Abstract: We used an unsupervised cortical algorithm to detect in real time anomalies in simulated and real seismic signals. The algorithm works on-line receiving continuous seismic acceleration signals from a streaming station. This could be a software source that is simulating the data stream from a suitable algorithm, or could be data timeseries from a database of real seismic events. Our system is based on the Numenta HTM engine that implements a cortical algorithm based on an innovative neural model in which dendrites act as coincidence detectors and synapses are formed continuously. The HTM networks learns and adapts to the seismic background noises and it is able to notice anomalies in real time and without the necessity of supervision. This approach to seismic detection could be useful not only to recognize earthquakes and contribute to early warning networks, but also could be of importance for detecting variations of the background noises characteristics that are not detectable with conventional methods or by conventional machine learning methods. We will show tests that demonstrate the ability of the HTM algorithm to recognize efficiently simulated and real earthquake waves we will report on our tests on seismic noises of specific types. Our approach points forward the possibility of the use of HTM networks to recognize signals form different kinds of seismic sources and types.
Media has not been submitted for this Presentation
Abstract: For more than two decades the Puerto Rico – Virgin Islands region has been monitored for crustal deformation using GPS geodesy. This objective was first done through a campaign style network that consisted of a dozen sites collecting one measurement annually during the nineties. Today, fifteen continuously operating GPS sites distributed throughout Puerto Rico and the Virgin Islands form the Puerto Rico Seismic Network (PRSN) GPS network. These sites provide data in two forms; 1Hz Real-time data available through an NTRIP Caster and 1Hz and 30sps BINEX archived files transferred from site to PRSN hourly and daily, respectively. Here we present data flow methods employed at PRSN to bring continuously operated GPS station data from the field back to PRSN, both in Real-time form and archived. This presentation seeks to encourage other GPS or GNSS networks in the Caribbean, particularly those hosting COCONet sites, to develop their own data acquisition methods based on the solutions that have been either adopted or developed at the PRSN. Empowering Caribbean networks to have control over their data is critical for capacity building and enables the region to develop scientifically as well as increase their education in matters of earthquake and tsunamis.
Media has not been submitted for this Presentation
Abstract: Long Valley in eastern California is sandwiched by the Sierra Nevada and the Basin and Range Province. There has been a stress perturbation in the vicinity of the caldera with respect to the regional stress field. Previous studies suggest that this perturbation is a result of the left-step offset in the Sierran range bounding normal faults instead of magma chamber inflation beneath Long Valley Caldera. In this study, we take advantage of the abundant seismic data and state-of-the-art finite-element numerical modeling to reinvestigate this local stress anomaly. We utilize the HASH program to compute focal mechanisms from P-wave first motion polarity observations for the relocated earthquakes between 1984 and 2014. The final ~42,000 good-quality focal solutions are then used to invert for the stress fields in 11 sub-areas by applying the SATSI algorithm. The orientations of the inverted minimum horizontal principal stresses (ShMIN) greatly agree with previous studies based on analyses of focal mechanisms, borehole breakouts, and fault offsets. The NE-SW oriented ShMIN in the resurgent dome and south moat of the caldera is in contrast to the dominating ~E-W orientation in the western Basin and Range province and Mammoth Mountain. In order to examine the source of this stress direction difference, we apply 3-D Finite Element Modeling with different parameters and combinations of tectonic and magmatic stress sources to search for the model that best fits the observed ShMIN orientations. Our preliminary results show that the local stress perturbation in the Long Valley area may be explained by the combination of a regional tectonic stress and an ellipsoidal magma source located next to the Mono Craters, indicating that the Mono Craters may play an important role in the local stress distribution in the area, although the current deformation is dominated by Long Valley Caldera.
