- © 2012 by the Seismological Society of America
In 2010, the Incorporated Research Institutions for Seismology (IRIS) Data Management Center (DMC) expanded the preexisting effort to generate, archive, and distribute data products derived from the extensive data archives of the center, which include the IRIS Program for Array Seismic Studies of the Continental Lithosphere (PASSCAL), Global Seismographic Network (GSN), EarthScope, and many other data sources. This expansion included dedicating two new full‐time staff to data products and to the development of a new system to manage and distribute them. With community guidance provided by a data products working group (DPWG), DMC has developed a number of freely available unique and increasingly popular data products for the seismological community. In addition, DMC is also archiving and distributing other data products that are produced by community members. A complete list of data products managed by DMC is maintained here: http://www.iris.edu/dms/products. Many of these are available from the Searchable Product Depository (SPUD, http://www.iris.edu/spud), the DMC data product management system.
Traditionally, data products result from specialized processing applied during research by the principal investigators and their teams. Products span the range of simple plots or animations derived from waveform data to highly refined models. Some waveform‐processing techniques (e.g., receiver functions) have gained wide community use and can be derived by widely accepted algorithms. Such first‐order data products make ideal candidates for routine generation to provide a standard for the user community. Data products are typically derived from raw data to serve as insightful summaries of the raw time series data, or as quality control metrics, or as foundations or stepping stones to more specialized products and research.
The overall goal of the DMC data product effort is to produce and manage a variety of community‐approved and utilized data products that complement the raw data managed at the DMC and ultimately serve as useful and time‐saving resources for research and education and outreach. Many factors are considered by the DPWG regarding which products to produce or archive at the DMC, including use as new fundamental data sets, characterization of the raw data, and education and public outreach uses.
Here, we describe data products developed at the DMC and are now available. We start by describing event‐based products. Event‐based products developed at the DMC are automatically triggered, generated, and distributed without human review. They are typically generated approximately an hour after the origin time and use the hypocenter and magnitude estimates available at the time (typically U.S. Geological Survey National Earthquake Information Center). We follow this with a description of other products, including community‐developed data products, available from DMC. The descriptions reported are for the products currently being produced. Although we do not anticipate significant changes, parameters may be adjusted in the future, and any such changes will be reported on the product description webpages.
USARRAY GROUND MOTION VISUALIZATION (http://www.iris.edu/dms/products/usarraygmv)
Dense arrays provide opportunities for earthquake wave‐field visualization using time sequences of collective ground motions. The Transportable Array component of the EarthScope USArray (http://www.usarray.org) consists of 400 broadband stations on a roughly uniform 70‐km grid that is currently migrating from the west to the east coast of the conterminous United States, with individual sites being occupied for approximately 2 years. This very large aperture array, along with other stations from the USArray, is spatially unaliased at very long periods and is especially well suited for visualizing continental‐scale broadband seismic‐wave fields.
USArray‐based ground motion visualizations (GMVs) were first developed by Charles Ammon (Ammon and Lay, 2007) and are now a standardized and automated animation‐based data product of IRIS DMC for vertical‐component ground velocities for all large events (M≥6 globally or M≥5.5 within the United States; see Fig. 1). For events with M≥7.0, three‐component GMVs are also generated (see Fig. 2). GMVs are available within 4 h after the earthquake origin time.
GMVs use a red–blue color scale to represent the amplitude of normalized vertical ground velocity at each station. From one frame to the next, the symbol color tracks the amplitude at each seismometer. In addition to using color to represent vertical motion, three‐component GMVs (Fig. 2) use tailed symbols to show the direction and amplitude of the normalized horizontal motion. GMV processing parameters are given in Table 1.
To complement the ground motion depicted by the colored symbols (Figs. 1 and 2) and to provide a sense of the ground displacement, the GMVs also show a plot of a representative vertical‐displacement seismogram from a reference station, where a yellow circle on the map highlights the reference station location. In addition, the three‐component GMVs display the horizontal‐displacement seismograms for the north–south and east–west components. Selected teleseismic‐phase arrivals are also indicated on the GMV seismograms; the MatTaup toolkit (http://www.ess.washington.edu/SEIS/FMI/matTaup.htm) is used to calculate predicted phase‐arrival times. The estimated minor great circle (R1) and major great circle (R2) Rayleigh wave‐arrival times are also marked when appropriate using a reference velocity of 3.9 km/s.
