2009 AGUFall Meeting
San Francisco, California Dec. 14-18, 2009
Real-time Estimation of Fault Rupture Extent for Recent Large Earthquakes
Yamada, M. and Mori, J. J.
Current earthquake early warning systems assume point source models for the rupture. However, for large earthquakes, the fault rupture length can be of the order of tens to hundreds of kilometers, and the prediction of ground motion at a site requires the approximated knowledge of the rupture geometry. Early warning information based on a point source model may underestimate the ground motion at a site, if a station is close to the fault but distant from the epicenter. We developed an empirical function to classify seismic records into near-source (NS) or far-source (FS) records based on the past strong motion records (Yamada et al., 2007). Here, we defined the near-source region as an area with a fault rupture distance less than 10km. If we have ground motion records at a station, the probability that the station is located in the near-source region is; P = 1/(1+exp(-f)) f = 6.046log10(Za) + 7.885log10(Hv) - 27.091 where Za and Hv denote the peak values of the vertical acceleration and horizontal velocity, respectively. Each observation provides the probability that the station is located in near-source region, so the resolution of the proposed method depends on the station density. The information of the fault rupture location is a group of points where the stations are located. However, for practical purposes, the 2-dimensional configuration of the fault is required to compute the ground motion at a site. In this study, we extend the methodology of NS/FS classification to characterize 2-dimensional fault geometries and apply them to strong motion data observed in recent large earthquakes. We apply a cosine-shaped smoothing function to the probability distribution of near-source stations, and convert the point fault location to 2-dimensional fault information. The estimated rupture geometry for the 2007 Niigata-ken Chuetsu-oki earthquake 10 seconds after the origin time is shown in Figure 1. Furthermore, we illustrate our method with strong motion data of the 2007 Noto-hanto earthquake, 2008 Iwate-Miyagi earthquake, and 2008 Wenchuan earthquake.
The Slapdown Phase in the Strong Motion Record for the 2008 Iwate-Miyagi Nairiku Earthquake
Mori, J. J., Yamada, M. , Heaton, T. H.
The 2008 Iwate-Miyagi Nairiku earthquake (Mw 6.9, Mjma 7.2) produced strong shaking throughout northern Honshu, Japan with severe damage of buildings and extensive landslides. The shallow event occurred in southwestern Iwate prefecture (39.03N, 140.88E, depth 8 km) on June 13, 2008 at 23:43:45 GMT (JMA, 2008). This earthquake produced relatively high-frequency ground motions, which resulted in large values of PGA (peak ground acceleration). Station IWTH25 of KiK-net, located 3 km southwest of the epicenter, produced one of the largest strong-motion values of PGA (4278 cm/s/s for the vector sum of the three components) ever recorded. The surface acceleration record at the station IWTH25 shows an asymmetric amplification in the vertical components (Aoi et al., 2008). The upward vertical acceleration is much larger than the downward direction, although in the borehole record at a depth of 260 m at the same site, the upward and downward accelerations have symmetric amplitudes. On the other hand, the horizontal components do not show this asymmetric effect. This difference between the surface and borehole recordings for the vertical component implies a strong non-linear amplification. In this presentation, we will analyze these records and provide an explanation for the asymmetric amplification in the vertical acceleration. We interpret the large upward spikes in acceleration as slapdown phases, which are also typically observed in near-field recordings of nuclear explosion tests. The large upward acceleration is produced when a near-surface layer separates from the sublayer then returns, striking the separation surface. This effect is seen in a number of strong-motion records that have larger upward than downward accelerations. If we assume the near-surface layer returns to the original level, the separation gap is roughly 1 to 12 mm.
Preliminary Results of the Time-Dependent Earthquake Forecast Model MARFSTA Applied to Mainland Japan
Smyth, C. W. and Mori, J. J.
The Gutenberg-Richter frequency magnitude distribution (Gutenberg and Richter, 1944) plays a major role in earthquake forecasting and hazards analysis. The parameters of this distribution are often assumed to be stationary. As earthquake forecasts are specified over a long period of time, short term irregularities in the parameter values are not of interest. However, we show that over a particular study region these parameters are in fact non-stationary, and that models with temporally variant parameter values describe the data significantly better than models assuming stationarity (Smyth and Mori, 2009). We attempt to model these temporal fluctuations in the Gutenberg-Richter distribution with an autoregressive technique to forecast the short term rate of earthquakes within a region of Japan. Over time, our model can be trained to detect increasing or decreasing rates based over a long history. The predicted forecast rate and frequency magnitude distribution are overlaid on a density map of the area obtained using a multivariate mixture model. The mixture model clusters historical earthquakes with no restraints placed on the size or shape of clusters. Overlaying the two models gives a spatially and temporally variant forecast of seismicity. We repeat the procedure for many regions to obtain a short term forecast for the Japanese mainland. The forecast model differs from currently proposed models by its assumption of non-stationarity and its density estimation. We also incorporate a further time-dependency component in our model by assuming that as time passes since the last large earthquake the probability of another larger earthquake increases. This requires the knowledge of the repeat times of earthquakes along a fault in order to estimate the probability distribution of repeat times. We cannot estimate the repeat times empirically owing to the long history required, so here we use a simulation approach based on the available historical data. The overall product is an earthquake forecast algorithm, MARFSTA, ready for real time testing. We present preliminary results of the model applied to mainland Japan. These results show that recent large earthquakes in Japan fall into regions of high probability indicated by our forecast.
