Progress Report on the ISM Project 2003-2007

 

Statistical Seismology Research Group

Prediction and Knowledge Discovery Research Center

The Institute of Statistical Mathematics

Research Organization of Information and Systems

 

 

Table of Contents

Motivations behind this project  ₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯

2

(1) Hypocenter data  ₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯

2

(2) Other geophysical datasets  ₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯

2

(3) Point-process models  ₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯

2

(4) Earthquake / Aftershock forecasting  ₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯

3

(5) Exploration and modeling of the interface between physical and stochastic processes  ₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯

4

(6) Space-time point-process modeling  ₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯

5

References  ₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯

5

Objectives of the project  ₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯

7

Work Plan 2003-2007  ₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯

8

Project Members  ₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯

8

Research Accomplishments 2003-2005  ₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯

9

Principal works during 2003 - 2005  ₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯

9

(1) Coseismic activation / quiescence triggered by a large earthquake and Coulomb stress changes  ₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯

9

(2) Relative quiescence in aftershock sequences and its mechanism  ₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯

10

(3) Seismicity anomalies preceding large earthquakes and crustal stress changes  ₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯

11

(4) Space-time ETAS modeling  ₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯

13

(5) Modeling the interface between physical and stochastic process  ₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯

15

(6) Simultaneous estimation of b-values and detection rates of earthquakes for the application to aftershock

probability forecasting  ₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯

16

Published papers 2003-2005  ₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯

16

Refereed Journals  ₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯

16

Main Proceedings  ₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯

18

Future plans for the project  ₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯

20

(1) Examination of scenarios for predicting asperity-slip based on the seismicity anomalies  ₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯

20

(2) Effective space-time modeling of seismic activity and detection of seismicity anomalies  ₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯

21

(3) Predictive space-time-magnitude characterization of foreshocks  ₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯

21

(4) Prediction and inversion problem between seismicity changes and stress-changes  ₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯

22

(5) Bayesian Probability assessments for Long-term prediction  ₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯

22

(6) Statistical modeling for more effective use and quality improvements of datasets  ₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯

23

References  ₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯₯

23

 

 

Motivations behind this project

(1)  Hypocenter data

The Hypocenter catalog of the Japan Meteorological Agency (JMA) is one of the most valuable databases for earthquake prediction. This is because it contains records of a large number of earthquakes dating back to 1925. The catalog covers the whole of Japan and its vicinity, although the records are heterogeneous in detection rate in space and time. The capacity for detecting earthquakes has been significantly improved recently following the unification of the seismic observation records among the JMA, universities and national institutes. Nevertheless, the conventional seismicity study has been using primary graphical methods such as cumulative curves and magnitude of earthquakes against time, space-time plot of earthquakes, etc., without complex statistical analysis, which does not convey all of the essential information included in the catalog. Indeed, we believe that an effective analysis of seismic activity can only be made by applying point-process models that are constructed based on physical speculations or hypotheses on seismic processes. For example, the data of earthquake mechanisms such as those in the F-net earthquake catalog is becoming increasingly useful and important in order to discuss stress changes of the crust.

(2)  Other geophysical datasets

There are many precise geophysical records that are useful in discussing the relationship with seismicity such as records of extensometers, tiltmeters, volmetric strainmeters and the GPS. However, these records are usually affected by various noises or ancillary geophysical signals caused by the earthtides, and in particular, by meteorological factors such as barometric pressure, precipitation, temperature and humidity. Therefore, it is important to model the causal relationships and response functions of these effects in order to calibrate the records for genuine quantities of interest. For example, the computer program BAYTAP-G [Ishiguro et al., 1984] used to implement the Bayesian deconvolution procedure for removing tidal effects has been used with much frequency by many researchers in seismology and geodesy in Japan. Additionally, Kitagawa and Matsumoto [1988] modeled the time series of observed water level, which is well decomposed into the four effects of convolutions of present and past values of the barometric pressure, earth-tides, the precipitations and error term. Much works on the similar modeling or nonlinear modeling of time series are required, especially with regard to GPS data.

(3)  point-process models

Point processes. A point-process is a mathematical model of the stochastic occurrence of a series of events. The modeling of point-processes became a powerful tool in the field of applied statistics in the 1980's, for three reasons. Firstly, the concept of the conditional intensity function provided us with an extensive free hand with which to produce models describing the detailed interface between the physical mechanism of occurrences and some stochastic factors. This is the predictive occurrence rate function (roughly, the differential of the occurrence conditional probability) of the present time, the occurrence times of the past events and other relevant time series data such as magnitudes and other available geophysical records. Use of this function also made available a general effective simulation method using the thinning operation [Ogata, 1983]. Secondly, it became possible to write the likelihood function directly in terms of the conditional intensity function. The third reason was due to the availability of practical algorithms for optimizing non-linear functions using a computer, enabling us to obtain the maximum likelihood estimate (MLE), their error estimates, and the likelihood-ratio or the AIC to examine the goodness-of-fit of models. Together with these revolutionary bases, a diagnostic analysis of the model and data became available with the time-transformation using the integral of the conditional intensity function [Ogata, 1983, 1988; Ogata and Shimazaki, 1984; Ogata, 1999]. Benefiting from these, a substantial number of applications became available, including the program packages TIMSAC84 [Akaike et al., 1985] and SASeis [Utsu and Ogata, 1997] in the IASPEI Software Library.   

 

The ETAS model. The epidemic type aftershock sequence (ETAS) model [Ogata, 1986, 1988, 1989] is one such model. It is generally accepted that each earthquake changes the probability of successive earthquakes in a region, the size of which scales with its magnitude, and by an amount that can be estimated using the Gutenberg-Richter magnitude distribution and the Omori-Utsu law for the rate of aftershocks, where aftershocks are allowed to be larger than the mainshock. The ETAS model forms the basis of many current probabilistic earthquake prediction schemes. Inherent in these models is the assumption that the probability of a large earthquake is completely determined by the sum of stresses transferred by prior earthquakes within a considered region. By a set of parameter values the ETAS model is well adapted to various seismicity patterns including the mainshock-aftershock type and swarm type.

 

Diagnostic analysis using the ETAS model. Since a sequence of aftershocks is triggered by complex mechanisms under fractal random media, it is difficult to calculate the transferred stresses within and near the rupture fault. That is, triggering mechanics within an aftershock sequence are too complex for us to calculate the effect of stress changes. Therefore, the statistical empirical laws of aftershocks are useful as a practical representation of the outcome due to the complex interaction of the self- or proximate triggering. On the other hand, the ETAS model is a statistical model constructed based on the empirical laws of aftershocks. Therefore, the model itself hardly describes the mechanism of affecting stress changes behind the seismicity changes. Diagnostic analyses of the model can reveal such new knowledge included in the data. That is to say, the activation and quiescence relative to the model's prediction could suggest exogenous stress-changes in the regions.

(4)  Earthquake / Aftershock forecasting

Primary models for aftershock forecasting. It has been about a decade since aftershock probability forecasting [Reasenberg and Jones, 1989] was implemented to inform the public in California and Japan. This is based on a combination of Gutenberg-Richter's law of magnitude frequency and a Poisson process model with the Omori-Utsu's decaying rate of aftershock occurrences [Omori, 1894; Utsu, 1961] where the characteristic parameters of both laws are adjusted from early observations by the maximum likelihood estimates [Utsu, 1965; Aki, 1965; Ogata, 1983]. It is known that the ETAS model usually fits better than the Omori-Utsu formula [Guo and Ogata, 1997], so the ETAS model will be taken as the primary model through which to compare the performance of forecasts with possible sophisticated physics-based models.

 

Anomalous aftershock activity. Precursory seismic quiescence as a predictor of large earthquakes has attracted much attention amongst seismologists in the last several decades, ever since Inouye [1965] first proposed the concept. This suggests that we need much more research into the relation between the quiescence and subsequent earthquake activity in order to obtain an effective prediction. We should also explain how quiescence can take place in a much wider area than the rupture source [Inouye, 1965; Ogata, 1992]. Studying many aftershock sequences should provide us with a better understanding of the physical mechanism of the precursory utility. For predicting large aftershock, Matsu'ura [1986] noted the utility of the quiescence in aftershock occurrences relative to the Omori-Utsu decaying formula.

   Using the ETAS model and on the basis of the proposed procedure, 259 aftershock sequences of various threshold magnitudes are investigated for the 76 main shocks of M6 class or over that occurred in and around Japan during the last three-quarters of the last century [Ogata, 2001]. Relative quiescence is revealed in ~40% of the aftershock sequences, and almost all of the others are developed normally; relative activation is rarely found. We have seen that the aftershock activity provides useful information for assessing the probability of a following large event (M6 class or over) in the neighbourhood. Namely, if the aftershock activity from the first event becomes relatively quiet compared to the expected normal decay, the occurrence rate of a larger event in the neighbourhood (within a distance of 3o) is a few times higher during the first decade after the main shock than would be the case for normal aftershock activity. However, it is not yet clearly understood the physical mechanism of the causal relationship between the relative quiescence and the forthcoming occurrences of large aftershocks or even large earthquakes of the similar size to the mainshock. The recently accumulated earthquake mechanism data may help better our understanding.

 

Seismicity rate change and Coulomb's failure stress change. We are concerned with the precise prediction of time- and history-dependent occurrence rates of an earthquake sequence, particularly, aftershock sequences, in order to test the hypothesis that abrupt stress-change due to a seismic or an aseismic slip triggers a seismicity-rate-change in the surrounding area. This is because stress changes in a region are very frequently affected by nearby events, which trigger further aftershock clusters. In principle, seismic activity should be enhanced in the zones where an increment of Coulomb's failure stress (CFS) is positive, and also activity should be reduced (seismic quiescence) in the stress-shadow zones. For example, some retrospective case studies have shown the stress shadow [e.g., Harris, 1998; Toda and Stein, 2002] due to large earthquakes to coseismically inhibit the activity in some neighboring seismic regions in California, in addition to many cases of activated seismicity triggered by an large event [e.g., Toda et al., 1998]. A similar phenomena can be expected by the activation and quiescence relative to the predicted seismicity by the ETAS model.

