PEER has just published Report No. 2020/02: "Data Resources for NGA-Subduction Project." Chapters of the report have multiple co-authors from various institutions. The project Principal Investigator is Yousef Bozorgnia and the report Editor is Jonathan P. Stewart, both from UC Los Angeles.
The NGA-Subduction (NGA-Sub) project is one in a series of Next Generation Attenuation (NGA) projects directed towards database and ground-motion model development for applications in seismic-demand characterization. Whereas prior projects had targeted shallow crustal earthquakes, active tectonic regions (NGA-West1 and NGA-West2), and stable continental regions (NGA-East), NGA-Sub is the first to address specifically subduction zones, which are a dominant source of seismic hazard in many regions globally, including the Pacific Northwest region of the United States and Canada.
This report describes the development of data resources for the NGA-Sub project. Agreements were formed with many owners and providers of ground-motion data and metadata worldwide to support data collection. Prior NGA projects organized the data collected into a series of spreadsheets. The enormous amount of the collected data for NGA-Sub required abandoning that strategy and ultimately the data was organized into a relational database consisting of 23 tables containing various data, metadata, and outputs of various codes required to compute desired quantities (e.g., intensity measures, distances, etc.). A schema was developed to relate fields in tables to each other through a series of primary and foreign keys. As with prior NGA projects, model developers and others largely interact with the data through flatfiles specific to certain types of intensity measures (e.g., pseudo-spectral accelerations at a certain oscillator damping level); such flatfiles are a time-stamped output of the database.
The NGA-Sub database contains 70,107 three-component records from 1880 earthquakes from seven global subduction zone regions: Alaska, Central America and Mexico, Cascadia, Japan, New Zealand, South America, and Taiwan. These data were processed on a component-specific basis to minimize noise effects in the data and remove baseline drifts. Component-specific usable period ranges are identified. Various ground-motion intensity measures (IMs) were computed including peak acceleration, peak velocity, pseudo-spectral accelerations for a range of oscillator periods and damping ratios, Fourier amplitudes, Arias intensity, significant durations, and cumulative absolute velocity-parameters.
Source parameters were assigned for earthquakes that produced recordings. Some of the 1880 earthquakes were screened out because of missing magnitudes or hypocenter locations, which decreased the number of potentially usable earthquakes to 1782. Further screening to remove events without an assigned event type (e.g., interface, intraslab, etc.) or distances reduced the number of events to 976. For those 976 events, source parameters of two general types are assigned: those related to the focus (including moment tensors) and those related to finite-fault representations of the source. A series of source-to-recording site distances and other parameters are provided using finite-fault representations of seismic sources. Finite-fault models of sources were developed from literature where available and from a simulation procedure otherwise. As part of the NGA-Sub project, the simulation procedure was revised and more fully documented. In addition, all events are reviewed to assign event types, event classes (mainshock, aftershock, etc.), and event locations relative to volcanic arcs.
Quality assurance (QA) of ground-motion data and source/path metadata was an important component of NGA-Sub. For ground motions, QA procedures included visual checks of records prior to processing, checks of records from each network that recorded each earthquake to check for systematic outliers (perhaps indicative of gain problems), and checks of limiting distances beyond which data sampling for a given event is likely to be biased by data approaching noise thresholds. Source/path QA procedures largely involved checking that information in database fields accurately reflects source documents.
Site metadata was compiled into a site table containing time-averaged shear-wave velocities in the upper 30 m of sites (VS30), basin depths, and related uncertainties. Major efforts were undertaken during the project to develop shear-wave velocity profile databases and to use those data to develop regional predictive models for site parameters when site-specific measurements are unavailable. Many of those predictive relations were published in journal or conference papers over the course of the NGA-Sub project (i.e., for Alaska, Cascadia, Chile, and Taiwan); those results are reviewed only briefly. Rather, emphasis in this report has been placed on procedures used for other regions. In addition to site parameters, all sites are also assigned a location relative to local volcanic arcs.