Is There a Relation Between b-values and Asperities?
Yui, Y. and Mori, J. J.
b-value is an important parameter, representing a statistical characteristic of the size distribution of earthquakes. Low b-values are sometimes reported to be associated with areas of large slip (asperities) and might be used to identify where the slip occurs prior to an earthquake. The 1999 Chi-chi, Taiwan earthquake is one of the best examples to study this issue, because there is a clear region of large slip which can be compared to the seismicity of the past 30 years. In this study, we look at the spatial relation between the b-value before and after the Taiwan Chi-chi earthquake using the seismicity data recorded in Taiwan from January 1973 to August 2003, a total number of 278,102 earthquakes. For estimating b-values, we determine the minimum magnitude and use the maximum likelihood method. We divide the aftershock region of the earthquake into smaller areas to compare the local b-values with the slip during the mainshock. We test several size regions from 230 to 3400 km2. We also look at the effect of depth by using grouping events using only epicentral locations and by grouping events by hypocenters that are close to the dipping fault plane of the mainshock. The results show that there are generally high b-values (1.37 to 1.80) in the northern part of the aftershock area where the slip was large during the earthquake. The b-value near the epicenter appears to be significantly lower (0.52 to 1.18). These results show that the area of the asperity had relatively high b-values during 10 to 20 years prior to the mainshock. Also, the spatial distribution of b-value is similar between the seismicity prior to the mainshock and the aftershocks.
Seismic Travel Time Tomography around the Yamasaki Fault System of Southwest Japan
Nugraha, A. D., Ohmi, S. , Mori, J. J., Shibutani, T.
We carried out a high-resolution tomographic inversion to determine crustal structures around the Yamasaki fault system. The Yamasaki fault system is one of the seismically most active onshore faults in southwestern Japan, although large earthquakes around the fault system have not been occurred for about 1100 years. We used a combination of data from a dense seismic network operated by DPRI Kyoto University and the Japan Meteorological Agency catalog. Totally, 16,800 events with 174,271 P and 154,406 S arrival times from 91 stations were used for the tomographic inversion in the study area. We set grid nodes with horizontal spacing of 10 x 10 km2 and vertical intervals of 2- 5 km. We determined three-dimensional Vp and Vp/Vs simultaneously with hypocentral adjustments. We show our results for the Vp, Vs, Vp/Vs ratio, and Poissonfs ratio structures of the crustal heterogeneity along the Yamasaki fault system. The low velocities and high Poissonfs ratio anomalies are imaged in the lower crust at depths of 15 - 25 km along the central portion of the northwestern part of the Yamasaki fault. These anomalies may be associated with fluid-rich regions.