(5)  Exploration and modeling of the interface between physical and stochastic processes

External stress changes and swarms. It is well known that volcanic swarms and other swarms are affected by magma intrusion [Dieterich et al., 2000; Toda et al., 2002] and water migration [Matsu'ura and Karakama, 2005]. According to the Coulomb failure criterion, the variation of both, stress and pore pressure, can result in earthquake rupture. Aftershock sequences characterized by the Omori-Utsu law are often assumed to be the consequence of varying stress, whereas earthquake swarms are supposed to be triggered by fluid intrusions. Statistical models for describing the relationship are desired. Also, there are some papers reporting that occurrence rates of some swarms and aftershocks are correlated to certain components of earth-tide's time series [e.g., Iwata, 2002]. It is desirable to have models examining the causal relationship between mechanisms of earthquakes and stress tensors of stresses due to earth-tide.

 

Inversion. On the other hand, it will be informative to produce an image of the geophysical quantities in time and space, by making an inversion of the parameters of the statistical models of earthquake occurrences. Such examples include b-values of the G-R magnitude frequency [Ogata et al. 1991 and 1993], and p-values of the Omori-Utsu model for the aftershock decay [Mogi, 1967]. More examples should also be considered, making use of models such as the extended space-time ETAS model, making use of a Bayesian framework. The geophysical quantities could include the stress distribution and temperatures of the crust, and the friction coefficients of the fault interface (asperities) etc.

 

Quiescence and activation relative to the ETAS prediction and crustal stress changes. Seismic quiescence and activation have attracted much attention amongst seismologists as the precursors to a large earthquake. Of particular interest is that the stress-changes transferred from a far-field rupture or silent slip can cause seismic changes in a region. On the other hand, since a sequence of aftershocks is triggered by complex mechanisms under heterogeneous fractal media, it is difficult to precisely describe the transfer of stresses both within and near to the field. In other words, triggering mechanics within an aftershock sequence are too complex to calculate the effect of stress changes. Nevertheless, we can use statistical empirical laws as a practical solution to aftershock triggering. That is to say, fitting and extrapolating a suitable statistical model for normal seismic activity in a situation without exogenous stress changes provides us with an alternative method through which to see the seismicity changes explicitly. Thus, our motivation is to show the possibility that the diagnostic analysis based on fitting the ETAS model, and its space-time extension, fitting to regional seismicity can be helpful in detecting small exogenous stress changes. Indeed, these changes are so slight that the geodetic records from the GPS network can barely recognize systematic anomalies in the time series of displacement records.

(6)  Space-time point-process modeling

The relative quiescence before great earthquakes is discussed in wide seismic regions in Ogata [1992] for high threshold magnitudes of more than M5. However, these are too high to discuss intermediate strong earthquakes. The locations of earthquakes from hypocenter catalogs are now accurate enough to discuss the spatial aspect of seismicity, such as the clustering of aftershocks and seismicity gaps. Ogata [1998] considers several possible extensions of the ETAS model to space-time data, based on classical empirical studies of aftershocks, and also on a number of contrasting speculative hypotheses about the physical nature of the space-time clustering. Their goodness-of-fit is compared by the aid of the AIC for two data sets from tectonically distinctive areas in and around Japan. Further practical extensions of the models are suggested for the realistic but complex features such as non-homogeneous background seismicity and occasionally anisotropic clusters that is closely related to the mechanism of the mainshock.

   However, as the data size increases, spatial heterogeneity of the seismic activity becomes conspicuous. For example, shape of the anistropic clustering becomes more complex, differing from place to place. Even cluster sizes are significantly different from one another, especially between those in offshore and inland area [Utsu, 1969; Ogata, 2001].

   Furthermore, according to the restricted trigger model [Ogata, 2001], secondary aftershocks of large cluster sizes are located around the boundary of the main rupture zone of the 1995 Kobe earthquake. To describe seismicity more accurately requires that the several parameters characterising the above space-time ETAS model should depend on the location of the considered region. Thus we need to make use of the objective Bayesian estimation procedure in a similar manner to Ogata et al. [1993] and Ogata and Katsura [1993].

 

References

Akaike, H., Ozaki, T., Ishiguro, M., Ogata, Y., Kitagawa, G., Tamura, Y., Arahata, E., Katsura, K. and Tamura, R. (1985) Time Series and Control Program Package, TIMSAC-84 (, Computer Science Monograph, No. 22/23, The Institute of Statistical Mathematics, Tokyo, Japan.

Aki, K. (1965) Maximum likelihood estimate of b in the formula log N = a-bM and its confidence limits, Bull. Earthq. Res. Inst., Univ. Tokyo, 43, 237-239.

Dieterich, J., Cayol, V. and Okubo, P., The use of earthquake rate changes as a stress meter at Kilauea volcano, Nature 408, 457-460 (2000).

Guo, Z. and Ogata, Y. (1997) Statistical relations between the parameters of aftershocks in time, space and magnitude, Journal Geophysical Research, Vol. 102, No. B2, pp. 2857-2873.

Harris, R. A. (1998) Introduction to special section: Stress triggers, stress shadows, and implications for seismic hazard, J. Geophys. Res., 103, 24,347–24,358.

Inouye, W. (1965) On the seismicity in the epicentral region and its neighborhood before the Niigata earthquake (in Japanese), Kenshin-jiho (Q. J. Seismol.), 29, 139-144.

Ishiguro, M., Akaike, H., Ooe, M. and Nakai, S. (1981). A Bayesian approach to the analysis of earth tide, Proceedings of the 9th International Simposium on Earth Tides, (ed. J.T. Kuo), E. Schweizerbart'sche Verlagsbahhandlung, Stuttgart.

Iwata, T. (2002) Tidal stress/strain and acoustic emission activity at the underground research laboratory, Canada, Geophy. Res. Let..

Kitagawa, G. and Matsumoto, N. (1996). Detection of coseismic changes of underground water level, J. Amer. Stats. Assoc., 91, 521-528.

Matsu'ura R.S. (1986) Precursory quiescence and recovery of aftershock activities before some large aftershocks, Bull. Earthq. Res. Inst., Univ. Tokyo, 61, 1-65.

Matsu'ura R.S. and Karakama I. (1986) A point-process analysis of the Matsushiro Earthquake Swarm sequence: The effect of water on earthquake occurrence, Pure and Applied Geophysics, 162, pp. 1319-1346.

Matsu'ura R.S., Hirata, N. and Urabe, S. (1995) Quasi-real-time watch of the aftershock activity change of Hyogoken-nanbu Earthquake – Prediction of Jan. 23 23h 16m M4.7 aftershock (in Japanese), Report of the Coordinating Committee for Earthquake Prediction, 54, pp. 600-607.

Mogi, K. (1967) Earthquakes and fractures, Tectonophysics, 5, 35-55.

Ogata, Y. (1978) The asymptotic behavior of maximum likelihood estimators for stationary point processes, Annals of the Institute of Statistical Mathematics, Vol. 30, No. 2, A, pp. 243-261.

Ogata, Y. (1981) On Lewis' simulation method for point processes, IEEE Transactions on Information Theory, Vol. IT-27, pp. 23-31.

Ogata, Y. (1983) Estimation of the parameters in the modified Omori formula for aftershock frequencies by the maximum likelihood procedure, Journal of Physics of the Earth, Vol. 31, pp. 115-124.

Ogata, Y. and Shimazaki, K. (1984) Transition from aftershock to normal activity, Bull. Seismo. Soc. Am., 74, 5, pp. 1757-1765.

Ogata, Y. (1986) Statistical models for earthquake occurrences and residual analysis for point processes, Mathematical Seismology (I), 228-281, Institute of Statistical Mathematics, Tokyo.

Ogata, Y. (1988) Statistical models for earthquake occurrences and residual analysis for point processes, Journal of American Statistical Association, Application, Vol. 83, No. 401, pp. 9-27.

Ogata, Y. (1989) Statistical model for standard seismicity and detection of anomalies by residual analysis, Techtonophysics, Vol. 169, pp. 159-174.

Ogata, Y. (1992) Detection of precursory relative quiescence before great earthquakes through a statistical model, J. Geophys. Res., 97, 19845-19871.

Ogata, Y. (1998) Space-time point-process models for earthquake occurrences, Ann. Inst. Statist. Math., 50, 379-402.

Ogata, Y. (1999) Seismicity analyses through point-process modelling: A review, in Seismicity Patterns, Their Statistical Significance and Physical Meaning M. Wyss, K. Shimazaki and A. Ito eds. Birkhauser Verlag, Basel, Pure and Applied Geophysics, 155, pp. 471-507.

Ogata, Y. (2001) Exploratory analysis of earthquake clusters by likelihood-based trigger models, Festscrift Volume for Professor Vere-Jones, J. Applied Probability, 38A, pp. 202-212.

Ogata, Y., Imoto, M. and Katsura, K. (1991) 3-D spatial variation of b-values of magnitude-frequency distribution beneath the Kanto District, Japan, Geophys. J. Int., 104, 135-146.

Ogata, Y. and Katsura, K. (1993) Analysis of temporal and spatial heterogeneity of magnitude frequency distribution inferred from earthquake catalogs, Geophys. J. Int., 113, pp. 727-738.

Ogata, Y. and Zhuang. J. (2001) Statistical examination of anomalies for the precursor to earthquakes, and the multi-element precision formula: hazard rate changes of strong earthquakes (M>=4) around Beijing area based on the ultra-low frequency ground electric observation (1982-1997) (in Japanese), Report of the Coordinating Committee for Earthquake Prediction, Vol. 66, pp. 562-570.

Omori, F. (1894) On the aftershocks of earthquakes, J. Coll. Sci. Imp. Univ. Tokyo 7, 111 -200.

Reasenberg, P.A. and Jones, L.M. (1989) Earthquake hazard after a mainshock in California, Science, 243, 1173-1176.

Toda, S. Stein, R.S., Reasenberg P.A. and Dieterich J.H. and Yoshida, A. (1998) Stress transferred by the Mw=6.9 Kobe, Japan, shock: Effect on aftershocks and future earthquake probabilities, J. Geophys. Res., 103, pp. 24,543-24,565.

Toda, S. and Stein, R.S. (2002) Response of the San Andreas Fault to the 1983 Coalinga-Nuñez Earthquakes: An Application of Interaction-based Probabilities for Parkfield, J. Geophys. Res. 107, 10.1029/2001JB000172.

