@phdthesis{Lilienkamp2024,
  author    = {Lilienkamp, Henning},
  title     = {Enhanced computational approaches for data-driven characterization of earthquake ground motion and rapid earthquake impact assessment},
  doi       = {10.25932/publishup-63195},
  url       = {http://nbn-resolving.de/urn:nbn:de:kobv:517-opus4-631954},
  school      = {Universit{\"a}t Potsdam},
  pages     = {x, 145},
  year      = {2024},
  abstract  = {Rapidly growing seismic and macroseismic databases and simplified access to advanced machine learning methods have in recent years opened up vast opportunities to address challenges in engineering and strong motion seismology from novel, datacentric perspectives. In this thesis, I explore the opportunities of such perspectives for the tasks of ground motion modeling and rapid earthquake impact assessment, tasks with major implications for long-term earthquake disaster mitigation. In my first study, I utilize the rich strong motion database from the Kanto basin, Japan, and apply the U-Net artificial neural network architecture to develop a deep learning based ground motion model. The operational prototype provides statistical estimates of expected ground shaking, given descriptions of a specific earthquake source, wave propagation paths, and geophysical site conditions. The U-Net interprets ground motion data in its spatial context, potentially taking into account, for example, the geological properties in the vicinity of observation sites. Predictions of ground motion intensity are thereby calibrated to individual observation sites and earthquake locations. The second study addresses the explicit incorporation of rupture forward directivity into ground motion modeling. Incorporation of this phenomenon, causing strong, pulse like ground shaking in the vicinity of earthquake sources, is usually associated with an intolerable increase in computational demand during probabilistic seismic hazard analysis (PSHA) calculations. I suggest an approach in which I utilize an artificial neural network to efficiently approximate the average, directivity-related adjustment to ground motion predictions for earthquake ruptures from the 2022 New Zealand National Seismic Hazard Model. The practical implementation in an actual PSHA calculation demonstrates the efficiency and operational readiness of my model. In a follow-up study, I present a proof of concept for an alternative strategy in which I target the generalizing applicability to ruptures other than those from the New Zealand National Seismic Hazard Model. In the third study, I address the usability of pseudo-intensity reports obtained from macroseismic observations by non-expert citizens for rapid impact assessment. I demonstrate that the statistical properties of pseudo-intensity collections describing the intensity of shaking are correlated with the societal impact of earthquakes. In a second step, I develop a probabilistic model that, within minutes of an event, quantifies the probability of an earthquake to cause considerable societal impact. Under certain conditions, such a quick and preliminary method might be useful to support decision makers in their efforts to organize auxiliary measures for earthquake disaster response while results from more elaborate impact assessment frameworks are not yet available. The application of machine learning methods to datasets that only partially reveal characteristics of Big Data, qualify the majority of results obtained in this thesis as explorative insights rather than ready-to-use solutions to real world problems. The practical usefulness of this work will be better assessed in the future by applying the approaches developed to growing and increasingly complex data sets.},
  language  = {en}
}
@phdthesis{Bora2015,
  author    = {Bora, Sanjay Singh},
  title     = {Regionally adaptable ground-motion Prediction Equations (GMPEs) for seismic hazard analysis},
