japan Meteorological Agency (JMA) in 1999 [Kusano and Yokota, 2011]. The latest version of the chart which was used from 1987 to 1999 determines three boundary lines of tsunami advisory/warning levels and x- and y-axis of the chart are distance from the epicenter to any forecast area, and seismic magnitude, respectively (Fig. 1). JMA had been operating tsunami forecast for 18 tsunami forecast areas in Japan using the focus location and magnitude determined by real-time. 2.3.3 Verification and Modification of TFC Because the latest version of TFC (hereafter TFC-J) did not experienced large near-field tsunami during its life time from 1987 to 1999, the chart has verified by maximum tsunami height observed during a different period (Fig. 2). Verification suggests that using TFC-J would have mislead underestimation of tsunami height by the 2011 off the Tohoku earthquake even under assumption of availability of final moment magnitude (Mw9.0). In addition, TFC-J cannot be applied to Chilean tsunami warning system as itself, because tsunami advisory/warning category for larger than magnitude 9.0 and further than 700km in epicentral distance is not defined. Then alternative two charts, or sets of threshold lines, are proposed. One is TFC-A, which is derived from definition of tsunami magnitude [Abe, 1979] and equations for tsunami height forecasting using seismic or tsunami magnitudes [Abe, 1989]; the other set TFC-M is the modified by using scaling law on large earthquakes by Murotani et al. (2013) with consideration of the standard aspect ratio of a seismic fault and a possible location of a hypocenter within its seismic fault (Fig. 2). Three lines for each set mean maximum epicentral distance where maximum tsunami height become 0.25m, 0.75m, or 2.5m, when the definition of tsunami magnitude is applied. Both TFC-A and TFC-M better explain the maximum tsunami height by the 2011 off the Tohoku earthquake than TFC-J (Fig. 2). Finally, TFC-A and -M for Chilean use are redefined so that the boundary lines indicate the epicentral distance for maximum tsunami height of 0.3m, 1m, and 3m, which are boundary values of tsunami warning ranks (small or no tsunami, tsunami watch, warning, warning (major tsunami)) under planning by SHOA (Fig. 3). As the typical extent of an Magnitude 8 or larger earthquake is estimated to exceed 100 km from scaling laws. TFC for large tunamigenic earthquakes is considered to involve uncertainty more than a 100 km. 2.3.4 Conclusions After modification of Tsunami Forecast Chart which used to be applied in Japan during non-database period, new TFCs for Chilean use (Fig. 3) are proposed. Because of the accuracy of TFC, in case of introducing tsunami forecast area in Chile, one or a few area for each state is desirable. Usage of TFC in real-time is simple; it is just to look up the warning rank for a certain forecast area in the chart using the Mw (or substitute seismic magnitude) as y-value and the minimum distance from the epicenter to the concerning forecast area as the x-value. Newly proposed TFCs can be used as one of backup methods (e.g. for unexpected earthquakes) of oncoming Chilean tsunami database system (Section 2.1); and also they can be used as a benchmark for determining thresholds to issue precaution evacuation based on Modified Mercalli Intensity (Section 4.1). Selection between TFC-A and TFC-M by using tsunami observation data in Chile, comparing to results of forecast experiments by using Chilean tsunami database (Section 2.1), and/or exchanging the equation using local seismic scaling laws may more improve the reliability of the chart. REFERENCES Abe, K. (1979): Size of great earthquakes of 1837-1974 inferred from tsunami data. J. Geophys. Res., Vol. 84, 1561-1568. Abe, K. (1989): Estimate of tsunami heights from magnitudes of earthquakes and tsunami. Bull. Earthq. Res. Inst., Vol. 64, 51-69. (in Japanese with English abstract) Kusano, F., and T. Yokota (2011): History of tsunami warning services in Japan. Quart. J. Seism., Vol. 74, 35-91. (in Japanese) Murotani, S., K. Satake, and Y. Fujii (2013): Scaling relations of seismic moment, rupture area, average slip, and asperity size for M~9 subduction-zone earthquakes. Geophys. Res. Lett., Vol. 40, 5070-5074. doi:10.1002/grl.50976. Tatehata, H. (1997): The new tsunami warning system of the Japan Meteorological Agency. in “Perspectives on tsunami hazard reduction" ed. G. Hebenstreit, Kluwer Academic Publishers, Dordrecht, The Netherlands, pp. 175-188. Fig. 1. Tsunami Forecast Chart used by Japan Meteorological Agency from 1987 to 1999 (translated from [Kusano and Yokota, 2011]). level to each forecast area. The chart indicates the guidelines for the tsunami warning Fig. 2. Maximum Tsunami Height Observation Data and Tsunami Forecast Chart based on JMA's categories. Maximum tsunami heights observed at tidal stations along the Pacific coast of Japan from earthquakes in the Pacific Ocean during 1999-2013 are classified into four and plotted (circles). Data observed in Tokyo, Ise or Osaka Bay are excluded. Three sets (TFC-J, -A, and -M) of boundary lines for neighboring tsunami forecast categories (0<h<0.25m, 0.25<=h<0.75m, 0.75m<=h<2.5m, 2.5m<=h) by JMA are indicated; TFC-J was used by JMA until 1999 [Kusano and Yokota, 2011], TFC-A was derived from definition of tsunami magnitude and corresponding equations in Abe (1989); and TFC-M which is modified by using the scaling law between faults areas and moment magnitudes by Murotani et al. (2013) and consideration of a bilateral fault rupture under assumption of standard aspect ratio (length / width = 2). Lines of TFC-A and TFC-M for far-field earthquakes are depend on the location of the epicenter; the uppermost lines (Mw=9.2, 8.7, 8.2) for earthquakes occurred in Aleutian-Alaska area, the middle lines (Mw=8.9, 8.4, 7.9) for South America, and the lowermost lines (Mw=8.7, 8.2, 7.7) for earthquakes along Kurile-Kamchatka. Fig. 3. Proposal of Tsunami Forecast Chart (TFC) for Chile as a backup for Database System. Data and threshold lines are based on SHOA's tsunami ranks. Maximum tsunami heights data used in this figure are same as the Fig. 2. TFC-A and -M are shifted from those of Fig. 2 so that lines indicate the threshold values (0.3m, 1m, 3m for the lower limit of tsunami alert, tsunami alarm, and major tsunami) for tsunami warning ranks by SHOA. 2.4 Instrumental Modified Mercalli Intensity Akio KATSUMATA, Yutaka HAYASHI, Kazuki MIYAOKA, Hiroaki TSUSHIMA (MRI, JMA), and Toshitaka BABA (Tokushima Univ.) 2.4.1 Introduction Seismic intensity is usually measured by sense of human or degree of damages. If seismic intensities are reported through the network like as in Japan (Hachimine, 1989), it is considered that rapid grasp of possible damaged areas can be done. Some methods have been proposed as instrumental modified Mercalli scale (e.g., Trifunac and Brady 1975; Wald et al., 1999; Shabestaria and Yamazaki, 2001). It is considered that low-coast equipment of current technology is enough to get seismic intensity value. The seismic intensity itself could be a trigger of evacuation from tsunami (Hayashi et al., 2014). A former tsunami alarm in Japan was based on seismic intensity observed by human. Similar way could be used for any country in the world. 