Papers by Matthias Ohrnberger
Nature, 2005
On 26 December 2004, a moment magnitude M w 5 9.3 earthquake occurred along Northern Sumatra, the... more On 26 December 2004, a moment magnitude M w 5 9.3 earthquake occurred along Northern Sumatra, the Nicobar and Andaman islands, resulting in a devastating tsunami in the Indian Ocean region 1 . The rapid and accurate estimation of the rupture length and direction of such tsunami-generating earthquakes is crucial for constraining both tsunami wave-height models as well as the seismic moment of the events. Compressional seismic waves generated at the hypocentre of the Sumatra earthquake arrived after about 12 min at the broadband seismic stations of the German Regional Seismic Network (GRSN) 2,3 , located approximately 9,000 km from the event. Here we present a modification of a standard array-seismological approach and show that it is possible to track the propagating rupture front of the Sumatra earthquake over a total rupture length of 1,150 km. We estimate the average rupture speed to be 2.3-2.7 km s 21 and the total duration of rupture to be at least 430 s, and probably between 480 and 500 s.
Geophysical Research Letters, 2005
1] We test the capability of broadband arrays at teleseismic distances to image the spatio-tempor... more 1] We test the capability of broadband arrays at teleseismic distances to image the spatio-temporal characteristics of the seismic energy release during the Dec 26, 2004 Sumatra earthquake at early observation times. Using a non-plane-wave array location technique previously reported values for rupture length (about 1150 km), duration (about 480 s), and average rupture velocity (2.4 -2.7 km/s) are confirmed. Three dominant energy releases are identified: one near the hypocenter, a second at 6°N 94°E about 130 s later and a third one after 300 s at 9°N 92-93°E. The spatio-temporal distribution of the radiated seismic energy in the source region is calculated from the stacked broadband recordings of two arrays in Germany and Japan and results in rough estimates of the total seismic energy of 0.55 Á 10 18 Nm (GRSN) and 1.53 Á 10 18 Nm (FNET) respectively. Changes in the relative ratio of energy as function of spatio-temporal location indicate a rotation of the focal mechanism during the rupture process. Citation: Krüger, F., and M. Ohrnberger (2005), Spatio-temporal source characteristics of the 26 December 2004 Sumatra earthquake as imaged by teleseismic broadband arrays, Geophys. Res. Lett., 32, L24312,
Near Surface Geophysics, 2004
Passive recordings of seismic noise are increasingly used in earthquake engineering to measure in... more Passive recordings of seismic noise are increasingly used in earthquake engineering to measure in situ the shear-wave velocity profile at a given site. Ambient vibrations, which are assumed to be mainly composed of surface waves, can be used to determine the Rayleigh-wave dispersion curve, with the advantage of not requiring artificial sources. Due to the data uncertainties and the non-linearity of the problem itself, the solution of the dispersion-curve inversion is generally non-unique. Stochastic search methods such as the neighbourhood algorithm allow searches for minima of the misfit function by investigating the whole parameter space. Due to the limited number of parameters in surface-wave inversion, they constitute an attractive alternative to linearized methods. An efficient tool using the neighbourhood algorithm was developed to invert the one-dimensional V s profile from passive or active source experiments. As the number of generated models is usually high in stochastic techniques, special attention was paid to the optimization of the forward computations. Also, the possibility of inserting a priori information into the parametrization was introduced in the code.
EGU General Assembly …, 2010
We present a Machine Learning approach aiming for improving the accuracy of automatic detections ... more We present a Machine Learning approach aiming for improving the accuracy of automatic detections of noise and signal at 3-component seismic stations. Using supervised learning in conjunction with the multivariate framework of Dynamic Bayesian Networks (DBNs) we make use of historical data obtained from the LEB bulletin to train a classifier to capture the intrinsic characteristics of signal and noise patterns appearing in seismic data streams. On a per station basis this yields generative statistical models that essentially summarize and generalize the information implicitly contained in the LEB allowing for classifying future an previously unseen seismic data of the same kind. Also, the system provides a numerical value reflecting the classification confidence potentially aiding the analyst is correcting or identifying events that are non-typical. The system has the potential for being implemented in real time: both feature computation/extraction as well as classification work on data segments/windows and seismic patterns of varying length, e.g., 12 sec. Various features are considered including spectral features, polarization information and statistical moments and moment ratios. All features are derived from a time-frequency-(amplitude) decomposition of the raw waveform data for each component, taking the 6 frequency bands currently in use at IDC into account. These different feature sets give rise to different DBN structures (model-feature scenarios) that probabilistically relate the features to each other depending on empirical observations and physical knowledge available. 1 week of waveform data is considered for training both the signal and noise classes. The performance of the classifier is measured on a separate test set from the same week of data but also on a 1-month data set, where 4 weeks of data is distributed over a one year period. In the system evaluation both a static approach as well as a sliding-window approach is taken. Binary classification accuracy, sensitivity and specificity is measured as well as a comparison to the SEL3 and LEB bulletins. Our results suggest that a model-feature scenario with spectral features performs well, with a relatively high accuracy. Moreover, when using the DBN trained on 1 week data to classify noise throughout the year, our analysis suggests that there is a seasonal noise variation/dependence (the degree depending on the station in question). Spectral analysis also confirms this observation. In effect this means that an adaptive approach to training the generative DBN for seismic noise needs to be employed in order to correctly identify noise at any given time.
