Rapid Earthquake-Induced Damage Detection Using Satellite Imagery and Machine Learning Algorithms for the September 19, 2017, M7.1 Central Mexico Earthquake

Abstract:

Rapid detection and classification of damage after earthquakes is important for loss estimation, rapid response, and research. We use optical satellite imagery after the recent M 7.1 Central Mexico earthquake on September 19, 2017 to develop a damage catalog for Mexico City. The satellite imagery before and after the earthquake are freely available through the Digital Globe disaster program with a spatial resolution of less than 1-meter. The pre-event image was taken on June 15, 2017 through GeoEye-1 sensor and the post-event image was taken on September 20, 2017 through WorldView-2 sensor. Machine learning algorithms (MLAs) including feed-forward and radial basis artificial neural networks (ANNs) are used to detect and classify building damage after the earthquake. MLAs work with nonlinear datasets, learn from limited training data, and have been successfully used in other classification problems. In addition to the spectral information of the imagery, textural and structural features such as dissimilarity and Laplacian of Gaussian (LoG) filter are used as the inputs to the MLAs and result in an improvement of the overall accuracy of the classification. Terrestrial images taken by individuals from the damaged buildings after the earthquake, which are available online and through reconnaissance reports, are used as ground truths to develop both training and testing data for the region. There are more than 1000 buildings that are partially damaged while around 20 are reported as collapsed. As the spatial and spectral resolution of the imagery are not high enough to detect the partial damages to the buildings, this study only focuses on totally collapsed structures. The classification results are validated using 2-fold cross-validation with a confusion matrix and to evaluate the overall accuracy of the algorithms. The results of this work provide preliminary evidenced that collapse catalog using MLAs and high resolution (1-m) optical imagery can be developed to inform loss estimation, rapid response and research efforts after major earthquakes.

Slidecast:

https://vimeo.com/276977582

High Resolution Imaging of the Plate Interface in Central Alaskan Subduction Zone Using Autocorrelation with Local Earthquakes

Abstract:

The physical properties of subducting slabs at convergent plate boundaries impact megathrust seismicity and arc volcanism, but remain incompletely understood. Slabs features are primarily imaged using phases from teleseismic earthquakes at frequencies below 1 Hz, resulting in low-resolution images. Local earthquake sources with frequencies of ≥10 Hz on the other hand, provide a potential means for high resolution imaging of small-scale slab structure to depths through the seismogenic zone. Here we image the subducting Yakutat oceanic plateau in Alaska using autocorrelation of scattered energy within local earthquake coda and successfully produce a higher resolution image of the slab than visible in receiver functions. To validate our result, we also produced a synthetic autocorrelation section using the SPECFEM finite element method. Clear phases are observed at times that coincide with expected boundaries based on the a priori structural model based on earlier receiver function results. We associate these phases with a low-velocity zone atop the subduction oceanic plate, and we also image a later arrival observed only in the autocorrelation section associated with the base of the subducting Yakutat crust. Our results demonstrate the value of dense seismic arrays at tens to hundreds of meters spacing for mantle structures, providing a means to advance the characterization of subduction megathrusts.

Slidecast:

https://vimeo.com/276984103

Implementing Inter-Period Correlations into the SDSD Broadband Ground Motion Method

Abstract:

Earthquake ground motion records reveal period-dependent correlations, which has implication for seismic risk (Bayless and Abrahamson, 2017). The empirical inter-period correlations of epsilon (the residual between simulations and the mean of the simulations in Fourier Amplitude Spectra (FAS) space) using the Effective Amplitude Spectrum (EAS) computed from the PEER NGA-West2 database resemble a two-sided exponential function. We attempt to incorporate such correlation into the current San Diego State University (SDSU) Broadband (BB) ground motion generator module, which combines deterministic (low-frequency) and stochastic (high-frequency) components. Here, we assume that the Fourier amplitude at frequency f0 is correlated with the Fourier amplitude at f with correlation coefficient exp(-|f-f0|/a) and define a one-sided decaying exponential filter function g = H(f)exp(-f/a), where a is a constant. To improve the EAS correlation in the current SDSU module, we first generate uncorrelated uniformly distributed Fourier spectral amplitudes with unit mean for different realizations, and convolve them with g, which are then multiplied with the Fourier amplitude of the high-frequency ground motion synthetics calculated using Zeng et al. (1991)’s scattering theory. Using our improved method, the BB results for 7 western U.S. events and 2 Japan events with Mw5.0-7.2 show that the empirical inter-period correlations of EAS are well predicted in the SDSU module for a large number of realizations from a single event with unbiased goodness-of-fit of the spectral accelerations in the presence of correlated synthetics.

