NG41B-0515
Numerical Simulations for Earthquake Physics (Virtual California): Recent Results and Current Plans
Attention has recently been focused on the need for numerical simulations of interacting earthquake fault systems [1]. "Virtual California" is a topologically realistic model which has been used to simulate earthquakes on the San Andreas fault and its associated fault system. The Virtual California model includes elastic interactions among the faults in the model, driving at the correct plate tectonic rates, and frictional physics on the faults using the physics obtained from laboratory models with parameters consistent with the occurrence of historic earthquakes. An important consequence of the elastic interactions in the model is the appearance of correlations and space-time patterns of occurrence of events. Without the interactions, each fault element would behave independently; with the interactions, cooperative phenomena and patterns are observed. Previous numerical analysis of this simulation has focused on the distribution of recurrence times between large events. From this it is possible to compute the waiting times until the next great earthquake for California faults. Here we report on a variety of recent results. 1) We are investigating the stability of the numerical solutions and how these depend on the Greens function computations and other assumptions in the displacement discontinuity method; 2) We have developed a new method for generating and scoring a synthetic earthquake catalog utilizing Virtual California. The idea is to use paleoseismic data to identify intervals within the artificial data which most closely resemble the current seismic state of California; 3) We are working with collaborators to understand the interplay between fault system complexity and predictability. In this talk we summarize and discuss these issues, and indicate directions for the future.
NG41B-0516
Virtual California: Interplay Between Fault System Complexity, Dynamics, and Numerical Stability
Virtual California is a topologically realistic simulation of the interacting earthquake faults in California. Inputs to the model arise from field data, and typically include realistic fault system topologies, realistic long term slip rates, and realistic frictional parameters. Outputs from the simulations include synthetic earthquake sequences and space-time patterns together with associated surface deformation and strain patterns that are similar to those seen in nature. It is a type of model called a backslip model, so called because the loading in the model arises from negative slip or "backslip" applied to each fault boundary element at its geologically observed long- term rate of offset, V(x). We note that in this context, a fault element is regarded simply as a "degree of freedom", rather than a spatially coherent entity with geological meaning. All of the rectangular fault elements are embedded in an elastic half space, and they interact with each other by means of quasistatic elastic interactions, whose stress Greens functions are computed by means of a boundary element method. Frictional coefficients are assigned to each fault element, along with other frictional parameters, by means of a data assimilation technique. When the model is used to produce a simulation, the result is a history of slip on the fault elements in response to the driving forces. The interactions between the fault elements serve to organize the system so that, instead of a sequence of single elements breaking individually, multiple elements break simultaneously, producing large earthquakes. The dynamical evolution in the model is also produced by means of a stochastic, cellular automaton method, in which a random overshoot or undershoot component is added at the time of sliding of a element. The topology of the model fault system has evolved over time, becoming more complex, and using progressively finer scales of detail. Here we discuss the latest versions of the fault system topology and the associated numerical simulations. We focus in particular on the interplay between fault system complexity, and complexity in the dynamics. We find that the mesh scale of the boundary elements is strongly affected by the scale fault topology, and that the stability properties of the dynamics are likewise strongly affected.
NG41B-0517
Comparison of the submarine landslide by the sector collapse of Oshima-Oshima island in the northern part of Japan with the debris avalanche of off Kaimon volcano in the southern part of Japan and several landslide.
