Introduction¶
DISCLAIMER: This documentation is still being updated. It is based on the documentation of the previous version of LASIF and some parts might still only apply to the previous version right now. We are working on updating the documentation and once we feel confident that it is ready we will remove this disclaimer.
LASIF 2.0 (Larg-scale Seismic Inversion Framework) is a data-driven workflow tool to perform full waveform inversions. It is opinionated and strict, meaning that it enforces a certain data and directory structure. The advantage is that it only requires a very minimal amount of configuration and maintenance. It attempts to gather all necessary information from the data itself so there is no need to keep index or content files.
All parts of LASIF 2.0 also work completely on their own; see the class and function documentation at the end of this document. Furthermore LASIF 2.0 offers a project based inversion workflow management system which is introduced in the following tutorial.
LASIF 2.0 works with the notion of so called inversion projects. A project is defined as a series of iterations working on the same physical domain. Where possible and useful, LASIF 2.0 will use hdf5 files and SQL databases to store information. The reasoning behind this is twofold. The hdf5 files are of the ASDF (Adaptable Seismic Data Format) which make it possible to reduce the amount of files generally needed to store information and now all the information for each event and each iteration can be stored in one file which helps with file management. The SQL databases work very fast and are optimal for querying certain parameters.
LASIF 2.0 is data-driven, meaning that it attempts to gather all necessary information from the available data. The big advantage of this approach is that the users can use any tool they want to access and work with the data as long as they adhere to the directory structure imposed by LASIF. At the start of every LASIF operation, the tool checks what data is available and uses it. To achieve reasonable performance it employs a transparent caching scheme able to quickly register any changes the user makes to the data. Also important to keep in mind is that LASIF 2.0 will never delete any data.
The aim of the LASIF 2.0 project is to facilitate the execution of mid-to large-scale full seismic waveform inversion using adjoint techniques.
Supported Data Formats¶
This is a short list of supported data formats and other software.
Waveform Data: All file formats supported by ObsPy.
Synthetics: ASDF formats
Event Metadata: QuakeML 1.2 incorporated into ASDF
Station Metadata: dataless SEED, RESP and FDSN StationXML. We strongly recommend to use StationXML! also incorporated into ASDF
All these dataforms are put together and structured using pyasdf
Mesh Files: HDF5 meshes like Salvus makes. LASIF also works mesh free for simple domains
Waveform Solvers: Any, extra support for Salvus
Further Notes¶
QuakeML files¶
LASIF 2.0 is designed to work with valid QuakeML 1.2 event files. Please assure that the files you use actually are just that. If possible try to only use QuakeML files with one origin and one focal mechanism, otherwise LASIF will choose the preferred origin and/or focal mechanism (or the first of each, if no preferred one is specified). The origin time specified in the QuakeML file will be the reference time for each event! Times specified in SAC files will be ignored.
This also means that the raw data files have to have the correct time information.
Right now LASIF 2.0 works best if it is used to collect the events and the relevant data. A tutorial on how to use your own data to produce a dataset that LASIF 2.0 can work with will be on the website later.