2009 Seismological Society of Japan Fall Meeting
Kyoto, October 21-23, 2009
Subsurface Velocity Changes during Strong Shaking as Seen from Seismic Interferometry
Yamada, M., J. Mori, S. Ohmi
Short History of the Rebeur-Paschwitz type Tiltmeters Used in Kamigamo Observatory, Kyoto
Takemoto, S., J. Mori, L. Rivera, J. Frechet
Modeling Temporal Variations of Seismicity Parameters to Forecast Earthquake Rates in Japan
Smyth, C. and J. Mori
Finding the Rebeur-Paschwitz tiltmeter at Kamigamo Observatory
Mori, J., L. Rivera, J. Frechet, S. Takemoto
Crustal Structure around the Yamasaki Fault System in Southwestern Japan as Recorded from High-Resolution Travel Time Tomography
Nugraha, A.D., S. Ohmi, J. Mori, T. Shibutani
Improvement of the 3D Seismic Velocity Structure in Southwestern Japan Using Pronounced sP Arrivals Observed for Intra-Slab Earthquakes
Hayashida, T., F. Tajima, J. Mori
Is there a Relation between b-value, Earthquake Generation Processes and Heterogeneous Structure
Yui, Yuichi and J. Mori
Slip Distribution of Two Large Solomon Islands Earthquakes in 1975 Estimated Using WWSSN Pdif Waveforms
Park, S.-C. and J. Mori
Joint Inversion using Teleseismic Waveform Data and inSAR Data for the 2009 Papua Indonesia Earthquakes (Mw7.6 and Mw7.3)
Norimatsu, K., J. Mori, M. Hashimoto
Japan Geoscience Union Meeting 2009
Makuhari, Chiba, May 16-21, 2009
Analysis of the 2009 Papua Indonesia Earthquakes (Mw 7.6 and 7.3)
Norimatsu, K. and J. Mori
The Mw 7.6 earthquake of 3 January 2009 occurred at 19:43 (UTC) as thrust faulting on a plate boundary along the northwest coast of the New Guinea Island. and the Mw 7.3 earthquake occurred at 22:33 (UTC) slightly to the east with almost same mechanisms. This is a region of complex tectonics where with convergence between the Pacific and Australian plates. The Pacific plate is likely being subducted southward underneath New Guinea on the Australian plate. Assuming southwestward dipping fault planes, these two events are consistent with this subduction process. The 2009 Papua earthquakes represent a doublet sequence. Doublet earthquakes are characterised by pairs of events within a short amount of time occurring on nearby portions of the same fault or no separate faults. It is important to understand the relationship between the earthquakes by estimating the slip distributions, distributions of the aftershocks and induced stress around the hypocenter area of mainshocks. We calculated the slip distributions for the Mw 7.6 and Mw 7.3 earthquakes using the teleseismic waveform modeling. From our results, the depth of Mw 7.6 earthquake is approximately 15 km shallower than the reported depth from USGS (30 km), and the slip area is about 135 km along strike and 60 km along dip. The maximum slip is located west of the epicenter near the location of the largest aftershock, before the occurrence of the Mw 7.3 earthquake. The Mw 7.3 earthquake was apparently triggered by the Mw 7.6 earthquake. We study the details of the source process using seismic and geodetic data to estimate the relation between Mw 7.6 and Mw 7.3 earthquake.
Prediction of the Earthquakes in the Wenchuan Aftershock Sequence
Smyth, C. and J. Mori
The Slapdown Phase in the Strong Motion Record for the Iwate-Miyagi Nairiku Earthquake
Yamada, M. and J. Mori
The 2008 Iwate-Miyagi Nairiku earthquake (Mw 6.9, Mjma 7.2) produced strong shaking throughout northern Honshu, Japan with severe damage of buildings and extensive landslides. The shallow event occurred in southwestern Iwate prefecture (39.03N, 140.88E, depth 8 km) on June 13, 2008 at 23:43:45 GMT (JMA, 2008). This earthquake produced relatively high-frequency ground motions, which resulted in large values of PGA (peak ground acceleration). Station IWTH25 of KiK-net, located 3 km southwest of the epicenter, produced one of the largest strong-motion values of PGA (4278 cm/s2 for the vector sum of the three components) ever recorded. The surface acceleration record at the station IWTH25 shows an asymmetric amplification in the vertical components (Aoi et al., 2008).The upward vertical acceleration is much larger than the downward direction, although in the borehole record at a depth of 260 m at the same site, the upward and downward accelerations have symmetric amplitudes. On the other hand, the horizontal components do not show this asymmetric effect. This difference between the surface and borehole recordings for the vertical component implies a strong non-linear amplification. In this presentation, we will analyze these records and provide an explanation for the asymmetric amplification in the vertical acceleration. We interpret the large upward spikes in acceleration as slapdown phases, which are also typically observed in near-field recordings of nuclear explosion tests. The large upward acceleration is produced when a near-surface layer separates from the sublayer then returns, striking the separation surface. This effect is seen in a number of strong-motion records that have larger upward than downward accelerations. If we assume the near-surface layer returns to the original level, the separation gap is roughly 1 to 12 mm.
Seismic Attenuation and Thermal Structures in Western Shikoku and Bungo channel: Relationship to Low-frequency Earthquakes
Nugraha, A.D., J. Mori, S. Ohmi,
We conducted 3-D attenuation tomography using our previous 3-D velocity structure in western Shikoku and Bungo channel region. We then estimated thermal structure by applying parameters about relation of temperature to P-wave attenuation data in order to investigate source region of low-frequency earthquakes (LFEs). We analyzed clear waveform for 973 small earthquakes and 128 Hi-net stations to calculate t₯Λ attenuation operators through spectral fitting procedure. We have successfully obtained 19,000 P-wave t* values and 18,500 S-wave t* values to invert for 3-D frequency independent attenuation tomography. We used this P-wave attenuation to estimate thermal structure by following high temperature background method. The obtained structures show the subducting Philippine Sea (PHS) slab is clearly imaged as low attenuation (Qp and Qs of 400-800) and low temperature features (660-680 C), respectively. We look at relationship between obtained structures and LFEs. The LFEs occur in a region of high attenuation (Qp and Qs of 200-250) and moderate high temperature (680-700 C), respectively.