  url       = {http://nbn-resolving.de/urn:nbn:de:kobv:517-opus4-88806},
  school      = {Universit{\"a}t Potsdam},
  pages     = {xiv, 138},
  year      = {2015},
  abstract  = {Adjustment of empirically derived ground motion prediction equations (GMPEs), from a data- rich region/site where they have been derived to a data-poor region/site, is one of the major challenges associated with the current practice of seismic hazard analysis. Due to the fre- quent use in engineering design practices the GMPEs are often derived for response spectral ordinates (e.g., spectral acceleration) of a single degree of freedom (SDOF) oscillator. The functional forms of such GMPEs are based upon the concepts borrowed from the Fourier spectral representation of ground motion. This assumption regarding the validity of Fourier spectral concepts in the response spectral domain can lead to consequences which cannot be explained physically. In this thesis, firstly results from an investigation that explores the relationship between Fourier and response spectra, and implications of this relationship on the adjustment issues of GMPEs, are presented. The relationship between the Fourier and response spectra is explored by using random vibration theory (RVT), a framework that has been extensively used in earthquake engineering, for instance within the stochastic simulation framework and in the site response analysis. For a 5\% damped SDOF oscillator the RVT perspective of response spectra reveals that no one-to-one correspondence exists between Fourier and response spectral ordinates except in a limited range (i.e., below the peak of the response spectra) of oscillator frequencies. The high oscillator frequency response spectral ordinates are dominated by the contributions from the Fourier spectral ordinates that correspond to the frequencies well below a selected oscillator frequency. The peak ground acceleration (PGA) is found to be related with the integral over the entire Fourier spectrum of ground motion which is in contrast to the popularly held perception that PGA is a high-frequency phenomenon of ground motion. This thesis presents a new perspective for developing a response spectral GMPE that takes the relationship between Fourier and response spectra into account. Essentially, this frame- work involves a two-step method for deriving a response spectral GMPE: in the first step two empirical models for the FAS and for a predetermined estimate of duration of ground motion are derived, in the next step, predictions from the two models are combined within the same RVT framework to obtain the response spectral ordinates. In addition to that, a stochastic model based scheme for extrapolating the individual acceleration spectra beyond the useable frequency limits is also presented. To that end, recorded acceleration traces were inverted to obtain the stochastic model parameters that allow making consistent extrapola- tion in individual (acceleration) Fourier spectra. Moreover an empirical model, for a dura- tion measure that is consistent within the RVT framework, is derived. As a next step, an oscillator-frequency-dependent empirical duration model is derived that allows obtaining the most reliable estimates of response spectral ordinates. The framework of deriving the response spectral GMPE presented herein becomes a self-adjusting model with the inclusion of stress parameter (∆σ) and kappa (κ0) as the predictor variables in the two empirical models. The entire analysis of developing the response spectral GMPE is performed on recently compiled RESORCE-2012 database that contains recordings made from Europe, the Mediterranean and the Middle East. The presented GMPE for response spectral ordinates should be considered valid in the magnitude range of 4 ≤ MW ≤ 7.6 at distances ≤ 200 km.},
  language  = {en}
}
@phdthesis{Hakimhashemi2009,
  author    = {Hakimhashemi, Amir Hossein},
