



Statistical Seismology Seminars F
Updated on
28 September 2017
ϊ{κ
The 67th
NEW! SpeakerF@Dr.
Wu, Stephen iAssistant
Professor of ISMj DateF@3 October 2017@@@TimeF@16:3017:30 LocationF@Room D312B (Seminar Room), Institute
of Statistical Mathematics TitleF@Review of earthquake early warning
from an engineering perspective AbstractF@ After
the concept of earthquake early warning (EEW) first appeared in the 1980s, we
now have officially working EEW systems around the world, such as, Japan,
Taiwan, Mexico, USA, Italy, and so on. The algorithms of EEW have evolved to
a large variety, including both onsite, regional and some hybrid methods.
The underlying seismic model ranges from simple pointsource ground motion
prediction equations to sophisticated finite fault prediction models.
Recently, researchers have also proposed to develop realtime GPS based EEW
and purely datadriven seismic intensity prediction models. Besides the
scientific advances, engineering applications of EEW have became another
important research topic. In this talk, I will briefly go through all the
topics above in a practical implementation point of view, and highlight some
important challenge of EEW. The 66th
1 SpeakerF@Dr.
Wu, Jing iInstitute of
Geology and Geophysics, Chinese Academy of Science, China E Associate Professorj DateF@29 August 2017@@@TimeF@16:00 17:00 LocationF@Room D312B (Seminar Room), Institute
of Statistical Mathematics TitleF@Seismicity and Seismic Anisotropy
beneath eastern Tibet AbstractF@ Eastern
Tibet is one of the most tectonically active areas in Chinese Mainland.
SongpanGanzi Block, Longmenshan Orogenic Belt, and Sichuan Basin are located
in this area from west to east. The uplifting mechanisms of eastern Tibet are
hot debated in recent years. In addition, a series of great earthquakes in
eastern Tibet (2008 Wenchuan Mw7.9, 2013 Lushan Mw6.6, and the most recent
2017 Jiuzhaigou Mw6.5) show the urgent need for accurate seismicity
detection, as we are still not clear how aftershocks evolve because of the
poor station coverage and overlapping of aftershocks. Here,
I would like to present our studies in eastern Tibet, including seismic
anisotropy and seismicity detection. Crustal anisotropy are inversed
according to shearwave splitting of Pms phase from permanent station, and we
observed that tectonic escaping, crustal flow, and crustal shortening may
contribute to the tectonic evolution in various subareas in eastern Tibet.
We also concentrated on the seismicity detection of 2013 Lushan earthquakes,
and obtained details of the spatial and temporal aftershock evolution with
the help of matched filter technique, suggesting that afterslip is the
potential mechanism triggering Lushan aftershocks. In
order to understand more about eastern Tibet, we would keep on working in
this area by focusing on the SKS, SKKS, PKS (hereafter, XKS phase) splitting
and repeating earthquakes, which may reveal geodynamic processes in mantle
and fault slip rate respectively. 2
SpeakerF@Dr.
Mak, Sum iGerman
Research Centre for Geosciences (GFZPotsdam), Germany E Research Assistantj DateF@29 August 2017@@@TimeF@17:00 18:00 LocationF@Room D312B (Seminar Room), Institute
of Statistical Mathematics TitleF@Empirical Validation of Seismic Hazard
Models AbstractF@ Seismic
hazard, for applications such as engineering structural design and insurance
loss estimation, is represented as a probabilistic forecast. The most common
form of seismic hazard representation is in the probability for a certain
level of ground motion exceedance. The hazard also varies spatially, forming
a hazard map. As
the amount of observation accumulates, recently there are more and more
attempts to statistically evaluate the performance of probabilistic seismic
hazard prediction using ground motion observations. This talk presents the
general theory of this type of studies, using the United States Geological
Survey National Seismic Hazard Maps as an example. The 65th
SpeakerF@Prof.
Liu, JannYenq iInstitute of
Space Science, National Central University, Taiwan EProfessorj DateF@13 June 2017@@@TimeF@16:0017:00 LocationF@Room A508 (Seminar Room), Institute of
Statistical Mathematics TitleF@Statistical Analyses on seismoionospheric
disturbances and precursors of the 11 March 2011 M9.0 Tohoku Earthquake AbstractF@ Groundbased
observations of the GPS TEC (total electron content) and satellite probing of
radio occultation (RO) of FORMOSAT3/COSMIC (F3/C) are employed to study the
coseismic disturbances and precursors of the 11 March 2011 M9.0 Tohoku
earthquake. It is for the first
time the tsunami origin observed.
The horizontal propagation of seismotraveling ionospheric
disturbances (STIDs) induced by tsunami and seismic waves of the Tohoku
earthquake are observed by the GPS TEC, while the associated vertical
propagation is probed by multi groundbased observations and F3/C RO
sounding. The raytracing and
beamforming techniques are used to find the propagation and origin of the
STIDs triggered by the seismic and tsunami waves. Meanwhile, z test and the
Receiver Operating Characteristic (ROC) curve are employed to find the
characteristic of the temporal SIPs (seismoionospheric precursor) of the GIM
(global ionosphere map) TEC associated with earthquakes in Japan during
19982014. It is found that
anomalies appearing 3 days before the Tohoku earthquake well agree with the
characteristic, which suggests that the SIPs of the earthquake have been
observed. A global study on the
distribution of anomalies shows that the SIPs specifically and continuously
occur over the epicenter on 8 March 2011, 3 days prior to the Tohoku
earthquake. Finally, a physical
model of the ionosphere is used to reproduce the observed anomalies and find
possible causal of the Tohoku SIPs. The 64th
1 SpeakerF@Prof.
Jiang, Changsheng iInstitute
of Geophysics, China Earthquake Administration, China E Research Professorj DateF@29 March 2017@@@TimeF@13:30 14:30 LocationF@Room D312B (Seminar Room), Institute
of Statistical Mathematics TitleF@Assessment of earthquake monitoring
capability and score of seismic station detection capability in China Seismic
Network (2008~2015) AbstractF@ In order
to scientifically assess the earthquake monitoring capability of China
Seismic Network (CSN), we investigated the seismic observation date of CSN
with total 1001 stations considered during the period from 2008/10/01 to
2015/09/17. The distribution of seismic detection probability (PE) and the
minimum magnitude of completeness (MP) were analyzed by using the method of
"Probabilitybased magnitude of completeness" (PMC). In addition to
mapping the seismic monitoring capability for entire CSN, we developed a new
method named gseismic monitoring capability scaleh, and defined the seismic
detection capability scale Dscore to analyze the statistical characters and
spatial distribution of the seismic detection capabilities for each national
and regional stations, which based on the amplitude contour curves.
Additionally, the method of setting the "best objective function"
of seismic detection capability was used to simulate the seismic monitoring
capability improvement of CSN obtained by improving the conditions of
observation. 2
SpeakerF@Prof.
Chen, Shi iInstitute
of Geophysics, China Earthquake Administration, China E Research Professorj DateF@29 March 2017@@@TimeF@14:30 15:30 LocationF@Room D312B (Seminar Room), Institute
of Statistical Mathematics TitleF@Gravity changes before and after the
2015 Mw 7.8 Gorkha, Nepal and the 2008 Mw 7.9 Wenchuan, China earthquakes AbstractF@ Absolute
gravity measurements at four stations in southern Tibet show significant gravity
increase from 2011 to 2013, up to ~22 ΚGals at the Shigatse station. Here we
report new measurements at the Shigatse station conducted in 2016, which show
that the gravity increase ended after the 2015 Nepal Mw 7.8 earthquake.
Similar gravity changes are measured at the Pixian absolute gravimetry
station near the epicenter of the 2008 Wenchuan Mw 7.9 earthquake, where 17
absolute gravity measurements have been conducted since 2002, including four
preearthquake measurements that show ~30 ΚGals increase from 2002 to 2008.
The trend of gravity increase ended after the Wenchuan earthquake. We
analyzed the gravity effects from ground vertical motions using data from
continuous GPS stations collocated with these absolute gravimetry stations,
and surficial and hydrological processes using local hydrological data. We
found that these effects are much smaller than the observed gravity increase
before the earthquakes, and suggest that the preearthquake gravity increase
may be caused by strain and mass (fluid) transfer in broad seismic source
regions. Further studies are needed to validate such preearthquake gravity
changes, which however are difficult to be resolved from spacebased gravity
models. The 63rd
SpeakerF@Prof.
Zhou, Shiyong iSchool of Earth
and Space Sciences, Peking University, China EProfessorj DateF@18 January 2017@@@TimeF@16:0017:00 LocationF@Room A504 (Seminar Room), Institute of
Statistical Mathematics TitleF@Could the abnormal seismicity increase
triggered remotely by great earthquakes be used to judge the regional
earthquake risk? AbstractF@ We
study the possible dynamic triggering effect in Northern China, including
Tangshan area, when the Japan Tohoku M_w 9.0 earthquake happened at March
11th, 2011(In short, Japan Tohoku earthquake). We use TimeSpace Epidemic
Type Aftershock Sequence Model (TimeSpace ETAS model) as the seismicity
statistic model in this research, using Stochastic Declustering method and
Gauss Kernel function to get TimeSpace background seismicity variation image
on the target area. Thus this research may find out whether the area with
large coseismic displacement would have sudden abnormal seismicity increase.
As a result, the Japan Tohoku earthquake has little effect on the total and
background seismicity of Tangshan area, which means that the seismic
structure of Tangshan area is fundamentally stable. However, when we did
research on the possible dynamic triggering effect in Southwestern
China, we found that seismicity
on some place in Sichuan and Yunnan has sudden abnormally increased almost at
the same time when 2004 Sumatra M_w 9.2 earthquake (In short, 2004 Sumatra
earthquake) happened. That is the statistic phenomenon which shows the
existence of coseismic dynamic triggering. This research helps to find out
the exact position of the high abnormal seismicity area in its time image.
Besides, this time image can also help to detect whether this high gabnormalh
seismicity in the picture is really abnormal or is triggered by certain large
earthquake or not. The 62nd
1 SpeakerF@Dr.
Helmstetter, Agnès iInstitut
des Sciences de la Terre, France EResearch
fellowj DateF@26 October 2016@@@TimeF@15:00 16:00 LocationF@Room D312B (Seminar Room), Institute
of Statistical Mathematics TitleF@Repeating icequakes AbstractF@ We
have detected repeating icequakes on three different sites : an alpine
glacier (Argentière, massif du MontBlanc, France), near the base of the
western margin of the Greenland Ice Sheet, and on a rockglacier (Gugla, Valais,
Switzerland). Repeating icequakes are events with very similar waveforms,
located at the base of a glacier, with quasiperiodic recurrence times of the
order of minutes or hours, and progressive changes in magnitude. The activity
of each cluster is intermittent. Burstlike episodes can last for a few hours
or months, and then disappear. In greenland, temporal changes of interevent
times and magnitudes are correlated with temperature, because surface
meltwater yields an increase in basal water pressure and in glacier flow
velocity. But each cluster reacts differently to temperature changes,
probably because the connectivity to the subglacial drainage system is
different for each asperity. In contrast, we observed no correlation between
temperature and repeating icequakes at Glacier d'Argentière and at Gugla rock
Glacier. However, we observed bursts of repeating icequakes at Gugla
triggered by snow falls. We suggest that the snow weight may have induced a
transition between aseismic slip and unstable stickslip. In addition to
repeating basal icequakes, we also detected swarms of icequakes induced by
crevasse opening, probably promoted by meltwater flow. These swarms of
icequakes have very different statistical distributions in time, space and
magnitude compared with repeating icequakes. Their recurrence times are power
law distributed, their magnitudes obey the GutenbergRichter law, and the
size of each cluster is several tens of meters. These different patterns may
help to identify the triggering mechanisms of earthquake swarms, and to
discriminate between fluid flow and aseismic slip. 2
SpeakerF@Dr.
Harte, David iGNS
Science, New Zealand EStatistical
Seismologist and Hazard Modellerj DateF@26 October 2016@@@TimeF@16:00 17:00 LocationF@Room D312B (Seminar Room), Institute
of Statistical Mathematics TitleF@Determining the Uncertainty in
Earthquake Forecasts AbstractF@ Forecasts
based on a selfexciting model, like ETAS, are often produced by simulation. From
these simulations, an empirical probability distribution can be derived for a
forecast in a specified spacetimemagnitude volume. We
will show that the forecast distribution can be characterised by probability
generating functions. This shows how deeply complex the dependency structure
is in such a model. While of theoretical interest, they remain intractable to
me in a practical sense. We
then consider whether the forecast distribution can be approximated, using
less computation than that required for simulation, by a "standard
" multiparameter probability distribution. The multiple parameters
gives us the ability to at least fit a distribution with comparable mean and
variance to that of the forecast distribution. One of the main questions is
how to determine the forecast mean, and then given the mean, the variance. The 61st
SpeakerF@Dr.
Helmstetter, Agnès iInstitut
des Sciences de la Terre, France EResearch
fellowj DateF@11 October 2016@@@TimeF@16:0017:00 LocationF@Room A504 (Seminar Room), Institute of
Statistical Mathematics TitleF@Adaptive smoothing of seismicity in
time, space and magnitude for longterm and shortterm earthquake forecasts AbstractF@ We
present new methods for longterm and shortterm earthquake forecasting that
employ space, time, and magnitude kernels to smooth seismicity. These
forecasts are applied to Californian and Japan seismicity and compared with
other models. Our models are purely statistical and rely on very few
assumptions about seismicity. In particular, we do not use OmoriUtsu law.
The magnitude distribution is either assumed to follow the GutenbergRichter law or is
estimated nonparametrically with kernels. We employ adaptive kernels of
variable bandwidths to estimate seismicity in space, time, and magnitude
bins. For longterm forecasts, the longterm rate in each spatial cell is
defined as the median value of the temporal history of the smoothed
seismicity rate in this cell, circumventing the relatively subjective choice
of a declustering algorithm. For shortterm forecasts, we simply assume
persistence, that is, a constant rate over short time windows. Our longterm
forecast performs slightly better than our previous forecast based on
spatially smoothing a declustered catalog. Our shortterm forecasts are
compared with those of the epidemictype aftershock sequence (ETAS) model.
Although our new methods are simpler and require fewer parameters than ETAS,
the obtained probability gains are surprisingly close. Nonetheless, ETAS
performs significantly better in most comparisons, and the kernel model with a
GutenbergRichter law attains larger gains than the kernel model that
nonparametrically estimates the magnitude distribution. Finally, we show
that combining ETAS and kernel model forecasts, by simply averaging the
expected rate in each bin, can provide greater predictive skill than ETAS or
the kernel models can achieve individually. The 60th
SpeakerF@Dr.
Hasih Pratiwi iSebelas
Maret University, Surakarta, Indonesia E
Lecturerj DateF@30 August 2016@@@TimeF@16:0017:00 LocationF@Room A504 (Seminar Room), Institute of
Statistical Mathematics TitleF@ESTIMATING EARTHQUAKE RISK BY USING
EPIDEMIC TYPE AFTERSHOCK SEQUENCE MODEL APPROACH (Case Study in Java Island,
Indonesia) (Hasih Pratiwi and
Respatiwulan) AbstractF@ Physical
losses caused by earthquakes are death or casualties and damage to buildings
and areas. Therefore, efforts to reduce the risk of earthquake are very
necessary. Relating to risk or loss generated by earthquake it is of course
does not get out of insurance world. Insurance as nonbank financial
institution can give guarantee or protection as done by banking sector. This
research discusses a method to estimate earthquake risk by using epidemic
type aftershock sequence model. Calculation of earthquake risk can be
determined through a damage probability matrix. The information contained in
the damage probability matrix and in the damage ratios can be combined for
defining the mean damage ratio. Then, based on the estimation of intensity
function in epidemic type aftershock sequence model we can formulate the
expected annual damage ratio, and the existing method for calculating
earthquake risk is modified to obtain earthquake insurance premium rates. We
use earthquakes data in Java Island obtained from U.S. Geological Survey
which consists of time of occurrence, longitude, latitude, magnitude, depth,
and catalogue source. The time span of this research is from January 1, 1973,
to December 31, 2010. Zonation map of earthquake generated in this research
is different from the zonation map SNI 2010 issued by Indonesian Ministry of
Public Works. The difference lies on the distribution of earthquake zone,
especially in regencies and cities with high risk. The earthquake insurance
premium rates for high and medium intensities obtained from this research are
significantly greater than the premium rates issued by PT Reasuransi Maipark
Indonesia. The current premium rates are relatively small when compared with
the rates in Turkey and from this research. Keywords:
earthquake insurance, intensity function, epidemic type aftershock sequence
model, damage probability matrix. The 59th
SpeakerF@Prof.
