Dataset for Global Reference Seismological Data Sets: Multimode Surface Wave Dispersion

• How fast do surface waves travel globally after any earthquake?
• Do we get the same information from various measurement techniques?
• Which features in the Earth are robust and can be resolved by a reference model?

Reference data with uncertainties are useful for improving existing measurement techniques, validating models of interior structure, calculating teleseismic data corrections in local or multiscale investigations and developing a 3-D reference Earth model. This study was done in collaboration with 18 scientists from 16 institutions in 7 countries who actively participated in the REM3D project. The project assimilated, archived, reconciled and modeled big (>200 million measurements) and diverse surface-wave datasets for global subsurface structure.

The reference data set summarizes measurements of dispersion of fundamental-mode surface waves and up to six overtone branches from 44,871 earthquakes recorded on 12,222 globally distributed seismographic stations. Dispersion curves are specified at a set of reference periods between 25 and 250 s to determine propagation-phase anomalies with respect to a reference Earth model. Empirically determined observational uncertainties (1 sigma) for each wave type, branch number and period can be found in Table 3. 

Summary:

[I] Reconciled large and diverse catalogues of Love-wave (49.65 million) and Rayleigh-wave dispersion (177.66 million) from eight groups worldwide.
[II] Retrieved missing station and earthquake metadata in several legacy compilations and codified scalable formats to facilitate reproducibility, easy storage and fast I/O on HPC systems.
[III] Systematic discrepancies between raw phase anomalies can be attributed to discrepant theoretical approximations, reference Earth models and processing schemes.
[IV] Phase-velocity variations yielded by the inversion of the summary data set are highly correlated (R ≥ 0.8) with those from the quality-controlled contributing data sets, especially for long-wavelength variations (up to degree ∼25) in fundamental-mode dispersion (50–100 s).
[IV] Only 2ζ azimuthal variations in phase velocity of fundamental-mode Rayleigh waves are required; maps of 2ζ azimuthal variations are highly consistent between catalogues ( R = 0.6–0.8).

Feedback/Questions? Please contact Raj Moulik (rajmoulik.com) at moulik@caa.columbia.edu 

Reference:

Please cite the following work if you use this data or software.

• Moulik, P. et al., (2022) Global reference seismological data sets: multimode surface wave dispersion. Geophys J Int 228, 1808–1849, doi: 10.1093/gji/ggab418. pdf

You can also cite the dataset and software from this Zenodo page (Optional).

• Moulik, P. (2022) Dataset for Global Reference Seismological Data Sets: Multimode Surface Wave Dispersion. In Geophys. J. Int. (v1.0, Vol. 228, pp. 1808–1849). Zenodo. doi: 10.5281/zenodo.8371228

HDF5 Container Format

• Reference Love waves (Download All Periods and Branches as Summary.SW.Love.data.h5)
• Reference Rayleigh waves (Download All Periods and Branches as Summary.SW.Rayl.data.h5)

Summary (reference) data between pairs of 2562 evenly-spaced knot points with an average spacing of 4.33◦. These files store the data in the RSDF HDF5 container format. These can be read using standard HDF5 modules (e.g. h5py) or using AVNI. For example, to read the reference data for fundamental mode R1 waves at 100s into a Pandas Dataframe containing data (df['data']) and a dictionary with the metadata (df['metadata']), and thereafter write contents to an ASCII text file, enter the following in Python:

• from avni.data.SW import readSWhdf5,writeSWascii
• df=readSWhdf5(query='0/100.0/R1/REM3D',hdffile='Summary.SW.Rayl.data.h5',datatype='summary')
• writeSWascii(df,'test.txt')

ASCII (text) Format

These files contain the same reference data as the HDF5 files above but in gzipped ASCII files. The files are named according to the overtone branch, wave type and period as Summary.$overtone.$wave.$period.REM3D.gz Table A1 from the paper describes the various columns in the surface-wave RSDF ASCII format files.