2.4.2 Instrumental Seismic Intensity Some formulas have been proposed to relate modified Mercalli scale to instrumentally obtained values. Relationship among seismic intensity, peak ground acceleration and velocity were investigated by Trifunac and Brady (1975) and Wald et al. (1999). Shabestaria and Yamazaki (2001) used a filter for Japan Meteorological Agency (JMA) seismic intensity scale to obtain the value of the modified Mercalli scale. Figure 1 shows observed seismic intensity and instrumentally obtained seismic intensities for the earthquake on April 1, 2014 off Iquique. Whereas effective intensity range is limited in each method, attention is not paid to the applicable range of the formulas here. Since the seismic intensity is usually affected by surface geology at the observation point very much and the observation sites are different between observed and instrumental values, it would be not proper to compare instrumental values with the observed values here. Fig. 1. Comparison of various instrumental seismic intensities. MMI:Yama, Shabestaria and Yamazaki (2001); MMI:PGA-W, Wald et al. (1999) based on peak ground acceleration; PGV-W, Wald et al. (1999) based on peak ground velocity; MMI-PGA-T, Trifunac and Brady (1975) based on peak ground velocity; MMI-PGV-T Trifunac and Brady (1975) based on peak ground velocity; OBS:MMI, observation by ONEMI. 2.4.3 Prototype of Instrumental Seismic Intensity Meter The MEMS (micro electro mechanical systems) sensors are currently available with low coast. Small computer systems are also available. We developed a low-cost equipment using a MEMS sensor and a small computer which could tell instrumental seismic intensities with display and synthetic voice. PC LCD monitor USB accelerometer Keyboard Fig. 2. Prototype of instrumental seismic intensity meter. Relatively low-noise MEMS sensors are available today. lower limit of being felt. 0.2 gal. Ground shaking of 1 gal is about the In the prototype of the meter, we used one of which noise level is about The noise level can be used to detect the ‘felt’ ground shaking. 2.4.4 Further Applications of the Seismic Intensity Meter It is considered that one of the quickest ways of tsunami evacuation is movement to higher place soon after a strong and long ground shaking. Information from mass media could be helpless sometimes due to power failure. It would be helpful if there is a device which tells risk of coming tsunami. Strong ground motion means that the source area of the event would be close to the current location, and long ground shake means that the event magnitude would be large. Fig. 3 shows the relationship between magnitude and various instrumentally observed values. Here we considered strong-motion duration, ground velocity, ground displacement, and their combination. The velocity and displacement are obtained from acceleration with numerical integration. The period of the high-pass filter is set at 20 seconds here. Velocities or displacements corresponding to acceleration lower than 0.2 gal are ignored assuming the usage of the MEMS sensors. The strong motion duration is good for distinguishing the huge earthquakes such as the 2011 off the Pacific coast of Tohoku Earthquake. However, the duration is not suitable to distinguish M7-class earthquake and M8-class earthquake. If we use ratio of numbers of data plot as the index of suitability, the displacement is the most suitable in the case of Fig. 3 Fig. 3. Relationship between magnitude and observed values. Upper left, duration of the modified Mercalli scale value calculated with Wald et al. (1999); upper center, PGV, upper right, PGD, lower left, product between the duration and the PGV; lower right, product of the duration and the PGD. The MEMS sensor equipment is considered to be used also for the earthquake early warnings. For example, it is possible to estimate S-wave amplitude from the amplitude of P-wave of a single sensor. Other application is considered to be possible to install in the instrumental seismic intensity meter. 2.4.5 Conclusions Instrumental seismic intensity meter is proposed here. If the instrumental seismic intensities are reported through the network, it would be valuable for quick dispatch of rescue teams. Such instrument can be used for standalone tsunami warning or earthquake early warning. Acknowledgements Data from University of Chile were used in Section 2.4.2 in this document. Data of JMA seismic intensity meters are used in section 2.4.3 REFERENCES Hachimine, T., 1989, Instrumentation of the JMA seismic intensity scale, Quart. J. Seismol., 52, 43-68. (In Japanese) Hayashi, Y., A. Katsumata, K. Miyaoka, H. Tsushima, P. Catalan, J. Baquedano, C. Zelaya, V. Orellana, and T. Baba, 2015, A method to ensure consistency between tsunami forecast chart-based warnings and instrumental Mercalli scale intensity-based evacuation, IUGG2015, JP05p-021. Shabestaria, K. T. and Fumio Yamazaki, 2001, A Proposal of Instrumental Seismic Intensity Scale Compatible with MMI Evaluated from Three-Component Acceleration Records, Earthquake Spectra, 17, 711–723. Trifunac, M. D. and A. G. Brady, 1975, On the correlation of seismic intensity scales with the peaks of recorded strong ground motion. Bull. Seism. Soc. Am., 65, 139-162. Wald, D. J., V. Quintro, T. H. Heaton and H. Kanamori, 1999, Relationships between peak ground acceleration, peak ground velocity and Modified Mercalli Intensity in California, Earthquake Spectra, 15, 557-564. Chapter 2. Proposal on Tsunami Warning System Based on Seismic Observation 2.1 Developing Tsunami Database Patricio CATALAN (UTFSM) Chapter 3. Proposal on Tsunami Warning System Based on Tsunami Observation 3.1 Review on Measurement Technologies for Deep-Ocean Tsunami Toshitaka BABA1,2 1 Institute of Science and Technology, University of Tokushima 2 Japan Agency for Marine-Earth Science and Technology Abstract In recent years, deep-ocean tsunami measurements became to more play an important role in understanding of the physics of tsunami generation and propagation and early warning. Currently several types of measurement systems are working around the world for deep-ocean tsunami such as Deep-ocean Assessment and Reporting of Tsunami, cabled connected bottom pressure recorders and GPS buoys. These observation systems successfully provided high-precision tsunami data of the 2010 Chile and 2011 Tohoku tsunamis and others. Japanese GPS buoys contributed to modify the tsunami warning level during the 2011 Tohoku tsunami issued by Japan Meteorological Agency. New measurement technologies are also developed such as oceanographic radars and ocean-bottom electromagnetometers. We here compare these measurement technologies for the deep-ocean tsunami. 