Chen, YuhIng iInstitute
of Statistics, National Central University, Taiwan E Distinguished Professorj DateF@19 July 2016@@@TimeF@16:0017:00 LocationF@Room D312A (Seminar Room), Institute
of Statistical Mathematics TitleF@Statistical evaluation of shortterm
hazard of earthquakes after 1999 M 7.3 ChiChi shock in Taiwan AbstractF@ The temporalspatial
hazard of the earthquakes in a continental region of Taiwan after the 1999
September 21 MW =7.7 ChiChi shock is investigated. The ReasenbergJones (RJ)
model (Reasenberg and Jones, 1989) that combines the frequencymagnitude
distribution (Gutenberg and Richter, 1944) and timedecaying occurrence rate
(Utsu et al., 1995) is conventionally employed for assessing the earthquake
hazard after a large shock (Wiemer, 2000). However, it is found that the b
values in the frequencymagnitude distribution of the earthquakes in the
studyregion dramatically decreased from background values after the ChiChi
shock, and then gradually increased up. The observation of a timedependent
distribution of magnitude motivated us to propose a modified RJ model (MRJ)
to assess the earthquake hazard (Chen et al. 2015). To incorporate the
possible impact of previous large earthquakes on thefollowing ones, a
simplified epidemictype aftershock sequence (ETAS) model (Ogata, 1988, Ogata
and Zhunag, 2006) is further considered. A modified ETAS (METAS) model that
combines the simplified ETAS model and the timedependent distribution of
magnitude is then suggested for the hazard evaluation. The MRJ and METAS
models are further separately used to make oneday forecast of large
earthquakes in the study region. To depict the potential rupture area for
future large earthquakes, we also develop the spacetime MRJ and METAS models
and construct the corresponding relative hazard (RH) maps. The Receiver
Operating Characteristics (ROC) curves (Swets, 1988) demonstrate that the RH
map based on the MRJ model is as good as the one based on the METAS model for
exploring the spatial hazard of earthquakes in a short time after the ChiChi
shock. The 58th
SpeakerF@Dr.
Zhuang, Jiancang iAssociate Prof. of ISMj DateF@29 June 2016@@@TimeF@16:0016:40 LocationF@Room D313ED314 (Seminar Room), Institute of
Statistical Mathematics @ISM@Statistical
Mathematics Seminar 2016 TitleF@Replenishing missing data in the
observation record of mark point processes AbstractF@ This
presentation illustrates a fast approach for replenishing missing data in the
record of a temporal point process with time independent marks. The basis of
this method is that, if such a point process is completely observed, it can
be transformed into a homogeneous Poisson process by using a biscale
empirical transformation. This approach includes three key steps: (1) Obtain
the transformed process by using the empirical transformation and find a
timemark range that likely contains missing events; (2) Estimate a new
empirical distribution function based on the data in the timemark range
inside which the events are supposed to be completely observed; (3) Generate
events in the missing region. This method is tested on a synthetic dataset
and applied to the data missing problem in the JMA record of the Kumamoto
aftershock sequence, occurring from 2016415 in Japan. The influence of
missing data on the MLE of the ETAS parameters is studied by comparing the
analysis results on the original and replenished datasets. The results show
that the MLEs of the ETAS parameters vary when the ETAS model is fitted to
the recorded catalog with different cutoff magnitudes, while when the
replenished dataset is used the MLE of the ETAS parameters keep stable. The 57th SpeakerF@Dr.
Shcherbakov, Robert iDepartment
of Earth Sciences, Western University(Ontario), Canada E Associate Professor DateF@22 June 2016@@@TimeF@16:4017:20 LocationF@Room D313ED314 (Seminar Room), Institute of
Statistical Mathematics @ISM@Statistical
Mathematics Seminar 2016 TitleF@Statistics and physics of aftershocks AbstractF@ Aftershocks
are ubiquitous in nature. They are the manifestation of relaxation phenomena
observed in various physical systems. In the studies of seismicity,
aftershock sequences are observed after moderate to large main shocks.
Empirical observations reveal that aftershocks obey powerlaw scaling with
respect to their energies (seismic moments) which in magnitude domain can be
modelled by the GutenbergRichter law. The decay rate of aftershocks above a
certain magnitude is typically inversely proportional to the time since the
main shock and is approximated by the modified Omori law. The largest
aftershocks in a sequence constitute significant hazard and can inflict
additional damage to infrastructure that is already affected by the main
shock. Therefore, the estimation of the magnitude of a possible largest
aftershock in a sequence is of high importance. In this presentation, a
Bayesian predictive distribution and the corresponding confidence intervals
for the magnitude of the largest expected aftershock in a sequence are
derived using the framework of Bayesian analysis and extreme value
statistics. The analysis is applied to several wellknown aftershock
sequences worldwide to construct retrospectively the confidence intervals
for the magnitude of the subsequent largest aftershock by using the
statistics of early aftershocks in the sequences. In order to infer the
physical mechanisms of triggering and time delays responsible for the
occurrence of aftershocks, a nonlinear viscoelastic sliderblock model is
considered. It is shown that nonlinear viscoelasticity plays a critical role
in the triggering of aftershocks. The model reproduces several empirical laws
describing the statistics of aftershocks, which are observed in the studies
of systems with relaxation dynamics, specifically, for earthquakes. The 56th
SpeakerF@Dr.
Strader, Anne iGFZ German
Research Centre for Geosciences, Germany E
Postdoctoral fellowj DateF@7 June 2016@@@TimeF@16:0017:00 LocationF@Room A508 (Seminar Room), Institute of
Statistical Mathematics TitleF@Evaluation of Current CSEP Testing
Methods: Case Studies for Japan and California AbstractF@ The
Collaboratory for the Study of Earthquake Predictability (CSEP) was developed
to rigorously test earthquake forecasts retrospectively and prospectively
through reproducible, completely transparent experiments within a controlled
environment (Zechar et al., 2010).
Forecasts are individually evaluated using a set of likelihoodbased
consistency tests, which measure the consistency between the number, spatial
and magnitude distribution of the observed and forecasted seismicity during
the testing period (Schorlemmer et al., 2007; Zechar et al., 2010).
Additionally, the classical paired ttest and nonparametric wtest are used
to directly compare two forecasts' performances at target earthquake
locations. These tests rely on a hypothesis testing framework, resulting in a
final decision (to reject or not reject a forecast), rather than quantifying
the model's lackoffit or localized performance. Residual methods are
employed by the CSEP to discern spatial variation in model performance
compared to the observed seismicity distribution and other models, but are
not currently incorporated into decisionmaking processes. To illustrate what
can be learned from commonly utilized current CSEP tests, we present two case
studies. The first is a retrospective evaluation of a rateandstate forecast
for the Japan CSEP testing classes, where spatiotemporal seismicity rate
fluctuations are inverted for Coulomb stress changes. Although the model
underestimates the number of earthquakes following the M9.0 Tohoku mainshock,
it displays positive information gain over baseline ETAS seismicity rates
(Ogata, 2011) within the rupture region. The second forecasting experiment is
a continued prospective evaluation of the timeindependent California
earthquake forecasts tested in the Regional Earthquake Likelihood Model
(RELM) experiment, from 20112016. Additionally, we test two models developed
by the United States Geological Survey (USGS): the timedependent Uniform
California Earthquake Rupture Forecast (UCERF2) and timeindependent National
Seismic Hazard Mapping Project (NSHMP) models. To reduce bias from expertbased
decision making utilized in current testing methods, we introduce the
framework of a Dynamic Risk Quantification (DRQ) platform, that will be
developed to combine and optimize ensemble forecasts and hazard models using
a datadriven approach, and updated as new data become available. The 55th
1 SpeakerF@Dr.
Guo, Yicun iSchool of
Earth and Space Sciences, Peking University, China E PhD studentj DateF@22 March 2016@@@TimeF@16:00 17:00 LocationF@Room D312B (Seminar Room), Institute
of Statistical Mathematics TitleF@Iterative finiteETAS model and some results
of the histETAS model of the North China Craton AbstractF@ We
introduce a iterative algorithm to refine the finite sources of main shocks
in the finite ETAS model, in which the weight of triggering ability for each
subfault is its productivity divided by the whole productivity of the main
shock. Also we apply histETAS model to North China Craton. It turns out that
the b value and background seismicity patterns coincide with the static
coulomb stress change induced by historical big earthquakes, and p value
variation in space is in agreement with velocity structure of the lithosphere
under major fault zones. Therefore we infer the statistical characteristics
of seismicity reflect the properties of medium to some extent, and make some
discussion of future earthquake hazard. 2
SpeakerF@Prof.
Ogata, Yosihiko iEmeritus
Professor of ISM; Project Researcher of Earthquake Research Institute,
University of Tokyoj DateF@22 March 2016@@@TimeF@17:00 18:00 LocationF@Room D312B (Seminar Room), Institute of
Statistical Mathematics TitleF@3D spatial models for seismicity
beneath Kanto region AbstractF@ Development
of pointprocess models for the seismicity in 3D space (longitude, latitude
and depth) beneath Kanto area down to 100km depth is more required than for
seismicity in the rest of the world. This is because the three tectonic
plates meet beneath Kanto plain; and interactions among the interplate and
intraplate earthquakes are too complex to make detailed analysis and
forecasts in 2D space that ignores the depths. We
consider the 3D hierarchical spacetime ETAS (epidemictype aftershock
sequence) model. Among the characterizing parameters, the background
seismicity rate \mu and aftershock productivity K are highly sensitive to the
locations, so that these parameters should be locationdependent.
Furthermore, the impact of the 2011 TohokuOki earthquake of M9.0 to the
seismicity beneath the Kanto region has been so large that we need a
spacetime function for representing the amount of the induced seismicity
beneath Kanto by this giant earthquake. Specifically, we adopt the OmoriUtsu
function as the effect of induced earthquakes, started after the occurrence
time of the TohokuOki earthquake, where we assume that the aftershock
productivity parameter KM9 of the OmoriUtsu function is also
locationdependent. For forecasting future large earthquakes, we further need
to estimate the locationdependent bvalue of the GutenbergRichter law. The
spatial variations of the characteristic parameters \mu(x,y,z), K(x,y,z) ,
KM9(x,y,z) and b(x,y,z) of our model are inverted to visualize the regional
changes of the seismic activity. For this objective, we make 3D Delaunay
tessellation of the Kanto volume, where every earthquake belongs to vertices
of a tetrahedron. Each of the above mentioned parameter function is a
3dimensional piecewise linear function defined by the values at the four
Delaunay tetrahedral vertices. The
estimates of the focal parameter functions are obtained by an optimal
tradeoff between the goodness of fit to the earthquake data and the
smoothness constraints (or roughness penalties) of the variations of
parameter values. Strengths of the constraints of or the penalties to
respective parameter functions can be simultaneously adjusted from the data
by means of an empirical Bayesian method using the Akaikefs Bayesian
information criterion (ABIC). Key
words: ABIC, aftershock productivity, background seismicity rate, bvalues,
Delaunay function, Delaunay tessellation, empirical Bayesian method,
OmoriUtsu function for induced seismicity, penalized loglikelihood. The 54th
SpeakerF@Dr.
Wang, Ting iDepartment
of Mathematics and Statistics, University of Otago, New Zealand ELecturerj
DateF@9 February 2016@@@TimeF@16:00 17:00 LocationF@Room D312B (Seminar Room), Institute
of Statistical Mathematics TitleF@Identification of seismic phases using
Markovmodulated marked Hawkes processes AbstractF@ Based
on a temporal Markovmodulated Hawkes process that we developed earlier to
investigate longterm patterns of seismic activity with multiple mainshocks,
we made extensions to this temporal model to include spatial variation of the
seismic activity and the earthquake magnitudes. Our aim is to categorize
spatiotemporal seismic hazards holistically, using the entire earthquake
record in a selected region to identify patterns correlated with subsequent
large earthquakes, rather than the traditional way of selecting individual
foreshockmainshock or mainshockaftershock sequences. I will use several
case studies to illustrate how this model works and discuss about the
problems that we had with the model fitting. The 53rd
1 SpeakerF@Dr.
Yin, Fengling iInstitute
of Geophysics, China Earthquake Administration, China EAssistant Professorj DateF@27 January 2016@@@TimeF@13:30 14:30 LocationF@Room D312B (Seminar Room), Institute
of Statistical Mathematics TitleF@Coulomb stress evolution along the
middle segment of Redriver fault zone over the past 180 Years due to
coseismic, postseismic and interseismic deformation iYin,Fengling, Jiang, Changsheng and
Han, Liboj AbstractF@ The
Redriver fault zone, for it being as the boundary of SichuanYunnan rhombic
block and southeastern margin of the Tibetan plateau, and near the Central
Yunnan city group, its seismic activity deserves attention. The Redriver
fault zone within Yunnan has experienced at least 9 earthquakes of M≥6 in
recent 180 years. Using stratified viscoelastic lithospheric model, we
calculate the coulomb failure stress evolution along the redriver fault zone
over the past 180 years due to coseismic, postseismic and interseismic
deformation. By analyzing 25 earthquakes occurred along the Redriver fault
zone and ajacent faults, we find that the middle segment of Redriver fault
zone remains low seismic activity in recent two hundred years. This is
consistent with the observed eseismic gapf as earthquake catalog shows.
Assuming there is no earthquake within about 30 years around the Redriver
fault zone, this fseismic gapf may remain due to postseismic and interseismic
deformation. 2
SpeakerF@Dr.
Taroni, Matteo iIstituto
Nazionale di Geofisica e Vulcanologia, Rome, Italy E Postdoctoral fellowj DateF@27 January 2016@@@TimeF@14:30 15:30 LocationF@Room D312B (Seminar Room), Institute
of Statistical Mathematics TitleF@Some recent techniques to improve
earthquake forecasting (Taroni, Matteo, Marzocchi, Warner, Zechar, Jeremy and
Werner, Maximilian) AbstractF@ In
this presentation I will show some recent results regarding the earthquake
forecasting techniques. In particular I will show: i)
How to consider aftershocks and foreshocks in the seismic hazard computation,
with an application to the Italian case. ii)
How to merge different catalogues to obtain a better estimation of the
Tapered GutenbergRichter distribution parameters, with an application to the
global and Italian case. iii)
How to create an ensemble model to improve the performance of the shortterm
earthquake forecasting models, with an application to the New Zealand case. The 52nd SpeakerF@Dr.
Guillas, Serge iDepartment
of Statistical Science, University College London, U.K. E Readerj
DateF@17 November 2015@@@TimeF@16:0017:00 LocationF@Room A508 (Seminar Room), Institute of
Statistical Mathematics TitleF@Dimension reduction for the
quantification of uncertainties in tsunami and climate models AbstractF@ VOLNA,
a nonlinear shallow water equations solver, produces high resolution
simulations of earthquakegenerated tsunamis. However, the uncertainties in
the bathymetry (from irregularlyspaced observations) have an impact on
tsunami waves. We first employ a stochastic partial differential equation
(SPDE) approach to quantify uncertainties in these boundary fields. These
uncertainties are then parametrized to be used as inputs of an emulator of
VOLNA. However, the dimension of these boundary fields is large and must be
reduced. We apply the gradientbased kernel dimension reduction approach
(gKDR) by Fukumizu and Leng (2014) and construct an Gaussian Process emulator
on this reduced input space. We propagate uncertainties in the bathymetry to
obtain an improved probabilistic assessment of tsunami hazard. In a
separate climate application, we employ the Bayesian calibration of complex
computer models using Gaussian Processes, introduced by Kennedy and O'Hagan
(2001), that has proven to be effective in a wide range of applications.
However, the size of the outputs, such as climate models' spherical outputs,
leads to computational challenges in implementing this framework. Covariance
models for data distributed on the sphere also present additional challenges
compared to covariance models for data distributed over an Euclidean space.
