- © Seismological Society of America
The Incorporated Research Institutions for Seismology (IRIS) Data Management Center (DMC) offers and distributes 31 distinct data products derived from raw data. These data products are intended to serve the seismology community as precursory tools and standardized baseline data sets for further research. The number of distinct data products has more than doubled since that reported previously in Trabant et al. (2012). In this article, we provide an overview of IRIS‐DMC data products, timeline of IRIS‐DMC data product activities, updated usage statistics, and highlights from 16 new data products.
Electronic Supplement:Figures of user access statistics and source time functions.
OVERVIEW OF IRIS‐DMC DATA PRODUCTS
Since 2010, the Incorporated Research Institutions for Seismology Data Management Center (IRIS‐DMC) has expanded its services to the seismology community beyond data archiving and distribution by offering higher‐order data products derived from its vast repository of raw waveform data. These data products are intended to serve many purposes: precursory tools for future research projects, data visualization, data characterization, baseline data sets, research validation, as well as outreach and education material (see Data and Resources).
DATA PRODUCTS DEVELOPMENT AND STATISTICS
In its sixth year of data product development effort, the DMC hosts and distributes 31 distinct data products. A timeline of the DMC’s data product activities, the associated release dates, and data product categories is shown in Figure 1. These products are generated internally at the DMC (10 “internal” in Fig. 1), provided by the research community (8 “contributed”) or jointly created between IRIS and the research community through funding (7 “funded”), personal collaboration (4 “contributed & internal”), or both (2 “contributed & funded”). These data products span a wide range of frequency bands (0.001–50 Hz) and interests in the seismology community and are often suitable as refined data or reference results for researchers working in related disciplines (Fig. 2). Most products are hosted by the Searchable Product Depository (SPUD; see Data and Resources), the IRIS‐DMC’s primary data product management system.
In addition to developing new data products, the DMC has also expanded its services to the research community through the following activities:
Creating webpages highlighting Special Events to collect and share preliminary research results and background information for events of significant interest to the research community and often to the public and media as well. These pages are seeded with the DMC’s automatically generated products, which are typically created within a few hours of the event, and supplemented with contributions from the seismology community as they are made available.
Creating animations of seismic waves propagating across the entire continental United States by stitching together ground‐motion visualizations (GMVs) from USArray data from multiple collocated earthquakes.
Generating an animation of USArray microseism noise recorded during Hurricane Sandy as it moved north.
Enhancing existing products, where possible, to include new data, utilize new visualization tools, and provide more supplemental information to ensure that these products continue to serve the growing needs of the research community.
Making of select software freely available for creating custom‐processed data products and derivatives. This currently includes Python, MATLAB, and web‐based software bundles.
Among the data products we are currently serving, eight are event based and are available generally within a few hours following seismic events of sufficient magnitude or public interest. These include USArray GMV, Event Plots, Backprojections, Earthquake Energy, Rayleigh‐Wave Source Time Functions (STFs), Aftershocks, Global ShakeMovie synthetic seismograms, and Moment Tensors. The sharp increase in user inquiries and downloads of these event‐based products after significant earthquakes and nuclear tests highlights their utilization by the research community (Fig. 3). For example, peak access rates for all products following the 25 April 2015 M 7.8 Nepal earthquake spiked to nearly 300 unique visits per hour, and access level remained above normal for over 5 days after the event (see Ⓔ Fig. S1, available in the electronic supplement to this article).
Other IRIS‐DMC data products include valuable research databases, such as EarthScope Automated Receiver Survey, University of Washington Continuous Envelope Functions, The Missouri University of Science and Technology shear‐wave splitting (SWS) database, The Géosciences Montpellier SplitLab’s Shear‐Wave Splitting Database, Transportable Array (TA) Infrasound Reference Event Database (TAIRED), TA Infrasound Detections (TAID), Infrasound Automated Event Location Using a Mesh of Arrays (AELUMA), and Western U.S. Ambient Noise Cross‐Correlations. These extensive databases serve as standardized baseline data sets and critical stepping stones for future research projects (see Data and Resources).