Toda, S., Stein, R.S. and Sagiya, T., (2002) Evidence from the AD 2000 Izu Islands swarm that stressing rate governs seismicity, Nature, 419, pp. 58-61, 5 September issue.

Utsu, T. (1961), A statistical study on the occurrence of aftershocks, Geophys. Mag., 30, 521605.

Utsu, T. (1965) A method for determinig the value of b in a formula log n =‚ bM showing the magnitude- frequency relation for earthquakes, Geophys Bull Hokkaido Univ., 13, 99-103 (in Japanese).

Utsu,T. (1969).Aftershocks and earthquake statistics(I): some parameters which characterize an aftershock sequence and their interaction, Journal of the Faculty of Science, Hokkaido University, Ser. VII (geophysics), 3, 129-195.

Utsu, T., Ogata, Y. and Matsu'ura, R. S. (1995) The centenary of the Omori formula for a decay law of aftershock activity, J. Phys. Earth, 43, 1-33.

Utsu, T. and Ogata, Y. (1997) Computer program package: Statistical Analysis of point processes for Seismicity (SASeis), invited and accepted for IASPEI Software Library for Personal Computers, the International Association of Seismology and Physics of Earth's Interior, vol. 6, 13-94.

Zhuang, J., Ogata, Y. and Vere-Jones, D. (2002). Stochastic declustering of space-time earthquake occurrences, J. Amer. Statist. Assoc., 97, 369-380.

 

 

Objectives of the project were as follows:

(1)   To develop practical space-time models that are sufficiently close to the real seismic activity. Specifically, the model must elaborate enough to spatially adapt to various different clustering patterns. In order to represent various seismicity patterns, we will adopt a hierarchical space-time point-process model, in which the parameter values are dependent on the location of the earthquakes.

(2)   To detect and evaluate anomalous features of general seismicity and aftershock activity relative to the modeled rate by the ETAS model, to explore its space-time features using statistical diagnostic methods and to make modeling of seismicity anomalies.

(3)   To explore the relation between the seismicity-rate-change in a region and the stress-change in the crust, which varies due to the fault mechanisms of earthquakes.

(4)   To statistically model the relationship between the occurrence time series of focal earthquakes in a region and the time series records of anomaly events, in order to enhance the performance of probability forecasting.

(5)   To model the heterogeneity of various datasets which are affected by some instrumental and geophysical factors, such as various noises, missing events or values, and detection rate changes of earthquakes in time and space.

 

 

Work Plan 2003 – 2007

We will undertake as many of the followings as early as possible:

(1)   We will apply the ETAS model to a number of sequences of earthquakes from various regions over a recent period to examine any significant deviation of the activity from the rate predicted by the model. We will explore matching such an anomaly and a crustal stress-change due to a co-seismic and a pre-seismic slip somewhere in order to examine whether such a seismic anomaly can be a sensitive sensor of stress change.

(2)   We will particularly be concerned with the anomalously quiet aftershock activity preceding a large aftershock.

(3)   We will make use of fault mechanism data of earthquakes to calculate the Coulomb-stress-change estimation transferred from a pre-slip or a slow slip in order to discuss the significance of the causal relation between stress change and seismicity anomalies.

(4)   We will try to make a statistical point process model to explain the causality between geodetic anomalies and seismicity anomalies in the same field.

(5)   We will publish a software package to make programs available such as the estimation and diagnostic procedure of the ETAS model for the seismologists, after testing the stability of such programs.

(6)   We will make a practical model to enable real-time probability forecasting of aftershocks immediately after a large earthquake (say, within 24hours), when the detection rate of aftershocks is extremely low, due to contamination of arriving seismic waves. This is urgent since the majority of large aftershocks are most likely to occur during this period, and hence forecasting during this time is most critical for public in the affected area.

(7)   We will develop methods of diagnostic analysis for space-time seismicity data by introducing a space-time seismicity-ratio of real seismicity to the predicted by the space-time ETAS model, and modeling it in Bayesian framework.

(8)   We will prepare a software package to publish the MLE and Bayesian procedure of the space-time ETAS model, and test the stability of such programs with many datasets.

 

 

Project Members:

Statistical Seismology Research Group, Prediction and Knowledge Discovery Research Center

Yosihiko OGATA, Prof. of the Institute of Statistical Mathematics (ISM)
Shinji TODA, Visiting Prof. of the ISM and Geological Survey of Japan, AIST, 2005 -
Yasuaki MURATA, Geological Survey of Japan, AIST , Visiting Assoc. Prof. of the ISM 2003 - 2004
Jiancang ZHUANG, Research Fellow upon JSPS program, 2001 - 2005
Kazuyoshi NANJO, Research Fellow upon JSPS program, 2003 -

Takaki IWATA, Research Fellow of ISM, 2005 -
Masatsugu WAKAURA, Graduate Student, The Graduate University for Advanced Studies
Ushio TANAKA, Graduate Student, The Graduate University for Advanced Studies

Akiko KUTSUNA, part-time assistant

 

 

Research Accomplishments 2003 – 2005

Principal works during 2003 - 2005

(1)  Coseismic activation / quiescence triggered by a large earthquake and Coulomb stress changes

It is clearly seen that after a large earthquake, many off-fault activities with positive Coulomb stress increments are enhanced, while negative ones (stress shadow) are inhibited, in regions that include many earthquakes of similar fault-mechanisms detected down to small magnitudes. Such various examples are shown in western Japan, which was affected by the two great earthquakes of 1944 and 1946 in the Nankai trough, with other various ones being seen in Hokkaido inland and the southern offshore of Hokkaido which was affected by the 2003 Tokacho-Oki great earthquake. See also the published referred papers [4] , [23], [32], [A8], and  the main  reported manuscript [A15], [A16], [A23] and [A24] for further examples of coseismic triggering. These papers and reports are listed in the end of this section under Research Accomplishments.

 

Seismicity changes in western Japan affected by the great earthquakes in Nankai trough [12, A4, A28].

Significant changes in seismicity (both quiescence and activation) took place during the period of 1944-1946 in some regions in western Japan. They are well explained by the corresponding changes in Coulomb failure-stress caused by the 1944 Tonankai earthquake of M7.9 and the 1946 Nankai earthquake of M8.0, both of which occurred in the Nankai trough. From the seismicity changes, with exception of coseismic changes, we identified precursory anomalies such as the quiescence in several regions before the Tonankai event and also the enhanced activity in southern Kii peninsula before the Nankai event. These quiescence and enhanced activity may be due to precursory aseismic slips in an area on the plate interface down-dip of the Tonankai rupture and slips transferring from the Tonankai to Nankai rupture zone. The records of felt shocks at the Wakayama Observatory during 1900-1995 provide support for the quiescence in the Wakayama District during 1944-1946, as well as a scenario of seismicity cycle until the next great event(s) in the Nankai trough, to occur sometime this century.

 

Seismicity-changes and stress-changes triggered by the 2003 Tokachi earthquake [A21, A33]. Coseismic activation and quiescence are conspicuous in the eastern inland region of Hokkaido after the 2003 great earthquake of M8.0 in the space-time occurrence of microearthquakes detected from 2001 through 2003 (depth25km). The ƒ’CFF for the receiver faults with N75oE right lateral strike-slip at the depth of 10km and the Tokachi-Oki event's slip model takes positive and negative values in the western and eastern part, respectively, which is corresponds well to the regions of activation and quiescence in microseismicity. The exception is that the very active spot in the western region suddenly stopped the activity after the great event, but this is also well explained by the different alignment of the receiver fault from the rest of the western part of the microseismicity.  We found another triggered activation and quiescence in a complicated manner within the 3D volume down to 100km depth beneath the southern inland region of Hokkaido and its southern offshore area. This complex pattern of activation and quiescence is well explained by the receiver faults due to the tectonic structure of this area called the Hidaka collision zone model.

 

Coseismic activation and quiescence relative to the ETAS model [22, A19]. Fitting and extrapolating the ETAS model for normal seismic activity in a situation without exogenous stress changes provides us with an alternative method through which we can detect the relative changes of seismicity sensitively. Thus, the diagnostic analysis based on fitting the ETAS model is helpful in detecting small exogenous stress changes. For example, shallow earthquakes (M >=1.5) in the Tohoku inland region of largest ƒ’CFF values (ranging +5~+50 millibars) due to the 2003 May 26 Miyagi-Ken-Oki earthquake of M7.1 is seen to be activated relative to the predicted occurrence rate by the ETAS model. The foreshock activity during the first event of M5.5 and the mainshock of M6.2 (the both occurred at 26 July 2003) was more active than the predicted rate by the ETAS fitted in the preceding interval. But the aftershock activity the mainshock of M6.2 seems similar to the predicted rate.

(2)  Relative quiescence in aftershock sequences and its mechanism

In order to extract regional stress-changes transferred from the slip of a far-field fault, we have to remove the effect of the complex, proximate triggering mechanics occurring within aftershock clusters. As a practical solution, the ETAS model is fitted to the sequence of events from the region in order to precisely mimic the normal activity there. We are primarily concerned with seismicity-rate-changes (enhancement and reduction) relative to the rate predicted by the ETAS model, and explore matching them to the pattern of Coulomb's stress-changes that occur due to a rupture or a suspected silent slip. We have shown a number of such examples from recent seismic activities in Japan. These lead us to a summarized observation that even a small CFS increment of the order of millibars can trigger such seismicity-rate-change, which is also supported by the seismicity-rate-equation of Dieterich. Thus, we expect that the anomalous seismic activity relative to the ETAS rates is sensitive enough to detect and measure slight stress-changes.

 

Aftershock activity of large earthquakes in Southern California [2]. The Hector Mine aftershock activity has been normal, relative to the decay predicted by the ETAS model during the 14 months of available data and no further large event has taken place in the vicinity up until now. In contrast, although the aftershock sequence of the 1992 Landers earthquake (M=7.3), including the 1992 Big Bear earthquake (M=6.4) and its aftershocks, fits very well to the ETAS up until about 6 months after the mainshock, the activity showed clear lowering relative to the modeled rate (relative quiescence) and the anomaly lasted nearly 7 years, leading up to the Hector Mine earthquake (M=7.1) in 1999. Specifically, the relative quiescence occurred only in the shallow aftershock activity, down to depths of 5 - 6 km. The sequence of deeper events showed clear, normal aftershock activity that fitted well to the ETAS throughout the whole period. We argue several physical explanations for these results. Among them, we strongly suspect aseismic slips within the Hector Mine rupture source that could inhibit the crustal relaxation process within "shadow zones" of the Coulomb's failure stress change.  Furthermore, the aftershock activity of the 1992 Joshua Tree earthquake (M=6.1) sharply lowered in the same day of the mainshock, which can be explained by a similar scenario due to aseismic slip.

