Masumi Yamada

Temporal changes of subsurface velocities during strong shaking as seen from seismic interferometry

The deconvolved waveforms reflect the subsurface velocity structure, and their horizontal and vertical components correspond to Sand P-wave, respectively, traveling from the borehole to the ground surface. The strong motion records with smaller values of peak acceleration do not include significant non-linear effects, so the deconvolved waveforms of the observed accelerations can be well simulated by the program SHAKE91.

Real-time estimation of fault rupture extent using near-source versus far-source classification

To estimate the fault dimension of an earthquake in real time, we present a methodology to classify seismic records into near-source or far-source records. Characteristics of ground motion, such as peak ground acceleration, have a strong correlation with the distance from a fault rupture for large earthquakes. This study analyzes peak ground motions and finds the function that best classifies near-source and far-source records based on these parameters. We perform (1) Fisher’s linear discriminant analysis and two different Bayesian methods to find the coefficients of the linear discriminant function and (2) Bayesian model class selection to find the best combination of the peak ground-motion parameters. Bayesian model class selection shows that the combination of vertical acceleration and horizontal velocity produces the best performance for the classification. The linear discriminant function produced by the three methods classifies near-source and far-source data, and in addition, the Bayesian methods give the probability for a station to be near-source, based on the ground-motion measurements. This discriminant function is useful to estimate the fault rupture dimension in real time, especially for large earthquakes.

Real-Time Estimation of Fault Rupture Extent Using Envelopes of Acceleration

We present a new strategy to estimate the geometry of a rupture on a finite fault in real time for earthquake early warning. We extend the work of Cua and Heaton who developed the virtual seismologist (VS) method (Cua, 2005), which is a Bayesian approach to seismic early warning using envelope attenuation relationships. This article extends the VS method to large earthquakes where fault finiteness is important. We propose a new model to simulate high-frequency motions from earthquakes with large rupture dimension: the envelope of high-frequency ground motion from a large earthquake can be expressed as a root-mean-squared combination of envelope functions from smaller earthquakes. We use simulated envelopes of ground acceleration to estimate the direction and length of a rupture in real time. Using the 1999 Chi-Chi earthquake dataset, we have run simulations with different parameters to discover which parameters best describe the rupture geometry as a function of time. We parameterize the fault geometry with an epicenter, a fault strike, and two along-strike rupture lengths. The simulation results show that the azimuthal angle of the fault line converges to the minimum uniquely, and the estimation agrees with the actual Chi-Chi earthquake fault geometry quite well. The rupture direction can be estimated at 10 s after the event onset, and the final solution is achieved after 20 s. While this methodology seems quite promising for warning systems, it only works well when there is an adequate distribution of near-source stations.

Statistical Features of Short-Period and Long-Period Near-Source Ground Motions

Abstract This study collects recorded ground motions from the near-source region of large earthquakes and considers to what extent this historic record can inform expectations of future ground motions at similar sites. The distribution of observed peak ground acceleration (PGA) is well approximated by the lognormal distribution, and we expect the observed distribution to remain unchanged with the addition of data from future earthquakes. However, the distribution of peak ground displacements (PGD) will likely change after a well-recorded large earthquake. Specifically we expect future observations of PGD greater than those previously recorded. We use seismic scaling relations to motivate the expected distribution of PGD as uniform on the logarithmic scale, or at least fat-tailed. Because PGA does not scale with fault rupture area or slip on the fault, there are no such scaling relations to predict the observed distribution of PGA. The observed records show that there is essentially no correlation between PGD and PGA for near-source ground motions from large events. The large uncertainty in a future value of PGD in the near-source region of a large earthquake exists despite the ability of Earth scientists to accurately model long-period ground motions. In contrast, the relative certainty in a future value of PGA exists despite the inability to model short-period ground motions reliably. The stability of the observed distribution of PGA with respect to new ground-motion records enables us to predict the distribution of future PGA and to calculate the probability of exceeding the largest recorded PGA.

Bayesian Approach for Identification of Multiple Events in an Early Warning System

The 2011 Tohoku earthquake (M_w 9.0) was followed by a large number of aftershocks that resulted in 70 early warning messages in the first month after the mainshock. Of these warnings, a non‐negligible fraction (63%) were false warnings in which the largest expected seismic intensities were overestimated by at least two intensities or larger. These errors can be largely attributed to multiple concurrent aftershocks from distant origins that occur within a short period of time. Based on a Bayesian formulation that considers the possibility of having more than one event present at any given time, we propose a novel likelihood function suitable for classifying multiple concurrent earthquakes, which uses amplitude information. We use a sequential Monte Carlo heuristic whose complexity grows linearly with the number of events. We further provide a particle filter implementation and empirically verify its performance with the aftershock records after the Tohoku earthquake. The initial case studies suggest promising performance of this method in classifying multiple seismic events that occur closely in time.