HIGHLIGHT OF NEW PRODUCTS (SEPTEMBER 2012–FEBRUARY 2016)
Short‐Arc Rayleigh‐Wave STFs
For large earthquakes, short‐arc Rayleigh waves (R1) can be used to estimate the mainshock rupture length, duration, and directivity (e.g., Lay et al., 2009). The R1 STF product uses vertical‐component long‐period data (LHZ channel) along with point‐source synthetic seismograms (theoretical Green’s functions) to calculate STFs at Global Seismic Network (GSN) and other stations for all events with Global Centroid Moment Tensor (CMT) magnitudes 7.0 and greater.
In the preprocessing phase for each earthquake, a station list of available LHZ‐component data at IRIS with epicentral distances between 30° and 125° is formed using mostly GSN stations. GSN stations work well for this application because they are very broadband in their frequency response and are distributed evenly enough across the globe to cover as much of the focal sphere as possible for most earthquakes. The station waveforms are then demeaned, detrended, and deconvolved to displacement, and a complimentary vertical‐component fundamental‐mode synthetic seismogram is calculated using normal‐mode summation (L. Rivera, personal comm., 2013), with the Preliminary Reference Earth Model (Dziewonski and Anderson, 1981) as the reference model and the Global CMT. Both data and synthetic traces are filtered based on the event magnitude and then windowed to isolate the R1 phase from body waves and the long‐arc Rayleigh wave, R2.
To generate a synthetic seismogram suitable for comparison with data from large earthquakes, an STF must be convolved with the point‐source synthetic seismogram (see Ⓔ Fig. S2). The best‐fitting simple two‐source STF is found by sweeping through all combinations of a variable‐width triangle(s) and/or trapezoid(s) with variable relative offset and amplitudes and retaining the best‐fitting solution. Triangle widths vary from 0 to 10 s, and trapezoid widths vary from 11 to 120 s. Because R1 STFs from stations with similar azimuths tend to look similar, summary figures only plot a single well‐fitting result from evenly distributed azimuthal bins.
A second approach using an iterative time‐domain deconvolution similar to that from Lay et al. (2009) is also used, but is only intended to be supplemental. Because the solutions are often complicated and have significant long‐period noise due to, among other things, an imposed positivity constraint, the iterative deconvolution STF results are only included in the zip attachment of individual station results for completeness and are not plotted in summary figures. The waveform fits of this approach tend to match data better because these STFs are free to be overly and unrealistically complex.
Summary figures include select STFs‐versus‐station azimuth (Fig. 4), STFs‐versus‐directivity parameter (Lay et al., 2009), and others. For large unilateral (in one direction) ruptures, an azimuthal pattern emerges in which STFs will be compressed at stations in the direction of the rupture propagation and extended at stations away from the propagation direction. For events with magnitudes less than about 7.5, results will generally not show much directivity at periods longer than 30 s, as used in the processing here. Before interpreting summary results, users are encouraged to judiciously inspect individual station results bundled in the zip file available at the bottom of each STF page.
In 1996, Astiz et al. stacked over 33,000 waveform traces from GSN stations for over 1500 earthquakes between 1988 and 1994 to illuminate the global seismic wavefield. Data were binned according to their source–receiver distance, aligned on their respective event’s origin time, and ratio functions of the waveforms were stacked. The ratio functions used were a short‐ to long‐term average (STA/LTA) calculated at each point along envelope functions of the traces (Earle and Shearer, 1994).
Almost two decades later, the Global Stacks product aims to expand on that original effort using an order‐of‐magnitude more data by initially collecting a data set of about 2 million traces from the IRIS‐DMC for shallow (≤50 km depth) M≥5.4 events from January 1990 to June 2012 and then culling them to about a million traces. The culling uses various quality metrics such as event priority based on time and magnitude, presence of gaps in data, and signal‐to‐noise thresholds for various phases. The phases most often used for this were the first‐arriving P‐wave and SS. Further details are given on the product webpage. Subsets of narrowly filtered versions of the data set were formed and further refined using the same metrics but with thresholds tuned for the frequency band used. A dozen passbands were chosen and range from 1 to 64 s. Long‐period data sets were also formed similarly using vertical‐, radial‐, and transverse‐component data. The duration of vertical broadband stacks is 90 and 180 min for all of the long‐period stacks.
