Abstract: The 2004 earthquake on the Sumatra trench was followed by a series of large events, along the plate boundary and within the subducting plate. Post-2004, the Wharton Basin was unusually active, notably the 2012, largest strike-slip earthquakes ever recorded. Here we discuss two types of earthquakes, those close to the trench (intraslab) and those, away from the trench (intraplate). The 2005 and 2010 earthquakes were sourced closer to the trench and their mechanisms show strike-slip faulting with some oblique slip. With its shallow dip-slip component, the latter generated a mild tsunami whereas the former did not as it exhibited no dip-slip component. Sourced at ~ 16 km and 35 km depth respectively, the 2005 and the 2010 earthquakes are believed to be the result of reactivation of the N-S fossil fabric within the subducting slab, following the post-2004 redistribution of stresses. Located within the oceanic plate, the 2012 and the 2016 earthquakes failed by pure strike-slip mechanism. While the deep centroid estimates from routine long-period modeling using Preliminary Reference Earth Model (PREM) places the focal depths of these intraplate earthquakes at ~ 50 km depth, other models propose shallower focus, above the 600°C isotherm depth. Thus, the source model for the 2016 earthquake places its depth at ~ 35 km and the rupture is interpreted as N-S and bilateral, which transects the orthogonal ridge-parallel fabric. Models of the 2012 earthquake also argue for complex and conjugate ruptures that involve the E-W fossil ridges. The deeper geometry of the faults inferred from geophysical evidence in the source zone of the 2012 earthquakes however suggest ~north-striking, east-dipping structures which possibly act as slip planes following the Olivine crystal preferred orientations, a model applicable also for the 2016 event. The 12-year history of the Wharton Basin presents ample evidence for its various modes of reawakening, following the 2004 megathrust earthquake.
Slidecast:
https://vimeo.com/277702750
Abstract: Machine learning has been applied to a number of problems in geoscience for decades. In general, applications used only portions of available data due in part to limited data and computing power. For instance, considerable work has been done using earthquake catalogs. Catalogs are assembled by hand or in some automated manner applying classical data processing techniques. Thus, out of both necessity and lack of understanding, vast amounts of existing data have gone unused. We know that fundamental information is embedded in many conventional geophysical continuous data signals, but that our ability to extract this information is limited by several factors: (a) it is often not apparent to human inspection and/or (b) when it is apparent, it requires a protracted expert analysis to interpret; (c) and/or we may be asking the wrong question of the data (human bias or ignorance). Our strategy is based on lessons learned from previous attempts to apply data-analytics, such as failed attempts at earthquake prediction. While machine learning offers a powerful path to extracting information rapidly from complex datasets, it must be strongly coupled to a fundamental understanding of the physical system to be meaningful and believable. We provide an overview of our work applying machine learning approaches to continuous seismic data streams. The primary focus of the work described will be on faulting and earthquakes, however we are turning our attention to many geological phenomena at many scales.
Slidecast:
https://vimeo.com/277698700
Abstract: On January 10, 2018 at 2h51 UTC (January 9 between 8:51 and 10:51 PM local time) a Mw 7.6 earthquake occurred offshore Swan Island (Honduras). Fortunately there was no significant damage from the earthquake and the tsunami was small. This was the first time the Pacific Tsunami Warning Center (PTWC) issued a Tsunami Threat for the Caribbean since the enhanced tsunami products for the UNESCO IOC Tsunami and Other Coastal Hazards Warning System for the Caribbean and Adjacent Regions (CARIBE EWS) were adopted in 2015. It was also the first time a Tsunami Advisory was issued for Puerto Rico and the US and British Virgin Islands (PR/VI). While the first message that was issued was based on the earthquake, following messages integrated the tsunami forecast and observations. The earthquake was felt in some countries; others for whom the Tsunami Threat/Advisory was issued did not report feeling the earthquake which is often the first sign of a potential threat, making even more important the official alerting process. Both the international and domestic text products were timely sent by the PTWC to designated authorities and posted on the tsunami.gov website. For the international stakeholders it is their responsibility thru their designated Tsunami Warning Focal Point/National Tsunami Warning Center to determine the threat level. In the case of PR/VI, the alert level (Warning, Advisory, Watch or Information) is established by the PTWC. In either case, international or domestic, it is the responsibility of the national/local governments to decide on the actions to be taken (evacuate or not evacuate) and to notify the public. The event timeline, as well as the UNESCO IOC CARIBE EWS after action report will be presented. In addition to identifying the strengths and gaps of tsunami warning system, the event review is also important considering the impact that hurricanes Irma and Maria had on local communication and emergency response systems and the public.