Numerical Simulation of Remarkable Later Phases for Events Occurred Within the Subducting Philippine Sea Slab
Hayashida, T., F. Tajima, J. Mori,
Later phase arrivals help us to construct more realistic 2D/3D seismic velocity models. Several significant later phases are recorded at many stations in southwest Japan for events occurring within the subducting Philippine Sea slab (e.g. Okada et al, 1990; Ohkura, 2000; Miyoshi and Ishibashi, 2007). A distinct later phase, has been detected in seismograms between P- and Swave arrivals at epicentral distances of about 70 km or more for aftershocks of the 2001 Geiyo earthquake. Characteristics of the phase are as follows;
(1) The dominant period is about 1 Hz (nearly same or slightly slower than that of the P wave).
(2) The phase is predominantly observed on the radial and vertical components of velocity seismograms.
(3) The phase is clearly observed at borehole stations that are distributed of the stations covers a wide range.
(4) The apparent velocity is almost 6.1 km/s, corresponding to the P-wave velocity in the uppermost crust.
(5) The amplitude is slightly smaller or larger than that of the P wave.
(6) The amplitude on the radial component is nearly the same or slightly larger than that on the vertical component.
We attempt to explain the characteristics of this phase using the e3d finite difference code by Larsen and Shultz (1995). The constructed model covers a region that extends 200 km in the horizontal and 100 km in depth, with a grid interval of 0.2 km. Parameters of each layer are based on the models of Kakehi et al. (2004) and Iwata et al. (2008), and the depths of the Moho and Conrad discontinuities are inferred from the receiver function images of Shiomi et al. (2004) and Ueno et al. (2008). The simulated waveforms can explain characteristics of the later phase well and the snapshots of seismic wave propagation indicate that this phase corresponds to a depth phase of sP, which is a P wave reflected from a direct S wave at the Earthfs surface. The sP phase is clearly seen in the simulated waveforms at epicentral distances of about 70 km or more. This is in good agreement with the features of the observed data. We also explain the travel times of S conversions from the Moho and Conrad P refrections which are clearly observed in recordings of the several stations by adjusting the depths of the Moho and Conrad discontinuities.
Assessing Temporal Variations in the Gutenberg-Richter
Distribution for a Short-term Forecast Model
Smyth, C. and J. Mori
2nd China-Japan Science Forum
The 2008 Wenchuan Earthquake and Natural Disaster Mitigation
Beijing, China March 9-10, 2009
Rupture Process of the 2008 Wenchuan, China Earthquake using a Back-Projection Analysis
P waveforms from about 700 Hi-net stations in Japan were used in a back projection analysis of the rupture for the May 12, 2008 Wenchuan, China earthquake (Mw7.9). The stations in Japan are at a distance of about 25 to 30 degrees from the earthquake. The P waveforms have durations of about 100 sec and these data were used to estimate the rupture propagation for the earthquake.
The back-projection method tests a grid of points to determine the best location for the source of seismic radiation for a designated time window of the P wave. The initial arrival of the first time window was assumed to come from the grid point corresponding to the earthquake hypocenter. For each subsequent time window, the data were stacked assuming a source at each grid point. The grid point that corresponds to the stack with the highest amplitude is determined to be the source of the P-wave energy. Relative time shifts for each time series were calculated using the theoretical travel times from the station to the grid point, using the IASPEI91 model. The grid of the 290 tested source locations over a fault length of 400 km was used for each time window. Time windows of 10 to 30 seconds were used and produced generally similar results.
There may be some significant complications in the waveforms from the depth phases for a shallow strike-slip earthquake. Part of this problem may be overcome if large enough time windows are used that include the direct arrivals and depth phases, and the whole wavetrain is used.
The results show the rupture progresses from the epicenter in the southwest toward the northeast for a distance of over 300 km. The rupture velocity appears to be variable with an average speed of about 3 km/second. There may be a relation between the areas of large slip and rupture slip. Areas of large slip appear to have faster rupture propagation speed across the fault plane for the Wenchuan earthquake.
Model Ensembles for Prediction of Wenchuan Aftershock Activity
Smyth, C., J. Mori,