• Love waves (Download All Periods and Branches as Summary.SW.Love.data.zip)
  • Fundamental Modes
    • Minor Arc Arrivals (L1) at 25s, 27s, 30s, 32s, 35s, 40s, 45s, 50s, 60s, 75s, 100s, 125s, 150s, 175s, 200s, and 250s
    • Major Arc Arrivals (L2) at 150s, 175s, 200s, and 250s
    • Higher Obit Arrivals - L3 at 150s, 175s, 200s, and 250s; L4 at 150s, 175s, 200s, and 250s; L5 at at 150s, 175s, 200s, and 250s.
  • I^st Overtone at 40s, 45s, 50s, 60s, 75s, 100s, 125s, 150s, 175s,  and 200s
  • II^nd Overtone at 40s, 45s, 50s, 60s, 75s, 100s, 125s, and 150s
  • III^rd Overtone at 40s, 45s, 50s, 60s, and 75s
  • IV^th Overtone at 40s, 45s, 50s, and 60s
  • V^th Overtone at 40s, 45s, and 50s
• Rayleigh waves (Download All Periods and Branches as Summary.SW.Rayl.data.zip)
  • Fundamental Modes
    • Minor Arc Arrivals (R1) at 25s, 27s, 30s, 32s, 35s, 40s, 45s, 50s, 60s, 75s, 100s, 125s, 150s, 175s, 200s, and 250s
    • Major Arc Arrivals (R2) at 150s, 175s, 200s, and 250s
    • Higher Obit Arrivals - R3 at 150s, 175s, 200s, and 250s; R4 at 150s, 175s, 200s, and 250s; R5 at 150s, 175s, 200s, and 250s.
  • I^st Overtone at 40s, 45s, 50s, 60s, 75s, 100s, 125s, 150s, 175s,  and 200s
  • II^nd Overtone at 40s, 45s, 50s, 60s, 75s, 100s, 125s, and 150s
  • III^rd Overtone at 40s, 45s, 50s, 60s, and 75s
  • IV^th Overtone at 40s, 45s, 50s, and 60s
  • V^th Overtone at 40s, 45s, and 50s
  • VI^th Overtone at 40s, 45s, and 50s

Other Data Products:

• ReferenceSW_Moulik2022_Figures(.zip or .pdf) - contains all figures from the paper in .png format
• Scatter_Plots.zip - contains scatter plots similar to Figure 5 in the paper, which compares measurements between two sets of techniques. The files with the suffix *raw.png are comparisons for original raw datasets, while those with the suffix *.clean.png are comparisons after the entire workflow is completed to create the clean datasets (e.g. Figure 13, bottom row).
• Half_cycle.zip and Cycle_skips.zip - contains list of source-station paths where discrepancies were found between pairs of techniques. Half (±0.9–1.1 · π ) or full-cycle discrepancies (±0.9–1.1 · 2π ) identified in Section 4.5 are used during outlier analysis (Section 5.3) to create the clean summary dataset. Half- and full-cycle discrepancies identified in these files indicate potential polarity reversals and cycle skips respectively. Note that all of these discrepancies have not been checked for specific causes manually. 
• vflip-table.REM3D - an ASCII file containing station names and start/end times where polarity reversal issues have been confirmed through manual analysis. This is in contrast to the automated half-cycle discrepancies identified in Half_cycle.zip above.
• M1442 and B2562 - Files containing the knot locations of evenly-spaced points on the surface. B2562 has an average knot spacing of 4.33◦ and is used as the underlying grid for the homogenization process to get summary data (Section 5.1). In order to obtain 2-D variations in local phase slowness or velocity, we use 1442 splines with an average knot spacing of 5.77◦ (Section 6.1)
• Cleanhomo.SW.Love.data.h5 and Cleanhomo.SW.Rayl.data.h5 - Clean homogenized data for each research group obtained at the end of our workflow (Figure 2). The ASCII files containing the same data are provided in Cleanhomo.SW.Love.data.zip and Cleanhomo.SW.Rayl.data.zip. The summary dataset listed earlier represents the reconciled measurements, and should be preferred over those from individual groups in most applications.
• Inversion_Example.zip - Contains an example of a 2D slowness map inversion with 2ζ azimuthal variations using the reference summary dataset at 100s for fundamental-mode minor-arc Rayleigh waves (R1). Also provided are plots for anistropic variation (Anisotropy_Plots), spline coeffients of 1442 evenly-spaced spherical splines (Spline_Coefficients), and corresponding values at every 1X1 degree pixel in extended pixel format (Maps_epix). The aim of this study is to provide dispersion measurements of surface-wave arrivals, not to provide detailed 2D phase velocity/slowness models. 

Note about Data Format

The underlying philosophy and format of data files are discussed in the reference seismic data format (RSDF) project. Table A1 from the GJI paper describes the various columns in the surface-wave RSDF format files above.