3.1.1 Introduction Measurement technologies of tsunamis in the deep ocean were greatly developed in the last decade so that we acquired many high-quality data of tsunami off shore. A satellite altimeter attached to Jason-1, for example, has detected the 2004 Sumatra tsunami in the midst of the Indian Ocean (Gower, 2005). Deep-ocean Assessment and Reporting of Tsunami (DART, e.g. Bernard and Meinig, 2011) buoys distributed in the Pacific Ocean recorded inherent tsunamis caused by the 2010 Chile earthquake and the 2011 Tohoku earthquake. These high-quality records were incredibly accurate enough to expose incompleteness in the numerical tsunami simulation using the conventional shallow water theory. There were significant discrepancies between the observed and computed tsunami waveforms in terms of the time of tsunami arrival and a small drawdown prior to the first-arriving positive peak during the Chile and Tohoku tsunamis (e.g. Fujii and Satake, 2013). But soon after, following studies (e.g. Allgeyer and Cummins, 2014, Watada et al., 2014) successfully improved the tsunami calculation model by adding effects of deformation of the Earth due to tsunami load and seawater density stratification to the equation of motion. We were able to obtain the advanced model to simulate the far-field tsunami thanks to progress on the tsunami measurement technologies in the deep ocean. There are two advantages in the deep-ocean tsunami measurement. Tsunami is largely deformed in the shallow sea region by effects such as nonlinearly, bottom friction, refraction, and reflection at the coast. We can ignore them in the deep ocean so as to get inherent tsunami signals. It gives useful information to tsunami source analysis. The second advantage is rapidness in tsunami detection. The tsunami stations off shore can detect tsunami before it attacks to the coast. It must be very important in terms of tsunami early warning. Tsunami early warning systems are basically based on analyses of earthquake hypocenters and magnitude derived from seismic observations, because seismic waves travel much faster than tsunamis. However, we have already known tsunami earthquakes generate large tsunamis with weak seismic shaking (Kanamori, 1972). Submarine landslides generate tsunami not accompanied with seismic wave (e.g., Tappin et al., 2001). For giant earthquakes such as the 2010 Chile and 2011 Tohoku earthquakes, the maximum amplitude of the short-period seismic wave would be saturated, resulting in underestimation of the tsunami warning level via underestimation of the earthquake magnitude (Hoshiba and Ozaki, 2012). Tsunami measurements off shore can predict tsunami accurately even if that caused by tsunami earthquakes, submarine landslides or giant earthquakes because it monitors tsunami signal itself, not seismic wave. It is clearly needed to embed the real-time data monitored by the deep-ocean tsunami measurements in the tsunami early warning systems. Chapter 3 leads the readers to several new real-time basis methods for tsunami prediction that effectively use the offshore tsunami data. But, for introduction, we here summarize the tsunami measurement technologies in the deep ocean to discuss their merits and demerits. 3.1.2 Measurement Technologies for Deep-Ocean Tsunami 3.1.2.1 Bottom Pressure Based Technologies Bottom pressure recorders (BPRs) are common for the tsunami observation in the deep ocean. The hydrostatic bottom pressure fluctuations during tsunami are converted to the sea level change by assuming a linear pressure-water height relationship. We have to resolve tsunami of 1-2 cm at least in the open ocean because tsunami is grown up by approaching the coast so as to have a potential casing damage in the coastal area. BPR is very sensitive to pressure change. It can resolve pressure change corresponding to less than a millimeter of water column. DART developed by Pacific Marine Environmental Laboratory (PMEL), National Oceanic and Atmospheric Administration uses the BPR for tsunami observation. DART is composed of a BPR and a surface buoy which receives the data from the BPR and send them to a land station via the Iridium satellite in real time (e.g. Bernard and Meinig, 2011). This is a battery-powered system. The endurance of deployed system is two years for the surface buoy and four years for the BPR. DART buoys are many distributed along the Pacific Rim, near Hawaii, east off USA, Caribbean Sea and Indian Ocean (Figure 1). PMEL assimilates the DART data in real-time into a tsunami forecast system for far-field tsunami (Titov, 2009). Japan has conducted cabled BPR observations since 1979. Japan Meteorological Agency (JMA) installed two BPRs off Tokai region where a large subcudtion earthquake is anticipated to occur (Meteorological Research Institute, 1980). This was followed by installations of cabled systems off Boso (Fujisawa, 1986), in Sagami Bay (Eguchi et al., 1999), off Sanriku (Kanazawa and Hasegawa, 1997), off Muroto (Monma et al., 1999), off Tokachi (Hirata et al., 2002). JMA again deployed a new cable system equipped with 3 BPRs in Tokai region in 2008. Two large-scale comprehensive cabled systems has been deployed in recent years. One of them named as “Dense Ocean Network system for Earthquakes and Tsunamis (DONET, Kawaguchi et al., 2012) was deployed on the sea bottom in the Nankai subduction zone. DONET has 51 BPRs in total to cover the seismic source regions of 1944 Tonankai and half of the 1946 Nankai earthquakes (Fig. 2). “Seafloor observation network for earthquakes and tsunamis along Japan Trench (S-net)” will install about 150 BPRs on the sea bottom where the 2011 Tohoku earthquake occurred (Uehira et al., 2012). The connection of those BPRs by a submarine fiber-optic cable to stations on land enables us to feed power to the systems and acquire submarine data in real time. Many ocean-bottom seismometers are also attached to these cable systems. We will be able to acquire precious data recorded directly above the source regions of the subduction zone earthquakes by a large number of cabled BPRs and seismometers. The cabled BPR observations are also carried out in Canada (Barnes et al., 2008) and Taiwan (Hsiao et al., 2014). 3.1.2.2 GPS Based Technology The GPS buoy (Fig. 3, Kato et al., 2005) employs the real-time kinematic (RTK) GPS technique to detect a tsunami before it reaches the coast. It measures the buoy altitude above the Earth ellipsoid WGS84 every second using RTK-GPS instruments on the buoy and at the base station on land. The GPS buoy system relies on an approximation that the height from the water surface to the GPS receiver on the top of the buoy is constant, but this height actually changes owing to the tilting of the buoy. The system therefore measures also the tilting angle. The 17 GPS buoys were deployed around Japan (as of July, 2015). The GPS buoys have successfully observed many tsunami including the 2010 Chile and 2011 Tohoku tsunami. During the 2011 Tohoku tsunami, the GPS buoys played an important role to modify the tsunami warning level issued by JMA. However, the observation accuracy is lower than BPR measurements. The root-mean-squares of the errors in vertical positioning is a few centimeters in the case of the fix solution for a distance of 20 km between the buoy and the base station, which is also affected by several factors, such as the GPS satellite number and geometry. This is the reason why deployment location is limited within 20 km from the coast. Displacement of the base station due to crustal deformation during the earthquake also decrease the accuracy of observation. A new GPS buoy system using precise point positioning technique is under development in order to solve these problems. If it is accomplished, the base station is not needed, we can install it anywhere you want in the ocean. Real-time tsunami forecast methods using the data recorded by the GPS buoy cooperating with BPRs are developed by, for example, Tsushima et al. (2014) and Takagawa (This issue, Chapter 3.4). 