Geophysical Journal …, 2012
Tsunami early warning (TEW) is a challenging task as a decision has to be made within few minutes... more Tsunami early warning (TEW) is a challenging task as a decision has to be made within few minutes on the basis of incomplete and error-prone data. Deterministic warning systems have difficulties in integrating and quantifying the intrinsic uncertainties. In contrast, probabilistic approaches provide a framework that handles uncertainties in a natural way. Recently, we have proposed a method using Bayesian networks (BNs) that takes into account the uncertainties of seismic source parameter estimates in TEW. In this follow-up study, the method is applied to 10 recent large earthquakes offshore Sumatra and tested for its performance. We have evaluated both the general model performance given the best knowledge we have today about the source parameters of the 10 events and the corresponding response on seismic source information evaluated in real-time. We find that the resulting site-specific warning level probabilities represent well the available tsunami wave measurements and observations. Difficulties occur in the real-time tsunami assessment if the moment magnitude estimate is severely over-or underestimated. In general, the probabilistic analysis reveals a considerably large range of uncertainties in the near-field TEW. By quantifying the uncertainties the BN analysis provides important additional information to a decision maker in a warning centre to deal with the complexity in TEW and to reason under uncertainty. Downloaded from 1280 L. Blaser et al.
Geophysical Journal …, 2011
The various uncertainties in the earthquake-triggered tsunami threat assessment are difficult to ... more The various uncertainties in the earthquake-triggered tsunami threat assessment are difficult to quantify and/or integrate into the tsunami early warning process. Uncertainties in the (seismic) input parameters and the lack of knowledge about the earthquake slip distribution contribute most to the total uncertainty in real-time evaluated tsunami assessment. We present a method how to integrate and quantify these uncertainties in the warning process by evaluating a tsunami warning level probability distribution with a Bayesian network (BN) approach. As soon as an earthquake is detected, the seismic source parameter estimates are evaluated and a probabilistic overview on different tsunami warning levels is provided, feasible to support a decision maker at a warning center with important additional data.
… to Reasoning with …, 2009
Early warning systems help to mitigate the impact of disastrous natural catastrophes on society b... more Early warning systems help to mitigate the impact of disastrous natural catastrophes on society by providing short notice of an imminent threat to geographical regions. For early tsunami warning, real-time observations from a seismic monitoring network can be used to estimate the severity of a potential tsunami wave at a specific site. The ability of deriving accurate estimates of tsunami impact is limited due to the complexity of the phenomena and the uncertainties in seismic source parameter estimates. Here we describe the use of a Bayesian belief network (BBN), capable of handling uncertain and even missing data, to support emergency managers in extreme time critical situations. The BBN comes about via model selection from an artifically generated database. The data is generated by ancestral sampling of a generative model defined to convey formal expert knowledge and physical/mathematical laws known to hold in the realm of tsunami generation. Hence, the database implicitly holds the information for learning a BBN capturing the required domain knowledge.
EGU General Assembly …, 2010
Tsunami early warnings are based on co-seismic evidences being the earliest available information... more Tsunami early warnings are based on co-seismic evidences being the earliest available information from a hazardous earthquake with the potential of causing a tsunami. Evaluations are generally done by applying rules derived from historic observation and making use of seismological expertise regarding regional tectonic contexts, faulting styles, occurrence frequency of large earthquakes and more. However, the co-seismic generation of a tsunami as well as the estimation of a potentially tsunamigenic event is prone to various uncertainties. As Bayesian networks (BNs) allow for integration and quantification of the uncertainties within the framework of probabilistic graphical models, we propose the usage of BNs for evaluating the imminence of a tsunami based on real-time seismic source parameter estimates. Earthquake parameter estimates (including uncertainties) are evaluated in real-time and the probabilities of tsunami threat levels are calculated and updated whenever new co-seismic evidence is available. The fast and efficient method gives an important additional information for the staff members at tsunami warning centers as it provides a probabilistic overview on the imminence of a tsunami for some particular costal region. In our work, we have developed a preliminary BN tsunami warning system for the region of Sumatra by extracting knowledge from a set of formulas describing the physical process from earthquake rupture to sea-floor deformation to tsunami wave propagation and finally shoaling at the coast. The physical knowledge was transformed by ancestral sampling to a synthetic database and thereof BNs were learned for several sites of interest along the Sumatran coast and the fore-arc islands. To determine the conditional probability of the tsunami amplitude a set of seven co-seismic variables was defined: epicenter, centroid, magnitude, hypocentral depth, rupture direction, rupture length and width. We illustrate the advantages of this approach by case studies of recent tsunamigenic earhtquakes offshore Sumatra with particular focus on the probabilistically sound treatment of uncertainties relevant to tsunami early warning problem.