Oshima-Oshima island is an active volcano located in the eastern margin of the Japan Sea off Hokkaido. Oshima-Oshima consists of Higashi-yama somma, Nishi-yama somma, and the central cone. The latest eruption occurred at the 18th century. In a huge eruption on August 1741 , Nishiyama of western part of Oshima- Oshima collapsed toward the northern submarine slope , and the horseshoe shape caldera was formed. It is proposed by Katsui et al.(1977) , Satake and Kato(2001) that Japan Sea tsunami in 1741 was generated by this collapse. Detailed swath bathymetry surveys have been conducted around Oshima-Oshima by Hydrographic and Oceanographic Department of Japan in 1993. As a result, a large area of debris avalanche deposits has been discovered on the northern submarine flanks of Oshima-Oshima island. In addition, sidescan sonar surveys was also conducted by Hydrographic and Oceanographic Department of Japan and University of Tokyo in 1995. In 1997, the lower part of the debris avalanche deposit was investigated using submersible 'Shinkai 2000' by JAMSTEC(Japan Agency for Marine-Earth Science and Technology). It was confirmed that those deposits were Oshima origins (Kato,1997). We compiled and analyzed using these detailed bathymetry data and sidescan data. As a result , we clarified a detailed geographical features of debris avalanche and the limit of their distribution. Scarp of caldera rim continues to approximately 1100m under the sea. Oshima-Oshima has diameter of approximately 16km. Oshima-Oshima has also an estimated total edifice volume (subaerial and submarine) of 127km3 and rises about 2200m from its base in 1500m depth of water. Northern part of Oshima-Oshima, the scarp of caldera rim on the subaerial area consecutively continues up to about 1100m of depth. The scarp has 100m-300m high, and width of landslide valley is about 2km. Hammocky surface starts from 1100m depth of water. Sea mount of like spur is composed of the collapse deposits has almost extended to the whole area in the trough deeper than 1100m. Debris avalanche deposits have been identified up to 2200m depth and 24km from Oshima-Oshima island, H/L is approximately 0.12. Maximum sizes of debris avalanche block is up to 1-2km width and 100m high. We divided the debris avalanche deposits into three areas (Type A-C) by the topography. Type-A is the main sedimentary area in front of the collapse area and the form is like a spur with 100 to 130m height. Type-B has a form like a knoll or block. Type-C shows gradual slope containing a small rise. We compared these feature with debris avalanche of off Kaimon volcano in the southern part of Japan and the other debris avalanche (landslide, submarine landslide).
NG41B-0518
Pattern Informatics Analysis of the Pingtung (Taiwan) offshore doublet earthquakes with magnitude of 7 on December 26, 2006
The retrospective pattern informatics (PI) analysis of the Pingtung (Taiwan) offshore doublet earthquakes with magnitude of 7 has been made in this paper. Using the PI method, the epicenters of the Pingtung doublet earthquakes were found to exhibit signatures of anomalous activity associated with the precursory seismic activation and quiescence. In addition, the self-organizing spinodal (SOS) model of earthquakes can explain some observations on the temporal evolution in the frequency-magnitude distributions of seismicity preceding the Pingtung doublet earthquakes. Our study thus suggests that, by inspecting temporal variations in the frequency-magnitude distributions of seismicity and the hotspot locations with intense seismicity changes in the PI map, it is possible to shed light on the relevant preparation processes of the forthcoming catastrophic earthquake.
NG41B-0519
Recurrence-time and frequency-slip statistics of slip events on the creeping section of the San Andreas fault
The frequency-size statistics and recurrence time interval statistics at a point on a fault are important questions in seismic hazard assessments. Does a point on a fault obey the same statistics as earthquakes in a region do? This is a difficult question to answer because the number of repetitive earthquakes on a particular fault that have been observed is small. In order to overcome this difficulty we consider slip events on the creeping section of the San Andreas fault in central California. Sequencies of up to 100 events are obtained from creepmeter records. We compare the statistical distribution of recurrence times with the Brownian passage-time, lognormal, and Weibull distributions and using goodness-of-fit tests find that the Weibull is the preferred distribution. We also consider the frequency-amplitude distribution of slip events. We find that the data clearly do not obey a Gutenberg- Richter distribution. Instead there is a uniform distribution of event sizes for a large fraction of the events.