Once the data set was reduced within each frequency band, multiple stacks were generated by stacking traces in 0.5°‐wide distance bins using the filtered traces, envelope functions, and STA/LTA functions. The STA/LTA functions were saturated at a value of 25 to minimize the impact of especially high signal‐to‐noise ratio P waves. Most stacks were simple linear stacks, though the STA/LTA functions were separately stacked using square‐root stacking (McFadden et al., 1986) to further enhance weak phases. Figures available on the product webpage plot square‐root stacked STA/LTA functions. To generate a smoothly varying image, the stacked trace from each distance bin was normalized by the peak amplitude. Furthermore, the traces were raised to the ninth root, and then the median was removed. Finally, a sliding 120‐s wide amplitude‐normalizing window was applied which slightly enhanced faint phases. The product webpage contains multiple plots of the various stacks using different frequency bands, components, and plot types (0°–180° and 0°–360°, with and without travel‐time curves) and includes a poster‐sized composite image of multiple plots. Figure 5 (inset) shows the composite image along with the composite image from Astiz et al. (1996). Although the frequency bands used in the two images differ slightly, they are similar enough to highlight the enhancement achieved by adding over an order of magnitude more data. For example, PKPPKPPKP (a P wave thrice reflected at the surface after traveling through the core, also known as P′P′P′) at around 60 min and 80° is clearly visible in the updated stacks. Even P′P′P′P′ at around 80 min and 130°, though faint, is also visible. Although direct measurement of amplitude of such exotic phases (e.g., Gutenberg, 1960) is not meaningful with these nonlinear stacks, comparison with similar stacks formed from synthetic seismograms using different attenuation models may help refine future models. Plots and bundled text files containing the envelope and STA/LTA stack values are available on the product and product supplement pages for each frequency band analyzed. Text files with lists of the initial ∼1 million‐trace data set, which may serve as a useful starting point for large global seismology studies, are also available and easily parsed.
In a related analysis, vespagram stacks (Rost and Thomas, 2002) were made using data from USArray and the BK, CI, US, and UW networks. Data were analyzed for each event in marching 5°‐wide distance bins every 0.5°. Within each distance bin and in different frequency bands, data were culled according to previously described quality metrics, then each trace was shifted slightly based on a multichannel cross correlation (VanDecar and Crosson, 1990) using a window around the first P arrival. Each vespagram was normalized by the peak amplitude and then further median normalized as described in detail on the webpage. This final step of processing ensures a common noise level before stacking vespagrams from all events. Finally, vespagrams from each event within each frequency band are stacked. Bundles of the vespagrams as text files, and animations showing vespagrams from 10 narrow frequency bands ranging from 1 to 48 s sweeping through different distances are available on the product supplement page (see Data and Resources).
Earth Model Collaboration (EMC) 3D Visualization Tool
The IRIS‐DMC EMC data product has served as a community‐supported repository of Earth models since November 2011. This product facilitates access to various Earth models under a uniform format (NetCDF) and provides multiple 2D visualization tools for model visualization and comparison. The EMC repository currently holds
38 seismic‐velocity models (3D);
7 reference Earth models (1D);
1 crustal thicknesses model (2D);
1 seismic‐attenuation (Q) model (3D); and
2 electrical conductivity models (3D).
The IRIS EMC’s web‐based 2D visualization tools are also available as a downloadable bundle. By installing this bundle, users can locally host the same visualization tools that are currently available from the IRIS EMC website and use their own local NetCDF Earth model files as well.
As a complement to EMC’s 2D visualization tools, an open‐source Python desktop application for 3D visualization is now available. The IRIS EMC 3D Visualizer with its simple interactive 3D visualization capabilities (Fig. 6) is based on the Mayavi Data Visualizer 3D application. EMC 3D Visualizer’s simple user interface (UI), minimum reliance on external packages, and ability to work directly with the Earth models in NetCDF format make it an effective 3D tool to bridge the gap between the existing EMC’s 2D visualization tools and complex viewers with steeper learning curves such as Unidata Integrated Data Viewer and ParaView.
The EMC 3D Visualizer’s flexible configuration parameters allow users to create region‐specific parameter files that can be used to create customized views of the models. The UI also allows users to easily change models and parameters in the same scene to facilitate comparison. In addition to displaying horizontal and vertical slices and isosurfaces from the loaded model, the script is configured to plot earthquakes from an International Federation of Digital Seismograph Networks (FDSN)‐event web service, the 5‐min resolution topography model from the National Oceanic and Atmospheric Administration (ETOPO5) slab models for subduction zones, plate boundaries, and present‐day hot spots (see Data and Resources). Changing the source of the auxiliary data or adding customized auxiliary data in many cases is simple and requires minimum coding and modification.
