ALBERT

Description: One of the most commonly used parameters to describe seismic attenuation is the high-frequency spectral decay parameter Kappa (κr), yet the physics behind it remain little understood. A better understanding of potential factors that lead to large scatter in estimated values of κr constitutes a critical need for ground-motion modeling and seismic hazard assessment at large. Most research efforts to date have focused on studying the site-to-site and model-to-model variability of κ, but the uncertainties in individual κr estimations associated with different events at a selected site (which we refer to as the within-station variability of κr) remain uncharacterized. As a direct corollary, obtaining robust estimates of the site-specific component κ0, and their corresponding interpretation become a challenge. To understand the sources of the variability observed in κr (and κ0) at a single site, we select 10 Japanese Kiban–Kyoshin network (KiK-net) downhole arrays and investigate the systematic contributions from ground-motion directionality. We observe that κr estimated from a single horizontal component is orientation dependent. In addition, the influence of ground-motion directionality is a function of local site conditions. We propose an orientation-independent κr-value, which is not affected either by ground-motion directionality or by the events’ azimuths. In addition, we find that focal depth of events used in κr calculations affects the estimation of the regional attenuation component κR, which, in turn, influences the within-station variability in the κ0 model.

Description: We apply a spectral decomposition approach to isolate the source spectra from propagation and site effects and, in turn, to estimate the source parameters of small-to-moderate earthquakes that occurred in central Italy. The data set is composed of about 400,000 waveforms relevant to 4111 earthquakes in the moment magnitude range 1.5–6.5, recorded by a high-density network of stations installed in the study area. We first investigate the reliability of the source parameters for small magnitudes through numerical simulations. We generate synthetic spectra for different source scaling models and near-surface attenuation effects, considering the source–station geometry and the data availability of the central Italy data set. Our analysis with synthetics shows that the spectral decomposition is effective in isolating the source contributions from other factors. Moreover, the analysis of the residual distributions suggests that moment magnitude 1.8 is the lower bound for the retrieval of reliable Brune’s source parameters, although we observe an increase of residual’s variability below magnitude 3, and the estimated source parameters could be biased below magnitude 2.3. Remarkably, the assessment of the stress drop Δσ for small events is strongly hampered by site-specific attenuation near the surface. In view of the results with synthetics, we analyze the source parameters of earthquakes recorded in central Italy. The corner frequency versus seismic moment relationship describes a source scaling in which Δσ increases with increasing moment magnitude Mw, the mean Δσ varying from 0.1 MPa for Mw5. In particular, Δσ increases mainly for Mw in the ranges 2.5–3 and 4.5–5.2. The corner frequencies estimated from the apparent source spectra do not show any dependence on hypocentral distance and magnitude, confirming that uncorrected anelastic attenuation effects do not significantly bias the results.

Description: We investigate the dependence of event-specific ground-motion residuals in the Ridgecrest region, California. We focus on the impact of using either local (ML) or moment (Mw) magnitude, for describing the source scaling of a regional ground-motion model. To analyze homogeneous Mw, we compute the source spectra of about 2000 earthquakes in the magnitude range 2.5–7.1, by performing a nonparametric spectral decomposition. Seismic moments and corner frequencies are derived from the best-fit ω−2 source models, and stress drop is computed assuming standard circular rupture model. The Brune stress drop varies between 0.62 and 24.63 MPa (with median equal to 3.0 MPa), and values for Mw〉5 are mostly distributed above the 90th percentile. The median scaled energy for Mw

Description: A method is proposed to identify within seismic catalogs those earthquakes that are most relevant to the seismic hazard. The approach contrasts with the classical approach to decluster the seismic catalog with the expectation that the remaining main shocks will be the relevant events for the seismic hazard analysis. We apply a time window like in the window declustering approach of Gardner and Knopoff, but the time window is motivated by relevance to engineering. A ground motion criterion replaces the spatial window. An event in the time window is included in the “Maximum Shaking Earthquake Catalog (MSEQ catalog)” if the median ground motion at its epicenter exceeds the predicted median ground motion there from the main shock, using a locally appropriate ground motion prediction equation. Ground motion can be measured by any parameter that is estimated by a ground motion prediction equation. We consider peak acceleration and spectral amplitude (SA) at periods of 0.2, 1.0, and 3.0 s. The longer period parameters systematically remove more small events. The purpose is not to produce a declustered catalog, in which each group of physically related earthquakes is represented by its largest event. Statistical properties of the MSEQ catalog somewhat resemble the corresponding declustered catalog in three tested regions, but the MSEQ catalogs all retain more large-magnitude earthquakes. The MSEQ catalog may better represent the potential hazard in a region, and thus might be considered as an alternative to a declustered catalog in developing the seismicity model for probabilistic seismic hazard analysis.

Description: This study presents the coupling of the spectral decomposition results for anelastic attenuation, stress drop, and site effects with the Graves-Pitarka (GP) hybrid ground-motion simulation methodology, as implemented on the Southern California Earthquake Center (SCEC) broadband platform (BBP). It is targeted to applications in the Upper Rhine graben (URG), which is among the seismically active areas in western Europe, yet a moderate seismicity area. Our development consists of three main steps: (1) calibration of regional high-frequency (HF) attenuation properties; (2) modification of the hybrid approach to add compressional waves in the HF computation and examine various strategies to evaluate site amplification factors in the Fourier domain (e.g., VS30-based or site-specific factors); (3) testing of the simulations using earthquake records from the URG (3.7