Trajectory of the August 7, 2010 Biwako fireball determined from seismic recordings

The Biwako fireball on August 7, 2010, produced a strong sonic boom throughout central Japan around 17:00 JST (UTC+9). There were many visual observations and reports of the sound in the Tokai and Kinki regions at that time. We have estimated the trajectory of this fireball and the location of its termination point by analyzing seismograms recorded on a dense local network. The isochrons of the arrival times are close to concentric circles, which suggest that the fireball disappeared due to fragmentation during entry. The fireball trajectory which explains the arrival times of the signal has a relatively high incident angle (55 degrees relative to the horizon) and the source is thought to disappear at a height of 26-km east of Lake Biwa. The azimuthal angle and velocity of the fireball are difficult to determine uniquely from this dataset. We identified an event thought to be due to fragmentation, with a location 3-km ENE and 9-km higher than the termination point. This location is consistent with the trajectory determined from the signal arrival. Based on this trajectory model, the source of the signal spans a horizontal range of 26 to 70 km, and the altitude of the source producing the sonic boom is about 30 to 50 km.

Estimation of Fault Rupture Extent Using Near-Source Records for Earthquake Early Warning

This chapter presents a methodology to estimate fault rupture extent in real time for the earthquake early warning. This approach identifies the fault rupture geometry by classifying stations into near source and far source. Suppose there is a sufficiently dense seismic network, the distribution of the near-source station can be used for identifying the fault geometry. In this chapter, we improved a discriminant function to classify seismic records into near-source or far-source records proposed in the previous work. We added the earthquake dataset obtained after 2007, and updated the discriminant function. Furthermore, we integrate the information on each station and proposed a methodology to display the fault rupture surface from the distribution of near-source stations. The probability that a station is near-source obtained from this optimal discriminant function shows the extent of the near-source area reasonably well, suggesting that the approach provides a good indicator of near-source and far-source stations for real-time analyses. After applying interpolation, we successfully displayed the fault rupture surface from the distribution of near-source stations.

Consecutive vibration characteristics monitoring of high-rise steel building

This paper reported the continuous seismic observation for a super-high-rise steel building in Osaka basin, Japan. We observed 16 earthquake records including the Tohoku earthquake on March 11, 2011. The signal to noise ratio is sufficiently good for all records. We estimated natural period and impulse response from those seismograms, and showed slight frequency shift due to nonlinear effect. We used microtremor records to estimate the natural period and damping of the building. By stacking sufficiently long records, we can extract those building property from microtremor records. Finally, the analysis models of the steel building used in the structure design are demonstrated based on the observed results. stories for the penthouse, and four stories under the ground level (Fig. 1(b)). The structure is steel moment-resisting frame above ground and steelreinforced concrete frame with shear walls under the ground level. The building is supported by the direct foundations with diaphragm walls. The depth of the direct foundation is 18 m from the ground and the diaphragm walls reaches to 35 m from the soil surface. The predominant natural periods of the design model are 2.28 seconds and 2.30 seconds in NS and EW direction, respectively (see Table 2). The designed damping factor is 2%.

Investigating the Distributions of Differences between Mainshock and Foreshock Magnitudes

Previous research produces seemingly contradictory statements about the distribution of differences between mainshock and foreshock magnitudes. Specifically, some authors find that the magnitude difference between the mainshock and the foreshock is equally likely to be large as to be small. However, other authors find that the distribution of the magnitude differences between the foreshock- mainshock pairs is not uniform. We consider foreshock-mainshock pairs within the recent Japanese earthquake catalog and worldwide data to explore the discrepan- cies between these seemingly contradictory studies. The results of the previous studies differ because of the different foreshock-mainshock earthquake pairs that are consid- ered by the two sets of authors. We show that using the definitions employed by either type of study permits the found distributions to be derived analytically and further explains how the results are dependent upon the assumed definition of foreshocks and the data selection.

Repeating Seismic Events Indicate Possible Stick-slip Behavior Before the Rausu Landslide†

A precursory sequence of repeating earthquakes was recorded before the Rausu landslide in Hokkaido, Japan on April 24, 2015. There were two seismic sequences with each consisting of very similar waveforms and leading up to significant landslide movements. The nearly-identical waveform shapes indicate similar source locations and mechanisms, so repeated events originated on a particular small area. This sequence is interpreted as stick-slip movement on a small patch leading up to the larger landslide failure. Our observations show that heterogeneous structure, such as asperities on the slip surface, can play an important role in the initiation of landslides, adding a new aspect to the conventional understanding of mechanisms controlling large mass movements.A precursory sequence of repeating earthquakes was recorded before the Rausu landslide in Hokkaido, Japan on April 24, 2015. There were two seismic sequences with each consisting of very similar waveforms and leading up to significant landslide movements. The nearly-identical waveform shapes indicate similar source locations and mechanisms, so repeated events originated on a particular small area. This sequence is interpreted as stick-slip movement on a small patch leading up to the larger landslide failure. Our observations show that heterogeneous structure, such as asperities on the slip surface, can play an important role in the initiation of landslides, adding a new aspect to the conventional understanding of mechanisms controlling large mass movements.