Media has not been submitted for this Presentation
Abstract: The mechanism of intermediate-depth earthquakes occurring within subducting slabs remains controversial. Two proposed mechanisms for these earthquakes are dehydration embrittlement and thermal shear instability. By analyzing the stress drops and radiated energies determined from the recorded waveforms of these earthquakes, we can constrain the radiated efficiencies of the events and hence determine to what degree the mechanisms causing them are dissipative. We use the method of empirical Green’s functions (eGfs) and spectral ratios to obtain stress drops for moderate to large (M4.0 and greater) intermediate-depth earthquakes since 2006 in the Kyushu and Hokkaido regions of Japan using data from the Hi-Net, F-Net, and Kiban seismic networks. With events at intermediate depth, the quality of small events usable as eGfs is a major limiting factor of the spectral ratio method. We examine the robustness of stress drop estimates determined from different source models, varying falloff rate and number of corner frequencies to better constrain the uncertainty in these estimates. We also look at the effects of near-source heterogeneities on the eGf waveforms, as in practice the data for intermediate-depth earthquakes often does not allow us to select smaller events within one source radius of the larger event. We are able to constrain the source parameters for the larger events, and can place bounds on them for the smaller events. The two loci of intermediate-depth earthquake activity differ in the character of the subducting plate, in that the Kyushu events occur in a plate that is younger and warmer. While we find generally larger stress drops and lower radiated efficiencies than shallow events in both regions, the difference is not as marked as has been found in previous studies of regions with anomalously high intermediate-depth activity, such as the Bucaramanga Nest in Colombia.
Media has not been submitted for this Presentation
Abstract: Seismic velocity models are becoming increasingly important due to their use in simulation, and other applications in seismology and engineering. We describe the development of a three-dimensional seismic velocity model for the Bengal basin, covering Bangladesh and the neighboring regions of India and Myanmar. The model encompasses an area of 760 x 780 km2. It is bordered by the stable Indian continent on the west and north-west, the Himalayas on the north and north-east, and the Burma plate on the east and south-east. The basin is shaped by the Ganges-Brahmaputra-Meghna delta—the largest delta system in the world—and rests over a sedimentary structure with thicknesses that can reach up to 20 km in depth. This is a region under high seismic hazard due to the collision between the Indian, Eurasian, and Burmese plates. The velocity model is developed based on available sedimentology and geological investigations. We collected and digitized information in the form of geological borehole logs; seismic, aeromagnetic and gravity profiles; and geologic mapping and contours developed from sedimentary deposition studies. Most of the borehole logs came from natural gas and other mineral exploration studies concentrated in the north-western and north-eastern regions. Aeromagnetic and gravity profiles were used to build contours of different sedimentary deposits and major velocity contrasts in the deep basin. The model is built as a computer code application interface (or API) designed to provide values of seismic velocities (Vp and Vs) and density. We have used the model to produce low frequency (<0.5 Hz) ground motion simulations of a few local earthquakes, and compared the results with available seismic data to test the model’s reliability. In the future, as we continue to make improvements, we expect to use the model in simulations of past and scenario earthquakes to gain insights about the region’s seismic hazard and the basin’s response.
Media has not been submitted for this Presentation
Abstract: The September 19th, 2017 Puebla earthquake caused life loss and severe property damage in Mexico City, even though the epicenter was located ~100 km away from the city. Mexico City is built on a thick clay-rich sedimentary sequence and, hence, is susceptible to land subsidence (at rates as high as 350 mm/yr), surface faulting, and seismic acceleration during earthquakes. The earthquake damage in the eastern part of the city, characterized by the collapse of several buildings, can be explained by seismic amplification. However, the damage in the southern part of the city, characterized by the collapse of small houses and surface faulting, requires a different explanation. We present here geodetic observations suggesting that the surface faulting in Mexico City triggered by the Puebla earthquake occurred in areas already experiencing differential displacements. Our study is based on Sentinel-1A satellite data from before and after the earthquake. We process the data using Interferometric Synthetic Aperture Radar (InSAR) to produce a coseismic interferogram. The results of our analysis reveal the locations and patterns of coseismic phase discontinuities, mainly in the piedmont of the Sierra de Santa Catarina, which agree with the location of earthquake’s damage reported by official and unofficial sources (GCDMX, 2017; OSM, 2017), the location of preexisting, subsidence-related faults (GCDMX, 2017), and differential displacements identified using a Fast Fourier Transform residual technique on high-resolution InSAR results from 2012 (Solano-Rojas et. al, 2017). We propose that the seismic energy released by the 2017 Puebla earthquake induced fast soil consolidation, which remotely triggered slip on the preexisting subsidence-related faults. The slip observed during this earthquake represents a hazard that needs to be considered in future urban development plans of Mexico City.
Media has not been submitted for this Presentation