NG41B-0520
Spatial Orientation and Persistence Variability in Seismicity Patterns Related to Volcanic Activity in Hawaii: Applications to Hazard Assessment
Volcanic activity, which often includes a complex interplay between different factors such as magmatic processes, changes in crustal stress, earthquake events, etc., may lead to remarkable challenges regarding the identification of practically useful hazard patterns. These patterns may change over time at variable speed and in ways that are difficult to predict. Nevertheless, it is very important for hazard management to achieve a reasonable characterization of such patterns at different spatial and temporal scales. Events thread analysis (ETA) has the advantage of grasping spatial and temporal aspects of earthquake patterns together with their relation to scale. In this paper we apply ETA in three-dimensional physical space to study orientation-dependent scaling properties of seismicity patterns related to hot spot volcanism in Hawaii. The events threads are obtained by connecting series of hypocentres of successive earthquakes; projections of events threads corresponding to all the possible spatial orientations with a given angular resolution are constructed and subjected to detrended fluctuation analysis. The resulting isopersistence diagrams characterize space-time correlations in seismicity patterns for different time windows. One can thus apply a rigorous tool to identify and compare persistence in the scaling behaviour of seismicity patterns for any orientation. Isopersistence diagrams provide "dynamic fingerprints" of the system under investigation, which we discuss in the light of the eruption history and of information from geological and geophysical investigations. We show that in the case of Hawaii one can distinguish relatively long-lived, stable dynamic regimes, interrupted by departures that can be interpreted in terms of the changing relationships between different parts of the geosystem. Eventually, due to a new type of phase space based on isopersistence diagrams, one can follow the evolution of dynamic fingerprints of Hawaiian volcanism and obtain timely information about relevant changes in hazard patterns.
NG41B-0521
Bayesian Analysis of the Pattern Informatics Technique
The pattern informatics (PI) [Rundle et al., 2000; Tiampo et al., 2002; Holliday et al., 2005] is a technique that uses phase dynamics in order to quantify temporal variations in seismicity patterns. This technique has shown interesting results for forecasting earthquakes with magnitude greater than or equal to 5 in southern California from 2000 to 2010 [Rundle et al., 2002]. In this work, a Bayesian approach is used to obtain a modified updated version of the PI called Bayesian pattern informatics (BPI). This alternative method uses the PI result as a prior probability and models such as ETAS [Ogata, 1988, 2004; Helmstetter and Sornette, 2002] or BASS [Turcotte et al., 2007] in order to obtain the likelihood. Its result is similar to the one obtained by the PI: the determination of regions, known as hotspots, that are most susceptible to the occurrence of events with M=5 and larger during the forecast period. As an initial test, retrospective forecasts for the southern California region from 1990 to 2000 were made with both the BPI and the PI techniques, and the results are discussed in this work.
NG41B-0522
Clustering Earthquake Migration on the Tibetan Plateau
Most earthquakes occur from an accumulation of stress, so investigating the migration of earthquakes can help us in understanding the tendency of stress adjustments and accumulations for deducing the tendency of future earthquakes. In this work, we use a clustering technique to analyze the spatial migration of earthquakes so that we can derive a trend to the changes in stress and adjustments of earthquakes through time. We look at 38 years of earthquake data on the Tibetan plateau at discrete time slices ranging from 7 to 15 years. Nonlinear trends discovered in earthquake migration since 1970 are compared to historical data of migration in the Tibetan plateau for deriving patterns. Visualization techniques are used to enhance understanding of cluster relationships and behaviors since viewing clusters in three-dimensions at different orientations and rotations allows for necessary complex multidimensional analysis and correlation of earthquake data.