To overcome these various challenges, we make use of the spherical harmonics
(SHs) decomposition of the computer model output, and then apply a Gaussian
process assumption to the coefficients in the decomposition. Furthermore,
using the SPDE approach, we can capture nonstationarity in the spatial
process. Hence, we generalize further the spherical correlation framework by
expanding the SPDE parameters used to quantify the nonstationary behavior in
the functional space spanned by the SHs. We illustrate our findings on
several synthetic examples. In particular, our method can outperform the
calibration based on principal components. Finally we show that our technique
has the potential to calibrate the Whole Atmosphere Community Climate Model
(WACCM). The 51st
1 SpeakerF@Dr.
Gerstenberger, Matthew iGNS
Science, New Zealand ERisk and Engineering
Team Leader, Senior Seismologistj
DateF@1 September 2015@@@TimeF@16:0017:00 LocationF@Room D312B (Seminar Room), Institute
of Statistical Mathematics TitleF@The New Zealand National Seismic
Hazard Model: Rethinking PSHA AbstractF@ We
are currently revising the New Zealand National Seismic Hazard Model. In this
revision we are exploring some of the fundamental assumptions of the model
and investigating how uncertainties in earthquake source and ground motion
estimation propagate through to the end uses of the model. Uncertainties
related to the source modelling that come from a paucity of data and from
different methods that can be used to model the seismic sources are currently
not fully quantified in the way we model seismic hazard. Additionally,
seismic sources are generally assumed to be a stationary Poisson process and
earthquake clustering is ignored. Including these uncertainties in the way
risk is modelled based on the outputs of the National Seismic Hazard Model
will likely lead to more robust estimates of risk for use by industry and in
the development of design standards. Notice:
This email and any attachments are confidential. If received in error please
destroy and immediately notify us. Do not copy or disclose the contents. 2
SpeakerF@Dr.
Chen, Shi iInstitute
of Geophysics, China Earthquake Administration, China EAssociate Professorj DateF@1 September 2015@@@TimeF@17:0018:00 LocationF@Room D312B (Seminar Room), Institute
of Statistical Mathematics TitleF@A study on the regional gravity
changes before large earthquakes from the statistical perspectives AbstractF@ The
repeated gravity surveys, also called mobile gravity measurements, have been
carried out for decades in the Chinese mainland. Significant gravity changes
have been detected before some cases of great earthquakes, such as the 1976
Tangshan Ms7.8 Earthquake, 2008 Wenchuan Ms8.0 earthquake, etc. The main aim
of the repeated gravity surveys is to monitor the geophysical field
variations in some major seismic hazard zones. By this sort of insitu
gravimetric network, the yearly changes of regional gravity can be obtained.
Through the Molchan Error Diagram tests, we found that observed gravity
changes are statistically correlated to the occurrence of future large
earthquakes, i.e., the gravity changes are more powerful than a seismicity
rate model in forecasting large earthquakes. These results imply that gravity
changes before earthquake include precursory information of future large
earthquakes. Key
words: Gravity changes, Earthquake prediction, Molchan error diagram,
Repeated gravity measurement, Chinese mainland. The 50th
SpeakerF@Dr.
Kagan, Yan Y. iDepartment
of Earth and Space Sciences, University of California, Los Angeles (UCLA) E Researcherj DateF@4 August 2015@@@TimeF@16:0018:00 LocationF@Room D208 (Conference room), Institute
of Statistical Mathematics TitleF@Statistics of earthquake focal
mechanisms AbstractF@ I.
Doublecouple earthquake source: symmetry and rotation We
analyze earthquake focal mechanisms and their forecast both analytically and
statistically. This problem is complex because source mechanisms are
tensorvalued variables, thus their analysis requires applying sophisticated
mathematical and statistical tools, many of which are not yet fully
developed. We describe general and statistical properties of the seismic
moment tensor, in particular, its most important form  the doublecouple
(DC) mechanism. We establish a method for the analysis of a DC source, based
on the quaternion technique, and then apply quaternions for the statistical
analysis of earthquake catalogs. The important property of the focal
mechanism is its symmetry. We describe the classification of the mechanism
symmetry and the dependence of the DC orientation on its symmetry. Four
rotations exist in a general case of a DC with the nodalplane ambiguity,
there are two transformations if the fault plane is known, and there is one
rotation if the sides of the fault plane are known. A statistical analysis of
symmetrical objects has long been the subject of crystallographic texture
investigations. We describe the application of crystallographic methods to
focal mechanism analysis and consider theoretical statistical distributions
appropriate for the DC orientation approximation. Uniform random rotation
distributions for various DC sources are discussed, as well as two
nonuniform distributions: the rotational Cauchy and von MisesFisher. We
discuss how the parameters of these rotations can be estimated by a statistical
analysis of earthquake source properties in global seismicity using the GCMT
catalog. We also show how earthquake focal mechanism orientations can be
displayed on the Rodrigues vector space. II.
Statistical earthquake focal mechanism forecasts In
the focal mechanism forecast, the sum of normalized seismic moment tensors
within a 1000 km radius is calculated and the P and Taxes for the predicted
focal mechanism are evaluated on the basis of the sum (Kagan and Jackson
1994, JGR). Simultaneously we calculate an average rotation angle between the
forecasted mechanism and all the surrounding mechanisms. This average angle
shows tectonic complexity of a region and indicates the accuracy of the
prediction. Recent interest by CSEP and GEM has motivated some improvements,
particularly a desire to extend the previous forecast to polar and nearpolar
regions. The major problem in extending the forecast is the focal mechanism
calculation on a spherical surface. In a modified program focal mechanisms
have been projected on a plane tangent to a sphere at a forecast point. A
comparison with the old 75S75N forecast shows that in equatorial regions the
forecasted focal mechanisms are almost the same, and the difference in the
forecasted focal mechanisms rotation angle is close to zero. However, closer
to the 75 latitude degree the difference in the rotation angle is large
(around a factor 1.5 in some places). The Gammaindex was calculated for the
average focal mechanism moment. A nonzero Index indicates that earthquake
focal mechanisms around the forecast point have different orientations. Thus
deformation complexity displays itself both in the average rotation angle and
in the Index. However, sometimes the rotation angle is close to zero, whereas
the Index is large, testifying to a large CLVD presence. Both new 0.5x0.5 and
0.1x0.1 degree forecasts are posted at
http://eq.ess.ucla.edu/~kagan/glob_gcmt_index.html III.
Evaluating focal mechanisms forecast skill We
discuss the ways to test the focal mechanism forecast efficiency. We start
with several verification methods, first based on adhoc, empirical
assumptions. However their performance is questionable. In the new work we
apply a conventional likelihood method to measure the skill of a forecast.
The advantage of such an approach is that an earthquake rate prediction can
be adequately combined with a focal mechanism forecast, if both are based on
the likelihood scores. This results in a general forecast optimization. To
calculate the likelihood score we need to compare actual forecasts or
occurrences of predicted events with the null hypothesis that the mechanism's
3D orientation is random. To better understand the resulting complexities we
calculate the information (likelihood) score for two rotational distributions
(Cauchy and von MisesFisher), which are used to approximate earthquake
source orientation patterns. We then calculate the likelihood score for
earthquake source forecasts and for their validation by future seismicity
data. Several issues need to be explored when analyzing observational
results: their dependence on the forecast and data resolution, internal
dependence of scores on the forecasted angle, and random variability of
likelihood scores. We propose a preliminary solution to these complex problems,
as these issues need to be explored by a more extensive theoretical and
statistical analysis. IV.
Future focal mechanisms studies 1.
Statistical earthquake focal mechanism forecasts, rigorous likelihood methods for
evaluating forecast skill. 2.
Likelihood analysis of GCMT catalog, including focal mechanisms. 3.
Focal mechanism clustering. 4.
Collapsing focal mechanism patterns. 5.
Influence of Earth surface on focal
mechanisms interaction (Morawiec,Ch8). 6.
Integrating Cauchy distribution on Rodrigues space, Morawiec, pp.116119. 7.
Calculating Cauchy and von MisesFisher distributions for 120 degrees
rotation limit. 8.
Investigating the signchange symmetry of a DC earthquake source. 9.
Studies of statistics of earthquake focal mechanisms in the Rodrigues space. 10.
Rotationtranslation distribution of doublecouples as different from
arbitrary symmetric deviatoric secondrank tensor. The 49th
SpeakerF@Prof.
Künsch, Hans R. iDepartment
of Mathematics, ETH Zurich, Switzerland E
Emeritus Professorj
DateF@7 April 2015@@@TimeF@16:0017:00 LocationF@Room D312B (Seminar Room), Institute
of Statistical Mathematics TitleF@Data assimilation in seismology ? AbstractF@ The
accuracy of weather forecasts has increased substantially over the past
decades. This is due to at least three factors: The increase in computing
power which allows a higher accuracy in solving the equations of the
underlying physical model, a denser set of observations of the state of the
atmosphere and better methods for data assimilation, that is the use of these
observations to adjust the initial conditions of the model sequentially. In
order to represent the uncertainty in assimilation and forecasting, ensembles
are used whose members represent different states of the atmosphere that are
compatible with the observations and the physical dynamics. Such ensemble
methods should be viewed as Monte Carlo methods which provide the link to
statistics. Recently
there has been interest in using these data assimilation tools also in
seismology in order to improve forecasts and quantify their uncertainty. In
this talk I will discuss some of the attempts in this direction. Time
permitting, I will also present some new ideas for ensemble data
assimilation. The 48th
SpeakerF@Dr.
Ishibe, Takeo iEarthquake
Research Institute, The University of Tokyo E
Project Researcherj DateF@24 February 2015@@@TimeF@16:0017:00 LocationF@Room A508 (Seminar Room), Institute of
Statistical Mathematics TitleF@Overview of Seismicity Changes in
Inland Japan after the 2011 TohokuOki Earthquake and Its Interpretation AbstractF@ In
this presentation, I overview the widespread changes in seismicity rate and
distribution of focal mechanism after the TohokuOki earthquake (Mw9.0) and
summarize the possible contributing factors. In Tohoku, westward from the
TohokuOki source, significant increases in seismicity rate were observed in
N. and S. Akita, SW off Oga peninsula, and Yamagata/Fukushima and
Ibaraki/Fukushima boundary regions as well as other surrounding areas. On the
other hand, aftershock activities in the source regions of recent large
earthquakes such as the 2008 IwateMiyagi earthquake have been suppressed. In
Kanto, southwest of the TohokuOki source, interplate earthquakes were
typically activated, while beltlike seismicity along the western edge of
slabslab contact zone and shallow earthquakes in some areas were also
activated. The
most plausible factor is the static changes in the Coulomb stress, which
seems to be valid for retrospectively forecasting the changes in seismicity
on some level, while some activated seismicity showed clear counterevidence.
Remotely triggered local events, whose origin times are well coincided with
the arrivals of mainshock seismic waves, suggest that dynamic stress changes
also contribute. Some swarmlike activities, showing temporal expansion of
the focal area which is attributed to fluid diffusion, suggest that changes
in pore fluid pressure should be another possible factor. The contribution of
indirectly triggered earthquakes might be important in some regions because
stress changes imparted by neighboring indirect aftershocks could be
comparable with those from a distant mainshock. Postseisimc slip and
viscoelastic effect would play an important role for longterm hazard
assessments. The 47th
SpeakerF@Dr.
Segou, Margaret iNational
Observatory of Athens (Greece) E
Researcherj DateF@10 February 2015@@@TimeF@16:0017:00 LocationF@4F Lounge, Institute of Statistical
Mathematics TitleF@The Future of Earthquake
Predictability AbstractF@ The
last decade dense seismological networks around the world provide the
opportunity to study more aftershock sequences in seismically active areas
across the world such as California (San Andreas Fault), Japan, New Zealand
(Canterbury Fault, Christchurch) and continental rift systems (Corinth Gulf,
Greece). The importance behind that is evident; the 2008 M7.9 Sichuan event
continues having catastrophic aftershocks (2013 Lushan M6.6) after five
years. The above provide the necessary motivation for geophysicists to
develop short and longterm earthquake forecasts for providing to scientists
and the public authoritative information on seismic hazard and answer
ultimately the question When the next big earthquake will occur. Static and
dynamic triggering are often described as the two primary mechanisms for
earthquake clustering in time and space. Recent studies have provided
evidence that physicsbased earthquake forecast models, combining fault aging
laws and the static stress triggering hypothesis, can accurately predict
(80%) transient seismicity rates. Static triggering plays an important role
in spatial clustering at distances 23 rupture lengths away from the seismic
source whereas dynamic triggering studies usually focus on larger distances
(>1000 km). But how dependent are our calculations on our incomplete
knowledge of the ambient stress of a region? What are the implications behind
the time dependent fault behavior? The last two questions are the key for
reducing the uncertainties of physical forecast models. Quite often the
development of such quantitative and testable models is followed by extensive
statistical performance evaluation, which is critical for understanding their
merits and pitfalls. In
this seminar I focus on recent development on physicsbased earthquake models
using worldwide examples and how they compare with statistical models.
Furthermore, I discuss how we can reduce their uncertainties and sketch the
future of our scientific predictability. Is it possible to hope on higher
information gains in the near future? and, How these forecast models could be
most effective in Japan? The 46th
1 SpeakerF@Dr.
Kumazawa, Takao iProject
Researcher of ISMj DateF@27 January 2015@@@TimeF@15:0016:00 LocationF@Room D312B (Seminar Room), Institute
of Statistical Mathematics TitleF@Predicting Offshore Swarm Rate by
Volumetric Strain Changes in Izu Peninsula, Japan AbstractF@ The
eastern offshore of Izu peninsula is one of the well known volcanic active
regions in Japan, where magma intrusions have been observed several times
since 1980s monitored by strainmeters located nearby. Major swarm activities
have been synchronously associated with coseismic and preseismic significant
sizes of volumetric strain changes (Earthquake Research Committee,
2010). We investigated the
background seismicity changes during these earthquake swarms using the
nonstationary ETAS model (Kumazawa and Ogata, 2013, 2014), and have found the
followings. The volumetric strain change data, modified by removing the
effect of earth tides and precipitation as well as removing coseismic jumps,
have much higher crosscorrelations to the background rates of the ETAS model
than to the whole seismicity rate change of the ETAS. Furthermore the strain
changes precede the background seismicity by lag of about half a day. This
relation suggests an enhanced prediction of earthquakes in this region using
volumetric strain measurements. Thus we propose an extended ETAS model where
the background seismicity rate is predicted by the time series of preceding
volumetric strain changes. Our numerical results for Izu region show
consistent outcomes throughout the major swarms. 2
SpeakerF@Dr.
Wang, Ting iDepartment
of Mathematics and Statistics, University of Otago, New Zealand ELecturerj
DateF@27 January 2015@@@TimeF@16:3017:30 LocationF@Room D312B (Seminar Room), Institute
of Statistical Mathematics TitleF@Marked point process modeling with
missing data in volcanic eruption records AbstractF@ Despite
ongoing efforts to compile new data, eruption records, particularly those of
earlier time periods, are pervasively incomplete. The probability of missing
an ancient eruption is much higher than a recent eruption. We consider
modeling both the times and sizes of the eruptions using a marked point
process. We propose to model the marks (the sizes of the events) as having a
timevarying distribution which takes the higher proportion of missing
smaller events in earlier records into consideration. We then estimate the
proportion of detected events over time based on the assumption that the most
recent record is complete and that the record of eruptions with the largest
size in the considered catalog is complete. With this information, we can
then estimate the true intensity of volcanic eruptions. The 45th
SpeakerF@Dr.