 

Consecutive earthquakes in Miyagi Prefecture and its offshore [22, A19]. Anomalous seismicity such as quiescence and activation is defined by a systematic deviation of seismic activity from the predicted rate by the ETAS model that represents the normal occurrence-rate of earthquakes in a region indicating the empirical triggering effect by the previous events. The model is fitted to a dataset of origin-times and magnitudes of earthquakes or aftershocks during May-August 2003 in and around northern Japan. The detected quiescence and activation relative to the predicted seismicity rate are consistent with the coseismic changes of Coulomb failure stress (CFS) in the corresponding regions, transferred from certain strong earthquakes. Few results in the present manuscript agree with the claim that there should be a threshold value of DCFS capable of affecting seismic changes. Thus, we expect that significant deviation of actual activity from the predicted rates is sensitive enough to detect a slight stress-change. Furthermore, we offer a similar interpretation of the detected seismicity lowering relative to the modeled rates preceding the strong earthquakes, assuming some aseismic slips.

 

Anomalous aftershock activity of the 2004 Niigata-Ken-Chuetsu earthquake of M6.8 [A30]. We are concerned with the drastic shift of the depth distribution of aftershocks against time, around 0.5 days after the mainshock when we have no major aftershock. The cross-section of ƒ’CFF diagram for the aftershocks of similar mechanisms to the mainshock's on the plane of fault's strike direction of the mainshock (reverse faulting) against depth shows positive and negative values in shallow and deep parts of the aftershock volume, respectively, by the assumed precursory slip of the large M6.1 event that occurred 3.7 days after the mainshock. The precursory slip can be triggered by the mainshock rupture with a large positive ƒ’CFF.

 

Relative quiescence reported before the occurrence of the largest aftershock (M5.8) with likely scenarios of precursory slips considered for the stress-shadow covering the aftershock area [30, A27, A35]. Monitoring of aftershock sequences to detect lowering activity, relative to the modeled rate (the relative quiescence), becomes realistic and practical in predicting the enhancement of the likelihood of having a substantially large aftershock, or even another earthquake of similar size to the mainshock or larger. A significant relative quiescence in the aftershock sequence of the 2005 March earthquake of M7.0 off the western coast of Fukuoka, Japan, was reported two weeks before the occurrence of the largest aftershock of M5.8 that also hit the Kyushu District. The relative quiescence was discussed in relation to the stress-shadowing as inhibiting the activity due to probable precursory slips. The reason of the shadowing was also retrospectively speculated in more detail in comparison with the 3D space-time feature of the main aftershocks by assuming a preslip in a zone between the main and secondary fault. The same slip should transfer the stress-shadow covering the active off-fault clusters which drastically lowered the activity. Likewise, another preslip around the largest aftershock can explain space-time feature of the relative quiescence preceding M5.0 event in the secondary aftershock sequence.

(3)  Seismicity anomalies preceding large earthquakes and crustal stress changes

An earthquake prediction scenario from the implication of the relative quiescence can be based on the asperity hypothesis in the sense that precursory slip around asperities applies more shear stress to the asperities which promotes the rupture of the main fault. On the other hand, aseismic slips in a particular region are not necessarily a precursor to the large event but may be aseismic slips that are repeated in the same region with no subsequent large events. Therefore, identification of an aseismic slip leading to the rupture of an asperity remains a further difficult research theme in earthquake prediction. At present, this issue should be considered in terms of probabilistic prediction proactively making likely scenarios of precursory slips.

 

Intermediate-term seismicity anomalies preceding the rupture around the focal region of the 2004 Niigata-Ken-Chuetsu earthquake of M6.8 [A27, 29]. The ETAS model is applied to four sequences of earthquakes (M>=2), which occurred from 1997 through to October 2004 in four respective regions divided around the source of the 2004 Niigata-Ken Chuetsu earthquake. The four regions are divided North, East, South and East around the source by the boundaries of positive and negative CFS increments (i.e., the counters of neutral CFS increment) for the receiver faults (10km depth) of dominating angles in this region. The actual cumulative number of events deviates upward in region North and South, but downward in East and West, from the predicted cumulative curve after the change-point, consistently with the regions of the Coulomb increments. These show that precursory slip may have taken place in the Chuetsu mainshock fault plane.

 

Synchronous seismicity changes in and around the northern Japan preceding the 2003 Tokachi-oki earthquake [21, A21, A25]. Preceding the 2003 Tokachi-oki earthquake, we have observed the synchronized onset of quiescence and activation in four different seismic regions relative to the rate predicted by the ETAS model. Then, such activation and lowering of the seismicity relative to the predicted rates, in addition to the change in the distribution of fault mechanisms of moderate earthquakes during 1972-1996 and 1997-2003 in northern Japan, have been well matched with the pattern of the Coulomb's stress changes due to the possible precursory slow slip. These results have led us to a summarized observation that even a small size of CFS changes of the order of millibars can trigger such lowering and activation, which is also supported by the seismicity-rate-change equation of Dieterich. Thus, the investigation of seismicity changes by the ETAS can be a very sensitive stress change sensor.

 

Features of seismic activities in and around Tohoku District, northern Japan, prior to the large interplate earthquakes off the coast of Miyagi Prefecture [29, A20]. This paper is concerned with the intermediate-term prediction of the forthcoming M7.4 ~ 8.2 earthquake on the plate boundary, off the east coast of Miyagi Prefecture, northern Japan, which has the highest occurrence probability among the long-term forecasted events announced to the public. Seismicity and aftershocks in the regions of stress-shadow show significantly lower activity than the rate predicted by the ETAS model (the relative quiescence) during some years preceding each of the previous ruptures in 1936 and 1978, whereas the seismicity is normal or even activated in the regions of neutral or increasing Coulomb failure stress (CFS), which leads to the scenario based on the likely precursory slip within or near to the source.

   Assuming such a scenario, a number of sequences of earthquakes or aftershocks during 1979-2004 from various regions in northern Japan are selected to analyze them by fitting the ETAS model. Then the results are examined in relation to the CFS increments in the considered regions using the source models of the 1793, 1936 and 1978 interplate ruptures, as well as the source model of the recently occurred 2003 Miyagi-Ken-Oki intra-slab earthquake of M7.1. It is likely that the results of the normal activity and relative quiescence in the respective activities are due to the preslip of the intra-slab earthquake rather than the preslip of the expected rupture on the plate boundary.

 

Anomalies in the aftershock sequences of the 2003 Tokachi-Oki earthquake of M8.0 and the 2004 Kushiro-Oki earthquake of M7.1, and seismicity changes in the eastern Hokkaido inland [A12, A27, A28, A33, A36]. The aftershock sequence (M2.5) of the 2003 Tokachi-Oki earthquake of M8.0 become quiet relative to the ETAS five months after the mainshock. Latitude coordinates of the aftershock epicenters against the transformed time by the ETAS model show that the quiescence took place in the southern part of the aftershock area whereas the northern part was normal. This is explained by the shadowing of the DCFS assuming slips in or near the source of the 2004 Kushiro-Oki earthquake of M7.1. At the same time, this slip should affect contrasting DCFS pattern in eastern Hokkaido inland compared to the DCFS pattern due to the 2003 Tokachi-Oki rupture that has been mentioned above. Coincidently, the contrasting activation and quiescence of the microseismicity in the area took place compared to the previous coseismic triggered activities as mentioned above. The similar space-time analysis of aftershocks of the 2004 Kushiro-Oki earthquake of M7.1, using the transformed time, showed quiescence in the western part of the aftershock region whereas the eastern part was normal. Afterslip in the southwestwern part of the Kushiro-Oki source may explain the space-time features of the aftershock activity.

 

Seismicity in and around the Kyushu District, preceding the 2005 earthquake of M7.0 at the western offshore region of Fukuoka Prefecture [A34]. The ETAS model is applied to the sequences of events with M1.5 during the 10 year period from1995 to March 23, 2005 in the 10 regions in which DCFS values are calculated for the most frequent angles of receiver faults for the respective depths, assuming stress transfer by the rupture on the fault model of the M7.0 earthquake. The activity is normal or activated relative to the ETAS prediction in the regions where DCFS values are positive, and the relative quiescence in the stress-shadow regions. This may support the hypothesis of precursory slips within the fault of the M7.0 earthquake.

 

On distributions of focal mechanisms [12, A11, A22, A25, A29]. We see a change of the normalized frequency distribution of DCFS values before and after 1995, calculated at the hypocenter of all earthquakes listed in the JMA earthquake mechanisms catalog in a wide area of northern Japan, assuming the suspected precursory slip of the 2003 Tokachi-Oki earthquake of M8.0. Normalized histograms are given for the arc-tangent transformation of ƒ’CFF. The ratio of the numbers of earthquakes with negative ƒ’CFF value to those with positive one drastically lowered after 1995. These indicate that either quiescence or activation occurred consistently with their receiver mechanisms overall in northern Japan. Similar changes took place in shallow earthquakes in central Japan before and after the 2000/2001 onset of slow slip beneath Lake Hamana.

   Next, we discussed the likelihood of three rupture models of the 2005 earthquake of M7.4 off the coast of the Kii Peninsula. The three models by the GSI, ERI and IISEE are based on GPS data, and seismograms at the near and far field stations, respectively, and we still do not know the right model due to the inaccurate hypocenters estimates due to the earthquake's occurrence far offshore. Relying on the proposition that aftershocks should have been triggered by the main rupture, ƒ’CFF of aftershocks that occurred outside of the main fault are calculated for each rupture model using the F-net mechanism data. The model by the IISEE is most likely given the ratio of the number of aftershocks with positive ƒ’CFF to the one with the negative ƒ’CFF.

