Abstract: Many inputs in the calculation of probabilistic seismic hazard analysis are known only with considerable uncertainty. Thus, it is important to capture the range of uncertain inputs in the PSHA results. This requires repeated hazard calculations with different sets of input parameters. It is also of interest which of the uncertain inputs have the greatest influence on the hazard results, which can be achieved by systematically varying the inputs and observing the variation in the hazard results. Repeated hazard calculations are computationally expensive, which can make large scale uncertainty quantification and sensitivity analysis prohibitive. We propose to use an emulator as a computationally efficient approximation to hazard calculations that can be used to address these questions. The emulator is estimated based on a few hazard runs for varying inputs, and is subsequently used to draw inferences about the distribution of hazard results for uncertain inputs. The emulator is cast in a Bayesian fashion, which allows to carry its associated uncertainties through the analysis. In particular, the emulator models the output, as a function of the input, as a Gaussian process (GP), and inferences about the distribution of the hazard outputs can be drawn from the posterior distribution of the GP. We illustrate the approach by performing uncertainty quantification and sensitivity analysis for a suite of simple hazard problems. The emulator achieves similar accuracy as a full Monte Carlo simulation, while requiring many fewer runs of the hazard code.
Media has not been submitted for this Presentation
Abstract: Earthquake-induced landslides cause a significant portion of earthquake-related fatalities and economic losses, and can have long-lasting negative societal impacts. We present a method to estimate the impact of seismically induced landslides on affected populations using the output of the USGS near-real-time earthquake products. Using a newly developed, comprehensive dataset of 196 historical earthquakes, including 127 events with known landslide fatality counts, we develop an empirical model that estimates the order of magnitude of potential fatalities based on the exposure of population to expected landslide occurrence. Using the grid of landslide probabilities output by the USGS Ground Failure earthquake product, we estimate population exposure by multiplying the predicted probability grid with a gridded global population database adjusted for population growth. We then sum over the entire grid for each event, and term this the predicted ‘landslide exposure index.’ We examine income level of the country of each earthquake as a secondary factor to represent relative vulnerability of the surrounding area. We compare these values to the number of actual fatalities for 91 training events in order to calibrate a model that can be used to predict the order of magnitude of potential fatalities due to plausible future earthquakes using scenario earthquakes. We observe a significant positive correlation between predicted and observed fatalities, with high variability in fatality rates for similar exposure levels. This suggests that other factors (e.g., building type, time of day, landslide density, effect of urbanization on population) may improve this estimate. Ultimately, the outputs of this method can be integrated into the USGS near-real-time earthquake information system. The outputs can provide input for use in the Ground Failure alert level designation and also an estimate of landslide fatalities, which are currently not included in PAGER’s loss estimates.
Media has not been submitted for this Presentation
Abstract: Chile host a great tsunamigenic potential along its coast, even with the large earthquakes occurred during the last decade, there is still a large amount of seismic energy to release. This permanent feature and the fact that the distance between the trench and the coast is just 100 km creates a difficult environment to do real time tsunami forecast. In Chile tsunami warnings are based on reports of the seismic events (hypocenter and magnitude) and a database of precomputed tsunami scenarios. However, because yet there is no answer to image the finite fault model within first minutes (before the first tsunami wave arrival), the precomputed scenarios consider uniform slip distributions. Here, we propose a scheme of processes to fill the gaps in-between blind zones due to waiting of demanding computational stages. The linear shallow water equations are solved to obtain a rapid estimation of the run-up distribution in the near field. Our results show that this linear method captures most of the complexity of the run-up heights in terms of shape and amplitude when compared with a fully non-linear tsunami code. Also, the run-up distribution is obtained in quasi real-time as soon as the seismic finite fault model is produced.