Schehr, Grégory iLaboratoire
de Physique Théorique et Modèles Statistiques, OrsayUniversity ParisSud E CNRSResearcherj DateF@14 October 2014@@@TimeF@11:0012:00 LocationF@Room D312B (Seminar Room), Institute
of Statistical Mathematics TitleF@Exact Statistics of the Gap and Time
Interval Between the First Two Maxima of Random Walks and Lévy Flights AbstractF (PDF)
The 44th
1 SpeakerF@Prof.
Zhou, Shiyong iSchool of
Earth and Space Sciences, Peking University, China EProfessor; Visiting Professor of ISM,
2014j DateF@5 August 2014@@@TimeF@16:0017:00 LocationF@Room D312B (Seminar Room), Institute
of Statistical Mathematics TitleF@Seismicity simulation in Western
Sichuan of China based on the fault interactions and its implication on the
estimation of the regional earthquake risk AbstractF@ Seismicity
over 10000 years in Western Sichuan of China has been simulated based on the
mechanical synthetic seismicity model we developed. According to the analysis
of the simulated synthetic seismic catalogue , the occurrence of strong
earthquakes with Ms ≥710 in the whole region of Western Sichuan is rather
random , very close to the Poisson process with seismic rate 010454Pyear , which
means it is reasonable to estimate the regional earthquake risk with Poisson
model in Western Sichuan. However, the occurrence of strong earthquakes with
Ms ≥710 on the individual faults of Western Sichuan is far from Poisson process
and could be predicted with a time2dependent prediction model. The fault
interaction matrices and earthquake transfer possibility matrices among the
faults in Western Sichuan have been calculated based on the analysis of the
simulated synthetic catalogues. We have also calculated the static change in
Coulomb failure stress (CFS) on one fault induced by a strong earthquake on
another fault in Western Sichuan to discuss the physical implications of the
earthquake transfer possibility matrices inferred from the synthetic catalogue. Keywords:
Simulation of earthquake generation , Poisson model , Coulomb stress ,
Seismic hazard 2
SpeakerF@Dr.
Wang, Dun iEarthquake
Research Institute, University of Tokyo E
JSPS Postdoctoralj DateF@5 August 2014@@@TimeF@17:0018:00 LocationF@Room D312B (Seminar Room), Institute
of Statistical Mathematics TitleF@Rupture speeds for recent large
earthquakes AbstractF@ Studying
the rupture speeds of earthquakes is of broad interesting for earthquake
research because it has a large effect on the strong nearfield shaking that
causes damage during earthquakes. Also rupture speed is a key observation for
understanding the controlling stresses and friction during an earthquake, yet
the speed and its variations are usually difficult to determine. Using only
farfield seismic waveforms, which is the only data available for many large
earthquakes, there are problems for estimating the rupture speed with
standard waveform inversions, due to tradeoff between the rupture speed and
the slip location. Here
we applied a back projection method to estimate the rupture speeds of Mw ≥
7.5 strikeslip earthquakes since 2001 which could be analyzed using Hinet
in Japan. We found that all events had very fast average rupture speeds of
3.06.0 km/s, which are near or greater than the local shear wave velocity
(supershear). These values are faster than for thrust and normal faulting
earthquakes that generally rupture with speeds of 1.03.0 km/s. Considering
the depthdependent shearwave velocity, the average propagation speeds for
all of the strikeslip events are closer to or greater than the shear wave
velocity. For large strikeslip events, transition from subshear to
supershear usually occurs within distances of 15 to 30 km from the
initiation, which is probably the reason for the scarcity of observed
supershear earthquakes for smaller magnitudes. Earthquakes
with supershear ruptures can cause more damage than events with subshear
ruptures because of the concentration of energy in the forward direction of
the rupture. Numerical modeling shows strong focusing and other effects of
energy at the rupture front which can intensify the ground motions. A recent
example is the April 13, 2010 Qinghai, China earthquake (Mw 6.9), where a
moderatesize event caused extensive damage in the Yushu region at the
southeastern end of the fault. Careful evaluation of long and straight
strikeslip faults should be emphasized for predicting strong ground motions
due to supershear rupture. The 43rd
SpeakerF@Dr.
Aiken, Chastity iGeorgia
Institute of Technology, U.S.A. E
National Science Foundation Graduate Fellow, ARCS Scholarj
DateF@8 July 2014@@@TimeF@16:0018:00 LocationF@Room D312B (Seminar Room), Institute
of Statistical Mathematics TitleF@Triggered Seismic Activity in Geothermal
Regions and on StrikeSlip Faults AbstractF@ Dynamic
stresses caused by large earthquakes are capable of triggering a wide range
of seismic/aseismic responses at remote distances. These responses include
instantaneously triggered microearthquakes, deep tectonic tremor, earthquake
swarms, slowslip events, and nearsurface icequakes. Systematic studies of these triggered
phenomena not only help us to understand how large earthquakes affect
seismic/aseismic processes at remote distances but also help improve our
understanding of the necessary physical conditions responsible for the
generation of seismic activity.
In this talk I present two recent studies: (1) a comparison of triggered
microearthquakes in three geothermal regions of California and (2) a
comparison of triggered tremor on four strikeslip faults in the Western
Hemisphere. Triggered seismic
activity is characterized as being triggered by the surface waves of large,
distant earthquakes. Triggered
earthquakes in geothermal regions are generally small magnitude (M<4) and
have distinct P and S waves, whereas triggered tremor is a lowamplitude,
emergent signal with no distinct P wave.
After identifying the large earthquakes that trigger seismic activity,
we analyze and compare the peak ground velocities, seismic wave incidence
angles, amplitude spectra of all distant earthquakes we examined, as well as
the background activity in each region to determine the factors that promote
triggering in geothermal regions and on strikeslip faults. The 42nd
SpeakerF@Dr.
Aranha, Claus iGraduate
School of Systems and Information Engineering, University of Tsukuba E Assistant Professorj
DateF@27 May 2014@@@TimeF@16:0017:00 LocationF@Room D312B (Seminar Room), Institute of
Statistical Mathematics TitleF@Using Evolutionary Algorithms to
optimize earthquake risk models: Early Ideas AbstractF@ Evolutionary
Algorithms are a class of metaheuristics that use genetic principles to
sample good solutions from a search space. They have shown great promise in a
wide variety of optimization problems, specially in problem domains where the
search space is multi modal and/or noncontinuous. However, evolutionary
algorithms have not yet seen a lot of use in the optimization of statistical
models for earthquake forecasting. Our goal is to explore this combination. In
this talk, we will (briefly) explain what evolutionary algorithms are, and
then proceed to outline our early proposals and results regarding their use
for the generation of an RELM based earthquake forecast model. The 41st
1 SpeakerF@Dr.
Wang, MinZhen iProject Researcher
of ISMj DateF@4 March 2014@@@TimeF@15:00 16:00 LocationF@Room D312B (Seminar Room), Institute
of Statistical Mathematics TitleF@Distributions on Torus, Cylinder and
Disc iWang, MinZhen and Shimizu, Kunioj AbstractF@ Statistics
for data which include angular observations is known as directional
statistics. Bivariate circular data such as wind directions measured at two
points in time are modeled by using bivariate circular distributions or
distributions on the torus. Likewise circularlinear data are modeled by
using distributions on the cylinder and disc. We propose some extensions of
distributions on the torus, cylinder and disc in the framework of directional
statistics.@A new
circular distribution (Wang and Shimizu, 2012) is also introduced, which is
obtained by applying the Mӧbius
transformation to a univariate cardioid random variable. The distribution
function, trigonometric moments, and conditions for unimodality and symmetry
are studied. Kato and Jones (2010) study a family of distributions which is
obtained by applying the Mӧbius
transformation to a von Mises random variable, and we discuss the
relationship between our model and the KatoJones model. The bivariate
circular case (Wang and Shimizu, 2012) which is generated from a
circularcircular structural model linked with Mӧbius
transformation or a method of trivariate reduction. The joint probability
density function, trigonometric moments and circularcircular correlation
coefficient are explicitly expressed. An illustration is given for wind
direction data at 6 a.m. and noon as an application of the bivariate cardioid
distribution. The distributions on the cylinder we proposed is generated from
a combination of von Mises and transformed Kumaraswamy distributions. It is
an extension of the Johnson and Wehrly (1978) model. The marginal and conditional
distributions of the proposed distribution are given. A distribution using
the method of generating a cylindrical distribution with specified marginals
is also proposed. We generate skew or asymmetric distributions on the disc by
using the Mӧbius transformation and modified Mӧbius
transformations as extensions of the Mӧbius
distribution proposed by Jones (2004). The new distributions called the
modified Mӧbius distributions have six parameters.
They can be reduced to the Mӧbius and uniform
distributions as special cases, but many members of the family are skew
distributions for both the linear and the angular random variables. Some
properties such as the joint probability and marginal density functions of
the proposed distributions are obtained. References:
[1] Johnson, R.
A. and Wehrly, T. E. (1978). Some angularlinear distributions and related
regression models. Journal of the
American Statistical Association, 73, 602–606. [2]
Jones, M. C. (2004). The Mӧbius
distribution on the disc. Annals of the
Institute of Statistical Mathematics, 56, 733–742. [3] Kato, S. and
Jones, M. C. (2010). A family of distributions on the circle with links to,
and applications arising from, Mӧbius
transformation. Journal of the American
Statistical Association, 105, 249–262. [4]
Wang, M.Z. and Shimizu, K. (2012). On applying Mӧbius
transformation to cardioid random variables. Statistical Methodology, 9, 604–614. 2 SpeakerF@Dr.
Llenos, Andrea L. iUS
Geological Survey, Earthquake Science Center E
Postdoctoralj DateF@4 March 2014@@@TimeF@16:30 17:30 LocationF@Room D312B (Seminar Room), Institute
of Statistical Mathematics TitleF@Statistical modeling and
identification of potentially induced seismicity rate changes AbstractF@ iTBAj The 40th
1 SpeakerF@Dr.
Varini, Elisa iInstitute
of Applied Mathematics and Information Technology, National Research Council
(IMATICNR), Italy E DateF@18 February 2014@@@TimeF@15:00 16:00 LocationF@Room D312B (Seminar Room), Institute
of Statistical Mathematics TitleF@Bayesian estimation of doubly
stochastic Poisson processes: a particle filtering approach AbstractF@ iTBAj 2 SpeakerF@Dr.
Harte, David iGNS
Science, New Zealand EStatistical
Seismologist and Hazard Modellerj
DateF@18 February 2014@@@TimeF@16:30 17:30 LocationF@Room D312B (Seminar Room), Institute
of Statistical Mathematics TitleF@Stochastic Earthquake Models: Ways to
Improve and Insights into the Physical Process AbstractF@ 1. Bayesian estimation of doubly
stochastic Poisson processes: a particle filtering approach We
aim to explore the hypothesis that the earthquakes of a seismic region occur
under different physical conditions, corresponding to as many seismicity
phases characterized by different occurrence rates. This
hypothesis can be modeled by doubly stochastic Poisson processes in which the
observed process of the occurrence times of the earthquakes is a point
process whose conditional intensity function is assumed to be dependent on
both the past history and the current hidden state. By
assuming some of the possible choices for the observed point process and the
hidden state process, a Bayesian analysis is carried out in which the
likelihood function is approximated by the particle filtering method. 2. Stochastic Earthquake Models: Ways to
Improve and Insights into the Physical Process We
present a version of the ETAS model where the offspring rates vary both
spatially and temporally. This is in response to deficiencies discussed in
[1]. This is achieved by distinguishing between those spacetime volumes
where the interpoint spacetime distances are small, and those where they are
considerably larger. In the process of modifying a stochastic earthquake
model, one needs to justify assumptions made, and these in turn raise
questions about the nature of the underlying physical process. We will use
this version of the ETAS model as the basis for our discussion, and by
focussing on aspects where the model does not perform so well, attempt to
find physical explanations for such lack of fit. Some possible discussion
points are as follows. What
is the nature of the so called background process in the ETAS model? Is it
simply a temporal boundary (t=0) correction or does it represent an
additional tectonic process not described by the aftershock component? Or are
these two alternatives on completely different time scales? An
epidemic (the basic analogy underpinning the ETAS model), or a living
organism, can evolve by reproducing offspring that are slightly different to
that of their parents  randomness or gene mutation. Certain
"modified" individuals will be able to adapt to the environment
better and tend to survive over others. In the ETAS context, a lower value of
$\alpha$ will cause more "generations" in the aftershock sequence.
This allows for a richer and more complex evolution of the process, both
spatially and temporally. Alternatively, if alpha is large, then more of the
aftershocks are direct offspring of the mainshock. In the epidemic context,
this implies that the mainshock contains much more of the "DNA"
which governs the evolution of the overall sequence. What
is the relationship between fractal dimension and clustering? Does the
fractal dimension provide a better discrimination between those spacetime
volumes with higher offspring rates and the others? If so, does the fractal
dimension provide a more obvious physical description of the difference
between these high rate volumes and the lower rate volumes, and hence a
suggestive physical explanation? [1]
Harte, D.S. (2013). Bias in Fitting the ETAS Model: A Case Study Based on New
Zealand Seismicity. Geophys. J. Int. 192(1), 390412. The 39th
1 SpeakerF@Prof.
Matsufura, Mitsuhiro iVisiting
Researcher of ISMj DateF@14 January 2014@@@TimeF@13:00 14:00 LocationF@Room D312B (Seminar Room), Institute
of Statistical Mathematics TitleF@Inversion of GPS Data using ABIC AbstractF@ To
monitor crustal movements of the Japanese Islands, a nationwide dense GPS
network (GEONET) has been operated by Geographical Survey Institute of Japan
(now Geospatial Information Authority of Japan) since 1996. We developed an
inversion method to estimate unbiased interseismic slipdeficit rates at
plate interfaces from GPS displacement rate (velocity) data with an elastic dislocation
model. In this method, first, we subtract theoretical surface velocities due
to known steady relative plate motion from the observed GPS data, and presume
the residuals to be caused by slip deficit at plate interfaces. However, the
observed GPS data always include rigid block translation and rotation, which
cannot be explained by the elastic dislocation model. We treated the rigid
block translation and rotation as systematic errors in the analysis, and
removed them by transforming the velocity data into the average strain rates
of triangle elements composed of adjacent GPS stations. By this
transformation, original information about intrinsic deformation is
preserved. Applying a general inversion formula using ABIC to the GPS strain
data, we can obtain unbiased slipdeficit rate distribution. We demonstrate
the applicability of the GPS strain data inversion method through the
analyses of coseismic and interseismic GPS data in the Japan region, where
the North American, Pacific, Philippine Sea, and Eurasian plates are
interacting with each other in a complicated way. 2 SpeakerF@Prof.
Zhou, Shiyong iSchool of
Earth and Space Sciences, Peking University, China EProfessorj DateF@14 January 2014@@@TimeF@14:10 15:10 LocationF@Room D312B (Seminar Room), Institute
of Statistical Mathematics TitleF@Detecting the regional tectonic stress
variations in background seismicity data through statistical earthquake
modeling
iYajun
Peng, Shiyong Zhou, Jiancang Zhuang and Jia Shij AbstractF@ Large
earthquakes could perturb the stress field in regions even thousands of
kilometers away, leading to abrupt changes in background seismicity. We have
developed a probability based approach, based on the epidemictype aftershock
sequence model and the stochastic declustering method, to invert the smoothed
temporal variation of background seismicity rate and to extract useful
physical signals from complex seismicity patterns. An iterative algorithm is
constructed to estimate the background seismicity simultaneously by using a
combination of maximum likelihood estimate and weighted variable kernel
estimate. We verify this approach through simulations and confirm that it can
sensitively recover the onset of dynamic triggering. The
algorithm is applied to an earthquake catalog in Yunnan Province, China, and
successfully identifies a rapid increment of background seismicity rate
following the occurrence of the 2004 Sumatra Mw 9.2 earthquake, whereas no
remote triggering effect is detected following the occurrence of the 2005
Sumatra Mw 8.7 earthquake. This phenomenon agrees with GPS observations. It
is found that the elevated seismic activity within 15 d after the Sumatra
earthquake is mostly composed by shallow events, and direct triggering
relationship is well established. We
also studied the possible dynamic triggering effect in Northern China,
including Tangshan area, when the Japan Tohoku Mw 9.0 earthquake happened at
March 11th, 2011 and found out whether the area with large coseismic
displacement would have sudden abnormal seismicity increase. As a result, the
Japan Tohoku earthquake has little effect on the total and background
seismicity of Tangshan area, which means that the seismic structure of
Tangshan area is fundamentally stable. 3 SpeakerF@Dr.