(4)  Space-time ETAS modeling

Stochastc clustering / declustering by the space-time ETAS model [13, 25, 33, A17, A18, A27]. On the basis of the space-time ETAS model and the thinning procedure, this paper gives the method of how to classify the earthquakes in a given catalogue into different clusters stochastically. The key points of this method are the probabilities of one event being triggered by another previous event and of it being a background event. Making use of these probabilities, we can reconstruct the functions associated with the characteristics of earthquake clusters to test a number of important hypotheses about the earthquake clustering phenomena.

   We applied this reconstruction method to the shallow seismic data in Japan and also to a simulated catalogue. The results show the following assertions: (1) The functions for each component in the formulation of the space-time ETAS model are good enough as a first-order approximation for describing earthquake clusters; (2) a background event triggers less offspring in expectation than a triggered event of the same magnitude; (3) the magnitude distribution of the triggered event depends on the magnitude of its direct ancestor; (4) the diffusion of the aftershock sequence is mainly caused by cascades of individual triggering processes, while no evidence shows that each individual triggering process is diffusive; and (5) the scale of the triggering region is still an exponential law, as formulated in the model but not the same one for the expected number of offspring. Thus, modification of the space-time ETAS has been made [25, 31].

   Taiwan hypocenter data obtained during the 20th century is analyzed and discussed on the basis of the background and clustering seismicity rates. Specifically, we find that the areas of the highest clustering ratio correspond to the major strike-slip fault traces in and around Taiwan. Additionally, in the Taiwan inland region, during the period 1960-1990, the outputs for the declustered catalogue show a clear quiescence in background seismicity preceding the recovery of activation and the occurrences of the 1999 Chi-Chi earthquake of ML7.3, while the other active regions show stationary background activity. This could be interpreted as an effect of the aseismic slip in the Chi-Chi rupture fault, whereby the inland region around the Chi-Chi source becomes a stress shadow.

 

Hierarchical Space-Time Model for Regional Seismicity [1, A2, A5, A6, A9]. A space-time point-process model is specified in which earthquake intensity is modelled as a function of previous earthquakes' occurrence times, locations, and magnitudes. Specific forms of the function of locations and times are based on the established empirical laws, such as the modified Omori formula and the Utsu-Seki scaling law of aftershock area against magnitude, but their parameter values are known to be different from place to place. Thus the parameters are further considered to be functions of spatial locations (but not time) represented by linear interpolation over a tessellation based on observed locations of earthquakes. Using the smooth representation for each parameter function in the model, a penalized log-likelihood method is used for the fitting, where the optimal weights of the penalties are objectively tuned by an empirical Bayesian method. Having done that, our final goal is to detect the temporal deviation of the actual seismicity rate from that of the modeled occurrence rate. For this procedure we estimate the space-time residual function represented by linear interpolation over a 3-dimensional tessellation based on observed times and locations of earthquakes, carrying out the similar penalized log-likelihood method. According to the estimated residual function, there are a number of zones where temporal deviation from the fitted model, with quiet periods, occurred before large earthquakes.

 

Hierarchical Space-Time Model for characterizing regional seismicity and anomaly detection [11, A2, A5, A6, A9]. The regional earthquake occurrence rate is modeled as a function of previous activity for which the specific form is based on empirical laws, such as the modified Omori formula and the Utsu-Seki scaling law of aftershock area against magnitude. Its parameters, including the p-value of the aftershock decay rate, can vary from place to place, showing some geophysical feature appearing correlated with the crustal temperature. This model is used to visualize features of the regional seismic activities in and around Japan. Among the parameters of the model, the present paper is particularly concerned with spatial variation of the normalized K-values. This takes high values around the boundaries of asperities that are estimated in the wide area off the East Coast of Tohoku District, Japan, by the inversion of the historical strong motion seismograms.

   Furthermore, this space-time model enables us to magnify anomalous periods and regions where the actual occurrence rates deviate systematically from the modeled one. Thus, the activation and quiescence relative to the model's prediction could provide sensitive detection of stress-changes in the regions. We are concerned with relative activation and lowering of the seismicity to explore the regions matching the pattern of Coulomb's stress changes due to a rupture or silent slip elsewhere. For example, such anomalies as those seen in the seismic activity in most central Japan (M 2.5) during 1995-1999 and during 2001 are likely the consequence of the 1995 Kobe

rupture and the interplate aseismic slip during 2001 beneath the western Tokai region, respectively.

(5)  Modeling the interface between physical and stochastic process

Detecting fluid signals in seismicity data through statistical earthquake modeling [14]. According to the Coulomb failure criterion, the variation of both stress and pore pressure can result in earthquake rupture. Aftershock sequences characterized by the Omori-Utsu law are often assumed to be the consequence of varying stress, whereas earthquake swarms are supposed to be triggered by fluid intrusions. The role of stress triggering can be analyzed by modeling the 3D elastic stress changes in the crust. However, fluid flows initiating seismicity cannot be identified without dealing with both pore pressure variations and earthquake connected stress field changes resulting in complex seismicity patterns.

   We show that the ETAS model is an appropriate tool to extract the primary fluid signal from such complex seismicity patterns. We analyze a large earthquake swarm that occurred in the year 2000 in Vogtland/NW-Bohemia, Central Europe. By fitting the stochastic ETAS model, we find that stress triggering is dominant in creating the seismicity patterns. This explains the observed fractal interevent-time distribution. External forcing identified with pore pressure changes due to fluid intrusion is found to directly trigger only about 1% of the total activity. However, a temporal deconvolution unveils a pronounced fluid signal initiating the swarm.

   Model simulations are performed in which earthquakes are triggered by fluid intrusion as well as coseismic and postseismic stress changes on a fault plane embedded in a 3D elastic half-space. They reproduce the observed swarm characteristics including the temporal power-law increase of the seismic moment release. Analyzing these simulations, we find that the proposed deconvolution procedure is able to reveal the underlying pore pressure variations. This model may be applied to the swarm owing to magma intrusion [e.g., A7].

 

Analysis of observations on the ultra-low frequency electric field in the Beijing Region [24]. This paper presents a preliminary analysis of observations on ultra-low frequency ground electric signals from stations operated by the China Seismological Bureau over the last 20 years. The data analyzed consists of estimates of the total strengths (cumulated amplitudes) of the electric signals during 24-hour periods. The thresholds are set low enough so that on most days a zero observation is returned. Non-zero observations are related to electric and magnetic storms, occasional man-made electrical effects, and, apparently, some pre-, co-, or postseismic signals.

   To investigate the extent that the electric signals can be considered as preseismic in character, the electric signals from each of five stations are jointly analyzed with the catalogue of local earthquakes within circular regions around the selected stations. A version of Ogata's Lin-Lin algorithm is used to estimate and test the existence of a pre-seismic signal. This model allows the effect of the electric signals to be tested, even after allowing for the effects of earthquake clustering. It is found that, although the largest single effect influencing earthquake occurrence is the clustering tendency, there remains a significant preseismic component from the electrical signals. Additional tests show that the apparent effect is not postseismic in character, and persists even under variations of the model and the time periods used in the analysis. Samples of the data are presented and the full data sets have been made available on local websites.

 

Microseismicity and Earthtide [15, A1]. This paper analyzes the microseismicity in the Tanba region during the two year period following the 1995 Kobe earthquake of M7.1 which was activated by the Kobe rupture. From the histogram of the number of superposed earthquakes on the interval of the lunar period, some correlation is strongly suspected between the earthtide related lunar motion. However, the seismicity includes the clustering and some decreasing trend after the Kobe event. The ETAS model that also includes the components of polynomial for the trend and Fourier series for the periodicity is applied to confirm the significance of the relationship. The paper [15] discusses the correlation between the b-value change of G-R magnitude-frequency of the underground AE data and tidal stress/strain due to lunar phase.

 

(6)  Simultaneous estimation of b-values and detection rates of earthquakes for the application to aftershock probability forecasting [A31, A32].

It is known that the detection rate of aftershocks is extremely low during the period immediately following a large earthquake due to the contamination of arriving seismic waves. This has resulted in considerable difficulty in obtaining an estimate of the empirical laws of aftershock decay and the magnitude frequency immediately after the main shock. This paper presents an estimation method for predicting the underlying occurrence rate of aftershocks of any magnitude range, based on the magnitude frequency model that combines Gutenberg-Richter's law with the detection rate function. This procedure enables us to announce real-time probability forecasting of aftershocks immediately after the mainshock, when the majority of large aftershocks is likely to occur.

 

 

Published papers 2003 - 2005

Refereed Journals:

2003:

[1] Ogata, Y., Katsura, K. and Tanemura, M. (2003) Modelling of heterogeneous space-time occurrences of earthquakes and its residual analysis, Applied Statistics (J. Roy. Stat. Soc. Ser. C.),
Vol. 52, Part 4, pp. 499-509 (2003).

[2] Ogata, Y., Jones, L. and Toda, S. (2003) When and where the aftershock activity was depressed: Contrasting decay patterns of the proximate large earthquakes in southern California, J. Geophys. Res., 108, No. B6, 2318, doi:@10.1029/2002JB002009.

[3] Ogata, Y. (2003) Examples of statistical models and methods applied to seismology and related earth physics, International Handbook of Earthquake and Engineering Seismology, International Association of Seismology and Physics of Earth's Interior, Vol. 81B, HandbookCD#2, Chapter 82.

[4] Toda, S. and Stein, R.S. (2003) Toggling of seismicity by the 1997 Kagoshima earthquake couplet: A demonstration of time-dependent stress transfer, J. Geophys. Res., 108, B12, 2567, doi: 10.1029/@2003JB002527,

[5] Vere-Jones, D. and Ogata, Y. (2003) Statistical principles for seismologists, International Handbook of Earthquake and Engineering Seismology, International Association of Seismology and Physics of Earth's Interior, Vol. 81B, pp. 1573-1586.

 

2004:

[6] Cho, I., T. Iwata, T. Shiga, T. Tokunaga, M. Fukuyo, and Y. Shinozaki, Data processing of circular arrays for an exploration method using microtremors: Applicability of Henstridge's method and new methods, BUTSURI-TANSA (Bulletin of Society of Exploration Geophysics of Japan), 57, 501-516 (in Japanese).

