![Hier klicken, um den Treffer aus der Auswahl zu entfernen](images/unchecked.gif) |
Titel |
ASDF - A Modern Data Format for Seismology |
VerfasserIn |
Lion Krischer, James Smith, Wenjie Lei, Matthieu Lefebvre, Youyi Ruan, Elliot Sales de Andrade, Norbert Podhorszki, Ebru Bozdag, Jeroen Tromp |
Konferenz |
EGU General Assembly 2017
|
Medientyp |
Artikel
|
Sprache |
en
|
Digitales Dokument |
PDF |
Erschienen |
In: GRA - Volume 19 (2017) |
Datensatznummer |
250151367
|
Publikation (Nr.) |
EGU/EGU2017-15938.pdf |
|
|
|
Zusammenfassung |
Seismology as a science is driven by observing and understanding data and it is thus vital to
make this as easy and accessible as possible. The growing volume of freely available data
coupled with ever expanding computational power enables scientists to take on new and
bigger problems. This evolution is to some part hindered as existing data formats have not
been designed with it in mind.
We present ASDF (http://seismic-data.org), the Adaptable Seismic Data Format, a novel,
modern, and especially practical data format for all branches of seismology with particular
focus on how it is incorporated into seismic full waveform inversion workflows. The format
aims to solve five key issues:
Efficiency: Fast I/O operations especially in high performance computing
environments, especially limiting the total number of files.
Data organization: Different types of data are needed for a variety of tasks.
This results in ad hoc data organization and formats that are hard to maintain,
integrate, reproduce, and exchange.
Data exchange: We want to exchange complex and complete data sets.
Reproducibility: Oftentimes just not existing but crucial to advance our science.
Mining, visualization, and understanding of data: As data volumes grow,
more complex, new techniques to query and visualize large datasets are needed.
ASDF tackles these by defining a structure on top of HDF5 reusing as many existing
standards (QuakeML, StationXML, PROV) as possible. An essential trait of ASDF is that it
empowers the construction of completely self-describing data sets including waveform,
station, and event data together with non-waveform data and a provenance description of
everything. This for example for the first time enables the proper archival and exchange of
processed or synthetic waveforms.
To aid community adoption we developed mature tools in Python as well as in C and
Fortran. Additionally we provide a formal definition of the format, a validation tool, and
integration into widely used tools like ObsPy (http://obspy.org), SPECFEM GLOBE
(https://geodynamics.org/cig/software/specfem3d_globe/), and Salvus (http://salvus.io). |
|
|
|
|
|