Slidecast:
https://vimeo.com/277700711
Abstract: The North Anatolian Fault Zone in the Sea of Marmara did not generate a M>7.0 earthquake since 1766. This fault section stores ~20 mm annual slip deficit and therefore is expected to accommodate at least one at most three M>7.0 in near future. In this study, we continuously monitor this critically strained fault section using seismo-geodetic stations that are equipped with 100Hz sampling seismographs and 1Hz sampling GPS recorders. This configuration allows covering a broad spectral band and is sensitive to both fast/slow tectonic motions at large/small temporal and spatial scale, from milliseconds to years, from centimeters to tens of kilometers. Therefore, recorded seismo-geodetic data will be used to identify (1) along-fault variation of the slip deficit, (2) fault segmentation, (3) interaction between slip-deficit and background seismicity, (4) pre-seismic seismo-geodetic symptoms and (5) co-seismic slip in case of M 7.0 earthquake(s).
Slidecast:
https://vimeo.com/278051069
Abstract: The eastern North America-Caribbean (NA-CAR) plate boundary near the islands of Hispaniola and Puerto Rico is a complex transition zone in which strain is accommodated by two transform fault systems and oblique subduction. In 2013, scientists from Baylor University, the Autonomous University of Santo Domingo, and the Puerto Rico Seismic Network deployed 16 broadband stations on the Dominican Republic to expand the local permanent network. The goal of the Greater Antilles Seismic Program (GrASP) is to combine its data with that from permanent networks in Puerto Rico, Haiti, Cuba, the Cayman Islands, and Jamaica to develop a better understanding of the crust and upper mantle structure in the Northeastern Caribbean (Greater Antilles). One important goal of GrASP is to develop robust velocity models that can be used to improve earthquake location and seismic hazard efforts. In this study, we focus on obtaining Rayleigh wave group velocity maps from ambient noise tomography. Here we use vertical-component broadband data recorded at 53 stations between 2010 to present, to obtain Green’s functions between 1165 pairs of stations. From these, we obtain dispersion curves by the application of FTAN methods with phase-matched filtering. Group velocity maps are generated between 4 to 40 s. One-dimensional shear wave velocity models are generated for selected sites by convolving group velocity results with depth kernels produced by forward modeling. The Markov Chain Monte Carlo method is applied to estimate the maximum likelihood S wave velocity model and, specifically, Moho depth. Results show strong correlations with large-scale geological and tectonic features for periods between 4 – 40 s, such as the Cordillera Central in both the Dominican Republic and Puerto Rico, the Mona Passage, and the NA-CAR subduction zone. Preliminary 1D shear wave velocity models suggest that the Moho underneath the Dominican Republic and Puerto Rico is in the range of 28 to 32 km.
Slidecast:
https://vimeo.com/278019459
Abstract: A moment tensor is a symmetric matrix that expresses the source for a seismic event. Uncertainty characterization of moment tensors is vital for any interpretation about the moment tensor, such as whether an event is likely to be an earthquake or not. We provide a method for characterizing and visualizing the uncertainty for a full moment tensor. Our uncertainty summary has four components: (1) variation in waveform misfit for the best-fitting moment tensor at each lune point; (2) probability density p(v,w) for moment tensor source type; (3) confidence curve Pcon(V); (4) confidence parameter Pav, which is the area under the confidence curve, with large values representing high concentration of probability near the best-fitting moment tensor. These characterizations are facilitated by a uniform parameterization of moment tensors with fixed magnitude. The parameters describe the source type, with v and w, and the orientation with strike angle, cosine dip angle, and rake angle. The parameterization is uniform in the sense that a uniform distribution of 5-tuples in the coordinate domain corresponds to a uniform distribution in the moment tensor space. Stated otherwise, volumes (i.e. 5-volumes) in the coordinate domain are proportional to the corresponding volumes of moment tensors. Uncertainty in source type and be derived from the probability density p(v,w) in source type, which is depicted easily on the vw rectangular domain. We discuss how this approach could be applied to event screening.
Slidecast:
https://vimeo.com/278017773
