Abstract: The geothermal field of Humeros Puebla, Mexico, is one of the most important in the country due to its production of 85 Mwh. Previous studies indicate that most of the seismicity is concentrated in two large areas mostly located in the northern part of the field, near the areas of injection and re-injection of fluids (Urban et al, 2013, Lermo et al, 2001, 2002, 2005, 2008). In the framework of the Working Package 5.2 of the GEMex bilateral project between Mexico and the European Union, a seismic network has been installed consisting of 26 short-period and 17 broad band stations. This seismic network represents the largest monitoring network installed in a geothermal field in Mexico. The seismic noise of the stations has been analyzed to determine i) the quality of the seismograms and ii) the level of noise in the high frequency range ( >1 Hz), which could be attributed to the extraction plants, steam pipelines or the effect of human activity. For this purpose, we analyzed the power spectral density function, through the implementation of routines in free software such as Seismic Analysis Code (SAC) and Jupyter. Likewise, the spectral ratios have been calculated for each station of the network in order to characterize and determine the site effects. We also present the location of the seismic events registered in the field since September 2017, calculated using a 1D velocity model presented by Urban and Lermo (2013). SEISAN software is used to locate the seismicity (Havskov and Ottemoller, 2003) and HypoDD double difference method (Waldhauser, 2001) to refine the event location. Our preliminary results indicate that the noise levels are acceptable compared with the model of Peterson (1993), which allows us to characterize local seismicity and low magnitude within the geothermal field.
Poster:
SSA_FINAL
Abstract: We develop an empirical ground-motion model (GMM) for Fourier amplitude spectra (FAS) using the PEER NGA-W2 database of ground motions from crustal earthquakes (Ancheta et al., 2014). GMMs have traditionally been developed for 5%-damped response spectra (PSA), however, there are some available models for FAS (e.g. Bora et al., 2015; Stafford et al., 2006; Yenier and Atkinson, 2015). There are two key advantages of developing FAS model over PSA models: the FAS is more closely related to the physical behavior so that it is easier to apply seismological constraints to the GMM; and the Fourier spectrum is linear which allows ground motion from small magnitude earthquake to be used to constrain the linear path and linear site effects without the spectral shape dependence of PSA. The GMM features a magnitude scaling formulation guided by Chiou and Youngs (2014). Large-magnitude near-fault saturation to capture the effects of extended ruptures is included and guided by broadband simulations up to M8.0. The GMM features stronger near-fault magnitude scaling (less saturation) than the CY14 model for PSA. The model includes terms for anelastic attenuation, site effects (Vs and kappa), style of faulting, hanging wall, and depth to top of rupture. The model coefficients are smoothed in a series of steps to assure smooth spectra and to constrain the extrapolation. An essential feature of forward-application for GMMs is their ability to reliably extrapolate to the key large magnitude and short distance ranges that often control the hazard. This allows for the GMMs to be applied beyond the ranges well constrained by empirical data. Our model is applicable to crustal earthquakes in the range M3-8, with rupture distances 0-300 km, for sites with Vs30 between 150-2,000 m/s, and over the frequency range 0.1-100 Hz. Models for the between-event, within-event, and total standard deviation are provided.
Poster:
SSA_2018_Poster
Abstract: The concept of picking earthquake phase arrivals and associating them to a common event is simple, but hard to implement in a robust and reliable manner when there are multiple overlapping events and false arrival time picks. Related issues include determining precisely if and when an event has occurred when noisy and false arrivals are included in the data set, and determining how many different unique events have occurred when move-out times are possibly overlapping. Interestingly when these challenges are posed as a precise mathematical problem, simple and accurate solutions arise through the use of basic multilayer perceptron (MLP) supervised classification and regression frameworks. For a pre-determined velocity model, we obtain distinct (shallow) MLP networks to classify an arbitrary set of arrival times on $n$ stations between a true P or S phase recorded event and noise (i.e. a false set of arrivals), determine the distinct number of unique events in a time window, determine which set of arrivals associate to a common event, and also, using a denoising autoencoder framework, predict arrival times on stations missing arrival time detections during an event. Discrete (i.e. vectors of arrival time data) and continuous (i.e. encoding of arrival time data as time-dependent Gaussian pulses) implementations are used. Synthetic tests of the method and applications to field data from the IPOC network of Northern Chile are presented.