Media has not been submitted for this Presentation
Abstract: The Mw7.8 Pedernales earthquake is associated with the subduction of the Nazca Plate beneath the South American Plate. The mainshock caused many casualties and widespread damages across the Manabi province. The 150 km-long coseismic rupture area is found beneath the coastline, near 25 km depth. The rupture propagated southward and involved the successive rupture of two discrete asperities, with a maximum slip (~ 6 m) on the southern patch. The rupture area is consistent with the highly locked regions observed on interseismic coupling models, overlaps the 7.2 Mw rupture zone, and terminates near where the 1906 Mw 8.8 megathrust earthquake rupture zone ends. Two neighboring highly coupled patches remain locked: (A) southern to and updip of the coseismic rupture zone and (B) northern and downdip. In this study, we relocate the aftershocks and compare the seismicity distribution to the interseismic coupling and the rupture area. We use continuous seismic traces recorded on the permanent network partly installed in the frame of the collaboration between l’Institut de Recherche pour le Développement (IRD-France) and the Instituto Geofísico, Escuela Politécnica Nacional, Quito, Ecuador. Detections are conducted using Seiscomp in play-back mode. Arrival-times are manually picked. To improve earthquake location, we use the MAXI technique and a heterogeneous a priori P-wave velocity model that approximates the large velocity variations of the Ecuadorian subduction system. Aftershocks align along 3 to 4 main clusters that strike perpendicularly to the trench, and mostly updip of the co-seismic rupture. Aftershock seismicity develops indifferently over portions of plate interface that are known to be strongly locked or almost uncoupled. The seismicity pattern is similar to the one observed during a decade of observation during the interseismic period with swarms such as the Galera alignment, Jama and Cabo Pasado, and between Manta and Puerto Lopez.
Media has not been submitted for this Presentation
Abstract: The high school Physics programs in Mexico do not consider the immediate application of the concepts learned by the students. According to some pedagogical theories, most of the acquired knowledge is assimilated when experimenting and developing projects. It is in high school when young people are exploring for experiences to decide the area in which they want to focus their studies. There is little interest in Earth sciences, reflected by the number of students in those areas, which may be due mainly to the lack of exposure and examples. We are working on a project that seeks, through the preparation of high-school teachers, to awake student curiosity in seismology. Based on the above, and taking as examples the successful programs “Seismographs in Schools” from IRIS and “Geoscience Information For Teachers” from EGU, the Mexican Servicio Sismológico Nacional (National Seismological Service, SSN) has launched a project that includes three stages:1) design and delivery of a diploma for high-school teachers; 2)installation of short period seismographs at each teachers’ schools; and 3)active participation of teachers and their students in research projects based on the data collected in their schools. We are in the first stage, which has the support of the project DGAPA-PAPIME PE107317. The diploma for high-school teachers is a 170 hours course that offers an introduction in topics related to seismology. It starts with basic concepts of physics, and examples of what they can offer their students in their classroom. The content seeks to apply some of the learning, and concepts of the physics courses in high school by the analysis of the information recorded in seismograms. In addition to the material presented to the participant through an online platform, and during sessions, they work on an integrating project, which objective is to prepare them in the use of seismographs and their data in their classes and begin in scientific research themselves and their students.
Media has not been submitted for this Presentation
Abstract: The main purpose of this study is to develop ground motion prediction equations (GMPEs) for the Gulf Coast region of the United States using a hybrid empirical method (HEM). This project contributes toward development of a new set of GMPEs for the Gulf Coast region, which will be consistent with the available recordings. This research supplement recent GMPEs developed by the Pacific Earthquake Engineering Research Center (PEER) Next Generation Attenuation (NGA-East) GMPEs for the Central and Eastern North America (CENA) regions. Recently, a number of GMPEs for CENA are developed as part of NGA-East project conducted by the PEER. However, in majority of them, ground motions recorded in the Gulf Coast region were excluded due to considerably different attenuation attributes in this region (EPRI, 1993). The Gulf Coast region exhibits significantly different ground-motion attenuation because of the thick sediments in the region (Dreiling et al. 2014). The purpose of this study is to develop specific GMPEs for use in the Gulf Coast region using the HEM. Because the strong motion data set is sparse in the Gulf Coast region, the hybrid empirical method represents an appropriate and robust approach which has been generally accepted to develop GMPEs.
Media has not been submitted for this Presentation