TA Infrasound Reference Event Database
Based on the station metadata available at IRIS, throughout 2011 infrasound sensors were added to hundreds of existing USArray TA sites and became a standard component at all new sites. By late 2012, there were over 400 sites with infrasound sensors with an average spacing of 70 km. In 2012, to promote and facilitate the use of these infrasound data, the IRIS‐DMC introduced two data products: an infrasound reference‐event database, known as TAIRED, and an infrasound signal detector, known as TAID. However, because of the TA infrasound single‐sensor arrangement, no systematic event locations were determined at that time.
Recently, de Groot‐Hedlin and Hedlin (2015) developed a new event detection and location technique AELUMA that relies on dividing sufficiently dense networks, such as TA, into a mesh of three‐element arrays (triads). Envelopes of waveforms from stations within each triad are cross correlated with each other to detect coherent signals. Then, sources are identified using a grid search to match coherent signals from clusters of triads with a common single origin. The AELUMA codes were contributed to the DMC Data Products team who utilized them to build an infrasound event catalog from TA infrasound data recorded between 2011 and 2015. This catalog contains over 2500 events that have been added to the Infrasound Event Database. These ALEUMA events, along with documented infrasonic events gathered from media reports (TAIRED), are merged into a single database and are available for download individually as record sections (Fig. 7) or as a whole as text files via SPUD (see Data and Resources).
Following all global earthquakes of magnitude 7.0 and greater, a package of maps and animations related to the event aftershock sequence is automatically generated and made available in SPUD. The package comprises an aftershock animation (Fig. 8), background seismicity maps, an aftershock map, plots of magnitude versus time, and distance along strike versus time using both strikes from the Global CMT focal mechanism, if available. The animation and related figures are automatically updated multiple times a day for 10 days following the mainshock. When available, Global CMT magnitudes and origins are preferred, and their focal mechanisms are included in the plots. Results are backfilled to 1980 using the Global CMT catalog as reference. The radius of aftershocks included in the plots increases with magnitude from 100 km for magnitude 6.0 events to 1000 km for magnitude 9.0 events.
The IRIS‐DMC’s automated Quality Assurance team has been gathering spectral characteristics of broadband seismic‐station data arriving in real time since 2004. Currently, these spectral characteristics are available from IRIS’s modular utility for statistical knowledge gathering system in the form of power spectral density (PSD) estimates and probability density functions (PDFs) for seismic channels. These PSDs and PDFs are computed based on predetermined parameters for seismic channels.
Different studies relying on spectral characteristics of ambient noise at a particular station rely on custom tuning of the frequency band, smoothing parameters, and the time window for PSDs, PDFs, and other spectral characteristics. To address these needs, the IRIS‐DMC has created the Noise Toolkit (NTK) data product (see Data and Resources). This toolbox of three highly configurable open‐source Python bundles facilitates customized spectral characterization of waveform data, including those from nonseismic channels. The bundles included in the NTK are as follows.
1. PDF/PSD bundle: The scripts in this bundle compute PSDs of waveform data using customized parameters. With this bundle, users can:
• request waveforms and response data for given stations/channel(s) using the ObsPy FDSN client or read the user’s own waveform data files (in SAC, MSEED, CSS, etc., format) and only request response information from IRIS.
• compute and populate an hourly file‐based PSD database.
• extract PSDs and PDFs from the PSD database.
2. Microseism energy bundle: The scripts in this bundle compute microseism energy (ME) from PSDs. These ME plots can reveal coherent variability of microseism noise across multiple stations as well as those caused by major storms. These scripts can help users:
• calculate power from existing PSDs over selected period bands (bins).
• calculate median power for a given time window from the computed PSD powers.
• plot temporal variation of the median power.
3. Frequency‐dependent polarization analysis bundle: The standalone scripts in this bundle perform polarization analysis of seismic noise recorded by three‐component seismometers:
• request waveforms and response data for given stations with three‐component data (ZNE) using the ObsPy FDSN client or read user’s own waveform data files (in SAC, MSEED, CSS, etc., format) and only request response information from IRIS;
• compute and populate an hourly file‐based three‐channel polarization attributes database;
• and extract hourly and daily polarization PDFs from the polarization database.