  title     = {Time-dependent occurrence rates of large earthquakes in the Dead Sea fault zone and applications to probabilistic seismic hazard assessments},
  url       = {http://nbn-resolving.de/urn:nbn:de:kobv:517-opus-52486},
  school      = {Universit{\"a}t Potsdam},
  year      = {2009},
  abstract  = {Die relativ hohe seismische Aktivit{\"a}t der Tote-Meer-St{\"o}rungszone (Dead Sea Fault Zone - DSFZ) ist mit einem hohen Gefahrenpotential verbunden, welches zu einem erheblichen Erdbebenrisiko f{\"u}r die Ballungszentren in den L{\"a}ndern Syrien, Libanon, Pal{\"a}stina, Jordanien und Israel f{\"u}hrt. Eine Vielzahl massiver, zerst{\"o}rerischer Erdbeben hat sich in diesem Raum in den letzten zwei Jahrtausenden ereignet. Ihre Wiederholungsrate zeigt Anzeichen f{\"u}r eine zeitliche Abh{\"a}ngigkeit, insbesondere wenn lange Zeitr{\"a}ume in Betracht gezogen werden. Die Ber{\"u}cksichtigung der zeitlichen Abh{\"a}ngigkeit des Auftretens von Erdbeben ist f{\"u}r eine realistische seismische Gef{\"a}hrdungseinsch{\"a}tzung von großer Bedeutung. Ziel der vorliegenden Arbeit ist es, anhand des zeitabh{\"a}ngigen Auftretens von Erdbeben eine robuste wahrscheinlichkeitstheoretische seismische Gef{\"a}hrdungseinsch{\"a}tzung am Beispiel der DSFZ zu entwickeln. Mittels dieser Methode soll die zeitliche Abh{\"a}ngigkeit des Auftretens von großen Erdbeben (Mw ≥ 6) untersucht und somit eine Gef{\"a}hrdungseinsch{\"a}tzung f{\"u}r das Untersuchungsgebiet getroffen werden. Prim{\"a}r gilt es zu pr{\"u}fen, ob das Auftreten von großen Erdbeben tats{\"a}chlich einer zeitlichen Abh{\"a}ngigkeit unterliegt und wenn ja, inwiefern diese bestimmt werden kann. Zu diesem Zweck werden insgesamt vier zeitabh{\"a}ngige statistische Verteilungen (Weibull, Gamma, Lognormal und Brownian Passage Time (BPT)) sowie die zeitunabh{\"a}ngige Exponentialverteilung (Poisson-Prozess) getestet. Zur Absch{\"a}tzung der jeweiligen Modellparameter wird eine modifizierte Methode der gewichteten Maximum-Likelihood-Sch{\"a}tzung (MLE) verwendet. Um einzusch{\"a}tzen, ob die Wiederholungsrate von Erdbeben einer unimodalen oder multimodalen Form folgt, wird ein nichtparametrischer Bootstrap-Test f{\"u}r Multimodalit{\"a}t durchgef{\"u}hrt. Im Falle einer multimodalen Form wird neben der MLE zus{\"a}tzlich eine Erwartungsmaximierungsmethode (EM) herangezogen. Zur Auswahl des am besten geeigneten Modells wird zum einem das Bayesschen Informationskriterium (BIC) und zum anderen der modifizierte Kolmogorow-Smirnow-Goodness-of-Fit-Test angewendet. Abschließend werden mittels der Bootstrap-Methode die Konfidenzintervalle der gesch{\"a}tzten Parameter berechnet. Als Datengrundlage werden Erdbeben mit Mw ≥ 6 seit dem Jahre 300 n. Chr. herangezogen. Das Untersuchungsgebiet erstreckt sich von 29.5° N bis 37° N und umfasst ein ca. 40 km breites Gebiet entlang der DSFZ. Aufgrund der seismotektonischen Situation im Untersuchungsgebiet wird zwischen einer s{\"u}dlichen, zentralen und n{\"o}rdlichen Subzone unterschieden. Dabei kann die s{\"u}dliche Subzone aus Mangel an Daten nicht f{\"u}r die Analysen herangezogen werden. Die Ergebnisse f{\"u}r die zentrale Subzone zeigen keinen signifikanten multimodalen Verlauf der Wiederholungsrate von Erdbeben. Des Weiteren ist kein signifikanter Unterschied zwischen den zeitabh{\"a}ngigen und dem zeitunabh{\"a}ngigem Modell zu verzeichnen. Da das zeitunabh{\"a}ngige Modell vergleichsweise einfach interpretierbar ist, wird die Wiederholungsrate von Erdbeben in dieser Subzone unter Annahme der Exponentialverteilungs-Hypothese abgesch{\"a}tzt. Sie wird demnach als zeitunabh{\"a}ngig betrachtet und betr{\"a}gt 9.72 * 10-3 Erdbeben (mit Mw ≥ 6) pro Jahr. Einen besonderen Fall stellt die n{\"o}rdliche Subzone dar. In diesem Gebiet tritt im Durchschnitt alle 51 Jahre ein massives Erdbeben (Mw ≥ 6) auf. Das letzte Erdbeben dieser Gr{\"o}ße ereignete sich 1872 und liegt somit bereits 137 Jahre zur{\"u}ck. Somit ist in diesem Gebiet ein Erdbeben dieser St{\"a}rke {\"u}berf{\"a}llig. Im