Figures 1 and 2 show a single frame of the vertical‐component and three‐component GMVs of the Mw 9.0 Tohoku earthquake of 11 March 2011 05:46:24 coordinated universal time (UTC), respectively. These frames show the ground motion across USArray at 06:26:39 UTC (approximately 40 min 16 s after the origin time). In Figure 2, one can see that the tangentially polarized Love wave (with a back azimuth of ∼315°) has passed the reference station and the slower and radially polarized Rayleigh wave trails close behind.
For users who would like to generate their own GMVs or to view custom GMVs generated at the DMC using user‐specified stations and processing parameters, a customized GMV web interface is also available (http://www.iris.edu/dms/products/usarraygmv/customize) along with the MATLAB codes needed to make the animations.
EVENT PLOT SUITE (http://www.iris.edu/dms/products/eventplot)
The Event Plot product is a suite of up to 80 plots that are automatically generated following all global M≥6.0ln earthquakes. These are meant to give users a cursory look at essential aspects of the data. Some quality control has been applied so that egregiously bad data and results are omitted, which can result in plots with seemingly few data or certain plot types omitted. The plots are generated using all open broadband data available at the IRIS DMC at the time the product was produced. Different plot types will be described later.
A basic map of all broadband high‐gain vertical (BHZ) stations is available at IRIS DMC, as well as an additional equidistant and equiazimuthal map, with colors representing the first P‐arrival (P/Pdiff/PKP) signal‐to‐noise ratio.
Global Surface‐ and Body‐Wave Record Sections
Multicomponent record sections using long‐period high‐gain (LH) data from GSN are shown, as well as body‐wave record sections using all BH data available. The body‐wave record sections are combed to show stations with a regular distance distribution that meet a minimum signal‐to‐noise ratio criteria (record sections that are not combed would appear as solid black regions at distances where many stations are present [e.g., USArray]). All record sections are aligned on the event origin time. Body‐wave record sections are accompanied by an equivalent version, with travel‐time curves of major phases overlain. All record sections plot instrument‐response‐corrected data; body‐wave record sections plot velocity, and surface‐wave record sections plot displacement. Different filtering and components are plotted to highlight different phases.
Phase‐Aligned Record Sections
Body‐wave record sections similar to those described previously in this paper but with data aligned on the first P wave (P, Pdiff, or PKPdf), S, or SKS. The plots aligned on S and SKS plot both the radial (blue) and tangential (red) components of displacement.
EarthScope USArray Body‐ and Surface‐Wave Record Sections
Body‐wave record sections are combed and show solely USArray BH data. The surface‐wave plots show combed USArray LH data filtered between 20 and 125 s and share the same amplitude scaling for all three seismogram components. The enhanced record sections that follow use the same traces but display the amplitudes using a nonlinear color scale, which can show a wider dynamic range of amplitudes.
Virtual Array P‐Waves and Vespagrams
First P arrivals aligned using multichannel cross‐correlation coefficient (MCCC) (e.g., VanDecar and Crosson, 1990) with windowing from −4 to +10 s around P and accompanying vespagrams are shown for all data, as assembled from a virtual regional network formed from all available stations within a region. Predefined virtual arrays are as follows: TA, central United States, mostly USArray data; UW, Pacific Northwest; CA, California and the southwestern United States; AK, AK and AV networks in Alaska; AU, Australia; and EU, central/western Europe. Data with average cross‐correlation coefficients (CCCs) below 0.6 are discarded. Two frequency bands are attempted for each virtual array, 0.3–1.0 and 0.1–0.5 Hz. The station average lags and CCCs from the MCCC are available as ASCII files. The vespagrams, which are stacks of seismogram amplitudes across the arrays as function of slowness and time, use a reference distance based on the mean source–receiver distance of all traces within the virtual array. Some of the virtual regional array apertures span up to ∼1500 km to maximize the number of traces. This can have the effect of smearing in the slowness (vespagram y‐axis) direction. Shallow or long‐duration earthquakes with extended coda tend to smear phases in the time (vespagram x‐axis) direction. Three vespagrams are constructed using linear, cube root, and a nonlinear stacking technique. The highly nonlinear version uses a nonlinear color scale unique to each vespagram so that features with amplitudes in the 99th and 50th percentiles are both visible. This is helpful for vespagrams using data that have a very impulsive and large first P arrival that may otherwise obscure all other features in the image.