3.1.2.3 Rader Based Technology Oceanographic radar remote sensing technology may have great potential for tsunami observation in the open ocean, which uses radio waves to measure the flow directions and velocities of the surface layer. We would be able to compute tsunami height distribution form surface current velocities by assuming the long wave theory. It has a wide field of view, which extends approximately ±45° in direction and beyond 50km in distance while the BPRs and GPS buoys conduct point-based measurements. Installation and maintenance of oceanographic radar is relatively easy because of land-based so as to make continuous observations of broad area over long periods. The first observation of tsunami by the oceanographic radar was carried out during the 2011 Tohoku tsunami. The radars captured the 2011 Tsunami not only in Japan (Hinata et al., 2011), and but also in California and Chile (Lipa et al., 2011, HELZEL, 2011). A satellite altimeter can measure sea surface height with resolution of a few centimeters for the average of a 5-km circle ocean. By using this instrument, Jason-1 has detected the 2004 Sumatra tsunami in the midst of the Indian Ocean (Gower, 2005). It showed that the ocean raised up above the height measured in previous cycles, to a maximum of about 70 cm, and the subsequent 30- to 40-cm drop below the average. Because satellites are continuously moving, timing and location of the observation strongly depends on satellite tracks. It is unlikely an optimal instrument for the tsunami early warning. However, it must provide important data for the study of propagation of trans-ocean tsunami as well as BPRs. 3.1.2.4 Electromagnetic Based Technology A flow of electrically conducting seawater induces the secondary magnetic field on the Earth’s primary magnetic field. It has been discussed if tsunami flows can produce measureable magnetic field (e.g. Manoj et al., 2010). Sugioka et al. (2014) developed high-precision ocean-bottom electromagnetometer (OBEM) array network deployed in the mid of the Pacific Ocean and reported on the detection of deep-seafloor electromagnetic perturbations of 10-micron-order induced by the 2010 Chile tsunami. BPRs were also attached to their OBEMs. The time series of recorded vertical magnetic field indicated very good agreement with that recorded by BPRs. The resolution limit of OBEM of 0.01 nT can resolve tsunami with cm-order heights. Theoretical relationship also demonstrated that observation of the EM field with two electric and three magnetic field components, even at a single station, can extract information about variations in not only sea level, but also the propagation directions of tsunami waves, the phase velocity of tsunami propagation and its frequency dependency. 3.1.3 Discussion and Conclusion The BPRs and GPS buoys are operationally used in tsunami early warning. The BPR can detect tsunami of a few millimeters and more, while the GPS buoy extract tsunami larger than a few centimeters. Currently, the installation position of GPS buoy should be located in about 20 km or less from the coast because the positioning accuracy becomes much worse if the baseline length exceeds this distance, although they have been improving it to remove the distance limit. BPRs with DART can be installed in anywhere in the ocean. For the versatility, the BPR observation sounds like better than the GPS buoy. However, we have to again notice that BPR observes the water pressure at the bottom of ocean, not sea level. The BPR data is also disturbed by hydro-dynamic pressure vibrations from seismic Rayleigh waves and so on. When the station is located near the source, the short-period pressure vibration overlaps on the long-period hydrostatic components (tsunami). We have to apply an appropriate filter to extract tsunami component on the pressure data. When the station is located just above the source, a more serious problem is seen in the hydrostatic component. BPR can detect changes hydrostatic bottom pressure, which is equivalent to total depth (depth of the sea + tsunami height). We consider the hydrostatic pressure change at a point on the seafloor uplifted by an earthquake. Almost no hydrostatic pressure change is expected during earthquakes because the sea surface and the ocean bottom are dislocated equally, the result being no change in the total depth. The hydrostatic pressure suddenly decreases afterward as the tsunami propagates. A change in hydrostatic pressure corresponding to the vertical displacement of the seafloor remains after the tsunami has passed. Thus, the hydrostatic pressure is simultaneously affected by the tsunami and the vertical seafloor displacement. It would be difficult to monitor sea surface fluctuations associated with a near-field tsunami based on hydrostatic pressure changes immediately after the occurrence of an earthquake. On the other hand, the GPS buoy doesn’t matter where it is relative to the tsunami source because of direct monitoring of the height of sea surface. For early warning of near-field tsunamis, we accordingly emphasis necessity of further development of the tsunami measurement technologies in the deep ocean that enable to observe the initial sea surface deformation due to the earthquake, that is tsunami initial condition. We look forward to the new GPS buoy system using precise point positioning technique that doesn’t have the distance limit. We can distributed it in the region far from the coast where is just above the large subduction zone earthquakes could occur accompanied by tsunami so that the initial tsunami can be directly observed immediately after the earthquake. On the other hand, there may be good methods to retrieve the sea level change from the BPR data even for near-field earthquakes. It might be possible to the estimate amount of uplift at the seafloor by analyzing the short-period pressure fluctuations recorded by BPRs or data recorded by hydrophone, or ocean-bottom seismometer. These kind of analysis methods should also be investigated and developed more. The point-based measurements in the ocean using the BPRs and the GPS buoys require on-going maintenance and management with great efforts and cost because we need anyway research vessels to maintain them. Oceanographic radar, on the other hand, uses only land-based instruments expecting easy access and maintenance. It can monitor the wide region which extends approximately ±45° in direction and beyond 50km in distance by one station. We would like to ask the oceanographic radar measurements and the data analysis methods to be developed for the operational use of the tsunami early warning. Acknowledgements Figure 1 was provided by NOAA/PMEL via internet (http://www.ndbc.noaa.gov/dart.shtml). Figure 3 was provided by the Ministry of Land, Infrastructure, Transport and Tourism. We thank