NG41B-0523
Hybrid zoneless probabilistic seismic hazard analysis: Test and first application to SW Germany
Zoneless approaches are one option to assess probabilistic seismic hazard (PSHA). They rely simply on smoothing the epicenter locations of past events according to the fractal distribution of earthquakes in space without defining seismic source zones. However, some shortcomings are connected with these zoneless approaches. Sometimes, one cannot apply one single magnitude completeness, especially when the study area covers different catalogue areas, where different completeness patterns are valid. Furthermore, the characteristic depths of the strongest earthquakes generally differ over larger areas. With respect to these two reasons, a hybrid approach for zoneless PSHA is introduced in this study; i.e. a combination of a zoneless approach with a seismic source model based solely on the large scale geological architecture. This large scale model meets the requirements of introducing different magnitude completeness. But the probability of the earthquake occurrence is based on a pure zoneless approach. Then, instead of a time set of magnitude completeness for the entire studying area, as it is the standard procedure for zoneless approaches, various time sets for different large zones are considered. Another innovation of our approach is to take into account the actual depth of each earthquake rather than assuming a certain weighting set of depths. We apply the hybrid zoneless approach to the SW part of Germany, which represents one of the highest seismicity areas in the country. The use of area specific magnitude completeness and actual depths of all events clearly results in a more realistic PSHA. Furthermore, we compare the results of the hybrid zoneless approach with the results of a recent study using a zone-based methodology. The zone-based approach shows more significant contrasts in the study area; i.e the hybrid zoneless approach provides a rather smoothed pattern. This corresponds to comparisons between zoneless and zone-based approaches by previous studies.
NG41B-0524
Failure of Rock Masses from Nucleation and Growth of Microscopic Defects and Disorder
Observations of rock masses over a range of spatial scales indicate that the failure modes of these systems, such as fracture, demonstrate scale invariant deformation, or power law behavior, characteristic of complex non- linear systems. These are observed in both laboratory settings in acoustic emission experiments, as well as in large scale field settings associated with tectonic faults (Gutenberg-Richter magnitude-frequency relation; Omori relation for aftershocks). One important reason for this behavior is that driven threshold systems of rock masses in which defects interact with long range interactions display near mean field dynamics and ergodic behavior. This result, which was first proposed on the basis of simulations and theory, was subsequently observed in field observations on the tectonic scale. We are investigating the failure of rock masses resulting from the complex physics of microscopic dynamical processes in rocks, as manifested in the nucleation and growth of defects, microcracks, damage, and macroscopic fracture. These processes are a result of the complex emergent dynamics of self-organizing geological materials which we analyze using the methods of statistical physics and large scale simulations employing both molecular dynamics and Monte Carlo methods. Fully interacting fields of defects and damage are generally not included in most current models for material deformation. Instead, defect density and damage fields are assumed to be non-interacting or dilute, implying a strictly mean field approach. We use statistical physics methods to understand the dynamics of interacting defect and damage fields, made possible by the construction and use of statistical field theories, to greatly improve our predictive capability for the macroscopic failure of materials. In this talk I will summarize our current understanding of rock deformation as a response to stresses driving the evolution of damage and defects within the rock mass. The important quantities to compute are the nucleation rate, or its inverse, lifetime to failure. I will also discuss applications of this research to rock deformation across a range of spatial and temporal scales.
NG41B-0525
Earthquake Scaling Laws and Seismic Hazard and Risk Analyses
In flood and wind hazard analyses, a hazard level and associated recurrence interval, such as peak discharge of a 100-year flood and 3-second gust wind speed of a 50-year storm, must be known at a specific site. Similarly, ground motion hazard with a recurrence interval at a specific site can be derived from earthquake spatial and temporal scaling laws. In seismology, earthquake magnitude and occurrence frequency follow the Gutenberg- Richter (temporal) relationship between the cumulative number of earthquakes and magnitude. Observed ground motion at a site with distance from an earthquake can be described by the ground-motion attenuation relationship (spatial). Combining the Gutenberg-Richter and ground-motion attenuation relationships results in a relationship between ground motion with an uncertainty at a site of interest and recurrence interval; i.e., a seismic hazard curve. In comparison with probabilistic seismic hazard analysis (PSHA), a widely used method for seismic hazard and risk analyses, seismic hazard derived directly from the scaling laws has a clear physical and statistical meaning. The seismic hazard has a similar meaning as those from the flood and wind hazard analyses and can be used in a similar way for risk analysis.