The above bundles could be used in many teaching and research applications such as station quality control, background noise characterization, generating weather maps of seismic noise, creating near‐real‐time microseism index databases, and creating plots as well as monthly summaries for data mining. Figure 9 shows an example of combined application of the PDF/PSD and ME bundles to look at the ME variations at the GSN station Davao, Philippines (IU.DAV) during the passage of super typhoon Haiyan in November 2013.
Other New IRIS‐DMC Data Products (Developed 2012–2016)
EQEnergy: Earthquake energy and rupture durations are estimated following all earthquakes with initial magnitude above Mw 6.0 and a Global CMT (following Convers and Newman, 2011).
Shake Movie Synthetics: 1D normal‐mode and 3D AxiSEM (Nissen‐Meyer et al., 2014) synthetic seismograms from the Global ShakeMovie project are available from the IRIS‐DMC archive using network code SY and location codes S1 and S3, respectively (Tromp et al., 2010).
ASWMS: Do‐it‐yourself automated surface‐wave tomography using the MATLAB‐based ASWMS package developed by Jin and Gaherty (2015). This product provides the ASWMS software package and weekly updated USArray surface‐wave tomography maps using ASWMS.
SWS‐DB‐MST: The Missouri University of Science and Technology SWS database. A teleseismic SWS database based on all available data from all broadband seismic stations in North America that contains over 16,000 pairs of well‐defined (and manually checked) SWS measurements for the western and central United States (west of 90° W).
Envelope Functions: A collection of continuous envelope functions to search for tectonic tremor and for data quality analysis. Envelope functions are computed automatically at the University of Washington utilizing select stations in the Pacific Northwest, Alaska, and Chile and are distributed by the IRIS‐DMC.
ANCC‐CIEI: An ambient noise cross‐correlation (ANCC)‐based database of empirical Green’s functions of the western United States using USArray TA data generated at the Center for Imaging the Earth’s Interior at the University of Colorado in partnership with the IRIS‐DMC following procedures outlined in Bensen et al. (2007).
Global Empirical Greens Tensors: A nested global empirical Green’s tensor database derived from three‐component continuous data at global‐, continental‐, and local‐length scales following Shen et al. (2012). The open database is intended for anyone to contribute to as long as data processing follows a consistent best practice. Inquiries for contributing should be directed to the DMC Products team.
SeisSound: The SeisSound Visualization is an audio/video‐based IRIS‐DMC data product that conveys the frequency and amplitude content of seismograms both visually and audibly to produce a better understanding of their spectral content. This follows processing from Kilb et al. (2012).
EMERALD: A complete open‐source software server‐based system for requesting and processing large sets of event‐based seismic data from a web browser. Data sets containing millions of seismic waveforms can easily be managed, reviewed, and processed. The system can automatically check for metadata updates and alert the user to metadata changes.
The success and wide variety of the DMC’s data products are in large part due to the generous contributions from the research community. The scope of the contributions that would be considered is very broad, ranging from archiving data sets that would otherwise be lost, one‐off derivative data sets of a unique nature, routine calculation of well‐established research products, to distribution of active research products.
The intention is to continue to develop new products in collaboration with the research community for the foreseeable future and to continue to maintain the current suite. In consultation with our governance committee, we will continuously review usage patterns and expand, refine, or deprecate products as needed to best use the finite resources available. In cases where products may be discontinued, we expect to continue to offer what has already been calculated in some form. We look forward to continuing to serve and expand our data‐product offerings. New contributions and suggestions for new data products from the community are welcome and may be submitted using the contact information on our main data product webpage.
DATA AND RESOURCES
Waveform data used to generate the highlighted data products were retrieved from the Incorporated Research Institutions for Seismology (IRIS) Data Management Center (DMC) at http://ds.iris.edu. All data retrieved from the IRIS‐DMC are openly available. All data products and software tools offered as data products are available from the IRIS‐DMC at http://ds.iris.edu/ds/products. Many of the data products are available from the Searchable Product Depository (SPUD) at http://ds.iris.edu/spud. Earth models, Earth Model Collaboration (EMC) 2D visualization tools, and the EMC 3D Visualizer are openly accessible from the EMC website http://ds.iris.edu/ds/products/emc/. Data used by the EMC visualization tools include: ETOPO5 topography (http://www.ngdc.noaa.gov/mgg/global/etopo5.HTML), slab models for subduction zones (http://earthquake.usgs.gov/data/slab), plate boundaries (http://www.ig.utexas.edu/research/projects/plates), and present‐day hot spots (https://www.ngdc.noaa.gov/nndc/struts/form?t=102557&s=5&d=5). Webpages created to highlight Special Events are available at http://ds.iris.edu/ds/nodes/dmc/specialevents. The “super GMV,” a combination of multiple ground‐motion visualizations (GMVs), is available at http://ds.iris.edu/ds/products/usarraygmv-super. The animation of USArray microseism noise recorded during Hurricane Sandy can be found at http://ds.iris.edu/ds/products/hurricanesandy. Related to the Global Stacks data product, vespagrams as text files and animations showing vespagrams from 10 narrow frequency bands are available at http://ds.iris.edu/ds/products/globalstacks_supplement.