Wang, Ting iUniversity
of Otago, New Zealand ELecturerj DateF@14 January 2014@@@TimeF@15:30 16:30 LocationF@Room D312B (Seminar Room), Institute
of Statistical Mathematics TitleF@Estimating the likelihood of volcanic
eruptions with incomplete eruption record AbstractF@ iTBAj
4 SpeakerF@Prof.
Ogata, Yosihiko iEmeritus
Professor of ISM; Visiting Professor of Institute of Industrial Science,
University of Tokyoj DateF@14 January 2014@@@TimeF@16:40 17:40 LocationF@Room D312B (Seminar Room), Institute of
Statistical Mathematics TitleF@Foreshocks and shortterm forecasting:
comparisons between in real seismicity and synthetic catalogs
iYosihiko
Ogata and Koichi Katsuraj AbstractF@ Some
statistical characteristics of foreshocks in the JMA earthquake catalog are
quantitatively different from those in the catalogs simulated by the
spacetime epidemictype aftershock sequence (ETAS) model associated with the
GutenbergRichter (GR) law. Also, the information gain of a foreshock
probability forecasting in the real seismicity is significantly large in
comparison with in synthetic catalogs. The 38th
SpeakerF@Prof.
Huang, Qinghua iDepartment
of Geophysics, Peking University, China EProfessorj
DateF@12 November 2013@@@TimeF@16:0017:00 LocationF@Room A508 (Seminar Room), Institute of
Statistical Mathematics TitleF@Seismicity changes revealed by the
RegionTimeLength (RTL) algorithm AbstractF@ The
RegionTimeLength (RTL) algorithm, which takes into account the epicenter,
time, and magnitude of earthquakes, is an effective technique in detecting
seismicity changes, especially the seismic quiescence. Based on the RTL
algorithm and the Qparameter (an average RTL parameter over a certain time
window), we can quantify the spatiotemporal characteristics of seismicity
changes. In order to reduce the possible ambiguity due to the selection of
model parameters in the RTL algorithm, we proposed an improved technique of
searching for the optimal model parameters. The applications of the RTL algorithm
in various tectonic regions indicated that seismic quiescence anomalies
generally started a few years prior to the occurrence of the mainshock. The
linear dimension of the seismic quiescence zone could be a few to several
times of the rupture dimension of the mainshock. The significance of the
seismic quiescence anomalies revealed by the RTL algorithm was supported by
the close investigations of model parameter effects and the stochastic test
based on randomized earthquake catalogs.
The 37th
SpeakerF@Herrmann,
Marcus iETH
Zurich, Swiss Seismological Servise E
PHD studentj DateF@27 August 2013@@@TimeF@16:0017:00 LocationF@Room D312A (Seminar Room), Institute
of Statistical Mathematics TitleF@Forecasting Losses Caused by a M6.6 Scenario
Earthquake Sequence in Basel, Switzerland AbstractF@ When
people and their environment are not properly prepared, earthquakes pose a
serious threat. Recently, the SEISMO12 earthquake scenario exercise
simulated the repeat of the 1356 Basel earthquake. This gave officials,
organizations, and the general public an idea of what may be expected in case
of a M6.6 earthquake. The present work relates to the scenario and
contributes to loss reduction by expressing the potential impact through
seismic risk. Reducing the shortterm seismic risk requires the evacuation of
vulnerable buildings. However, one cannot always evacuate in times of an
ongoing seismic sequence. Based on information of the continuous seismicity,
probabilistic forecasts show increasing benefit for shortterm defense
against earthquakes. Forecast probabilities subsequently allow timevarying
seismic hazard calculation. Only another combination with timeinvariant loss
estimation permits the assessment of shortterm seismic risk. Seismic risk
delivers a more direct expression of the socioeconomic impact than seismic
hazard, but one must characterize vulnerability and exposure to estimate
risk. Risk assessment brings together a variety of data, models and
assumptions. Based on the specific earthquake scenario, I perform a
probabilistic forecast of human losses. Seismologists may not be responsible
for communicating shortterm risk information to the public, but they have to
support decisionmakers to take worthwhile actions that may save lives.
However, the lowprobability environment and the complexity of involved
processes challenge decisionmakers. A final costbenefit analysis
constitutes greater benefit than pure statistical approaches by providing
objective statements that may justify evacuations. To deliver supportive
information in the simplest reasonable form, I propose a warning approach 
in terms of alarm levels  which allows one to explore worthwhile
mitigation actions for each district of the Basel region. The 36th
SpeakerF@Dr.
Enescu, Bogdan iTsukuba
University, Japan E Associate
Professorj DateF@23 July 2013@@@TimeF@16:0017:00 LocationF@Room D313 (Seminar Room), Institute of
Statistical Mathematics TitleF@Dynamic triggering of earthquakes in
Japan due to the 2011 Tohokuoki earthquake: some observations, stress
modeling and interpretation AbstractF@ iTBAj The 35th
SpeakerF@Dr.
Marzocchi, Warner iIstituto
Nazionale di Geofisica e Vulcanologia, Rome, Italy E Chief scientistj
DateF@3 July 2013@@@TimeF@16:0016:40 LocationF@Room D313 (Seminar Room), Institute of
Statistical Mathematics TitleF@Operational Earthquake Forecasting and
Decision Making AbstractF@ Traditionally,
seismic risk reduction is achieved only through a sound earthquake building code.
Nonetheless, some recent seismic disasters have highlighted the need for
enlarging the range of risk mitigation actions beyond that. In particular,
the occurrence of a seismic sequence may increase the weekly probability of a
large shock by orders of magnitude, although the absolute probability usually
remains below 1/100. Here, we summarize the state of the art in shortterm
earthquake forecasting and discuss how these forecasts may be used to
mitigate seismic risk in this time horizon. Because of the low probabilities
and high false alarm rates of possible advisories, mandatory mitigation
actions would not be an effective practical strategy to reduce risk.
Alternatively, we propose some low cost strategies, such as increasing
vigilance and preparedness, for using probabilistic forecasting to mitigate
seismic risk. These are based on the enudgingf principle of devolving
decisionmaking down from civic authorities to the individual level. The 34th
SpeakerF@Dr.
Chen, Xiaowei iScripps
Institution of Oceanography, University of California, San Diego E PostDoctoral Fellowj
DateF@16 April 2013@@@TimeF@16:0017:00 LocationF@Room D312B (Seminar Room), Institute
of Statistical Mathematics TitleF@Aspects of earthquake triggering and seismicity
clustering AbstractF@ Earthquakes
strongly cluster in space and time, driven both by earthquaketoearthquake
triggering and underlying physical processes, such as tectonic stress
loading, increased pore pressure, etc. To
understand the general characteristics of earthquake clustering from a large
dataset of earthquakes, I analyze seismicity in southern California. I use a
highresolution earthquake catalog based on waveform crosscorrelation to
study the spatialtemporal distribution of earthquakes. Parameters based on
event location, magnitude and occurrence time are computed for isolated
seismicity clusters. Spatial migration behavior is modeled using a
weightedL1norm method. Aftershocklike event clusters do not exhibit
significant spatial migration compared with earthquake swarms. Two triggering
processes are considered for swarms: slow slip and fluid diffusion, which are
distinguished based on a statistical analysis of event migration. The results
suggest fluidinduced seismicity is found across southern California,
particularly within geothermal areas. In the Salton Sea geothermal field
(SSGF), a correlation between seismicity and fluid injection activities is
seen. Spatialtemporal variations of earthquake stress drops are investigated
in different regions, and a distancedependence of stress drop from the
injection source is found in the SSGF, suggesting the influence of increased
pore pressure. Temporal variation of stress drops within mainshock source
regions shows that foreshocks and earthquake swarms have lower stress drops
than background seismicity and aftershocks. These results, combined with the
spatial migration observed for some large foreshock sequences, suggests an
aseismic transient process is likely involved in foreshock triggering. The 33rd
SpeakerF@Dr.
Han, Peng iChiba
University, Japan EPh.D.
Studentj DateF@14 February 2013@@@TimeF@16:0018:00 LocationF@Room D312A (Seminar Room), Institute
of Statistical Mathematics TitleF@Investigation of ULF seismomagnetic phenomena
in Kanto, Japan during 2000 – 2010 AbstractF@ Earthquakes
are one the most destructive natural hazards, causing huge damages and high
casualties. Especially during the past decade, huge/mega earthquakes have hit
many countries. Thus, effective earthquake forecasting is important and
urgent. Since the end of last century, ULF seismomagnetic phenomena have
been intensively studied. Recently, it has been considered a promising
candidate for shortterm earthquake prediction as a number of case studies
have been reported. However, scientists also found that
sometimes a sizeable earthquake happened without magnetic anomalies and
sometimes magnetic anomalies followed by no earthquakes. Thus, the relation
between magnetic anomalies and seismicity has been queried. Moreover, there
are two essential problems puzzling the researchers: (1) what is the exact
waveform of electromagnetic signals associated with earthquakes or
underground activities; (2) how are the signals generated. These two
questions have not yet been answered clearly and fully. There are still
active debates in the geophysical community on the seismoelectromagnetic
phenomena. In order to verify, clarify, and evaluate the ULF seismomagnetic
phenomena, longterm continuous monitoring of ULF magnetic field in a
seismically active area is required. Therefore, a sensitive observation
network has been established in KantoTokai area since the year 2000. Based
on eleven yearsf observation, plenty of geomagnetic data have been
accumulated, which provides an excellent opportunity to find answers to the
questions above. Thus, in this study I have conducted an investigation of ULF
seismomagnetic phenomena in Izu and Boso Peninsulas, Japan, based on the
data observed from 20002010. First, case studies of major events
have been applied. Energy of ULF geomagnetic signals at the frequency around
0.01 Hz has been investigated by wavelet transform analysis. In order to
minimize the influences of artificial noises, only the midnight time data (LT
1:00 ~ 4:00) have been utilized. To indentify anomalous changes from
ionospheric disturbances, the standard station Memabutsu has been chosen as a
reference station. (1) Case studies of the 2000 Izu Islands earthquake swarm
have indicated that there are unusual geomagnetic energy enhancements in
vertical component before and during the earthquake swarm. (2) Case studies
of the 2005 Boso M 6.1 earthquake have also shown clear geomagnetic energy
enhancements in vertical component before the earthquake. (3) Case studies of
the 2002 and 2007 slow slip events have demonstrated that there are
geomagnetic energy enhancements in both vertical and horizontal components
during the slip events. Then,
to verify and clarify the relation between ULF geomagnetic anomalies and seismicity,
statistical studies by superposed epoch analysis (SEA) have been carried out.
The results have indicated that before a sizeable earthquake there are
clearly higher probabilities of ULF anomalies than after the earthquake: for
Seikoshi (SKS) station in Izu, about 20~30 days before, one week and few days
before, and one day after the event statistical results of daily counts are
significant; for Kiyosumi (KYS) station in Boso around two weeks before, few
days before, and one day after the event. Finally, to find out the detailed
waveform of anomalous magnetic signals, waveform analysis has been performed.
The results show that there are mainly two kinds of seismomagnetic
signature. (1) Noiselike signals: Compared with the background, the signals
exhibit small increases of amplitudes at a wide frequency range. (2)
Transient/quasirectangular signals: the signals have
transient/quasirectangular waveforms with amplitudes of several nT (~ 10^{9}
T). The noiselike signals usually persist for several days or even a few
weeks, and are mainly associated with large earthquakes; the
transient/quasirectangular signals have durations of few seconds to few ten
seconds, and are registered mainly during slow slip events. Based on the results obtained above,
we conclude that: (1) there is a correlation between ULF geomagnetic
anomalies and local sizeable earthquakes in Izu and Boso Peninsulas, Japan,
and the common period of significant results is few days before and one day
after a sizeable earthquake; (2) there are mainly two kinds of
seismomagnetic signature registered in Izu and Boso Peninsulas: noiselike
signals and transient/quasirectangular signals. The mechanisms of the
anomalous geomagnetic signals are still unclear and need further studies. The 32nd
SpeakerF@Prof.
Savage, Martha iSchool of
Geography, Environment and Earth Sciences, Victoria University of Wellington,
New Zealand EProfessorj
DateF@13 November 2012@@@TimeF@16:0018:00 LocationF@Room D312A (Seminar Room), Institute
of Statistical Mathematics TitleF@Towards Predicting Earthquakes and
Volcanic Eruptions using Statistical Techniques AbstractF@ Predicting
natural hazards is fraught and statistical techniques are necessary to put
such studies on a firm standing. Here
we discuss two methods that we have applied to volcanic areas. Analysis of the rates of earthquake
activity (CURATE) is used to determine the characteristics of earthquake
swarms to try to determine how they develop over time. The technique compares favourably to
other declustering techniques and allows us to consider whether some swarms
are triggered by underlying processes that create diffuse seismicity that is
not well modelled by Omorifs law. We also analyse waveforms of earthquakes to
determine seismic anisotropy, which depends upon stress orientation and
magnitude, which in turn can be influenced by earthquake and volcanic
activity. Seismic waves travel
faster when their particle motion is along the cracks, which orient with the
principal stress direction. At
volcanoes around the world, we discovered significant changes in seismic
anisotropy strength and orientation that correlate with magma movement. Detecting and evaluating such changes
is complicated by scattered measurements, which sometimes have 90 degree
ambiguities and we have been considering ways to make the techniques more
robust. These observations will
provide the data that may eventually lead to prediction tools. The 31st
SpeakerF@Dr.
Omi, Takahiro iUniversity
of Tokyo E JST Postdoctoral
Researcherj DateF@25 October 2012@@@TimeF@16:0017:00 LocationF@4F lounge, Institute of Statistical
Mathematics TitleF@A statespace model for estimating the
timevarying detection rate of earthquakes and its application to immediate
probabilistic prediction of aftershocks AbstractF@ After
a large earthquake, the detection rate of earthquakes temporarily decreases,
and a lot of earthquakes are missed from a catalog. Such incompleteness of
the catalog prevents us from estimating statistical models of aftershock
activity accurately. To overcome this difficulty, Ogata and Katsura (2005)
modeled the incomplete catalog by using a parametric model of a timevarying
detection rate of earthquakes. In
this talk, we propose a state space model for estimating the timevarying
detection rate. In our model, the estimation problem is recursively solved,
by using Kalman filter and a Gaussian approximation of the posterior
probability distribution. Thus our model has an advantage in realtime
computation. Finally our model is combined with the OmoriUtsu law to predict
the occurrence rate of underlying aftershocks. We present some results on the
immediate probabilistic prediction of aftershocks. The 30th
SpeakerF@Dr.
Wang, Ting iUniversity
of Otago, New Zealand E Lecturerj DateF@24 July 2012@@@TimeF@16:0017:00 LocationF@Room D312B (Seminar Room), Institute
of Statistical Mathematics TitleF@Hidden Markov models in modelling
earthquake data AbstractF@ Earthquakes
are processes in which the internal workings (such as the accumulation of
tectonic stress) are only observed indirectly, although the final effects are
all too observable! Hidden Markov models (HMMs, a general statistical
framework for modelling partially observed systems) are an intuitively
attractive idea for analysing seismicity. I will briefly introduce the idea
of using HMMs to investigate earthquake cycles, and then focus on one case
study incorporating GPS data into earthquake forecasting. A new
model we developed, the Markovmodulated Hawkes process with stepwise decay
(MMHPSD), can capture the cyclic parentgeneratingoffspring feature of the
temporal behaviour of earthquakes. The decomposition of the earthquake cycle
motivated the construction of a nonlinear filter measuring shortterm
deformation ratechanges to extract signals from GPS data. This filter was
applied to a) deep earthquakes in central North Island, New Zealand, and b)
shallow earthquakes in Southern California. The study examines the use of
HMMs to extract possible precursory information that indicates an elevated
probability of large earthquakes occurrence. This study is controversial and
still requires further tests. Japan is an ideal place to carry out this test.
The 29th
1 SpeakerF@Prof.
Daley, Daryl iDepartment
of Mathematics and Statistics, The University of Melbourne E Honorary Professorial Associatej DateF@6 July 2012@@@TimeF@14:30 15:30 LocationF@Room D312A (Seminar Room), Institute
of Statistical Mathematics TitleF@Dimension walks and Schoenberg
spectral measures for isotropic random fields AbstractF@ Schoenberg
(1938) showed how Bochner's basic representation theorem for positive
definite functions (e.g. correlation function of a stationary stochastic
process) `simplifies' for spatial processes (ddimensional random fields)
which are isotropic: the standard Fourier kernel function is replaced by the
characteristic function of a random direction in dspace and the spectral
measure, instead of being on dspace, is on the positive halfline. The
talk describes how Wendland's `dimension walks', which were defined earlier
by Matheron as Descente and Montee in studying relations between dD and
either (d+2)D or (d2)D correlation functions, are equivalent to simple
modifications of their dSchoenberg measures. Another
family of dimension walks arises from projections from unit dspheres to
lower dimensional spheres, first via the kernel functions in the Schoenberg
representation and then more generally, for dSchoenberg measures. 2 SpeakerF@Dr.