NG41B-0526
Non-Poissonian Distribution of Tsunami Waiting Times
Analysis of the global tsunami catalog indicates that tsunami waiting times deviate from an exponential distribution one would expect from a Poisson process. Empirical density distributions of tsunami waiting times were determined using both global tsunami origin times and tsunami arrival times at a particular site with a sufficient catalog: Hilo, Hawai'i. Most sources for the tsunamis in the catalog are earthquakes; other sources include landslides and volcanogenic processes. Both datasets indicate an over-abundance of short waiting times in comparison to an exponential distribution. Two types of probability models are investigated to explain this observation. Model (1) is a universal scaling law that describes long-term clustering of sources with a gamma distribution. The shape parameter (γ) for the global tsunami distribution is similar to that of the global earthquake catalog γ=0.63-0.67 [Corral, 2004]. For the Hilo catalog, γ is slightly greater (0.75-0.82) and closer to an exponential distribution. This is explained by the fact that tsunamis from smaller triggered earthquakes or landslides are less likely to be recorded at a far-field station such as Hilo in comparison to the global catalog, which includes a greater proportion of local tsunamis. Model (2) is based on two distributions derived from Omori's law for the temporal decay of triggered sources (aftershocks). The first is the ETAS distribution derived by Saichev and Sornette [2007], which is shown to fit the distribution of observed tsunami waiting times. The second is a simpler two-parameter distribution that is the exponential distribution augmented by a linear decay in aftershocks multiplied by a time constant Ta. Examination of the sources associated with short tsunami waiting times indicate that triggered events include both earthquake and landslide tsunamis that begin in the vicinity of the primary source. Triggered seismogenic tsunamis do not necessarily originate from the same fault zone, however. For example, subduction-thrust and outer-rise earthquake pairs are evident, such as the November 2006 and January 2007 Kuril Islands tsunamigenic pair. Because of variations in tsunami source parameters, such as water depth above the source, triggered tsunami events with short waiting times are not systematically smaller than the primary tsunami.
NG41B-0527
Modelling Aftershock Sequences With Missing Data
We extend the previous models for aftershock sequences (ETAS and BASS) by assuming that the magnitudes m occuring at a certain time t after a large shock of magnitude m0 can only be observed with probability p(mo,m,t)= 10γ (mc(mo,t)-m) where mc(mo,t) =m0 -4,5 -0.75*\log10(t) describes the time-dependent threshold of Helmstetter, Kagan and Jackson, above which the earthquake catalogues should be complete. We show that this ansatz yields a reasonable description of the time dependence of (a) the cumulated magnitude distribution and (b) the magnitude-dependent rate after large mainshocks when comparing our model results with 4 Californian aftershock sequences (Parkfield, Landers, Northridge and Hectormine.
NG41B-0528
Water Level Dynamics in the Great Lakes of North America
Anthropogenic as well as natural fluctuations such as precipitation, runoff, snowmelt, retention time, evaporation, and outflow all contribute to water levels observed in the Great Lakes. Verified hourly water level data for five stations in Lake Michigan and four stations in Lake Superior were obtained from NOAA and examined. For each station, an hourly time series ranging from 20 to 30 years in duration was decimated to produce a time series with four hour intervals. Four distinct regions of scaling are observed with inflection points at approximately 1 day, 5 days, and 30 - 60 days. For time scales of less than one day, the power-scaling exponent (β) ranges from 0.1 to 0.5, indicating a white noise. From 1 day to 5 - 7 days, β ranges from 1.5 to 2.6, indicating moderate to strong persistence which is probably due to frontal movements of weather systems. On timescales between 5 days and 30 - 60 days, β ranges from 0.1 to 0.4, again indicating a white noise which may be due to monthly and seasonal weather variations within the Great Lakes System. Beyond 30 - 60 days, all stations exhibit strong persistence, with β between 2.1 and 2.7. Barometric pressure and precipitation data also exhibit scaling with β different from the water level data, but with breaks in slope occurring over the same period of time. The variations observed in the changing β of water levels (environmental noise) are likely to have biological impacts on population dynamics of organisms, including rates of survival or extinction (Batchhelder and Powell 2002). Understanding biological-physical coupling and the impact of water level fluctuations is fundamental to ecosystem dynamics.