Global Body‐Wave Envelope Stacks
Global body‐wave envelope stacks are generated by stacking of envelope functions in 1°‐distance bins (e.g., Earle and Shearer, 2001). The envelopes have the mean pre‐P‐wave noise removed, and the square roots of the amplitudes are then summed. The stacks are nonlinearly weighted by the number of contributing seismograms to enhance visualization of the wave field. Similar to vespagrams, these plots are especially useful for quickly scanning events to determine whether specific phases are visible in the data or not.
P‐Wave Coda Stacks
P‐wave coda stacks process data from 0° to 95° in the same way as the global body‐wave envelope stacks do, except that the data are stacked in either azimuth or distance bins and are linearly stacked. For very large earthquakes, a cosine azimuthal dependence can indicate rupture directivity. Rupture duration is sometimes also discernable, although coda durations have been shown to correlate with source depth, with shallow events typically having a slower coda decay rate (Shearer and Earle, 2004). The summation trace at the top can indicate the basic source characteristics of the event, such as whether it was a single or a doublet impulsive event. This distinction can be complicated for shallow events. Depth phases have short lag times relative to P, so travel‐time curves for pP, sP, PcP, and PP are included.
BACK‐PROJECTION SOURCE IMAGING (http://www.iris.edu/dms/products/backprojection)
The back‐projection (BP) product shows the beamformed time history and geographic location of coherent short‐period P‐wave energy generated by large earthquakes observed at three regional arrays and across the GSN and are generated following all global M≥6.5↦ earthquakes. These BPs are meant to provide a standardized time‐evolving image of the radiation of high‐frequency P‐wave energy. Sometimes, overall rupture features (duration, directivity, and length) can be inferred for larger events (e.g., M>∼7.5) when two or more arrays produce BPs that are in general agreement. Azimuthally dependent Green’s functions, particularly arising from the interactions between depth and direct phases, will, however, typically produce somewhat different results for different imaging arrays. These automated BP results for smaller events with rupture lengths less than ∼30 km (M<∼7.0) usually mimic point sources and are thus below the resolution needed to interpret rupture characteristics. However, distinguishing between a simple rupture and an earthquake doublet is sometimes possible. BPs can also show the absolute location of earthquakes as imaged by the particular receiving array within the 1D reference model (Preliminary Reference Earth Model; Dziewonski and Anderson, 1981).
BPs are performed by beamforming (stacking) energy to a flat square grid at the hypocentral depth in the source region with variable spatial resolution. The size of the source region is scaled by the magnitude of the earthquake (Table 2). Data from the North America (NA), Europe (EU), and Australia (AU) virtual networks are filtered at 0.25–1.0 Hz, and data from the GSN network are filtered at 0.05–0.25 Hz. A sliding 5‐s wide cosine tapered window is used to average the beams (calculated every 0.25 or 0.5 s) to form BP images in increments of 1 s. Before generating the BP animations (Fig. 3), the images are averaged over a 10‐ to 30‐s tapered window, depending on the magnitude, to reduce artifacts that appear to swim in one direction that can that result from the limited slowness space sampling by the finite source‐receiver geometry (described subsequently in this paper). Static images of the cumulative BP stack are formed by summing the images over all times for which BPs are calculated. In the animations, warmer colors indicate greater beam power, and a red circle shows the location of the peak beam power when the absolute beam powers are low. When beamforming, many stacking approaches can be used. Linear stacking tends to be more true to amplitudes, although for many events, results can be very noisy. Nth‐root stacking offers greater enhancement of signal‐to‐noise ratio and highlights coherent bursts of high‐frequency energy. Square‐root (N=2) stacking is used by default for this product.
The BP virtual networks used are formed by choosing stations within 25° distance or azimuth to the center of three continents for NA, EU, and AU; GSN is also used. All available BHZ data are then aligned before filtering using an MCCC scheme. The length of the cross‐correlation window is based on a short‐term average/long‐term average sliding window and will be short (−5 to +10 s relative to P) for impulsive events and extended longer for emergent events. Finally, the algorithm resamples the network to only use stations with the highest average CCC and having a nearest station neighbor distance of at least 1°.
Each BP animation also shows a plot of peak beam power over all grid points as a function of time. Beam power does not directly correspond to the amount of slip or to the scalar moment rate function, although for many events, the inferred rupture duration from these is consistent with each other; many factors influence the beam power versus time curves, including processing parameters such as the window length used to average results. At the bottom of the BP page in SPUD, ASCII files showing peak beam power and the peak beam geographic location as a function of time relative to origin are available for both the data and the synthetic array response function (ARF; described subsequently in this paper).