[7] Iwata, T., and Nakanishi, I. (2004) Hastening of occurrences of earthquakes due to dynamic triggering: The observation at Matsushiro, central Japan, Journal of Seismology, 8, pp. 165-177.

[8] Nanjo, K., Nagahama, H. and Yodogawa, E.. (2004) Symmetry in the Self-organized Criticality, The Journal of the International Society for the Interdisciplinary Study of Symmetry (ISIS-Symmetry) Symmetry: Art and Science 2004 (Editors D. Nagy and G. Lugosi) ISIS-Symmetry, Budapest, Hungary, pp. 302-305.

[9] Nanjo, K. and Nagahama H. (2004) Fractal Properties of Spatial Distributions of Aftershocks and Active Faults, Chaos, Solitons and Fractals, 19, pp. 387-397, doi: 10.1016/S0960-0779(03)00051-1.

[10] Nanjo, K. and Nagahama, H. (2004) Discussions on Fractals, Aftershocks and Active Faults: Diffusion and Seismo-electromagnetism, The Arabian Journal for Science and Engineering, 2004, 29, 2C, pp. 147-167.

[11] Ogata, Y. (2004) Space-time model for regional seismicity and detection of crustal stress changes, J. Geophys. Res., Vol. 109, No. B3, B03308, doi:10.1029/2003JB002621.

[12] Ogata, Y. (2004) Seismicity quiescence and activation in western Japan associated with the 1944 and 1946 great earthquakes near the Nankai trough, J. Geophys. Res., 109, B4, B04305, doi:10.1029/2003JB002634.

[13] Zhuang, J., Ogata, Y. and Vere-Jones, D. (2004) Analyzing earthquake clustering features by using stochastic reconstruction Journal of Geophysical Research, 109, B5, B05301, doi:10.1029/2003JB002879.

 

2005:

[14] Hainzl, S. and Ogata, Y. (2005) Detecting fluid signals in seismicity data through statistical earthquake modeling, J. Geophys. Res., Vol.110, No.B5, B05S07, doi:10.1029/2004JB003247 (2005).

[15] Iwata, T. and Young, P. (2005) Tidal stress/strain and the b-value of acoustic emissions at the Underground Research Laboratory, Canada, Pure and Applied Geophysics, 162, pp. 1291-1308.

[16] Iwata, T., M. Imoto, and S. Horiuchi (2005) Probabilistic estimation of earthquake growth to a catastorophic one, Geophys. Res. Let., 32. L19307, 10.1029/2005GL023928.

[17] Nanjo, K.Z., Nagahama, H. and Yodogawa, E. (2005) Symmetropy of fault patterns: Quantitative measurement of anisotropy and entropic heterogeneity, Mathematical Geology, 37, 3, pp. 277-293, doi: 10.1007/s11004-005-1559-z.

[18] Nanjo, K.Z., Turcotte, D.L. and Shcherbakov, R. (2005) A model of damage mechanics for the deformation of the continental crust, J. Geophys. Res., 110, B7, B07403, DOI: 10.1029/2004JB003438.

[19] Nanjo, K.Z. and Turcotte, D.L. (2005) Damage and rheology in a fiber-bundle model, Geophys. J. Int., 2005, 162, pp. 859-866, doi:10.1111/j.1365-246X.2005.02683.x.

[20] Holliday, J.R., Nanjo, K.Z., Tiampo, K.F., Rundle, J.B. and Turcotte, D.L. (2005) Earthquake forecasting and its verification, Nonlinear Processes in Geophysics, 12, pp. 965-977, doi: 1607-7946/npg/2005-12-965.

[21] Ogata, Y. (2005) Synchronous seismicity changes in and around the northern Japan preceding the 2003 Tokachi-oki earthquake of M8.0, J. Geophys. Res., 110, B8, B08305, doi:10.1029/2004JB003323.

[22] Ogata, Y. (2005) Detection of anomalous seismicity as a stress change sensor, J. Geophys. Res., Vol.110, No.B5, B05S06, doi:10.1029/2004JB003245.

[23] Toda, S., Stein, R.S., Richards-Dinger, K. and Bozkurt, S. (2005) Forecasting the evolution of seismicity in southern California: Animations built on earthquake stress transfer, J. Geophys. Res.,110, B05S16, doi:10.1029/2004JB003415.

[24] Zhuang, J., Vere-Jones, D., Guan, H., Ogata, Y. and Ma, Li (2005) Preliminary analysis of observations on the ultra-low frequency electric field in the Beijing region, Pure and Applied Geophysics, 162, pp. 1367-1396.

[25] Zhuang, J.  Chang, C., Ogata, Y.  Chen, Y. (2005) A study on the background and clustering seismicity in the Taiwan region by using point process models, J. Geophys. Res., 110, B5, B05S18, doi:10.1029/ 2004JB003157.

 

In press or accepted:

[26] Nanjo, K.Z., Nagahama, H. and Yodogawa, E., Symmetropy of earthquake patterns: asymmetry and rotation in a disordered seismic source, Acta Geophysica Polonica, in press, Volume 54.

[27] Nanjo, K.Z., Rundle, J.B., Holliday, J.R. and Turcotte, D.L., Pattern informatics and its application for optimal forecasting of large earthquakes in Japan, Pure and Applied Geophysics, accepted.

[28] Chen, C.C., Rundle, J.B., Holliday, J.R., Nanjo, K.Z., Turcotte, D.L., Li, S.C. and Tiampo, K.F., The 1999 Chi-Chi, Taiwan, earthquake as a typical example of seismic activation and quiescence, Geophys. Res. Let., 2005 accepted.

[29] Ogata, Y., Seismicity anomaly scenario prior to the major recurrent earthquakes off the east coast of Miyagi Prefecture, northern Japan, and its implication for the intermediate-term prediction, Special Issue on Dynamics of Seismicity Patterns and Earthquake Triggering, eds. S. Hainzl, G. Zoler and I. Main, Tectonophysics, in press.

[30] Ogata, Y., Anomaly monitoring of aftershock sequence by a reference model: A case study of the 2005 earthquake of M7.0 at the western Fukuoka, Kyushu, Japan, Geophys. Res. Letters, in press.

[31] Ogata, Y. and Zhuang, J., Space-time ETAS models and an improved extension, Special Issue on Critical Point Theory and Space-Time Pattern Formation in Precursory Seismicity, eds. K. Tiampo and M. Anghel, Tectonophysics, in press.

[32] Toda, S. and Matsumura, S., Spatio-temporal stress states estimated from seismicity rate changes in the Tokai region, central Japan, Tectonophysics, in press.

[33] Zhuang J., Ogata Y. and Vere-Jones D., Diagnostic analysis of space-time branching processes for earthquakes. Chapter 15 of Case Studies in Spatial Point Process Models, Eds. Baddeley A., Gregori P., Mateu J., Stoica R. and Stoyan D. Springer-Verlag, New York, in press.

 

Main Proceedings:

2003:

[A1] Iwata, T. and Katao, H. (2003) Analysis of a correlation between the phase of the moon and the occurrences of microearthquakes in the Tanba plateau through the point-process modeling, Programme and Abstracts of the 2003, Fall Meeting of the Seismol. Soc. Japan, A062.

[A2] Ogata, Y. (2003) A practival space-time model for regional seismicity (invited), EGS-AGU-EUG Joint Assembly,Nice, France,  Geophysical Research Abstract , Volume 5, 2003,  CD-ROM, ISSN: 1029-7006

[A3] Ogata, Y. (2003) Sesimicity-change-analysis by a space-time point-process model (invited) The 3rd Statistical Seismology Workshop, Juriquilla, Mexico.

[A4] Ogata, Y. (2003) Seismicity quiescence and activation in western Japan associated with 1944 and 1946 great earthquakes near the Nankai Trough (in Japanese), Report of the Coordinating Committee for Earthquake Prediction, 70, pp. 378-383, Geographical Survey Institute of Japan.

[A5] Ogata, Y. (2003) Seismicity changes in western Japan (1995-2001) detected by a statistical space-time model (in Japanese), Report of the Coordinating Committee for Earthquake Prediction, Vol. 70, pp. 5-6 and pp. 361-363.

[A6] Ogata, Y. (2003) A space-time model for regional seismicity and detection of seismicity changes (in Japanese), Chikyu Monthly, 25, 10, pp. 783-787.

[A7] Toda, S. and Stein, R.S. (2003) Earthquake triggering by volcano-tectonic events: An example from the 2000 Izu Islands swarm (invited talk), XXIII Ceneral Assembly of the International Union of Geodesy and Geophysics, 2003.

[A8] Toda, S. (2003) A Fresh Look at the Triggering of Earthquake Pairs, Such as the Landers-Big Bear, Landers-Hector Mine, Izmit-Duzce, and Nenana-Denali, and March-May 1997 Kagoshima Events (invited talk), American Geophysical  Union 2004 fall meeting.

 

2004:

[A9] Ogata, Y. (2004) The 6th World Congress of the Bernoulli Society for Mathematical Statistics and Probability, and 67th Annual Meeting of the Institute of Mathematical Statistics, gSpace-time model for regional seismicity and detection of crustal stress changesh, July 25-29, 2004, Barcelona, Spain, (invited lecture)

[A10] Ogata, Y. (2004) Statistical models of point processes and prediction and discovery in seismic activity (in Japanese), Spring Meeting of the Mathematical Society of Japan, March 30, 2004, Tsukuba, Japan (specially organized invited lecture)

[A11] Ogata, Y. (2004) Static triggering and statistical modeling (in Japanese), The 156-th Meeting of  Coordinating Committee for Earthquake Prediction , Geographical Survey Institute, Kudan, Tokyo, February 16, 2004, Japan (topics invited lecture)

[A12] Ogata, Y. (2004) Synchronous seismicity changes in and around the northern Japan preceding the 2003 Tokachi-oki earthquake of M8.0 (invited talk) International Conference in Commemoration of 5-th Anniversary of the 1999 Chi-Chi Earthquake, Taipei, Taiwan.