Slidecast:

https://vimeo.com/276979821

Regional Evolution of Coda Waves’ Attenuation in the NW South America

Abstract:

Several studies in some regions of the world have shown during last decades that coda waves’ attenuation may vary as a response to different processes in the Earth interior, e.g. evolution of cracks due to stress variations, migration of fluids, and mobility of magmas. It means, that permanent monitoring of coda waves derived from seismological records from the same area, using identical lapse time since the S-wave onset, and application of similar signal processing techniques is an interesting approach for detecting changes related to the evolution of regions with tectonic or volcanic activity. Nevertheless, these observations are not detected in other regions, probably due to the concurrence of additional processes not yet identified or inadequately understood. In addition, contrasting seismological information among active regions may also promote diverse sensitivity and reduce the hope of success in the volcanic and earthquake forecasting or prediction studies. In this work is reported the estimation of approximately 180.000 Qc values using a lapse time of 10 s since the S-wave onset, several frequency bands and applying the Sato (1977) technique. Estimations were performed with seismological local records (hypocenter distances less than 200 km) detected by the National Seismological Network of Colombia in the period June 1993 to October 2017. Comparing time series of Qc and mL (>6.0) allow to detect a cyclic behavior of this inverse attenuation parameter, which presents a stage of mean value reduction with a high variability that matches with a period of approximately 5 years of high earthquake activity in the NW South America (1995-2000). The Qc cycle looks of approx. 25 year and suggests that a new seismic crisis coming soon. Other observations and explanations to this phenomenon are presented and suggested.

Slidecast:

https://vimeo.com/276979191

Contributions of Modern Seismic Imaging to Understanding the Andean Convergent Margin

Abstract:

Improved seismic images of the South American convergent margin are providing new insights to longstanding tectonic problems including: (1) the formation and destruction of thick continental crust, (2) the influence of flat slab subduction on the over-riding plate, and (3) the role of slab morphology on mountain building. We have used data from over 1000 seismic stations and multiple techniques to generate seismic images spanning ~3000 km of the South American convergent margin. In the central Andes, the Altiplano and Puna Plateaus have thick crust but have different upper mantle seismic signatures with nearly complete lithospheric removal beneath the Puna and more piecemeal removal beneath the Altiplano. Both the central Altipano and northern Puna show radial anisotropy in the crust that is likely related to crustal flow and magmatic processes respectively. The active arc and backarc of the northern Puna Plateau show evidence of crustal structures associated with magmatic processes suggesting that locally magmatic addition plays a role in modifying the crust. The South American subduction zone has two regions of flat slab subduction beneath Peru, and central Argentina. The Argentina flat slab has high rates of seismicity while the Peru flat slab has much less seismicity suggesting a difference in hydration. Both these flat slab segments show indications of strong coupling to the over-riding plate, upper plate deformation and associated slab tears. The subducting Nazca slab penetrates into the lower mantle along most of the length of the central Andes but with different dips and slab thicknesses. We image a well defined slab beneath the south central Andes where many global models lack a slab anomaly. North of the Bolivia orocline the slab is steeper and appears to thicken in the Mantle Transition Zone (MTZ) compared to south of the orocline where the slab has a shallower dip and continues into the lower mantle with minor deformation in the MTZ.

Slidecast:

https://vimeo.com/276983868

The Two Subduction Zones of the Caribbean-South American Plate Boundary

Abstract:

The Caribbean-South American (CAR-SA) plate boundary is a complex transform fault system connecting oppositely vergent subduction zones, the Antilles in the east, and a currently locked CAR-SA flat slab subduction zone in the west. Teleseismic P-wave tomography shows both the Atlantic (ATL) and the Caribbean (CAR) plates subducting in opposite directions to transition zone depths under northern South America. Receiver functions (RF) show a depressed 660 discontinuity and thickened transition zone associated with each subducting plate. In the east, the ATL part of the SA plate subducts westward beneath the CAR. The eastern end of the El Pilar-San Sebastian strike-slip system, a subduction-transform edge propagator (STEP) fault, lies above the point where the ATL tears away from SA as it descends into the mantle. The Paria cluster seismicity is the mechanical expression of the plate tear. Body wave tomography and LAB depth determined from RFs and Rayleigh waves suggest that the descending plate also viscously removes the bottom third to half of the SA continental margin lithosphere. This has left thinned continental lithosphere under northern SA as the subduction zone has migrated along its northern coast. The thinned lithosphere extends almost the entire length of the El Pilar-San Sebastian fault system, from ~65o to 69oW, and inland more than 100 km. In northwestern SA the CAR plate subducts at < 30o to the ESE under northern Colombia to about Lake Maracaibo, Vn, and extends laterally from northernmost Colombia to perhaps as far south as the Bucaramanga nest seismicity. The flat slab is associated with three Neogene, basement cored, Laramide-style uplifts: the Santa Marta block, the Perija Range, and the Merida Andes. To the SE under Lake Maracaibo and the Merida Andes the CAR descends steeply to the transition zone. The steep descent suggests that the CAR plate is internally torn, separating the subducting CAR from CAR forming the seafloor to the north.

Slidecast:

https://vimeo.com/276984598