Baddeley, Adrian iCSIRO
Mathematics, Informatics & Statistics E
Research Scientistj DateF@6 July 2012@@@TimeF@15:30 16:30 LocationF@Room D312A (Seminar Room), Institute
of Statistical Mathematics TitleF@Leverage, influence and residual
diagnostics for point process models AbstractF@ For a
spatial point process model fitted to spatial point pattern data, we develop
diagnostics for model validation, analogous to the classical measures of
leverage and influence and residual plots in a generalized linear model. The
diagnostics can be characterised as derivatives of basic functional of the
model. They can also be derived heuristically (and computed in practice) as
the limits of classical diagnostics under increasingly fine discretizations
of the spatial domain. We apply the diagnostics to example datasets where
there are concerns about model validity. 3 SpeakerF@Dr.
Shimatani, Kenichiro iAssociate
Prof. of ISMj DateF@6 July 2012@@@TimeF@15:30 16:30 LocationF@Room D312A (Seminar Room), Institute
of Statistical Mathematics TitleF@Inferring parameters in inhomogeneous
NeymanScott processes using the Palm likelihood AbstractF@ Plant
populations often exhibit spatially clustering distributions, in which the
two processes, limited seed dispersal and limited safe sites, are primary
mechanisms. The inhomogeneous NeymanScott process can combine and model
these two ecological processes. Estimating the model parameters allows
evaluation of the relative effects of dispersal and safe sites
retrospectively from spatial individual distribution data along environmental
gradients. Here we propose a likelihoodbased method for this spatial point
process by extending the recently developed method, the Palm likelihood. Our
approach was applied to evenaged black spruce forests in Canada. We obtained
a set of model parameters that well reproduced the observed spatial patterns,
and the fitted point processes predicted the reassembly pathway of the boreal
forests. The 27th
SpeakerF@Dr.
Jiang, Changsheng iInstitute
of Geophysics, China Earthquake Administration, China E Associate Professorj DateF@19 June 2012@@@TimeF@ LocationF@4th floor lounge, Institute of
Statistical Mathematics TitleF@Background sesismicty and its
application in the study of Accelerating Moment release (AMR) and Pattern
Informatics (PI) method
iChangsheng
Jiang, Zhongliang Wu and Jiancang Zhuangj AbstractF@ The 26th
SpeakerF@Dr.
Chan, ChungHan iDepartment
of Geosciences, National Taiwan University E
Postdoctoral Fellowj DateF@29 May 2012@@@TimeF@16:0017:00 LocationF@Room D312A (Seminar Room), Institute
of Statistical Mathematics TitleF@Shortterm earthquake forecasting
through a smoothing Kernel and the rateandstate friction law: Application
to Taiwan and the Kanto region, Japan AbstractF@ An
earthquake forecasting approach was employed for estimating the
spatiotemporal distribution of seismicity density in Taiwan and the Kanto
region, Japan. To evaluate longterm seismicity rate, a smoothing Kernel
function based on the distribution of past earthquakes was proposed. With the
use of the rateandstate friction model, shortterm rate evolution according
to the faultinteraction stress disturbance was forecasted. To test
feasibility of this model, it was applied using a catalog for the area
surrounding Taiwan. It leads to good agreement between the model forecast and
actual observations to prove its forecasting accuracy. To check its
stability, we estimated the deviation of the models according to different
parameters used in the approach. We conclude that deviations within each
parameter had an insignificant impact on forecasting stability. For the
application to the Kanto region, we proposed a 3D forecasting model due to
its complex tectonic setting. The seismicity patterns at various depths are
illustrated and the seismicity rate in the crust and along the subduction
zones can be distinguished. The high seismicity rate offshore in the east at
the depth of 2050 km can be associated with stress increase imparted by the
2011 Tohoku sequence. This phenomenon can be forecasted according to the
rateandstate friction model. The proposed approach, with verified
applicability for seismicity forecasts, could be useful for seismic hazard
mitigation. The application could provide a warning before the occurrence of
consequent earthquakes and would be valuable for consequent studies, e.g.,
probabilistic seismic hazard assessment. The 25th
SpeakerF@Dr.
Guillas, Serge iDepartment
of Statistical Science, University College London E Readerj
DateF@18 May 2012@@@TimeF@16:0017:00 LocationF@Room D312A (Seminar Room), Institute
of Statistical Mathematics TitleF@Earthquake occurrence: emulation and
climate forcing AbstractF@ In
earthquake occurrence studies, the socalled q value can be considered both
as one of the parameters describing the distribution of interevent times and
as an index of nonextensivity. Using simulated datasets, we compare four
estimators, based on principle of maximum entropy, method of moments, maximum
likelihood, and probability weighted moments (PMW) of the parameters of the
distribution of interevents times, assumed to be a generalized Pareto
distribution. We then use the unbiased version of PWM estimators to compute
the q value for the distribution of interevent times in a realistic
earthquake catalogue simulated according to the epidemic type aftershock sequence
(ETAS) model. Finally, we use these findings to build a statistical emulator
of the q values of ETAS model. We employ treed Gaussian processes to obtain
partitions of the parameter space so that the resulting model respects sharp
changes in physical behaviour. The emulator is used to understand the joint
effects of input parameters on the q value, exploring the relationship
between ETAS model formulation and distribution of interevent times. We
then present statistical evidence for a temporal link between variations in
the El Ni¬o¬Southern
Oscillation (ENSO) and the occurrence of earthquakes on the East Pacific Rise
(EPR). We adopt a zeroinflated Poisson regression model to represent the
relationship between the number of earthquakes in the Easter microplate on
the EPR and ENSO (expressed using the southern oscillation index (SOI) for
east Pacific sealevel pressure anomalies) from February 1973 to February
2009. We also examine the relationship between the numbers of earthquakes and
sea levels, as retrieved by Topex/Poseidon from October 1992 to July 2002. We
observe a significant positive influence of SOI on seismicity: positive SOI
values trigger more earthquakes over the following 2 to 6 months than
negative SOI values. There is a significant negative influence of absolute
sea levels on seismicity (at 6 months lag). We propose that increased
seismicity is associated with ENSOdriven seasurface gradients (rising from
east to west) in the equatorial Pacific, leading to a reduction in
oceanbottom pressure over the EPR by a few kilopascal. This relationship is
opposite to reservoirtriggered seismicity and suggests that EPR fault
activity may be triggered by plate flexure associated with the reduced
pressure. The 24th
SpeakerF@Dr.
Terakawa, Toshiko iEarthquake and Volcano Research Center, Graduate School of
Environmental Studies, Nagoya University EAssistant
Professorj DateF@11 May 2012@@@TimeF@14:0017:00 LocationF@Room D312A (Seminar Room), Institute
of Statistical Mathematics TitleF@Temporal Changes in Fault Strength and
Pore Fluid Pressures Following the 2011 off the Pacific Coast of Tohoku
Earthquake AbstractF@ Extensive
aftershocks and triggered seismic events are ubiquitous following large
earthquakes, but the controlling mechanisms are not yet understood. Focal
mechanisms of these events can provide insight into physical triggering
mechanisms because they reflect friction coefficient and pore fluid pressure
on the fault as well as the tectonic stress pattern. In the present study we
investigate physical processes triggering seismic events in inland region of
the northeast Japan following the 2011 off the Pacific Coast of Tohoku
earthquake (Mw = 9.0) by examining focal mechanisms to the tectonic stress
pattern and changes in the Coulomb failure function (DCFF). The local
excitation of seismicity rate in the regions with negative DCFF indicates
that these aftershocks would have been triggered by decrease of fault
strength due to the increase of pore fluid pressures. We show a plausible
explanation that seismic events on unfavourably oriented preexisting faults
relative to the tectonic stress pattern, or focal mechanisms inconsistent
with the tectonic stress pattern, would be evidence for drastic decrease of
fault strength due to increase of pore fluid pressures. The 23rd
SpeakerF@Prof.
Khmaladze, Estate V. DateF@12 December 2011@@@TimeF@16:00 17:00 LocationF@Room D312A (Seminar Room), Institute
of Statistical Mathematics (Tachikawa) TitleF@Infinitesimal analysis of setvalued
functions and applications to spacial statistics and image analysis AbstractF@¨PDF The
problem under consideration started with a study of spacial changepoint
problem, or changeset problem as we called it: suppose there is a set $K$,
such that our observations inside this set behave differently then anywhere
outside it. Given $n$ observations, how can we test that a given $K$ is the
correct one against its small perturbation $K(\varepsilon)$ as an
alternative? In huge amount of publications on the changeset
problem, almost all devoted to the estimation of the changeset, we could not
find such an object as an "alternative set" $K(\varepsilon)$. One
reason for this porbably is that it is not easy to realize how to describe
small changes in sets. We
developed an appropriate notion of the derivative of setvalued function in (Khmaladze, 2007) and used it to
build a version of contiguity theory in (Einmahl, Khmaladze, 2011) for
statistical problems where the parameter of interest is a set. One
single result here is that if $\Phi_n$ is a point process with increasing
intenstiy $n$, and the symmetric difference $K(\varepsilon)\Delta K$ shrinks
and "vanishes" as $\varepsilon \to 0$, then the sequence
$\Phi_n(K(\varepsilon)\Delta K)$ lives in the limit on the derivative set
$dK(\varepsilon)/d\varepsilon$: $$\Phi_n(
K(\varepsilon)\Delta K) \to \Psi (dK(\varepsilon)/d\varepsilon, {\rm as}
n\to\infty, \varepsilon\to 0.$$ In
the talk we present the main framework of this approach. We hope that some
discussions would lead to further applications. The 22nd
SpeakerF@Dr.
Iwata, Takaki iProject Associate Professor of ISMj DateF@25 November 2011@@@TimeF@13:3014:30 LocationF@Room D312B (Seminar Room), Institute
of Statistical Mathematics (Tachikawa) TitleF@The slip distribution of the 2011
Tohokuoki earthquake inferred from the spatial distribution of its
aftershocks AbstractF@ We
have developed a method to estimate the spatial slip distribution of a large
earthquake based on its onfault aftershock activity and the rate and
statedependent friction model (Dieterich, 1994, JGR). This talk will
represents the application of this method to the 2011 Tohokuoki earthquake.
The outline of the method is as follows. First, we divided the source area of
the destructive earthquake into 450 subfaults of which size is 20 x 20 km.
Next the slips of each subfault were optimized to make the expected spatial
distribution of aftershocks fit to the observed distribution during one day after
the mainshock. The expected distribution was derived from the Dieterichfs
formulation and the goodnessoffit between the expected and observed
distributions was evaluated by means of the likelihood for pointprocess
model. Then we constructed a Bayesian model incorporating smoothness
constraint on the spatial slip distribution, because optimizing such a huge
number of parameters is unstable; the incorporation of the constraint
enhances the stability of the optimization. To compute the posterior distribution
of the parameters in this Bayesian framework, the Markov Chain Monte Carlo
method was used. The result of this approach found an asperity close to the
hypocenter and some small asperities located to the southwest of the
hypocenter. This feature is consistent with the results of some slip
inversions based on seismograph, geodetic, and/or tsunami data, suggesting
that seismic activity data would play an important role in the estimation of
rupture process of an earthquake. The 21st 1 SpeakerF@Prof.
Schorlemmer, Danijel iGFZ German
Research Centre for Geosciences, Potsdam, Germany EProfessorj DateF@18 October 2011@@@TimeF@15:00 16:00 LocationF@Room D312A (Seminar Room), Institute
of Statistical Mathematics (Tachikawa) TitleF@Advancements in Probabilistic Seismic
Network Completeness Studies AbstractF@ An
important characteristic of any seismic network is its detection
completeness, which should be considered a function of space and time. Many
researchers rely on robust estimates of detection completeness, especially
when investigating statistical parameters of earthquake occurrence. We
present the Probabilitybased Magnitude of Completeness (PMC) method for
computing the spatial variation and temporal evolution of detection
capability of seismic networks based on empirical data only: phase data,
station information, and the network specific attenuation relation. New
developments are extending this method to complex 3D structures like mining
environments. We
present studies of regional networks from California, Switzerland, Italy,
Japan, New Zealand, and compare the result with estimated completeness levels
of other methods. We report on the time evolution of monitoring completeness
in these regions. Scenario computations show the impact of different possible
network failures and offer estimates of possible network optimization
strategies. Optimizations are reducing the necessary processor time for
computing. All presented results are published on the Completeness Web (www.completenessweb.org) from which
the user can download completeness data from all investigated regions,
software codes for reproducing the results, and publicationready and
customizable figures 2 SpeakerF@Prof.
Rundle, John B. iDepartment
of Physics and Geology, University of California, Davis, U.S.A. E Distinguished Professorj DateF@18 October 2011@@@TimeF@16:00 18:00 LocationF@Room D312A (Seminar Room), Institute
of Statistical Mathematics (Tachikawa) TitleF@Forecasting Large Earthquakes:
Problems, Pitfalls and Promise AbstractF@ Forecasting
the future behavior of a stochastic complex system is a necessary and
critical activity with wide applications. As the mathematician Edward Thorp
showed many years ago [1,2], forecasting has applications in games of chance
as well as in financial markets.
Both fields represent applications of statistics, stochastic random
walks, and probability theory.
Objective evaluation of forecasts by established tests and measures is
also a necessary and important component of a forecasting system. Many of the modern tests have been
tabulated at [3]. Earthquake
forecasts are a special case of the forecasting problem, particularly as
applied to large earthquakes such as the March 11, M9 Off the Pacific Coast
of Tohoku earthquake. Forecasts
of future events in complex systems are in general plagued by incomplete
information, a problem that must be considered in constructing forecasts. In addition to these problems,
delivery of realtime forecast information to the scientific community and to
the public is an issue as well.
Here, web 2.0 technology is helpful in allowing rapid dissemination of
information. In this lecture, I
shall discuss these general aspects of the forecasting problem as applied to
earthquake forecasting. I will
discuss ideas based on the Natural Time Weibull method of earthquake
forecasting, recently developed by our group. I will also discuss our experiences
with numerical earthquake simulations, as well as with public outreach using
the World Wide Web (see www.openhazards.com). [1]
E. Thorp and S. Kassouf, Beat the Market: A Scientific Stock Market
System, Random House (1967) [2]
E. Thorp, Beat the Dealer: A Winning Strategy for the Game of TwentyOne,
Random House (1962) [3]
http://www.cawcr.gov.au/projects/verification/ The 20th
SpeakerF@Dr.