[A13] Ogata, Y. (2004) Stress changes, seismicity changes and statistical models, Workshop on Seismic Activity and Probabilities of Major Earthquakes in the Kanto and Tokai Area , Central Japan, Wadati Memorial Hall, Institute for Earth Science and Disaster Prevention, Tsukuba, Japan (invited presentation),  http://kt-jisin.bosai.go.jp/WS/Program/index.html, (invited talk)

[A14] Nanjo, K.Z., Rundle, J.B. and Holliday, J.R.. (2004) Pattern Informatics and Its Application to Forecasting Large Earthquakes in Japan, Abstract for AGU 2004 Fall Meeting,  Eos Trans. AGU, 85(47), Fall Meet. Suppl., Abstract NG22A-07 (invited talk).

[A15] Toda, S. (2004) Recent progress in earthquake triggering study and possible applications to earthquake prediction (in Japanese), The 156-th Meeting of  Coordinating Committee for Earthquake Prediction , Geographical Survey Institute, Kudan, Tokyo, February 16, 2004, Japan (topics invited lecture).

[A16] Toda, S., and Matsumura, S. (2004) Spato-temporal stress states estimated from seismicity rate changes in the Tokai region, central Japan (invited talk), American Geophysical Union 2004 fall meeting.

[A17] Zhuang, J., Ogata, Y. and Vere-Jones, D. (2004) Diagnostic analyses of space-time branching processes for earthquakes, Spatial Point Process Modeling and its Applications, Benicassim, Castellon, Spain, Spatial Point Process Modelling and Its Applications, Col-Lecco Treballis D'Infomatica/Tecnologia, Num. 20, ISBN 84-8021-475-9 Publication de la Universitat Jaume-I, Castello de la Plana, Spain, pp. 273-292.

[A18] Zhuang, J., Ogata, Y. and Vere-Jones, D. (2004) Visualizing goodness-of-fit of point-process models for earthquake clusters., Analysis of Natural and Social Phenomena: Data Science and System Reduction; an international workshop of the 21st Century COE program at Keio University, http://coe.math.keio.ac.jp/english/event/cherry_bud/index.html, (invited talk).

[A19] Ogata, Y. (2004) Quiescence of the 2003 foreshock/aftershock activities in and off the coast of Miyagi Prefecture, northern Japan, and their correlation to the triggered stress-changes (in Japanese), Report of the Coordinating Committee for Earthquake Prediction, 71, pp. 260-267, Geographical Survey Institute of Japan.

[A20] Ogata, Y. (2004) Statistical analysis of seismic activities in and around Tohoku District, northern Japan, prior to the large interplate earthquakes off the coast of Miyagi Prefecture (in Japanese), Report of the Coordinating Committee for Earthquake Prediction, 71, pp. 268-278, Geographical Survey Institute of Japan.

[A21] Ogata, Y. (2004) Seismicity changes and stress changes in and around the northern Japan relating to the 2003 Tokachi earthquake of M8.0 (in Japanese), Report of the Coordinating Committee for Earthquake Prediction, 72, pp. 110-117, Geographical Survey Institute of Japan.

[A22] Ogata, Y. (2004) Static triggering and statistical modeling (in Japanese), Report of the Coordinating Committee for Earthquake Prediction, Vol. 72, pp. 631-637, Geographical Survey Institute of Japan.

[A23] Toda, S. (2004) Recent progress in earthquake triggering study and possible applications to earthquake prediction (in Japanese), Report of the Coordinating Committee for Earthquake Prediction, 72, pp. 624-626, Geographical Survey Institute of Japan.

[A24] Toda, S. (2004) Seismicity changes in inland activity before and after the 2003 Miyagi-Ken-Oki earthquake and its implications (in Japanese), Chikyu Monthly, 27, 1, 56-61.

[A25] Ogata, Y. (2004)  On changes of statistical distribution of focal mechanisms of events prior to the main ruptures, Programme and Abstracts of the 2005 Fall Meeting of the Seismol. Soc. Japan, S023.

 

2005:

[A26] Murata, Y. and Ogata, Y. (2005) Surficial density estimation from gravity data using Delaunay triangular network, Joint meeting for Earth and Planetary Science, J031-004, Makuhari, Chiba Prefecture, Japan, May 2005.

[A27] Ogata, Y.(2005) Seismicity anomalies measured by the ETAS model and stress changes (solicited), The General Assembly 2005 of the European Geosciences Union (EGU), , April 24-29, 2005, the Austria Center Vienna (ACV), Vienna, Austria (invited lecture).

[A28] Ogata, Y. (2005) Seismicity changes in western Japan associated with the great earthquakes near Nankai trough and their contemporary implications, Specially organized session S095, invited talk.

[A29] Ogata, Y. (2005) On the aftershock activity of the 2004 earthquake of M7.4 at the southeast off the coast of the Kii Peninsula, and constraints on the fault rupture models by the mechanisms and space-time pattern of the aftershocks (in Japanese), Report of the Coordinating Committee for Earthquake Prediction, 73, pp. 495-498 Geographical Survey Institute of Japan.

[A30] Ogata, Y. (2005) On an anomalous aftershock activity of the 2004 Niigata-Ken-Chuetsu earthquake of M6.8, and intermediate-term seismicity anomalies preceding the rupture around the focal region (in Japanese), Report of the Coordinating Committee for Earthquake Prediction, 73, pp. 327-331, Geographical Survey Institute of Japan.

[A31] Ogata, Y. (2005) Simultaneous estimation of b-values and detection rates of earthquakes for the application to aftershock probability forecasting (in Japanese), Report of the Coordinating Committee for Earthquake Prediction, Vol. 73, pp. 666-669, Geographical Survey Institute of Japan.

[A32] Ogata, Y. (2005) Toward urgent forecasting of aftershock hazard: Simultaneous estimation of b-value of the Gutenberg-Richterfs law of the magnitude frequency and changing detection rates of aftershocks immediately after the mainshock, preprint.

[A33] Ogata, Y. (2005) Anomalies in the aftershock sequences of the 2003 Tokachi-Oki earthquake of M8.0 and the 2004 Kushiro-Oki earthquake of M7.1 and seismicity changes in the eastern Hokkaido inland (in Japanese), Report of the Coordinating Committee for Earthquake Prediction, 74, pp. 83-88, Geographical Survey Institute of Japan.

[A34] Ogata, Y. (2005) Seismicity changes in and around Kyushu District before the 2005 earthquake of M7.0 in the western offshore of Fukuoka Prefecture (in Japanese), Report of the Coordinating Committee for Earthquake Prediction, 74, pp. 523-528, Geographical Survey Institute of Japan.

[A35] Ogata, Y. (2005) Relative quiescence reported before the occurrence of the largest aftershock (M5.8) in the aftershocks of the 2005 earthquake of M7.0 at the western Fukuoka, Kyushu, and possible scenarios of precursory slips considered for the stress-shadow covering the aftershock area (in Japanese), Report of the Coordinating Committee for Earthquake Prediction, 74, pp. 529-535, Geographical Survey Institute of Japan.

[A36] Ogata, Y. (2005)  Anomalies in the aftershock sequences of the 2003 Tokachi-Oki earthquake of M8.0 and the 2004 Kushiro-Oki earthquake of M7.1 and seismicity changes in the eastern Hokkaido inland, Programme and Abstracts of the 2005 Fall Meeting of the Seismol. Soc. Japan, S023.

[A37] Toda, S. (2005) Style of stress accumulation and release in northern Honshu Japan: A concept to explain the coexistence of destructive inland earthquakes and interplate thrust earthquakes (invited talk), Spatial and Temporal Fluctuation in the Solid Earth, 21COE International Symposium 2005, Sendai, Japan.

 

 

Future plan for the project

(1)  Examination of scenarios for predicting asperity-slip based on the seismicity anomalies

Anomaly monitoring of aftershock sequences to detect lowering activity, relative to the modelled rate (the relative quiescence), is now becoming realistic and practical in predicting the enhancement of the likelihood of having a significantly large aftershock, or even another earthquake of similar size or larger occurring. In order to predict the location of a large aftershock or another proximate large earthquake, we have to assume that a significant slip may have occurred within and near to the source of the suspected earthquake due to the acceleration of quasi-static (slow) slips on the fault as the time of rupture of the major asperity approached. This is indicated, for example, by the analysis of small repeating earthquake data. Thus, we should look carefully at the activity in the stress-shadow, transferred from the slip. Such scenarios for the prediction should be useful for examining and explaining any such anomalous features.

   In fact, given an anomaly of seismicity rate change, the difficulty lies in identifying the slip location and its imminence to a major rupture; most of them are unknown so far as no other data or constraints are available. For probable predictions, Ogata [2005c] explored and reported several scenarios of stress transfers from some thinkable slips, such as that they could have been triggered by the M7.0 rupture at 20th March 2005 off western Fukuoka city and moreover that the stress-shadows due to the slips should, in turn, have covered the majority of the aftershock region. These include a conjugate fault of the main fault rupture and several known active faults near the rupture fault. However, the largest aftershock that occurred two weeks later was not included in the suspected scenarios. The only exception was the predicted unlikely slip within the Kego Fault that runs through the urban area in the city of Fukuoka, since this results in a stress increase in the entire aftershock region, and no part of aftershock region can be the stress shadow.

(2)  Effective space-time modeling of seismic activity and the detection of seismicity anomalies

In any application of the (hierarchical) space-time ETAS model, the data has to be homogeneous in space and time. The threshold magnitude of completely detected earthquakes throughout a long period and wide region is high, and the number of earthquakes above such a threshold magnitude is very limited compared to the number of listed earthquakes in a catalog. This is because the detection rate of earthquakes for each magnitude is dependent on time and location. Especially, the detection rate of aftershocks is extremely low during the period immediately following the main shock, due to contamination of arriving seismic waves.

   For an estimation of the detection rate in time and space, we propose utilizing the statistical model introduced by Ogata and Katsura [1993] for the simultaneous estimation of the b-values of Gutenberg-Richter law together with the detection-rate (probability) of earthquakes of each magnitude-band from the provided data of all detected events, where both parameters are allowed to change in time. Thus, by using all the detected earthquakes in a given period and area, we can estimate the underlying ETAS rate of both the detected and undetected events and their b-value changes, taking the time-varying missing rates of events into account. It is our hope that it will become possible to give details of seismicity patterns such as aftershock productivity parameters to delineate asperities [Ogata, 2005a] over the period and region of the relative quiescence as given in Ogata et al. [2003].