In the BP animations, energy commonly appears to swim in the direction of the array. This happens because there is some tradeoff between origin time and distance from the receiving array for stacked energy. Ideally the true absolute location of bursts of energy will always be revealed by the BPs; however, the results are influenced by many factors including complicated Green’s functions (dipping slab and depth phases), radiation pattern effects, and variable slip dip/rake/direction as the rupture propagates (e.g., Lay et al., 2010). Limited source‐receiver geometry can also produce unintuitive constructive and destructive interference and is often the greatest source of noise, especially for smaller M<∼7.5 events.
To estimate the effect of the limited source‐receiver geometry for a given array configuration, an ARF is generated. To estimate an ARF, the same stations and processing are used, except that the seismograms are replaced with point‐source triangle source function synthetics. That is, traces whose amplitudes are zero everywhere, except at the predicted arrival time of the P wave, which has a 3‐s‐wide triangle function. BP results with very wide ARFs are automatically culled, and results with complicated ARFs that remain should be viewed judiciously and as being indicative of strong array‐related artifacts. ARFs are often asymmetric in space and in time due to interference and time averaging.
Included in the BP summary figures (Fig. 4) are the time and location of local maxima. The circle diameter corresponds to relative beam power and the colors corresponds to relative time. For well‐imaged events, the circles will outline the major subevents. Conversely, an image with numerous scattered circles usually indicates a poorly imaged rupture. These images offer only a quick static look at the rupture by the imaging array and are often only useful for very large earthquakes.
For reasons summarized previously in this paper, we reemphasize that artifacts can abound in these automatic BPs at their current state of development, and a growing number of advanced approaches and higher resolution results using careful manual processing appear in the literature (e.g., Meng et al., 2012). Poor BPs are often the result of bad data selection and alignments, which is difficult to automate. Egregious problems in data selection and alignment are easily discerned in the MCCC‐aligned and scaled traces in the lower right panel of the summary figures (Fig. 4). Koper et al. (2012) provided a useful overview of BP limitations and interpretation and described how processing details can affect results.
OTHER EVENT‐BASED DATA PRODUCTS
IRIS DMC is managing and distributing a number of other event‐based data products created by the IRIS community or other institutions including the following:
The EarthScope Automated Receiver Survey (Crotwell and Owens, 2005), a system developed to calculate bulk crustal properties and to generate a large catalog of receiver functions: http://www.iris.edu/dms/products/ears
The Global ShakeMovie 1D and 3D synthetic seismograms (Tromp et al., 2010) converted to SEED format and available using any of the DMC’s request mechanisms: http://www.iris.edu/dms/products/shakemoviesynthetics
Event bulletins produced by the USArray Array Network Facility and a University of Washington phase picking project supported by IRIS: http://www.iris.edu/spud/eventbulletin
THE EARTH MODEL COLLABORATION (http://www.iris.edu/dms/products/emc)
Advances in seismic modeling and the ever‐growing volume of data have brought about a rapid growth in the development and use of 2D and 3D velocity Earth models at regional and global scales. However, the file formats that various authors use to present and distribute the tomography models are diverse. To facilitate discovery and access to the Earth models, IRIS DMC has developed and maintains the Earth Model Collaboration (EMC), a website serving as a repository for contributed regional and global velocity models. The EMC repository uses the netCDF file format (http://geon.unavco.org/unavco/IDV_seismic_tomo_data.html) as a unifying container for contributed models. EMC guidelines and templates for authors to convert their Earth models to netCDF format and directions on how to contribute them to the repository are available at http://www.iris.edu/dms/products/emc/modelContribution.htm.
Each model in the EMC has its own model webpage that contains a description and model‐related references and links (http://www.iris.edu/dms/products/emc/models). A number of reference models are also hosted in the EMC repository (http://www.iris.edu/dms/products/emc/models/refModels.htm). The EMC currently hosts 15 tomography models and seven reference models (Table 3).
The Earth models contributed to the EMC, and their reference models are available for download in their original and/or netCDF formats from the EMC model overview pages (http://www.iris.edu/dms/products/emc/models) or from SPUD (http://www.iris.edu/spud/earthmodel).