Kagan, Yan Y. iDepartment
Earth and Space Sciences (ESS), UCLA, U.S.A. E
Researcherj DateF@2 September 2011@@@TimeF@16:00 17:00 LocationF@Room D312B (Seminar Room), Institute of
Statistical Mathematics (Tachikawa) TitleF@Statistical properties of earthquake
occurrence and their application for earthquake forecasting AbstractF@ Earthquake
occurrence exhibits scaleinvariant statistical properties: (a) Earthquake
size distribution is a powerlaw (the GutenbergRichter relation for
magnitudes or the Pareto distribution for seismic moment). Preservation of
energy principle requires that the distribution should be limited on the high
side; thus we use the generalized gamma or tapered Pareto distribution. The
observational value of the distribution index is about 0.65. However, it can
be shown that empirical evaluation is upward biased, and the index of 1/2,
predicted by theoretical arguments, is likely to be its proper value. The corner
(maximum) moment has an universal value for shallow earthquakes occurring in
subduction zones. We also determined the corner moment values for 8 other
tectonic zones. (b) Earthquake
occurrence has a powerlaw temporal decay of the rate of the aftershock and
foreshock occurrence (Omori's law), with the index 0.5 for shallow
earthquakes. The shortterm clustering of large earthquakes is followed by a
transition to the Poisson occurrence rate. In the subduction zones this
transition occurs (depending on the deformation rate) after 715 years,
whereas in active continents or plateinteriors the transition occurs after
decades or even centuries. (c)
The spatial distribution of earthquakes is fractal: the correlation dimension
of earthquake hypocenters is equal to 2.2 for shallow earthquakes. (d)
The stochastic 3D disorientation of earthquake focal mechanisms is
approximated by the rotational Cauchy distribution. On
the basis of our statistical studies, since 1977 we have developed
statistical short and longterm earthquake forecasts to predict earthquake
rate per unit area, time, and magnitude. The forecasts are based on smoothed
maps of past seismicity and assume spatial and temporal clustering. Our
recent program forecasts earthquakes on a 0.1 degree grid for a global region
90N90S latitude. For this purpose we use the PDE catalog that reports many
smaller quakes (M>=5.0). For the longterm forecast we test two types of
smoothing kernels based on the powerlaw and on the spherical Fisher
distribution. We employ adaptive kernel smoothing which improves our forecast
in seismically quiet areas. Our forecasts can be tested within a relatively
short time period since smaller events occur with greater frequency. The
forecast efficiency can be measured by likelihood scores expressed as the
average probability gains per earthquake compared to the spatially or
temporally uniform Poisson distribution. The other method uses the error
diagram to display the forecasted point density and the point events. As an
illustration of our methods, we are trying to answer a question: was the
Tohoku megaearthquake of 2011/3/11 a surprise? We consider three issues
related to the 2011 Tohoku earthquake: (1)
Why was the magnitude limit for the Tohoku region so badly underestimated,
and how can we estimate realistic limits for subduction zones in general? (2)
How frequently can such large events occur off Tohoku? (3)
Could shortterm forecasts have offered effective guidance for emergency
preparation? Two
methods can be applied to estimate the maximum earthquake size in any region:
statistical analysis of available earthquake records, and the moment
conservation principle  how earthquakes release tectonic deformation. We
have developed both methods since 1991. For subduction zones, the seismic
record is usually insufficient (in fact it failed badly for Tohoku), because
the largest earthquakes are so rare. However, the moment conservation
principle yields consistent estimates for all subduction zones. Various
measurements imply maximum moment magnitudes of the order 9.09.7. A
comparison of the interearthquake secular strain accumulation and its
release by the coseismic slip implies a similar maximum earthquake size
estimate. Beginning in 1999 we
used our statistical short and longterm earthquake forecasts, based on the
GCMT catalog, for the western Pacific, including Japan. We have posted them
on the web and included expected focal mechanisms as well. Longterm
forecasts indicate that the average frequency for magnitude 9 earthquakes in
the Tohoku area is about 1/400 years. We have archived several forecasts made
before and after the Tohoku earthquake. As expected, the Tohoku
megaearthquake changed the forecasted longterm rate by just a few percent.
However, the magnitude 7.5 foreshock increased the short term rate to about
100 times the longterm rate, and the magnitude 9 event increased it briefly
to more than 1000 times the longterm rate. These results could well justify
the development of an operational earthquake forecasting plan. The 19th DateF@4 March 2011@@@TimeF@15:00 16:00 LocationF@Room D312A (Seminar Room), Institute
of Statistical Mathematics (Tachikawa) 1 SpeakerF@Dr.
Talbi, Abdelhak iERI,
University of Tokyo E
Researcherj TitleF@Analysis of Earthquake Interevent
Times AbstractF@ Understanding
temporal behavior of earthquakes is a fundamental step towards building
reliable statistical model fitting observed seismicity. A successful class of
models assume two seismicity components corresponding to stable backgroundEand
varying triggeredErates or interevent times. In this
study, the distribution of interevent times is modeled assuming triggered
events governed by a nonhomogenous Poisson process, and background events
governed by different hypothetical distribution (Exponential, Gamma and
Weibull). The model is analytically introduced using PalmKhinchine equations
and fitted in practice to seismicity data from southern California, Japan and
Turkey. The analytic form of the distribution is discussed when different
priory hypotheses are adopted.In a second step, the temporal clustering of
events is studied using the distance between the whole distribution of
interevent times, and the residual distributions obtained using different declustering
approaches. Short and long range correlations are studied in space and time.
The residual background process is found dominant around the mean interevent
time and the mean interevent distance. The former analysis describes
seismicity as the accumulation of local perturbations related to a unique
mean field backgroundEprocesses characterized by the mean
interevent time and the mean interdistance. 2 SpeakerF@Prof.
Console, Rodolfo iNational
Institute of Geophysics and Volcanology Rome, Italy E Senior Scientific Advisorj TitleF@Shortterm earthquake forecasting
before and during the L'Aquila
(Central Italy) seismic sequence of April 2009 AbstractF@ The
M5.9 earthquake occurred on April 6th 2009, which caused more than 300
casualties in the city of L'Aquila and neighboring villages in Central Italy,
immediately generated a lot of discussions about the potential practical use
of foreshocks and other kind of information for mitigating seismic risk among
the population. These discussions triggered studies related to the validity
of statistical clustering models such as the ETAS model, not only for
forecasting aftershocks, but also mainshocks following potential foreshocks.
In the frame of the above mentioned studies, this presentation reports
preliminary results of the statistical analysis of the L'Aquila
seismic sequence by means of a version of the group of ETAS models. The free
parameters used in the algorithm are obtained through the maximum likelihood
method from a learning data set of instrumental seismicity collected from
2005 up to March 2009 in the region of L'Aquila.
Our method includes statistical declustering of the background seismicity by
an iterative process until the maximum likelihood of the learning data set
under the ETAS model is obtained. For testing purposes, an algorithm for
producing simulations of seismic series has been developed and applied to
produce synthetic catalogues, the statistical properties of which are
compared with those of the real one. Finally, the daily forecasts of
earthquakes at different threshold magnitudes were produced for a testing
period including the L'Aquila 2009 mainshocks and its largest aftershocks.
The results show that the probability of occurrence of an M5.9 computed from
the ETAS algorithm at the midnight preceding the L'Aquila
2009 mainshock, even if it was much higher than the background Poisson
probability, was quite low if compared with reasonable expectations for a
practical operational forecast. Moreover, the comparison between the daily
rate expected by the ETAS forecast method, and the real daily number of
aftershocks shows a systematic underestimation of such rate. The 18th SpeakerF@Dr.
Falcone, Giuseppe iNational
Institute of Geophysics and Volcanology Rome, Italy E Researcherj DateF@17 February 2011@@@TimeF@15:00 16:00 LocationF@Room D313 (Seminar Room), Institute of
Statistical Mathematics (Tachikawa) TitleF@Earthquake occurrence models and their
validation AbstractF@ This
presentation describes the tests and the forecast verification procedures of
three earthquake occurrence models applied to the various regions of the
globe (Italy, California, Japan) to assess the occurrence probabilities of
future earthquakes: two as shortterm (24 hour) models, and one as longterm
(5 and 10 years). The first model for shortterm forecasts is a purely
stochastic epidemic type earthquake sequence (ETES) model. The second
shortterm model is an epidemic ratestate (ERS) forecast based on a model
that is physically constrained by the application to the earthquake
clustering of the Dieterich ratestate constitutive law. The third forecast
is based on a longterm stress transfer (LTST) model that considers the
perturbations of earthquake probability for interacting faults by static Coulomb
stress changes. The forecast verification procedures have been carried out in
forwardretrospective and in real time way making use of statistical tools as
the Relative Operating Characteristics (ROC) diagrams, Loglikelihood,
NTest, LTest and Observed and forecasted number of events. The seismic
hazard modeling approach so developed, after a suitable period of testing and
refinement, is expected to provide a useful contribution to earthquake hazard
assessment, even with a possible practical application for decision making
and public information. These
models have been submitted to the Collaboratory for the Study of Earthquake
Predictability (CSEP) for forecast testing for Italy (ETH Zurich) and Japan
(ERI Tokyo). The 17th SpeakerF@Dr.
Chen, Kate Huihsuan iDepartment
of Earth Sciences, National Taiwan Normal University E Assistant Professorj DateF@28 January 2011@@@TimeF@15:00 16:00 LocationF@Room D313 (Seminar Room), Institute of
Statistical Mathematics (Tachikawa) TitleF@Triggering effect of small to large
earthquakes on earthquake cycle of small repeating events AbstractF@ Knowledge
of what governs the timing of repeating earthquakes is essential to
understanding the nature of the earthquake cycle and to determining earthquake
hazard, yet the variability and controls of earthquake recurrences are not
well established. Several unsolved problems regarding the recurrence
properties of natural earthquake sequences remain: How do the repeating
sequences respond to static and dynamic stress perturbations associated with
nearby earthquakes? To what degree does fault interaction influence the
timing of repeating earthquakes? Do spatially adjacent repeating sequences
communicate with each other in a way that is clearly evident in similar
occurrence times or recurrence patterns? Here
we use a large population of small, characteristically repeating earthquakes
at Parkfield provides to study how the interaction of nearby earthquakes
affects their recurrence properties. We analyze 114 M 0.4 ~ 3.0 repeating
earthquake sequences (RES) to examine the triggering effect from nearby
microseismicity. We find that closebyevents influence RESfs timing in a
matter of minutes or hours by shortterm triggering. Events that occurred
within less than 1 day of an RES often imposed or experienced high stress
changes. A stress increment of 10 kPa appears to be needed to produce such
effectively immediate triggering. The 16th SpeakerF@Prof.
Console, Rodolfo iNational
Institute of Geophysics and Volcanology Rome, Italy E Senior Scientific Advisorj DateF@17 January 2011@@@TimeF@15:00 16:00 LocationF@4F Lounge, Institute of Statistical
Mathematics (Tachikawa) TitleF@Renewal modeling and coseismic stress
transfer for seismic hazard assessment in the Corinth Gulf, Greece, fault
system AbstractF@ Earthquake
forecasts have always been a difficult task because they can be affected by
uncertainty in terms of the most appropriate model and the involved parameter
values. The models adopted in this study belongs to the category of the
renewal models, based on the characteristic earthquake hypothesis, the
necessary ingredients of which are a fixed geometry and the knowledge of the
slip rate on the faults. Both the BPT and the Weibull distribution have been
tested. The hazard rate so obtained is then modified by the inclusion of a
permanent effect due to the Coulomb static stress change caused by failure of
neighboring faults that occurred since the latest characteristic earthquake
on the concerned fault. I apply this method along the Corinth gulf extension
zone, a place that is rich with observations of strongearthquake recurrence
behavior, to assess their relative forecast applicability. The validity of
the renewal models is assessed in retrospective way on the data of the last
300 years by comparison with a plain time independent Poisson model. This is
done by means of statistical tools as the ROC diagram, the Rscore and the
loglikelihood ratio. I find that the renewal models perform better than the
Poisson hypothesis. It seems also that the BPT distribution works slightly
better than the Weibull distribution, while little advantage is achieved by
the introduction of the Coulomb static stress change in the forecasting
algorithm. The 15th SpeakerF@Dr.
Parsons, Tom iUnited
States Geological Survey, U.S.A. E
Research Geophysicistj DateF@4 January@2011@@@TimeF@14:00 15:00 LocationF@Room D312B (Seminar Room), Institute
of Statistical Mathematics (Tachikawa) TitleF@What causes aftershocks? AbstractF@ Despite
decades of research devoted to this question, we still do not know how
mainshocks cause aftershocks. In this presentation I show recent research
that attempts to isolate static and dynamic stressing signals so that we can
learn more about their abilities to trigger other earthquakes. We hope
someday to combine physical models with statistical models for forecasting,
but I show that the physical models still have problems when used
prospectively. The 14th SpeakerF@Dr.
Toda,
Shinji iVisiting
Prof. of ISM / Associate Prof. of The Disaster
Prevention Research Institute, Kyoto Universityj DateF@17 November 2010@@@TimeF@13:3014:30 LocationF@Room D312B (Seminar Room), Institute
of Statistical Mathematics (Tachikawa) TitleF@Rate/state Coulomb stress transfer
model for the CSEP Japan seismicity forecast AbstractF@ Numerous
studies retrospectively found that seismicity rate jumps (drops) due to
coseismic Coulomb stress increase (decrease). The Collaboratory for the Study
of Earthquake Predictability (CSEP) instead provides us an opportunity for
prospective testing of the Coulomb hypothesis. Here we adapt our stress
transfer model incorporating rate and state dependent friction law to the
CSEP Japan seismicity forecast. We demonstrate how to compute the forecast
rates of large shocks in 2009 using the large earthquakes during the past 120
years. The time dependent impact of the coseismic stress perturbations
explains qualitatively well the occurrence of the recent moderate size
shocks. Such ability is partly similar to that of statistical earthquake
clustering models. However, our model differs from
them as follows: the offfault aftershock zones can be simulated using finite
fault sources; the regional areal patterns of triggered seismicity are
modified by the dominant mechanisms of the potential sources; the imparted
stresses due to large earthquakes produce stress shadows that lead to a
reduction of the forecasted number of earthquakes. Although the model relies
on several unknown parameters, it is the first physics based model submitted
to the CSEP Japan test center and has the potential to be tuned for
shortterm earthquake forecasts. The 13th SpeakerF@Dr.
Bansal, Abhey Ram iNational
Geophysical Research Institute, Council of Scientific and Industrial Research
(NGRI, CSIR) (India)EScientist
E1^Fellow of
the JSPS Invitation Fellowship Programs for Research in Japan (a longterm
program); Visiting Researcher of the ISMj DateF@16 July 2010@@@TimeF@15:0016:00 LocationF@Room D312B (Seminar Room), Institute
of Statistical Mathematics (Tachikawa) TitleF@Statistical seismology of Sumatra:
before and after the megaearthquake of 26 December 2004 AbstractF@ We
examine the effect of the Special
Seminar SpeakerF@Prof.
VereJones,
David iStatistics
Research Associates Limited (SRA), New Zealand E
Directorj DateF@30 June 2010@@@TimeF@15:0016:00 LocationF@Room D312B (Seminar Room), Institute
of Statistical Mathematics (Tachikawa) TitleF@The Evolution of Statistical
Seismology AbstractF@ Some
account is given of the early days of statistical seismology, and some issues
raised for future consideration in this field. The 12th SpeakerF@Dr.
Sugaya, Katsunori iProject
Researcher of ISMj DateF@23 April 2010@@@TimeF@15:0016:00 LocationF@Room D312B (Seminar Room), Institute
of Statistical Mathematics (Tachikawa) TitleF@Coseismic change and recovery of
scattering environment in the crust after a large earthquake (Sugaya,
K., Hiramatsu, Y., Furumoto, M. and Katao, H.) AbstractF@ The
coda waves mainly consist of scattered S waves in the crust. The attenuation
property of coda waves, coda Q^1 or Qc^1, reflects the scattering and
absorption environments in the crust and is supposed to be a good tool to
investigate medium properties in the crust. Furthermore, Qc^1 is a good
indicator of the stress condition in the crust. We
observe a unique temporal variation in crustal heterogeneity from the
analysis of Qc^1 for 14 years in the Tamba region, northeast to the rupture
zone of the 1995 Hyogoken Nanbu earthquake (Mjma 7.3) in southwest Japan.
The values of Qc^1 at lower frequencies (1.54.0 Hz) that increased
coseismically due to the static stress change decreased back to the preevent
values in about two years. No such variations are found at higher frequencies
(5.024.0 Hz). We confirm that no tectonic events that cause a significant
stress change occurred during the recovery period. The time required for the
recovery of the scattering environment, such as the number density of cracks
and cracks opened by the stress change, observed here is consistent with
those of previous studies focused on the brittle shallower crust. This
suggests a possibility that a similar mechanism of the recovery operates both
in the brittle and the ductile parts of the crust. The 11th SpeakerF@Prof.