   As a primary step toward achieving this objective, Ogata [2005b] presents an estimation method for predicting underlying occurrence rate of aftershocks of any magnitude range. This procedure enables real-time probability forecasting of aftershocks immediately after the mainshock, when the majority of large aftershocks is likely to occur and when the forecasting is most critical for public in the affected area.

(3)  Predictive space-time-magnitude characterization of foreshocks

Descrimination of foreshocks and clustering / declustering algorithms as a short-term prediction. Discrimination of foreshock sequences from other clustering activity is an important problem in short term earthquake prediction. When sequential earthquake activity starts somewhere, it can be a swarm, a foreshock sequence, or simply a mainshock-aftershock sequence. Therefore, it is very desirable to know whether the activity is a precursor to a forthcoming significantly larger earthquake or not. Ogata et al. [1995, 1996 and 1999] and [Ogata, 1999b] investigated data sets of earthquake clusters to discriminate features of foreshocks from earthquakes of other cluster types in a statistical senseCand found several features of some predictive value, including the fact that foreshocks were more closely-spaced in time than either swarms or aftershocks, the fact that foreshocks were closer together in space than other types of events, and that foreshocks' sizes were more likely to increase chronologically than the other types. By modeling such discriminating features they developed probability forecasts of an earthquake cluster being of foreshock type.

   However the primary difficulty of this procedure is to find a suitable declustering algorithm for real-time forecasting of probability. There are two contrasting typical clustering algorithms, that is, the magnitude-based clustering (MBC) and the single-link-clustering (SLC) ones. The MBC based forecast performs very well, but a MBC cluster cannot formed until the occurrence of the main shock. The SLC algorithm is better for the real-time recognition of a cluster, but does not perform objectively. Recently, based on the space-time ETAS model, Zhuang et al. [2002, 2005] proposed a stochastic clustering algorithm, which appears quite useful in such a forecasting. Because of its stochastic nature, a Baysian predictive procedure will be useful.

(4)  Prediction and inversion problem between seismicity changes and stress-changes

It is becoming important to study how to make predictions and how to solve inverse problems based on the quantitative relationships between seismicity rate changes (due to the ETAS) and stress-changes. In particular, we are concerned with the theory of rate/state-dependent friction and its application to the seismicity-rate-change equation of Dieterich [Dieterich, 1994; Dieterich et al., 2000; Toda and Stein, 2003].

(5)  Bayesian Probability assessments for Long-term prediction

Inference of hazard of a fault rupture and its uncertainty from data of a small number of events, possibly with uncertain occurrence times. Conventionally, a probability of the next rupture during a future period is calculated by the predictive hazard function into which we plug the MLE values of BPT model [Matthews et al., 2002] estimated by the historical data of occurrence times. However, when the number of events in the data is small, the predictive hazard function based on the MLE usually causes a serious bias of hazard rates. Therefore, in order to see the uncertainty of estimated hazard functions, we use the Bayesian inference introducing a set of appropriate prior distributions. Furthermore, if a record of the slip-sizes of the events is available, Ogata [2001, 2002] proposes an extended renewal process for a stochastic version of the time-predictable model of Shimazaki and Nakata [1980] by assuming the same distribution of the ratio of the time interval of the successive pair of events to the slip size of the first event of the pair. Thus, we can estimate more effectively not only the hazard function of the next event but also its uncertainty based on the occurrence time data of the events associated with the records of corresponding slip sizes.

   When paleoearthquake data is analyzed, our further concern is the raw data in which occurrence times of events themselves are uncertain and given by confidence intervals or some chronological likelihood function [e.g., Sieh et al., 1989] inferred from geological evidences based on trench studies. For such records, we consider another Bayesian inference in which each uncertainty of occurrence time is interpreted as a prior distribution associated with the likelihood of a renewal process model [Ogata, 1999]. Integration of the posterior function with respect to the uncertain occurrence times and also the parameters of the renewal process model is implemented in order to compare the goodness-of-fit of competing renewal processes (e.g., log-Normal, Weibull and BPT models). Thus, the corresponding marginal posteriors of the model provide both estimates of distributions of the uncertain occurrence times and also the predictive hazard rates associated with the selected renewal process model. Particularly, in the case where occurrence time of the last event is uncertain, a natural assessment of current and future hazard of the forthcoming rupture can be provided.

(6)  Statistical modeling for more effective use and quality improvements of datasets

There are many precise measurements such as geodetic extensometers, tiltmeters, volmetric strainmeters, GPS and so on. However, these records are always affected by some noises and other signals of various kinds caused by the motions of the sun, moon, and particularly meteorological factors such as balometric pressure, precipitation, temperature and humidity. Therefore, it is important to model the causality and response functions of these effects in order to know the genuine records of interest. Special care should be put to the GPS time series data to detect geodetic anomalies sensitively, since this dataset is now playing an important role similar to the earthquake catalog in the sense that the stations are very densely located throughout Japan. With regards to the earthquake catalog, the homogeneity in magnitude by its calibration will be main concern due to the changes of the seismographs for such a long period of about 80 years. Modeling of the detection rate change of earthquakes is already discussed above.

 

 

References

Akaike, H. (1998) Selected Papers of Horotugu Akaike, E. Parzen, Tanabe, K. and Kitagawa, G. eds., Springer Series of Statistics – Perspectives in Statistics, Springer, New York, 434pp.

Aki, K. (1981) A probabilistic synthesis of precursory phenomena, in Earthquake Prediction: An International Review, edited by D.W. Simpson & P.G. Richards, A.G.U., Washington, D.C., 566-574.

Dieterich, J. (1994), A constitutive law for rate of earthquake production and its application to earthquake clustering, J. Geophys. Res., 99, 2601–2618.

Dieterich, J., Cayol, V. and Okubo, P., The use of earthquake rate changes as a stress meter at Kilauea volcano, Nature 408, 457-460 (2000).

Matthews, M., Ellthworth, W.L. and Reasenberg, P. (2002) A Brownian Model for Recurrent Earthquakes, Bull. Seismol. Soc. Am., 92, 6, 2233-2250, doi: 10.1785/0120010267.

Ogata, Y. (1999) Estimating the hazard of rupture using uncertain occurrence times of paleoearthquakes, J. Geophys. Res. 104, 17995-18014.

Ogata, Y. (1999b) Real time discrimination of forshocks (in Japanese), Chikyu Monthly, No. 24, pp. 167-173.

Ogata, Y. (2001) Biases and uncertainties when estimating the hazard of the next Nankai earthquake (in Japanese), Chigaku Zasshi (Journal of Geography), Vol. 110, No. 4, pp. 602-614.

Ogata, Y. (2002) Slip-size dependent renewal processes and Bayesian inferences for uncertainties, J. Geophys. Res., 107, B11, 2268, doi:10.1029/2001JB000668, 2002.

Ogata, Y. (2004) Space-time model for regional seismicity and detection of crustal stress changes, J. Geophys. Res., 109, B3, B03308, doi:10.1029/2003JB002621.

Ogata, Y. (2005a) Simultaneous estimation of b-values and detection rates of earthquakes for the application to aftershock probability forecasting (in Japanese), Report of the Coordinating Committee for Earthquake Prediction, 73, pp. 666-669.

Ogata, Y. (2005b) Toward urgent forecasting of aftershock hazard: Simultaneous estimation of b-value of the Gutenberg-Richterfs law of the magnitude frequency and changing detection rates of aftershocks immediately after the mainshock, priprint.

Ogata, Y. (2005c) Anomaly monitoring of aftershock sequence by a reference model: A case study of the 2005 earthquake of M7.0 at the western Fukuoka, Kyushu, Japan, Geophys. Res. Letters, in press.

Ogata, Y. and Katsura, K. (1993) Analysis of temporal and spatial heterogeneity of magnitude frequency distribution inferred from earthquake catalogs, Geophys. J. Int. 113, 727-738.

Ogata, Y. Utsu, T. and Katsura, K. (1995) Statistical features of foreshocks in comparison with other earthquake clusters, Geophys. J. Int., 121, pp. 233-254.

Ogata, Y., Utsu, T. and Katsura, T. (1996) Statistical discrimination of foreshocks from other earthquake clusters, Geophys. J. Int., 127, pp. 17-30.

Ogata, Y. and Utsu, T. (1999) Real time statistical discrimination of foreshocks from other earthquake clusters (in Japanese), Tokei-Suri (Proc. Inst. Statist. Math), Vol. 47, No. 1, pp. 223-241.

Ogata, Y., Katsura, K. and Tanemura, M. (2003) Modelling of heterogeneous space-time earthquake occurrences and its residual analysis, Applied Statistics (J. Roy. Stat. Soc. Ser. C), 52, Part 4, 499-509.

Shimazaki, K., and T. Nakata (1980) Time-predictable recurrence model for large earthquakes, Geophys. Res. Letters, 7, 279-282.

Sieh, K., M. Stuiver, and D.R. Brillinger (1989) A more precise chronology of earthquakes produced by the San Andreas fault in southern California, J. Geophys. Res., 94, 603-623.

Toda, S. and Stein, R.S. (2003) Toggling of seismicity by the 1997 Kagoshima earthquake couplet: A demonstration of time-dependent stress transfer, J. Geophys. Res., 108, B12, 2567, doi: 10.1029/@2003JB002527,

Utsu, T. (1977) Probabilities in earthquake prediction, Zisin(2) (J. Seismol. Soc. Jap.), 30, 179-185 (in Japanese).

Utsu, T. (1978) Calculation of the probability of success of an earthquake prediction (In the case of Izu-Oshima Kinkai Earthquake of 1978), Report of the Coordinating Committee for Earthquake Prediction, 31, 129135, Geographical Survey Institute of Japan (in Japanese).

Zhuang, J., Ogata, Y. and Vere-Jones, D. (2002) Stochastic declustering of space-time earthquake occurrences, J. Amer. Statist. Assoc., 97, 369-380.

Zhuang, J., Ogata, Y. and Vere-Jones, D. (2004) Analyzing earthquake clustering features by using stochastic reconstruction J. Geophy. Res., 109, B5, B05301, doi:10.1029/2003JB002879.

 

 


Back to Statistical Seismology Research Group Home Page


Back to Ogatafs English Home Page


Updated on 13 July 2006