statistischen Mittel liegt die Zeit zwischen zwei Erdbeben zu 96\% unter 137 Jahren. Zudem wird eine deutliche zeitliche Abh{\"a}ngigkeit der Erdbeben-Wiederauftretensrate durch die Ergebnisse der in der Arbeit neu entwickelten statistischen Verfahren best{\"a}tigt. Dabei ist festzustellen, dass die Wiederholungsrate insbesondere kurz nach einem Erdbeben eine sehr hohe zeitliche Abh{\"a}ngigkeit aufweist. Am besten repr{\"a}sentiert werden die seismischen Bedingungen in der genannten Subzone durch ein bi-modales Weibull-Weibull-Modell. Die Wiederholungsrate ist eine glatte Zeitfunktion, welche zwei H{\"a}ufungen von Datenpunkten in der Zeit nach dem Erdbeben zeigt. Dabei umfasst die erste H{\"a}ufung einen Zeitraum von 80 Jahren, ausgehend vom Zeitpunkt des jeweiligen Bebens. Innerhalb dieser Zeitspanne ist die Wiederholungsrate extrem zeitabh{\"a}ngig. Die Wiederholungsrate direkt nach einem Beben ist sehr niedrig und steigert sich in den folgenden 10 Jahren erheblich bis zu einem Maximum von 0.024 Erdbeben/Jahr. Anschließend sinkt die Rate und erreicht ihr Minimum nach weiteren 70 Jahren mit 0.0145 Erdbeben/Jahr. An dieses Minimum schließt sich die zweite H{\"a}ufung von Daten an, dessen Dauer abh{\"a}ngig von der Erdbebenwiederholungszeit ist. Innerhalb dieses Zeitfensters nimmt die Erdbeben-Wiederauftretensrate ann{\"a}hernd konstant um 0.015 Erdbeben/Jahr zu. Diese Ergebnisse bilden die Grundlage f{\"u}r eine zeitabh{\"a}ngige probabilistische seismische Gef{\"a}hrdungseinsch{\"a}tzung (PSHA) f{\"u}r die seismische Quellregion, die den n{\"o}rdlichen Raum der DSFZ umfasst.},
  language  = {en}
}
@phdthesis{Hainzl2011,
  author    = {Hainzl, Sebastian},
  title     = {Earthquake triggering and interaction},
  url       = {http://nbn-resolving.de/urn:nbn:de:kobv:517-opus-50095},
  school      = {Universit{\"a}t Potsdam},
  year      = {2011},
  abstract  = {Earthquake faults interact with each other in many different ways and hence earthquakes cannot be treated as individual independent events. Although earthquake interactions generally lead to a complex evolution of the crustal stress field, it does not necessarily mean that the earthquake occurrence becomes random and completely unpredictable. In particular, the interplay between earthquakes can rather explain the occurrence of pronounced characteristics such as periods of accelerated and depressed seismicity (seismic quiescence) as well as spatiotemporal earthquake clustering (swarms and aftershock sequences). Ignoring the time-dependence of the process by looking at time-averaged values - as largely done in standard procedures of seismic hazard assessment - can thus lead to erroneous estimations not only of the activity level of future earthquakes but also of their spatial distribution. Therefore, it exists an urgent need for applicable time-dependent models. In my work, I aimed at better understanding and characterization of the earthquake interactions in order to improve seismic hazard estimations. For this purpose, I studied seismicity patterns on spatial scales ranging from hydraulic fracture experiments (meter to kilometer) to fault system size (hundreds of kilometers), while the temporal scale of interest varied from the immediate aftershock activity (minutes to months) to seismic cycles (tens to thousands of years). My studies revealed a number of new characteristics of fluid-induced and stress-triggered earthquake clustering as well as precursory phenomena in earthquake cycles. Data analysis of earthquake and deformation data were accompanied by statistical and physics-based model simulations which allow a better understanding of the role of structural heterogeneities, stress changes, afterslip and fluid flow. Finally, new strategies and methods have been developed and tested which help to improve seismic hazard estimations by taking the time-dependence of the earthquake process appropriately into account.},
  language  = {en}
}