Kuensch, Hans R.iETH Zurich, Switzerland E Professorj DateF@9 April 2010@@@TimeF@15:0016:00 LocationF@Room D312B (Seminar Room), Institute
of Statistical Mathematics (Tachikawa) TitleF@Particle and Ensemble Kalman filtering
(Kuensch, Hans R. and Frei, Marco) AbstractF@ Ensemble
and particle filtering are two sequential Monte Carlo methods for
approximating the filter distributions in nonlinear and/or nonGaussian state
space models. They differ in the way a new observation is taken into account
in the update step. The particle filter is nonparametric and works by
weighting and resampling. The Ensemble Kalman filter (EnsKF) requires a
linear Gaussian observation model and proceeds by a Monte Carlo
implementation of the linear Kalman
filter update. In a different version of the ENSKF called the square root
filter, updates are done by a linear transformation of the prediction sample
(without any additional randomness). Even though the EnsKF is based on
unwarranted assumptions, it is extremely robust in many high dimensional
applications where the number of particles is small because of computational
limitations. In contrast, the particle filter degenerates quickly as the
dimension of the observations increases. I
will discuss some of the following issues: EnsKF for nonlinear observation
equations, the use of the EnsKF for parameter estimation, regularization of
the estimated prediction covariance, bias due to estimation of the prediction
covariance, localization of the EnsKF update, behavior of the EnsKF for
Gaussian mixture predictions and ideas for combining the EnsKF and the
particle filter. The 10th SpeakerF@Wang,
Qi iDepartment
of Earth and Space Sciences, University of California, Los Angeles, U.S.A.j DateF@19 February 2010@@@TimeF@15:0016:00 LocationF@Room D312A (Seminar Room), Institute
of Statistical Mathematics (Tachikawa) TitleF@An optimized fiveyear large
earthquake forecast in California based on smoothed seismicity (Qi Wang,
D.D.Jackson and Y.Y. Kagan) AbstractF@ Earthquake
forecasting models based on seismic, geologic, tectonic and geodetic
information have been discussed a lot. We tried to state a five year forecast
of California earthquakes with magnitude above 5.0 based on smoothed
seismicity. We used extended sources to represent large earthquakes. Kagan
and Jackson (2007) have presented a fiveyear forecast of southern California
earthquakes with magnitude 5 or larger. Here we extend the forecast region
from southern California to all of California using the new California
earthquake catalog with estimated moment magnitude, and in one case we
include historic earthquakes as well as instrumentally recorded ones after
considering catalog incompleteness. This forecast model differs from others
like it because larger events are represented by multiple point sources and
because the parameters in the spatial smoothing kernel have been optimized in
learning period by using maximum likelihood estimation. We tried to answer
two basic questions: (1) whether the anisotropic and magnitudedependent
smoothing kernels outperform the isotropic and magnitudeindependent ones;
(2) whether including large historical earthquakes could improve the
forecast. Moreover, earthquake
data have errors in location, magnitude and focal mechanism that can
influence the results of earthquake studies. Neglecting these errors, or
estimating them poorly, could cause valid hypotheses to be rejected or
invalid ones to be accepted. We tried to estimate how the uncertainty of
catalog data especially uncertainty of magnitude influences our testable
forecast results. The 9th SpeakerF@Dr.
Terakawa, Toshiko iGeodynamics,
SteinmannInstitute, University of Bonn, Germanyj DateF@29 January 2010@@@TimeF@15:0016:00 LocationF@Room D312A (Seminar Room), Institute
of Statistical Mathematics (Tachikawa) TitleF@Identification of the high fluid
pressure source driving the 2009 Lfaquila earthquake sequence (Toshiko
Terakawa, Anna Zoporowski, Boris Galvan, and Stephen A. Miller) AbstractF@ The
April 6, 2009 Lfaquila intraplate earthquake (Mw=6.3) in the Central
Apennines occurred at the boundary separating regions of diffuse CO_2
degassing and regions where degassing is not observed. The same tectonic and
geologic environment hosted the 1997 Colfiorito sequence to the north, which
was shown to be driven by degassing of a highpressure fluid source at depth
with a fluid diffusion model [Miller et al. 2004]. Here we show the 3D fluid
pressure field in the Lfaquila
region, applying a new analysis technique termed Focal Mechanism Tomography
[Terakawa et al. 2009 (submitted)] to actual seismic data. We identify three largescale pockets
of high fluid pressure at depths of 710 km, and show a very strong
correlation between these high fluid pressure regions and both foreshock and
aftershock hypocenters. The shape of overpressured regions and the evolution
of aftershock locations indicate that aftershocks are being driven in part by
fluid flow associated with volumetric compression from the mainshock acting
upon the overpressured poroelastic reservoir [e.g., Nur & Booker 1982;
Bosl & Nur 2002], and by fracturing and subsequent flow from trapped high
pressure pockets [Miller et al. 2004].
The mapped 3D fluid pressure field inferred from our analysis provides
an important boundary condition for forward modelling of fluid flow and
stress evolution for a mechanistic assessment of the continuing seismic
hazard in the region. These results also form a baseline hypothesis against
which other geophysical and geochemical measurements can be tested. The 8th SpeakerF@Dr.
Himeno, Tetsuto iProject
Researcher of Institute of Polar Sciencesj DateF@15 January 2010@@@TimeF@15:0016:00 LocationF@Room A508 (Seminar Room), Institute of
Statistical Mathematics (Tachikawa) TitleF@Time variation of background
seismicity on ETAS model AbstractF@ The
Epidemic Type Aftershock Sequence (ETAS) model introduced by Ogata (1988) is
one of statistical method for seismicity. This model classifies earthquake
sequence into aftershocks and background seismicity. For the frequency of
aftershocks, some empirical relations is assumed. On the other hand,
background seismicity rate is constant. Therefore, the main shock sequence is
distributed according to stationary Poisson process. In this study, we focus
the time variation of background seismicity and expand the background
seismicity. The 7th SpeakerF@Dr.
Chu, Annie iProject
Researcher of ISMj DateF@4 December 2009@@@TimeF@15:0016:00 LocationF@Room D312A (Seminar Room), Institute
of Statistical Mathematics (Tachikawa) TitleF@Comparison of ETAS parameter estimates
across different global tectonic zones AbstractF@ Branching
point process models such as the ETAS (EpidemicType Aftershock Sequence)
models introduced by (Ogata 1988, 1998) are often used in the description,
characterization, simulation, and declustering of modern earthquake catalogs.
The present work investigates how the parameters in these models vary across
different tectonic zones. After considering divisions of the surface of the
Earth into several zones based on the plate boundary model of Bird (2003),
ETAS models are fit to the occurrence times and locations of shallow
earthquakes within each zone. Computationally, the EMtype algorithm of Veen
and Schoenberg (2008) is employed for the purpose of model fitting. The fits and variations in parameter
estimates for distinct zones are compared. Seismological explanations for the
differences between the parameter estimates for the various zones are
considered, and implications for seismic hazard estimation and earthquake
forecasting are discussed. The 6th SpeakerF@Prof.
Zhou, Shiyong iDepartment
of Geophysics, Peking University, China E
Professorj DateF@30 October 2009@@@TimeF@15:0016:00 LocationF@Room D304 (Seminar Room), Institute of
Statistical Mathematics (Tachikawa) TitleF@Was the Ms 8.0 Long Men Shan
(Wenchuan) Earthquake in 2008@induced by
the Zipingpu Reservoir? AbstractF@ According
to the Coulomb failure criterion the variation of either shear stress, normal
stress, or pore pressure can result in earthquake occurrences. Abnormal
seismicity increases around reservoirs are often thought to be induced by the
water piled behind the dam, which leads to increases in crustal pore pressure
and Coulomb stress nearby, and so promote the nearby faults to fail. To
investigate how much the ZiPingPu reservoir, whose dam is just a few
hundreds of meters from the Long Men Shan fault, influenced the May 12, 2008
Wenchuan earthquake Ms8.0, we calculated the Coulomb stress variation induced
by the filling of the Zipingpu reservoir, beginning in December 2004, and
analyzed the correlation between the seismicity variations and the induced
Coulomb stress variations. Both the calculated Coulomb stress variations and
the observed seismicity analysis suggest that the probability that the huge
Long Men Shan earthquake Ms8.0 was induced by the Zipingpu reservoir is very
low. The filling of the Zipingpu reservoir could only result in a small
increase in the rate of shallow earthquakes with hypocenter depth smaller
than 5km near the reservoir region. The 5th SpeakerF@Nomura,
Shunichi
iPh.D. Student of The Graduate
University for Advanced Studies (SOKENDAI)j DateF@2 October 2009@@@TimeF@15:0016:00 LocationF@Room D312, Institute of Statistical
Mathematics (Tachikawa:
http://www.ism.ac.jp/access/index_j.html or
http://www.ism.ac.jp/access/index_e.html) TitleF@Composing prior distributions of BPT
distribution with slip rates in renewal process of recurrent earthquakes AbstractF@ Renewal
process with Brownian Passage Time (BPT) distribution is mainly used in the
longterm evaluation of active faults by the Earthquake Research Committee,
the Headquarters of Earthquake Research Promotion (ERC/HERP). In ERC/HERP (2001),
two parameters of BPT distribution, called mean parameter and aperiodicity
(shape) parameter, are estimated as follows: mean parameter is estimated by
either recurrence intervals or slip rates of active faults, and aperiodicity
parameter is commonly set at 0.24. The value of aperiodicity parameter is
derived from the active faults on which more than four recurrence intervals
are detected, but recent researches showed the value of aperiodicity
parameter is inappropriate to apply for all active faults. In
this seminar, we propose an alternative method to estimate the parameters of
BPT distribution of recurrence intervals. First, we normalize the intervals
by mean recurrence time derived from slip size and slip rate of the fault.
Then we can use both intervals and slip rate to estimate its mean parameter.
Second, we use Bayesian inference to estimate the parameters. Bayesian
predictive distribution gives stable prediction when there are few data in
the fault. Prior distributions and their parameters are estimated by
maximizing their marginal likelihood. We simulate recurrence intervals from a
virtual fault and compare our model and ERC/HERP's model by their relative
entropy. The 4th SpeakerF@Kumazawa,
Takao
iPh.D. Student of The Graduate
University for Advanced Studies (SOKENDAI)j DateF@28 August 2009@@@TimeF@15:0016:00 LocationF@Room 330, Institute of Statistical
Mathematics TitleF@Seismicity changes in Tohoku District
before the 2008 IwateMiyagi Inland Earthquake
i Kumazawa,
Takao (The
Graduate University for Advanced Studies), Ogata, Yoshihiko (The
Institute of Statistical Mathematics) and AbstractF@ The
2008 IwateMiyagi Nairiku earthquake of M7.2 (JMA) is one of the largest
inland earthquakes occurred in
the northern part of Honshu island. We examined seismicity rate changes in
the surrounding areas prior to this earthquake for around 10 years in
relation to the ∆CFF values of it. We selected a number of regions according
to the sign of the ∆CFF from the IwateMiyagi inland earthquake, then
examined the series of earthquakes in each region by fitting ETAS models and
examined if the changes of seismicity, if any, were significant and
consistent with the ∆CFF. We
confirmed the preseismicity by transient anomalies in GPS measurement of
several stations. Five
years prior to the event, a M7.1 earthquake of 2003 occurred at 71km. depth
in the subducting Pacific plate, beneath Sanriku coast, Miyagi
prefecture. This faultfs movement
activated the reverse fault system in inland Tohoku District including the
one responsible for IwateMiyagi inland earthquake, raising the seismicity
there. The activated area also covered the fault of M6ǒ5Ņinland earthquake
occurred two months afterward. It as well likely contributed to the
activation. Here
we hypothesize these two large earthquake in 2003 enhanced the precursory
slow slip in the fault of our interest, including its deeper extension, and
explain the observed preseismic patterns. This hypothesis is in concordance
with the GPS location measurement from several stations on and around the
fault. The rate of distance shrinkage between the two stations, one being
right on the fault, can only be well explained by assuming that the active
site moved deeper at some point. We
fitted ETAS model to the observed seismic activity in each selected area and
see if there are meaningful changing points of time, across which the
seismicity being changed significantly in terms of AIC. The search for the
changing points were done first by dividing the whole period either at points
of known events, or at suspicious points with some systematical shifting of
them, then to each divided period ETAS was refitted. Among these combined
ETAS models, the best performed model was compared with the single ETAS
without divisions and examined if its improvement was significant. We finally checked these changing
points are consistent with ∆CFFs. The 3rd SpeakerF@Prof.
Matsufura, Mitsuhiro (Specially Appointed Prof. of ISMj DateF@14 August 2009@@@TimeF@15:0016:00 LocationF@Room 330, Institute of Statistical
Mathematics TitleF@Inversion of seismic and geodetic data
to estimate tectonic stress fields in the Earth's crust AbstractF@ The
Earth's crust is macroscopically treated as a linear elastic body, but it
includes a number of defects. The occurrence of inelastic deformation such as
brittle fracture at the defects, which can be generally represented by moment
tensor, brings about elastic deformation in the surrounding regions. Since
the moment tensor is mathematically equivalent to the volume integral of
stress release over the whole elastic region surrounding the source, we can
quantitatively relate the centroid moment tensor (CMT) solutions of seismic
events with a true but unknown tectonic stress field. On the basis of such an
idea, we developed an inversion method to estimate tectonic stress fields
from CMT data using Akaike's Bayesian information criterion (ABIC) [1]. We
show the 3D tectonic stress pattern in and around Japan, obtained by
applying the CMT data inversion method to 15,000 seismic events within the
magnitude range of 3.55.0 in the period of 19972007. The inverted 3D
stress pattern illuminates the presentday (Quaternary) complex tectonic
motion of Japanese Islands. On the other hand, the crustal deformation
observed through geodetic measurements is the sum of the inelastic
deformation as source and the elastic deformation as effect. Representing the
sources by moment density tensor distribution, we created a theory of
physicsbased strain analysis, and developed an inversion method to
separately estimate 3D elastic and inelastic strain fields from GPS data
[2]. In this method, first, we determine the optimum distribution of moment
density tensor from observed GPS array data by using Akaike's information
criterion (AIC). Converting the optimum moment density tensor distribution
with elastic compliance tensor, we can directly obtain 3D inelastic strain
fields. Given the optimum moment density tensor distribution, we can
theoretically compute 3D elastic strain fields. We applied the inversion
method to GPS horizontal velocity data (19962000) in the NiigataKobe
transformation zone, central Japan, and succeeded in estimating 3D elastic
and inelastic strain
rate fields separately. References: 1. Terakawa, T.
and M. Matsu'ura (2008), CMT data inversion using a Bayesian information
criterion to estimate seismogenic stress fields, Geophys. J. Int., 2.
Noda, A. and M. Matsu'ura (2009), Physicsbased GPS data inversion to
estimate 3D elastic and inelastic strain fields, Geophys. J. Int.,
submitted. The 2nd SpeakerF Dr. Zhuang, Jiancang
iAssistant
Prof. of ISMj DateF@31 July 2009@@@TimeF@15:0016:00 LocationF@Room 330, Institute of Statistical
Mathematics TitleF@Gambling with reputation: On scoring
earthquake forecasts and predictions AbstractF@ This
seminar presents a new method, namely the gambling score, for scoring the
performance earthquake forecasts or predictions. Unlike most other scoring
procedures that require a regular scheme of forecast and treat each
earthquake equally, regardless their magnitudes, this new scoring method
compensates the risk that the forecaster has taken. Once a forecaster makes a
prediction or forecast, he is assume to have bet some points of his
reputations. The reference model, which plays the role of the house,
determines how many reputations the forecaster can gain if he succeeds,
according to a fair rule, and also takes away the reputations bet by the
forecaster if he loses. From the viewpoints of both the reference model and
the forecaster, the rule for rewarding and punishment is fair. This method is
also extended to the continuous cases
of point process models, where the reputations bet by the forecaster
become a continuous mass on the spacetimemagnitude range of interest. We
also calculate the upper bound of the gambling score when the true model is a
renewal process, the stress release model or the ETAS model and when the
reference model is the Poisson model. The
1st SpeakerF Dr. Harte, David iDirector
of Statistics Research Associates Limited (SRA), New Zealandj DateF@3 July /2009@@@TimeF@15:0016:00 LocationF@Room 330, Institute of Statistical
Mathematics TitleF@Using R for Modelling Marked Point
Processes Indexed by Time AbstractF@ A
unified approach will be described for the fitting and analysis of a large
class of point process models. The models discussed will be those that are indexed
by time and contain additional marks. These include many of the point process
models used for describing earthquake processes, e.g. stress release model,
ETAS model, spatial ETAS model and the linked stress release model. By
exploiting characteristics of their conditional intensity function, the
loglikelihood function can be reduced to a more simple form. This simplified
structure can also be used to calculate the residual process and perform
simulations. The method involves "decomposing" these models into
more elementary parts, and then utilising these parts in an object orientated
manner within the R programming language. The longer term purpose in such an
approach is to lessen programming effort involved in fitting such models.
This should give us a better ability to more easily make modifications to
existing models, and hence test alternative hypotheses. I will show R
programming code for some fitted models to earthquake data. 