for their kindness. REFERENCES Allgeyer, S. and P. Cummins (2014), “Numerical tsunami simulation including elastic loading and seawater density stratification”, Geophys. Res. Lett., 41, 2368-2375, doi:10.1002/2014GL059348. Bernard, E., C. Meinig (2011), “History and future of deep-ocean tsunami measurements”, in Proceedings of Oceans' 11 MTS/IEEE, Kona, IEEE, Piscataway, NJ, 19–22 September 2011, No. 6106894, 7 pp. Eguchi, T., Y. Fujiwara, E. Fujita, S. Iwasaki, I. Watanabe, and H. Fujita (1998), “A real-time observation network of ocean-bottom-seismometers deployed at the Sagami trough subduction zone, central Japan”, Mar. Geophys. Res., 20, 73–94. Fujii, Y., K. Satake (2013), “Slip distribution and seismic moment of the 2010 and 1960 Chilean earthquakes inferred from tsunami waveforms and coastal geodetic data, Pure Appl. Geophys., 170, 1493-1509, doi:10.1007/s00024-012-0524-2. Fujisawa, K., S. Tateyama, A. Funasaki (1986), “Permanent Ocean-Bottom Seismograph Observation System off the Boso Peninsula (in Japanese)”, Sokkou jihou, 53, 127-166. Gower, J. (2005), “Jason 1 detects the 26 December 2004 Tsunami”, EOS. Trans. Am. Geophys. Union, Vol. 86, 37-38. HELZEL (2011), “WERA in Chile observed tsunami signatures”, WERA Newsletter, August, 2011, 2-3. Hinata, H., S. Fujii, K. Furukawa, T. Kataoka, M. Miyata, T. Kobayashi, M. Mituzani, T. Kokai, and N. Kanatsu (2011), “Propagating tsunami wave and subsequent resonant response signals detected by HF radar in the Kii Channel, Japan”, Est. Coast. Shelf Sci., 95, 268-273. Hirata, K., et al. (2002), “Real-time geophysical measurements on the deep seafloor using submarine cable in the southern Kurile subduction zone”, IEEE J. Oceanic Eng., 27, 170– 181. Hoshiba, M., T. Ozaki (2012), “Earthquake early warning and tsunami warning of JMA for the 2011 off the Pacific Coast of Tohoku earthquake (in Japanese with English Abstract)”, Zisin, 64, 155– 168. Hsiao, N.C., T.W. Lin, S.K. Hsu, K.W. Kuo, T.C. Shin, P.L. Leu (2014), “Improvement of earthquake locations with the Marine Cable Hosted Observatory (MACHO) offshore NE Taiwan”, Mar. Geophys. Res., 35, 327-336, doi:10.1007/s11001-013-9207-3. Kanamori, H. (1972), “Mechanism of tsunami earthquake”, Phys. Earth Planet. Inter., 5, 129-139. Kanazawa, T., and A. Hasegawa (1997), “Ocean-bottom observatory for earthquakes and tsunami off Sanriku, north-east Japan using submarine cable”, International Workshop on Scientific Use of Submarine Cables, Inst. of Electr. and Electr. Eng., Okinawa, Japan. Kawaguchi, K., Y. Kaneda, E. Araki et al. (2012), “Reinforcement of seafloor surveillance infrastructure for earthquake and tsunami monitoring in western Japan”, Proc. Oceans2012 Mts/IEEE, Yeosu, May 21-24, 2012. Lipa, B., D. Barrick, S. Saitoh, Y. Ishikawa, T. Awaji, J. Largier and N. Garfield (2011), “Japan tsunami current flows observed by HF radars on two continents”, Remote Sensing, 3, 1663-1679. Manoj, C., A. Kuvshinov, S. Neetu, T. Harinarayana (2010), “Can undersea voltage measurements detect tsunamis?”, Earth Planet Space, 62, 353-358, doi:10.5047/eps.2009.10.001. Meteorological Research Institute (1980), “Permanent ocean-bottom seismograph observation system (in Japanese with English Abstract)”, Tech. Rep. MRI 4, pp. 1 – 233, Tsukuba, Japan. Monma, H., N. Fujiwara, R. Iwase, K. Kawaguchi, S. Suzuki, and H. Kinoshita (1997), “Monitoring system for submarine earthquakes and deep sea environment”, MTS/IEEE OCEANS’97, Mar. Technol. Soc., Columbia, Md. Sugioka, H., Y. Hamano, K. Baba, T. Kasaya, N. Tada, D. Suetsugu (2014), “Tsunami: Ocean dynamo generator”, Sci. Rep., 4, 3596, doi:10.1038/srep03596. Tappin, D.R., P. Watts, G.M. McMurtry, Y. Lafoy, T. Matsumoto (2001), “The Sissano, Papua New Guinea tsunami of July 1998-offshore evidence on the source mechanism”, Mar Geol. 175, 1–23. Titov, V.V. (2009), “Tsunami forecasting”, In; The sea, Vol. 15, Tsunamis, (Eds. A. Robinson and E. Bernard), Harvard University Press, Cambridge, USA, pp. 371-400. Tsushima, H., R. Hino, Y. Ohta, T. Iinuma, S. Miura, (2014) “tFISH/RAPiD: Rapid improvement of near-field tsunami forecasting based on offshore tsunami data by incorporating onshore GNSS data”, Geophys. Res. Lett., 41, 3390-3397, doi:10.1002/2014GL059863. Uehira, K., T. Kanazawa, S.I. Noguchi et al. (2012), “Ocean bottom seismic and tsunami network along the Japan Trench”, AGU 2012 Fall meeting, OS41C-1736. Watada, S., S. Kusumoto and K. Satake (2014), “Traveltime delay and initial phase reversal of distant tsunamis coupled with the self-gravitating elastic Earth”, J. Geophys. Res. Solid Earth, 119, 4287-4310, doi:10.1002/2013JB010841. Figure 1. Present DART station distribution (as of July, 2015). The map was provided by PEML/NOAA (http://www.ndbc.noaa.gov/dart.shtml) Figure 2. Schematic map of Dense Oceanfloor Network system for Earthquakes and Tsunamis (DONET). Figure 3. A GPS buoy. 3.3 Methodology of Optimizing Tsunami Observation Array Using Tsunami Inversion Hiroaki TSUSHIMA (MRI, JMA), Joaquín MEZA (UTFSM), César NÚÑEZ (SHOA), Patricio CATALÁN (UTFSM), Cecilia ZELAYA (SHOA), Tomohiro TAKAGAWA (PARI), Toshitaka BABA (Tokushima Univ.), and Yutaka HAYASHI (MRI, JMA), 3.3.1 Introduction As mentioned in the previous sections, two ocean-bottom pressure (OBP) sensors have been deployed in deep sea off the Chilean coast, and new ones plan to be deployed. Tsunamis can be observed at those offshore stations earlier than in the near-field coasts, and therefore real-time tsunami forecasting based on the data will contribute to improvement of tsunami early warning. Tsunami forecasting methods based on offshore tsunami data have been developed by several researchers, including Japanese WG3 members [e.g., Baba et al., 2004, 2013; Tsushima et al., 2009, 2012a, 2012b; Hayashi, 2010, Takagawa and Tomita, 2012a, 2012b, 2014] (more details are reviewed by Tsushima and Ohta, 2014). Here we focus on a method based on estimation of tsunami source (tsunami inversion) from the OBP data, because the tsunami data acquired in deep ocean are not distorted by complex topography, and thus they keep information of tsunami source signature. One of the key factors for tsunami forecasting based on offshore tsunami data is array configuration of tsunami stations, because time when sufficient tsunami data can be observed at offshore stations strongly depends on spatial relationship between a tsunami source and the stations. If tsunami sensors are located near a tsunami source, tsunami can be detected very quickly. In addition, when the many sensors surround the tsunami source area densely, we can obtain sufficient data for knowing detail of the tsunami. However, it may not be realistic to densely deploy OBP sensors densely over the whole seismogenic zone along the Chilean coast, because the seismogenic zone along the coast is very long (~3000 km) and an offshore tsunami sensor is not cheap. In perspective of cost-performance ratio, it is very important to improve forecasting performance by adding the small number of new offshore tsunami stations. In this section, we propose a methodology of optimization of offshore tsunami observation array for effective improvement of tsunami forecasting based on tsunami inversion. 