In addition to providing data, metadata, and references for each contributed Earth model, the EMC also provides a set of visualization tools that are based on the openly available and widely utilized Generic Mapping Tools (GMT) plotting package (http://gmt.soest.hawaii.edu). These visualization tools allow users to produce horizontal slices, vertical slices, and velocity profiles from the netCDF Earth model files. Users can further refine these plots by downloading a plot‐specific visualization bundle that includes both the model data and the GMT plotting scripts.
To enhance model visualizations, two of the EMC visualization tools, the Horizontal Slice Viewer (http://www.iris.edu/dms/products/emc/horizontalSlice.html) and the Generalized X‐section Viewer (http://www.iris.edu/dms/products/emc/gcross-section.html) provide access to some auxiliary data that can be plotted on the corresponding map or section, including the following:
Topography based on ETOPO5 elevation data
Earthquakes from the IRIS event database (http://www.iris.edu/ws/event)
Global CMTs from the IRIS database (http://www.iris.edu/spud/momenttensor)
Slab1.0, a 3D compilation of global subduction geometries (http://earthquake.usgs.gov/research/data/slab)
Plate boundaries (http://www.ig.utexas.edu/research/projects/plates)
Volcano locations (http://www.ngdc.noaa.gov/hazard)
All stations from the IRIS database (http://www.iris.edu/ws/station)
Named boundaries within the selected reference velocity model
Figure 5 shows a horizontal slice map at the depth of 150 km for the TX2011 model (updated model based on TX2000 from Grand, 2002), created using the Horizontal Slice Viewer. Google Maps is used to select the mapping region. The horizontal slice also includes auxiliary data to show the locations of volcanoes (red triangles), earthquakes (colored solid circles) for the depth range of 138–162 km (halfway to the model depths directly above and below the selected slice depth of 150 km), plate boundaries (magenta line), and four contour lines from Slab1.0 (Hayes et al., 2012) showing slab depth in the subduction zone at 20‐km intervals above and below the slice depth of 150 km. Background colors represent shear‐velocity perturbations in TX2011 relative to its reference model at 150‐km depth.
Figure 6 is a cross section along latitude 36.5°N of the TX2011 model created using the Generalized X‐section Viewer, which uses Google Maps to select the great circle arc of the record section. The profile is located in the middle of the area covered in Figure 5. The cross section also shows earthquakes (blue dots) and volcanoes (red triangles) in the vicinity of the profile (within −0.5°), with earthquake symbols scaled based on their magnitude. Location of the Japan trench and depth to the top of the subducting slab (dashed magenta line) from Slab1.0 are shown. The section colors indicate variations in shear‐velocity perturbation relative to its reference model and display faster velocities to the east of the subduction zone and slower velocities in the shallow portion of the subduction zone wedge.
NON‐EVENT‐BASED DATA PRODUCTS
IRIS DMC manages and distributes a number of other non‐event‐based data products created by the IRIS community or other institutions including the following:
Shear‐wave splitting measurement database, initially containing a mirror of the Splitting Database (Wüstefeld et al., 2009) maintained by the SplitLab group at Géosciences Montpellier: http://www.iris.edu/dms/products/sws-db
Magnetotelluric transfer functions produced by the MT component of USArray: http://www.iris.edu/spud/emtf
WWSSN Film Chip image archive: http://www.iris.edu/spud/filmchip
Station digests from the USArray Transportable Array, a summary of station‐specific details for each occupied TA site: http://www.iris.edu/spud/stationdigest
Calibration data for the IRIS/IDA (International Deployment of Accelerometers) component of the GSN: http://www.iris.edu/spud/calibration
The DMC data product effort represents a unique collaboration between the research community and IRIS. The strengths of each are leveraged, with the community providing domain expertise and guidance and the DMC focusing on noncontroversial data processing, product management, and long‐term distribution and duration.
DMC is continuing to develop new data products, with many new ones anticipated for release in 2012. New products will be added to the list on the products homepage, where we welcome suggestions from the community for new data products (http://www.iris.edu/dms/products).
We thank the current and past members of the Incorporated Research Institutions for Seismology (IRIS) data products working group, representing 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, the Global Centroid Moment Tensor project members, and the Global ShakeMovie project members. Development of data products at the DMC was supported by the National Science Foundation (NSF) Grant EAR‐0733069 (EarthScope USArray), and the DMC facility was supported by NSF Grants EAR‐0552316 and EAR‐1063471 (Instrumentation and Facilities).