Description: Strong ground motion records and free open access to strong-motion data repositories are fundamental inputs to seismology, engineering seismology, soil dynamics, and earthquake engineering science and practice. This article presents the current status and outlook of the Observatories and Research Facilities for European Seismology (ORFEUS) coordinated strong-motion seismology services, namely the rapid raw strong-motion (RRSM) and the engineering strong-motion (ESM) databases and associated web interfaces and webservices. We compare and discuss the role and use of these two systems using the Mw 6.5 Norcia (Central Italy) earthquake that occurred on 30 October 2016 as an example of a well-recorded earthquake that triggered major interest in the seismological and earthquake engineering communities. The RRSM is a fully automated system for rapid dissemination of earthquake shaking information, whereas the ESM provides quality-checked, manually processed waveforms and reviewed earthquake information. The RRSM uses only data from the European Integrated Waveform Data Archive, whereas the ESM also includes offline data from other sources, such as the ITalian ACcelerometric Archive (ITACA). Advanced software tools are also included in the ESM to allow users to process strong-motion data and to select ground-motion waveform sets for seismic structural analyses. The RRSM and ESM are complementary services designed for a variety of possible stakeholders, ranging from scientists to the educated general public. The RRSM and ESM are developed, organized, and reviewed by selected members of the seismological community in Europe, including strong-motion data providers and expert users. Global access and usage of the data is encouraged. The ESM is presently the reference database for harmonized seismic hazard and risk studies in Europe. ORFEUS strong-motion data are open, “Findable, Accessible, Interoperable, and Reusable,” and accompanied by licensing information. The users are encouraged to properly cite the data providers, using the digital object identifiers of the seismic networks.

Description: We present Rapid Assessment of MOmeNt and Energy Service (RAMONES), a service for disseminating through a web interface, the estimates of seismic moment (M0) and radiated energy (ER) for earthquakes occurring in central Italy with local magnitudes above 1.7. The service is based on a fully-automatic procedure developed for downloading and processing open seismological data from the European Integrated Data Archive, Italian Civil Protection repository, and Incorporated Research Institutions for Seismology (IRIS). In its actual configuration, RAMONES uses the seismic catalog generated through the event webservice of the Italian Institute of Geophysics and Volcanology (compliant with International Federation of Digital Seismograph Networks standards) to guide the data download. The concept of RAMONES is to estimate M0 and ER from features extracted directly from recordings, namely the S-wave peak displacement (PDS) and the integral of the squared velocity (IV2S) evaluated over the S-wave window at local distances. A data set composed of 6515 earthquakes recorded in central Italy between 2008 and 2018 was used to calibrate the attenuation models relating M0 to PDS and ER to IV2S, including station corrections. The calibration values for M0 and ER were extracted from the source spectra obtained by applying a decomposition approach to the Fourier amplitude spectra known as the generalized inversion technique. To test the capabilities of RAMONES, we validate the attenuation models by performing residual analysis over about 60 earthquakes occurring in 2019 that were used for the spectral decomposition analysis but not considered in the calibration phase. Since January 2020, a testing operational phase has been running, and RAMONES has analyzed about 800 earthquakes by September 2020. The distribution of the source parameters and their relevant scaling relationships are automatically computed and disseminated in the form of maps, parametric tables, figures, and reports available through the RAMONES web interface.

Description: Summary Precise real time estimates of earthquake magnitude and location are essential for early warning and rapid response. While recently multiple deep learning approaches for fast assessment of earthquakes have been proposed, they usually rely on either seismic records from a single station or from a fixed set of seismic stations. Here we introduce a new model for real-time magnitude and location estimation using the attention based transformer networks. Our approach incorporates waveforms from a dynamically varying set of stations and outperforms deep learning baselines in both magnitude and location estimation performance. Furthermore, it outperforms a classical magnitude estimation algorithm considerably and shows promising performance in comparison to a classical localization algorithm. Our model is applicable to real-time prediction and provides realistic uncertainty estimates based on probabilistic inference. In this work, we furthermore conduct a comprehensive study of the requirements on training data, the training procedures and the typical failure modes. Using three diverse and large scale data sets, we conduct targeted experiments and a qualitative error analysis. Our analysis gives several key insights. First, we can precisely pinpoint the effect of large training data; for example, a four times larger training set reduces average errors for both magnitude and location prediction by more than half, and reduces the required time for real time assessment by a factor of four. Second, the basic model systematically underestimates large magnitude events. This issue can be mitigated, and in some cases completely resolved, by incorporating events from other regions into the training through transfer learning. Third, location estimation is highly precise in areas with sufficient training data, but is strongly degraded for events outside the training distribution, sometimes producing massive outliers. Our analysis suggests that these characteristics are not only present for our model, but for most deep learning models for fast assessment published so far. They result from the black box modeling and their mitigation will likely require imposing physics derived constraints on the neural network. These characteristics need to be taken into consideration for practical applications.

Description: Summary Earthquakes are major hazards to humans, buildings and infrastructure. Early warning methods aim to provide advance notice of incoming strong shaking to enable preventive action and mitigate seismic risk. Their usefulness depends on accuracy, the relation between true, missed and false alerts, and timeliness, the time between a warning and the arrival of strong shaking. Current approaches suffer from apparent aleatoric uncertainties due to simplified modelling or short warning times. Here we propose a novel early warning method, the deep-learning based transformer earthquake alerting model (TEAM), to mitigate these limitations. TEAM analyzes raw, strong motion waveforms of an arbitrary number of stations at arbitrary locations in real-time, making it easily adaptable to changing seismic networks and warning targets. We evaluate TEAM on two regions with high seismic hazard, Japan and Italy, that are complementary in their seismicity. On both datasets TEAM outperforms existing early warning methods considerably, offering accurate and timely warnings. Using domain adaptation, TEAM even provides reliable alerts for events larger than any in the training data, a property of highest importance as records from very large events are rare in many regions.