Pure and Applied Geophysics, 2012
Knowledge Discovery in …, 2007
... 80(1), 170 186 (1990) 7. Kulesh, M., Holschneider, M., Diallo, MS: Geophysics wavelet librar... more ... 80(1), 170 186 (1990) 7. Kulesh, M., Holschneider, M., Diallo, MS: Geophysics wavelet library: Applications of the continuous wavelet transform to the polarization and dispersion analysis of signals. Comput-ers & Geoscience (Submitted 2007) 8. Kumar, P., Foufoula-Georgiou ...
Geophysical Research Abstracts, 2007
Geophysical Journal …, 2013
Statistics in Volcanology. …, 2006
... Merapi volcano is one of the most active and dangerous volcanoes of the earth. ... Due to the... more ... Merapi volcano is one of the most active and dangerous volcanoes of the earth. ... Due to the close relationship between the volcanic unrest and the occurrence of seismic events at ... of Merapi's seismicity plays an important role for recognizing major changes in the volcanic activity. ...
Journal of Volcanology and Geothermal Research, 2001
In order to monitor the seismic activity of Mt. Merapi (Indonesia) over a long period of time, we... more In order to monitor the seismic activity of Mt. Merapi (Indonesia) over a long period of time, we installed a permanent array of both broadband and short-period seismometers during the summer of 1997. Considering the requirements of an automatic classification and localization system for seismic monitoring and surveillance at active volcanoes, we split this network into three small-aperture arrays distributed around the volcano. We introduce here a newly developed method to determine the hypocenters in an automatic, non-linear manner using the coherence of seismic waves observed at the different arrays. To test this method, we analyze a swarm of VT-B events recorded by the network. The first step in this algorithm is based on a modified smoothed coherence transform. In the second step, we perform a semblance analysis applied to the 3D problem, evaluating the quality of the estimated relative onset-times. After more than one year of dormancy, Mt. Merapi renewed its activity at the end of June 1998. This gave us the opportunity to analyze all stages of dome growth, collapse and new intrusion of magma using the associated seismicity in a post-processing sense. This also allowed us to calibrate and test our newly developed automatic monitoring system using the more pronounced waveforms of VT-B events. By detecting and classifying different event types automatically, we are able to localize a large number of VT-B events, which occurred just before the initial eruption. We are also able to resolve some properties of the wavefield at Mt. Merapi, which are essential for further interpretations. Finally, the results show that the source region of the VT-B type seismicity just before the 1998 eruption is closely related to the region of subsequent high volcanic activity and therefore may represent a promising tool to forecast future eruptions.
Journal of volcanology …, 2004
... Merapi volcano [Lat. ... high sampling rate of the fumarole temperature enables us to compare... more ... Merapi volcano [Lat. ... high sampling rate of the fumarole temperature enables us to compare the temperature variation directly with the seismic activity. ... monitoring of fumarole temperatures offers a powerful tool for monitoring variations in degassing behavior at active volcanoes. ...
Earthquake …, 2010
Logic trees have become a popular tool to capture epistemic uncertainties in seismic hazard analy... more Logic trees have become a popular tool to capture epistemic uncertainties in seismic hazard analysis. They are commonly used by assigning weights to models on a purely descriptive basis (nominal scale). This invites the creation of unintended inconsistencies regarding ...
This study presents an unsupervised feature selection approach for the discovery of significant p... more This study presents an unsupervised feature selection approach for the discovery of significant patterns in seismic wavefields. We iteratively reduce the number of features generated from seismic time series by first considering significance of individual features. Significance testing is done by assessing the randomness of the time series with the Wald-Wolfowitz runs test and by comparing observed and theoretical variability of features. In a second step the in-between feature dependencies are assessed based on correlation hunting in feature subsets using Self-Organizing Maps (SOMs). We show the improved discriminative power of our procedure compared to manually selected feature subsets by cross-validation applied to synthetic seismic wavefield data. Furthermore, we apply the method to real-world data with the aim to define suitable features for earthquake detection and seismic phase classification in seismic recordings.
Uploads
Papers by Matthias Ohrnberger