Data products at the DMC are grouped below, based on their availability and application:
Event‐based products that are generated automatically within a few hours following triggering events: USArray GMV (http://ds.iris.edu/spud/gmv), Event Plots (http://ds.iris.edu/spud/eventplot), Backprojections (http://ds.iris.edu/spud/backprojection), Earthquake Energy (http://ds.iris.edu/spud/eqenergy), Rayleigh‐Wave Source Time Functions (http://ds.iris.edu/spud/sourcetimefunction), Aftershocks (http://ds.iris.edu/spud/aftershock), Global ShakeMovie synthetic seismograms (http://ds.iris.edu/spud/synthetic), and Moment Tensors (http://ds.iris.edu/spud/momenttensor).
Earth modeling and waveforms: The IRIS‐DMC EMC (http://ds.iris.edu/ds/products/emc), EarthScope Automated Receiver Survey (EARS) (http://ds.iris.edu/ds/products/ears), University of Washington Continuous Envelope Functions (http://ds.iris.edu/ds/products/envelopefunctions), The Missouri University of Science and Technology (Missouri S&T) shear‐wave splitting database (http://ds.iris.edu/ds/products/sws-db-mst), The Géosciences Montpellier SplitLab’s Shear‐Wave Splitting Database (http://ds.iris.edu/ds/products/sws-db), and Western U.S. Ambient Noise Cross‐Correlations (http://ds.iris.edu/ds/products/ancc-ciei).
Detections: Transportable Array (TA) Infrasound Reference Event Database (TAIRED) (http://ds.iris.edu/ds/products/infrasound-taired), TA Infrasound Detections (TAID) (http://ds.iris.edu/ds/products/infrasound-taid), and Infrasound Automated Event Location Using a Mesh of Arrays (AELUMA) (http://ds.iris.edu/ds/products/infrasound-aeluma).
Software packages: The IRIS‐DMC Noise Toolkit (NTK) (http://ds.iris.edu/ds/products/noise-toolkit), EMERALD—a software framework for seismic‐event processing (https://seiscode.iris.washington.edu/projects/emerald), and ASWMS—do‐it‐yourself automated surface‐wave tomography using the MATLAB‐based ASWMS package (http://ds.iris.edu/ds/products/aswms).
Other products: SeisSound—an Audio/Video Seismic Waveform Visualization (http://ds.iris.edu/ds/products/seissound/), EMTF—Magnetotelluric Transfer Functions (http://ds.iris.edu/ds/products/emtf/), Film Chip and Station Digest, Event Bulletins.
All URLs were last accessed on March 2017.
We thank the past members of the Incorporated Research Institutions for Seismology (IRIS) data‐products working group, which represented the research and educational community, for the guidance provided while developing these data products and generally for the assistance in this effort. We also thank contributors that have provided their data products to us for distribution, in particular, the authors of the tomographic models in the Earth Model Collaboration product, and those who contributed codes for Infrasound Automated Event Location Using a Mesh of Arrays, the Noise Toolkit, R1 Source Time Functions, and Earthquake Energy. We further thank the many organizations that contribute their data to the IRIS‐Data Management Center (DMC) for open distribution and use. Finally, we thank Debi Kilb and an anonymous reviewer for their constructive reviews which improved the article. The initiation of data‐product development at IRIS was supported by the National Science Foundation (NSF) Award EAR‐0733069 as an EarthScope/USArray activity, with support of the DMC facility through NSF Awards EAR‐0552316 and EAR‐1063471, and continuing support through NSF Award EAR‐1261681.