3.3.2 Real-time tsunami forecasting method based on tsunami inversion A tsunami forecasting method we use was developed by Tsushima et al. [2009]. The method consists of two steps to predict tsunamis using offshore tsunami data. Firstly, we perform waveform-inversion analysis to estimate spatial distribution of sea-surface displacement soon after an earthquake occurrence (i.e. initial sea-surface displacement distribution) by using offshore tsunami data. Then, we synthesize tsunami waveforms to forecast arrival times and wave heights of the first tsunamis along the near-field coast (Fig. 1). In the successive calculation, seismic magnitude and fault geometry are never used, and therefore it can be applicable to tsunami earthquakes [Kanamori, 1972] and non-interplate earthquakes such as outer-rise normal faulting events. Since the real-time calculations are based on linear approximation, the computation time is about a few minutes. In advance of an earthquake occurrence, we calculate tsunami waveforms from unit tsunami sources (i.e. Green’s functions) to perform inversion and waveform synthesis based on linear combination of those Green’s functions. The good performances of the method have been shown by its application to actual tsunami observation data from large earthquakes around Japan [Tsushima et al., 2011, 2012]. Fig. 3.3.1 Schematic diagram of tsunami forecasting method based on tsunami inversion. 3.3.3 New methodology for optimization of offshore tsunami observation array We propose a new methodology to determine an optimal array configuration of offshore tsunami sensors for tsunami forecasting. The methodology is based on numerical simulations of tsunami forecasting by assuming various offshore tsunami arrays. The outline of the methodology is as follows: (1) we assume a certain array configuration of offshore tsunami stations; (2) tsunami waveforms at the offshore and coastal observing points were computed numerically by assuming fault motion of an earthquake of interest. The resultant waveforms are regarded as observation data at each stations (hereafter, synthetic observations); (3) tsunami forecasting calculation (inversion and waveform synthesis) is carried out by using the synthetic observation data at the offshore stations; (4) evaluation of forecasting accuracy is made by comparison between synthetic observations (results from step 2) and predictions (results from step 3) at coastal points. For the evaluation, we define forecasting-accuracy indicators relating to wave heights, arrival times, and waveform of the first tsunamis in the near-field coasts (here we call score), because these tsunami parameters are very important for tsunami disaster mitigation; (5) we return to step 1 and assume different offshore tsunami array configurations, and then steps 2 and 3 are repeated to investigate score for each array configuration. After a lot of these simulations, we compare the scores to determine an optimal array configuration. 3.3.4 Effectiveness of our new methodology To know performance of our new optimization methodology, we applied it to the northern Chile sea area. Here, we assume a tsunamigenic earthquake that is located in the north of the 2014 Mw 8.2 Pisagua earthquake (Fig. 3.3.2a). In the region, no large earthquakes occur since the 1877 great earthquake: i.e. seismic gap. We produced synthetic observation data by using a fault model that was proposed by XXXX et al. as a possible earthquake scenario in the seismic gap. For the inversion analysis, synthetic observation data at two OBP sensors were used in all the tsunami-forecasting simulations, but a combination of offshore stations were changed every simulation. One OBP sensor is fixed at outer-rise region off Iquique. The other is different in each simulation, chosen from the pre-defined offshore observing points inside or near the source (triangle symbols in Fig. 3.3.2a). Here we focus on tsunami forecasting 10 min after an earthquake occurs, because at XX min after the earthquake the resulting tsunamis arrive at the coasts close to the source, and thus the forecast at this elapsed time is very important for tsunami disaster mitigation. Figure 3.3.2b shows spatial distribution of score for costal site Arica (shown by AR in Fig. 3.2.2b), which is located near the tsunami source. High score, shown by light red in Fig. 3.3.2b, is distributed around the epicenter (star symbol) and perpendicular to the trench. This means that deployment of the two OBP sensors, one is inside the high-scored area and the other is off Iquique (i.e. fixed location in our simulations), contributes to improvement of forecasting accuracy for Arica. Fig. 3.3.2 Application of our new methodology to a tsunamigenic earthquake in the northern Chile. (a) Coseismic seafloor deformation due to fault motion. Star indicates an epicenter. Triangles show observing points at offshore (gray) and coast (red). (b) Distribution of forecasting score obtained by our optimization methodology for coastal site Arica (AR). Color variation indicates forecasting score. 3.3.5 Discussion and conclusion We proposed a methodology to obtain spatial distribution of tsunami-forecasting indicators, i.e. scores, for optimal placement of offshore tsunami observatories. The results of its test application show its effectiveness and we can know preferred locations of offshore tsunami sensors for improvement of tsunami forecasting (Fig. 3.2.2b). In future, more applications of our optimization methodology to other tsunamigenic earthquakes are important, because disastrous earthquakes often occur everywhere along the spatially long seismogenic zone off Chile. Also, investigation assuming various types of offshore sensors may be important (Fig. 3.3.3), because measurement physical values are different with sensor type. For example, when OBP sensors are located inside a tsunami source area, it is difficult to find tsunami signal in the OBP records soon after the tsunami occurs. This is because the measured OBP records are affected not only due to sea-surface movement (i.e. tsunami), but also due to permanent seafloor deformation, and thus these two effects cancel out the OBP fluctuation immediately after an earthquake [e.g., Tsushima et al., 2012; Baba et al., 2013]. On the other hand, buoy-type sensors such as GPS buoy [e.g., Kato et al., 2005] is not affected by the permanent seafloor deformation on the records, and thus tsunami signal will appear on the record soon after the earthquake occurrence. Thus, the score distributions based on OBP sensors and those based on GPS buoys may be different, even though the deployment locations are same. These detailed investigations based on our new methodology will contributes to determining placement of offshore tsunami sensors, resulting in great improvement of tsunami early warning in Chile. Fig. 3.3.3 Comparison of tsunami-forecasting results between different types of sensors (OBP sensors and GPS buoys in the same locations). Acknowledgements Bathymetric data off Chile and information of Chilean forecasting points were provided by SHOA. Tsunami Green’s functions applied in tsunami forecasting were calculated by using JAGURS [Baba et al., 2014]. Figures were prepared using Generic Mapping Tools (GMT) [Wessel and Smith, 1998]. REFERENCES Baba, T., K. Hirata, and Y. Kaneda (2004), Tsunami magnitudes determined from ocean bottom pressure gauge data around Japan, Geophys. Res. Lett., 31, L08303, doi:10.1029/2003GL019397. Baba, T., N. Takahashi, and Y. Kaneda (2013), Near-field tsunami amplification factors in the Kii Peninsula, Japan for Dense Oceanfloor Network for Earthquakes and Tsunamis (DONET), Mar. Geophys. Res., doi:10.1007/s11001-013-9189-1. Baba, T., N. Takahashi, Y. Kaneda, Y. Inazawa, and M. Kikkojin (2014), Tsunami inundation modeling of the 2011 Tohoku earthquake using three-dimensional building data for Sendai, Miyagi Prefecture, Japan, in Y. A. Kontar, V. S.