Poster:
IWM_Automatic_Association_Earthquake_Arrivals
Abstract: Using the earthquake catalogues from China and ISC (International Seismological Centre), the uniform catalogue of Himalaya and adjacent areas has been established for the seismic hazard analysis and seismic hazard map of Himalaya areas according to according to the following principles. 1) The earthquake parameters should be taken from China catalog if the events occurred in China. 2) The earthquake parameters should be taken from ISC catalog if the events occurred out of China. 3) The earthquake parameters should be checked if the events occurred on the China border areas based on the data of earthquake stations. There are 17886 events with event magnitude 4.0 and greater in the catalogue, among them, there are 65 ones with magnitude 7 and greater. Based on the catalogue, the seismicity of the Himalaya and adjacent areas has been analyzed, and the following conclusions are made. 1) The epicenters generally follow a NW-trending band distribution along the Himalayan arc. At the eastern and western ends, that is, the structural knots of the Himalayas, the epicenter strips are respectively turned to the NE-trending. 2) Intermediate-depth earthquake are distributed on both east and west ends of the Himalayan arc structure belt, that is, the Pamir at the west end and the Assam at the east end. In the Qinghai-Tibet Plateau and the front of the Himalaya arc, events are basically shallow earthquakes. 3) Earthquakes with magnitude greater than 4.5 are basically complete since 1960 A.D.. 4) The b-values of magnitude – frequency relationship for the study areas are between 0.9 and 1.0.
Poster:
SSA2018poster-XU-116
Abstract: Relative arrival times have been used in the past to produce relative relocations of earthquakes and explosions. Precise relocations of explosions have value in the context of the Comprehensive Nuclear-Test-Ban Treaty on-site inspection when data are available for a previous event. The Democratic People’s Republic of Korea (DPRK) has now conducted 6 declared nuclear tests from 2006-2017. After the declared test in 2009, many researchers applied relative relocation techniques to improve the accuracy and precision of the seismic event locations. Relative relocation methods require multiple events at the same location in order to utilize the differential arrivals between events. There are few locations in which such a scenario takes place with the same source type and similar location. However, there is still the overall unknown of the exact ground truth location for these events. We will demonstrate the application of a straight-forward master event methodology for doing relative relocations for DPRK events as well as a modified version of the master event method in a simultaneous, maximum likelihood solution. We also validate this methodology using a data set where we precisely know the locations of multiple, similarly-located explosive events with arrivals from common stations. Taking advantage of readily available waveforms, we are able to make relative picks for many stations at local, regional, and teleseismic distances, depending on the particular event and data set. Waveforms are manually-aligned on the first few cycles in order to match the initial arrival information. We will compare the relative relocations with those obtained using more recent techniques that involve simultaneous inversion of data from multiple events. The application of the standard master event method consistently provides high relative accuracy and precision, even when the master and test events are separated by tens of kilometers.
Poster:
SSA2018-Begnaud-DPRK_Relative_Relocation
Abstract: The seismicity of France is moderate in its Metropolitan part and related to a subduction context in its West Indies part. RESIF network (www.resif.fr) and, West Indies (IPGP-OVS) and bordering countries observatories provide us instrumental data. These accelerometric and velocimetric data are available in near real time. The BCSF-RENASS (EOST) provides Intensities, preliminary values are based on rapid citizen testimonies, frequently several thousands for Mw>≈4.5, and final values EMS98 are based on study including field investigations if damage. With about 1 city per 15 square km, we sample in details the territory when the about 400 seismic stations give irreplaceable precise ground motion parameters but very locally and most of the times at a farther epicentral distance. The USGS ShakeMap V3.5 run over the Pyrenees since 2012 (www.SisPyr.eu) and since 2016, it runs for the whole territory of France at BCSF-RENASS (www.franceseisme.fr). We summarise few feedback on V3.5 and expectation on V4. We had to adapt the ShakeMap V3.5 for region with moderate size events. We obtain a huge improvement at any distance by including intensities but we need to weight their contribution depending the quality of data, preliminary or final, and the possible lack of data near the epicenter. Despite we added new GMPE or IPE; we need to use different regional attenuation law versus the location of the epicenter. An expectation would be to select automatically the GMPE or IPE that fit at best with the observations. Moreover, the variation of the attenuation versus azimuth seems to be strongly smoothed by the GMPE and IPE used. The V3.5 helped us to detect important ML overestimation in few regions, but the ShakeMap has still a strong dependency to the hypocenter localization, which cannot be corrected for example by using collected macroseismic data. We built a ShakeMap WG that brings together the French researchers and observatories (transversal action seismicity RESIF).
Poster:
SSA-2018-Poster-red
