Data Services Products: globalstacks Global stacks of millions of seismograms


Global stacks of up to a million event-windowed seismograms using short-term to long-term averages (STA/LTA) in different frequency bands for vertical broadband data (1990-2012) available from the IRIS DMC. Long period versions include vertical and horizontal component data.


In 1996 Astiz et al. stacked over 33,000 traces from the IRIS FARM archive (1988-1994) to illuminate the global seismic wavefield. In a similar spirit almost two decades later, this product aims to expand on that original effort using an order of magnitude more data, much of it from USArray. An initial data set of about two million traces available from IRIS spanning shallow M≥5.4 events from 1990-2012 was collected and then culled using various QC metrics to about a million traces. Plots of global STA/LTA stacks using broadband (BH) vertical component data were made using different frequency band filters and various processing techniques aimed at enhancing the signal to noise ratio of weak phases. This product provides figures, the stacked STA/LTA functions used in the figures, other stacked time series (e.g. envelope stacks), as well as lists of the QCed traces used in the final data set.

Zip files of the broadband vertical stack images in different frequency bands.

Zip file of global stack images in all frequency bands for broadband vertical data, with and without travel time curves:

  • (13 MB) From 0-180 degrees distance
  • (25 MB) From 0-360 degrees distance
    Note: the plotting in these is less processed than the images with black background in order to better show subtle features.

Long period stacks: Vertical, Radial, Transverse

Long period data (LH) were processed and stacked in a similar fashion to broadband (BH) data, including horizontal component data rotated to their radial and transverse components. A slight difference was the long period analysis used P, S, SS and SKS as the reference phases for signal-to-noise ratio based quality control. The long period data were stacked out to 180 minutes after origin time. Figures are available from 0-180 and 0-360 degrees distance and from 0-90 and 0-180 minutes.
Note: Longer STA & LTA windows and filtering causes phase arrivals to appear slightly late, e.g. the first arriving P- or S-wave at 0 degrees in either image below. Filtering side lobes are also responsible for the coherent arrivals arriving at origin time from 0-30 degrees distance.

phase arrivals in the longer period stacks appear to arrive slightly later due to the longer STA & LTA windows and filtering used.

Zip files of long-period global stack images in different frequency bands for vertical, radial and transverse component data:

Travel time curves. See a phase without a curve? Email us and we’ll add it!

Travel time curves of identified phases have been overlain on the stack images, not all visible phases have been identified in this first release of this product. Help identify unmarked phases: If you are able to identify a phase, please send an email (contact at bottom of page) and it will be added in an update.

Data set & processing

  • use FetchEvent to query all shallow (z<50km) M≥5.4 events in ISC & USGS earthquake catalogs from 1990-2012.06
  • events are ignored if another event of similar or larger magnitude occurred within an hour of it’s origin time
  • collect 90 minutes of all available BHZ data from the archive at the IRIS DMC using FetchEvent
  • any records with spikes, offsets or gaps are eliminated
  • a rough estimate of SNR is calculated in windows around P,Pdiff,PKP,S in two frequency bands. If none of the SNRs ever exceed 3.0, the trace is eliminated.
  • about a million seismograms remain
  • data are filtered in narrow bands, and further QCing is applied within each band using median envelope and phase SNR metrics
  • an STA/LTA function is calculated for each trace within each frequency band. The length of the STA window is variable with frequency band; the length of the LTA window is nine times the length of the STA window.
  • STA/LTA functions are nonlinearly stacked within 0.5 degree distance bins
  • for plotting, stacks within each distance bin are normalized, amplitudes are raised to ^(1/9) and then the median is removed
  • for the majority of images shown here, a 120sec wide normalizing window has been applied to enhance faint phases

These panels highlight the processing workflow. The top panel shows a raw trace from station MD.BDTF.BHZ recording the 1999/04/20 18:55:33 M6.5 event in the Kermadec Islands. The next two panels show the results of applying a narrow band-pass filter around 8 sec and it’s STA/LTA function. For all traces used in this project the various QC metrics were applied to the filtered STA/LTA functions. After culling the dataset within each passband, the STA/LTA functions are truncated at a value of 25 in order to avoid saturation by especially high signal-to-noise arrivals. Square root stacking is then applied to all traces within 0.5 degree wide distance bins. Finally, a 120-sec amplitude normalizing window is applied.

The final stack traces for all distance bins are combined into a single image. The trace from 80 deg in the previous figure is overlain. Stacks within each distance bin are normalized for plotting resulting in darker or lighter vertical streaks due to varying signal-to-noise ratios of the peak arrival within each distance bin, e.g. the caustic in PKP around 140 degrees. Between 160-180 degrees, the signal-to-noise ratio of the peak arrival is relatively low, resulting in a grainy appearance in the stack image.

Supplements: QCed data set, all stack files & vespagrams

The QCed data set

Text files give lists of seismograms (source-receiver pairs) passing the QC criteria in this project. Lists are given for 1) all stations for each event as well as 2) all stations & event times within each distance bin. More details are given in the README file.
README.QCedTraceLists.txt (38 MB ~6000 text files)

Stack files & vespagram files

For those interested in the values or files for plotting themselves, files of all stacks are available. Included in the files are values for the STA/LTA N-th root stacks used in the figures as well as raw amplitude stacks and envelope stacks. Similarly, files are available for vespagrams made using USArray.

Global stacks supplements page


Figure showing effect of QCing

The figure shows that the QCing criteria applied enhances the quality of the stacks and more is not always better. The lower quality stacks in the right panel include 222,000 QC rejected seismograms, more than double the amount in the stacks of the middle panel which use only traces passing QCing.

Stacks in different frequency bands

Stacks for all twelve frequency bands are shown in this animation. At shorter periods, late core phases, e.g. those arriving later than 60 minutes, have high signal to noise ratios. At longer periods P & S wave multiples start to dominate the images.

Download full size animated gif of the stacks in different frequency bands

Citations and DOIs

To cite the IRIS DMC Data Products effort:

  • Hutko, A. R., M. Bahavar, C. Trabant, R. T. Weekly, M. Van Fossen, T. Ahern (2017), Data Products at the IRIS‐DMC: Growth and Usage, Seismological Research Letters, 88, no. 3, http://doi:10.1785/0220160190.

To cite the IRIS DMC Global Stacks data product:


Astiz, L., P. Earle and P. Shearer, Global stacking of broadband seismograms, Seis. Res. Lett., 67, 8-18, 1996..


  • Alexander Hutko
  • This project was made possible using waveform data from over 100 FDSN networks and earthquake catalog information from the ISC and USGS.


Global stacks released