-Fandi˜no, and T. Takahashi (Eds.) Tsunami Events and Lessons Learned, Environmental and Societal Significance, Advances in Natural and Technological Hazards Research, 35, 89-98, doi:10.1007/978-94-007-7269-43, Springer, Amsterdam. Hayashi, Y. (2010), Empirical relationship of tsunami height between offshore and coastal stations, Earth Planets Space, 62, 269-275, doi:10.5047/eps2009.11.006. Kanamori, H. (1972), Mechanism of tsunami earthquakes, Phys. Earth Planet. Inter., 6, 346–359, doi:10.1016/0031-9201(72)90058-1. Kato, T., Y. Terada, K. Ito, R. Hattori, T. Abe, T. Miyake, S. Koshimura, and T. Nagai (2005), Tsunami due to the 2004 September 5th off the Kii peninsula earthquake, Japan, recorded by a new GPS buoy, Earth Planets Space, 57, 297–301. Takagawa, T. and T. Tomita (2012a), Effects of rupture processes in an inverse snalysis on the tsunami source of the 2011 Off the Pacific coast of Tohoku Earthquake, Proc. of the Twenty-second (2012) Int. Offshore and Polar Engineering Conf., Rhodes, Greece, June 17-22. Takagawa, T. and T. Tomita (2012b), Tsunami Source Inversion with Time Evolution and Real-time Estimation of Permanent Deformation at Observation Points, J. Jpn. Soc. Civil Eng., Ser. B2 (Coastal Engineering), 68(2), I_311-I_315, doi:10.2208/kaigan.68.I_311 (in Japanese with English abstract). Takagawa, T. and T. Tomita (2014), Simultaneous inference of credible interval of inverted tsunami source and observation error by a hierarchical Bayes model, J. Jpn. Soc. Civil Eng., Ser. B2 (Coastal Engineering), 70(2), I_196-I_200, doi:10.2208/kaigan.70.I_196 (in Japanese with English abstract). Tsushima, H., R. Hino, H. Fujimoto, Y. Tanioka, and F. Imamura (2009), Near-field tsunami forecasting from cabled ocean bottom pressure data, J. Geophys. Res., 114, B06309, doi:10.1029/2008JB005988. Tsushima, H., R. Hino.Y. Tanioka, F. Imamura, and H. Fujimoto (2012a), Tsunami waveform inversion incorporating permanent seafloor deformation and its application to tsunami forecasting, J. Geophys. Res., 117, B03311, doi:10.1029/2011JB008877. Tsushima, H., K. Hirata, Y. Hayashi, K. Maeda, and T. Ozaki (2012b), Effect of offshore tsunami station array configuration on accuracy of near-field tsunami forecast, J. Jpn. Soc. Civil Eng., Ser. B2 (Coastal Engineering), 68(2), I_211-I_215, doi:10.2208/kaigan.68.I_211, (in Japanese with English abstract). Tsushima, H. and Y. Ohta (2014), Review on near-field tsunami forecasting from offshore tsunami data and onshore GNSS data for tsunami early warning, Journal of Disaster Research, 9(3), 339-357. Wessel, P., and W. H. F. Smith (1998), New, improved version of generic mapping tools released, Eos Trans. AGU, 79(47), 579, doi:10.1029/98EO00426. 3.4 Real-Time Inundation Estimation Using GPU Computing Tomohiro TAKAGAWA (PARI) Rapid estimation methods of tsunami sources are discussed in the previous section. If we simulate tsunami propagation and inundation from the estimated source rapidly enough, the simulation result would be used as a precise forecast of tsunami inundation. In this section, GPU computing is introduced as one of the potential solutions of a rapid tsunami simulation. GPU is a short expression of Graphic Processing Unit. It is originally developed for computer graphics, coming to be used for numerical simulations due to the high performance of parallel computing. In the latest supercomputer ranking of TOP500 lists June 2015, the No.2 system, Titan, and the No. 6 system, Piz Daint, use NVIDIA GPUs to accelerate computation (TOP500.Org, 2015). Here, the performance of a GPU accelerated tsunami simulation (GTS) code made by the author will be shown as a reference of the GPU performance for tsunami simulation. The GTS model calculates the tsunami propagation and inundation on the basis of non-linear long-wave equations (Goto et al., 1997). Comparing with a legacy single thread CPU model, the GPU code accelerated more than 700 times (Takagawa et al., 2013). A model simulation was conducted by using GTS. A target site is Iquique (Fig. 1). The computational domain is nested by grids with the different intervals of 10, 30, 90, 270, 810m. Two-way interaction between the grids is considered. We performed three calculations with different finest grid intervals of 10, 30, 90 m for comparison of the accuracy and computational efficiency. The number of grid cells and time steps are shown in Table. 1. The time step is selected to satisfy the CFL condition. A potential tsunami source due to the earthquake off of Iquique proposed by Yagi (xxxx) is assumed (Fig. 1). Calculation result and short discussion for real-time use will be described here. This figure will be replaced in the case of Iquique Fig. 1. Nested grids of the computational domain and the assumed tsunami source Table 1. Computational conditions and the results about calculation efficiency and accuracy Grid interval Number of Maximum Time step (s) Calculation Inundation (m) grid cells depth (m) efficiency* Area (m2) 810 XXX XXX N/A N/A 270 XXX XXX N/A N/A 90 XXX XXX XXX XXX XXX 30 XXX XXX XXX XXX XXX 10 XXX XXX XXX XXX XXX * Calculation efficiency indicates the ratio of computational time to real time. This figure will be replaced in the case of Iquique Fig. 2 This figure will be replaced in the case of Iquique Fig. 3 REFERENCES Goto, C., Y. Ogawa, N. Shuto, and F. Imamura (1997) IUGG/IOC time project: Numerical method of tsunami simulation with the leap-frog scheme, Manuals and guides, 35, UNESCO, Intergovernmental Oceanographic Commission, Paris. Takagawa, T., T. Tomita, D. Tatsumi (2013) Development and validation of real-time tsunami hazard mapping system, American Geophysical Union, Fall Meeting 2013, abstract #NH41B-1716. TOP500.Org. (2015) Top 500 Supercomputer Sites. Available on the WWW, August 2015. http://www.top500.org/. Yagi Y. (XXXX)….. Chapter 4. Proposal on Communication Protocol between Organizations Related to Tsunami Warning 4.1 Methods to Determine Thresholds to Issue Precaution Evacuation based on Modified Mercalli Intensity Yutaka HAYASHI, Akio KATSUMATA, Kazuki MIYAOKA, Hiroaki TSUSHIMA (MRI, JMA), Patricio CATALAN, Jose BAQUEDANO (UTFSM), ZELAYA Cecilia (SHOA), Victor ORELLANA (ONEMI), and Toshitaka BABA (Tokushima Univ.) 4.1.1 Background Three national organizations in Chile cooperate for tsunami early warning operations. Centro Sismológico Nacional, Universidad de Chile (CSN) operates real-time seismic analysis, Servicio Hidrográfico y Oceanográfico de la Armada (SHOA) evaluates the necessity of tsunami alerts or alarms by using CSN data, and Oficina Nacional de Emergencia del Ministerio del Interior y Seguridad Pública (ONEMI) is the only responsible in disseminating warnings and prompting residents to evacuate directly (Fig. 4.1.1). For example, after the Pisagua earthquake on April 1, 2014 (moment magnitude (Mw) 8.2 by CSN; Fig. 4.1.2a), a tsunami warning was issued for all coastlines of mainland Chile. Another example is the case of the earthquake on July 13, 2014 (at 4:56 p.m. local time; local magnitude (M) 5.6 by CSN; Fig. 4.1.2b); based on Modified Mercalli Intensity (MMI) report from coastal cities, ONEMI issued precautionary evacuation of the coastal edges of Iquique, in which high MMI was reported (Fig. 4.1.2). However, the evacuation was soon canceled. The false alarm of the first example for far areas from the seismic source is inevitable because tsunami forecast areas have not been defined in Chile yet; and the second example exposed an inadequate MMI-based procedure. Seismic intensity is the rapidest information obtainable in real-time and corresponding to the tsunami potential. ONEMI's real-time use of seismic intensity for decision of tsunami potential is considered to be a strong point of Chilean tsunami warning operation from the viewpoint of rapidness (Fig. 4.1.3). However, as the above-mentioned examples show, there are difficulties due to the necessity of consistency to the tsunami warning methodology based on magnitude and focus location. This section shows that deriving an empirical relationship among MMI, epicentral distance, and seismic moment will enable us to optimize the standard of decision of MMI-based precautionary evacuation, so that consistency between Tsunami Forecast Chart (TFC)-based warnings and MMI-based evacuation is assured. TFC is a tool to decide tsunami warning rank for each tsunami forecast area by using epicentral distance and seismic magnitude. Possible utilization of TFC for Chilean tsunami warning system with tsunami forecast area is proposed in Section 2.3. 4.1.2 Methodology The recipe to derive the adequate standards for improved determination of the area to issue precautionary evacuation is as follows [Hayashi et al., 2015]. (1) First, moment magnitude (Mw) data from the earthquake catalog by CSN, past tsunami observation data at the tide gauges by SHOA are collected; some tsunami numerical calculation results should be added if observation data is not sufficient. These data can derive TFC for Chilean use as proposed in Section 2.3. (2) Then, previous empirical equations between MMI, and peak ground acceleration or velocity (PGA or PGV) are validated by using MMI data collected by ONEMI, and PGV or PGA data by CSN's observatory and others. Equations in Wald et al. (1999b) which are used for rapid estimation of instrumental intensity in USA [Wald et al., 1999a] are examples of equations to be tested. (3) Next, previous empirical equations between a distance (Δ) from a seismic source to an observation point, and PGA or PGV are validated or modified by using strong motion data and focus location collected data by CSN for several large earthquakes. The empirical equation considering a fault size (e.g. Si and Midorikawa (1999, 2000)) should be selected for this purpose, because a fault size of a large earthquake estimated by scaling law (e.g. Murotani et al., 2013) is too large to be neglected. (4) After validation process of (2) and (3), the integrated empirical equation among MMI, Moment magnitude (Mw), and distance (Δ) can be derived. (5) Therefore, the lines indicating the condition where specific intensities, e.g. MMI=VII, VIII, and IX, should be observed can be superscribed on the TFC by using the equations derived in (4). (6) Finally, by comparing the lines in TFC, the minimum MMI and area for issuing precaution evacuation can be defined, so that the condition of MMI-based precautionary evacuation are expected to be most equivalent to TFC-based tsunami warning. For example, the condition for issuing precautionary evacuation can be defined as "within R km from any coastal city in which S or higher in MMI is observed", where R and S are parameters to be optimized. 4.1.3 Optimizing Parameters Among six steps indicated in 4.1.2, tentative results for validation steps (2) and (3) are indicated in Fig. 4.1.4. and Fig. 4.1.5, respectively. Other steps are under research now. 4.1.4 Conclusions This section indicates that optimization of parameters of the MMI-based method for consistency to the TFC method is promising approach to improve the reliability of early tsunami warnings in Chile. In addition, as described in Section 2.4, intensity-meter can observe objective instrument intensity together with estimated magnitude from duration of strong instrumental intensity. Combination use of objective intensity data might improve the reliability of MMI-based precaution evacuation. REFERENCES Hayashi, Y., A. Katsumata, K. Miyaoka, H. Tsushima, P. Catalan, J. Baquedano, C. Zelaya, V. Orellana, and T. Baba (2015): A method to ensure consistency between tsunami forecast chart-based warnings and instrumental Mercalli scale intensity-based evacuation. Abstract of 26th General Assembly of the International Union of Geodesy and Geophysics, JP05P-021. Murotani, S., K. Satake, and Y. Fujii (2013): Scaling relations of seismic moment, rupture area, average slip, and asperity size for M~9 subduction-zone earthquakes. Geophys. Res. Lett., Vol. 40, 5070-5074. doi:10.1002/grl.50976. Si, H., and S. Midorikawa (1999): Attenuation relations for peak ground acceleration and velocity considering effects of fault type and site condition, J. Struct. Construct. Eng., no. 523, 63-70 (in Japanese). Si, H., and S. Midorikawa (2000): New attenuation relations for peak ground acceleration and velocity considering effects of fault type and site condition. Proceedings of 12th World Conference on Earthquake Engineering, CD-ROM. Wald, D. J., V. Quitoriano, T. H. Heaton, H. Kanamori, C.W. Scrivner, and C. B. Worden (1999a) : TriNet "Shake Maps" : Rapid generation of instrumental ground motion and intensity maps for earthquakes in southern California, Earthquake Spectra, Vol. 15, 537-555. Wald, D. J., V. Quitoriano, T. H. Heaton, H. Kanamori (1999b): Relationships between peak ground acceleration, peak ground velocity, and Modified Mercalli Intensity in California. Earthquake Spectra, Vol. 15, 557-564. Fig. 4.1.1. Roles of Three National Organizations in Chile Cooperating for Tsunami Early Warning Operations, and Comparison to Japan. Fig. 4.1.2. Distribution of Modified Mercalli Intensity for Two Earthquakes with Issuing Tsunami Evacuation Operations. Mw, hypocenter, and origin times are analyzed by CSN, MMIs are collected by ONEMI. Fig. 4.1.3. Summary of Bases and Timings of Tsunami Warning in Chile and Japan. Fig. 4.1.4. Comparison of Attenuation Relation between Fault Distance and Peak Ground Velocity [Si and Midorikawa, 2000], and observed PGV data by seismometer during the 2010 Maule earthquake. (by J. Baquedano) Fig. 4.1.5. Comparison of Empirical Relation between PGV and Modified Mercalli Intensity [Wald et al., 1999b], and Observed MMI and PGV data from earthquakes in Chile. (by J. Baquedano)