MSNoise 1.5

Release date: 28 April 2017

Release type: major

Release notes:

Introduction

About 1 year after the last major release (MSNoise 1.4) we are proud to announce the new MSNoise 1.5. It is a major release, with a massive amount of work since the last one: in GitHub numbers , it’s over 120 commits and over 2500 lines of code and documentation changed or added!

MSNoise 1.5 introduces a series of new features :

  • We have started to move core math functions to ObsPy, currently the only one ready is linear_regression, a function I wrote to remove the dependency to statsmodels, required to move mwcs to ObsPy later.

  • The preprocessing routine has been isolated, rewritten and optimized. It is now a standalone script, callable by plugins. It returns a Stream object with all the data needed for the analysis.

  • This change in preprocessing was done mostly to allow cross-component, auto- correlation and cross-correlation, with or without rotation, to be done with the same code. CC, SC and AC are now supported in MSNoise with proper whitening (possible to disable spectral whitening for specific cases).

  • This documentation is now available in PDF too (easier for offline usage) and it also includes a new tutorial for setting up the MySQL server and Workbench.

  • Last but not least: MSNoise is “tested” automatically on Linux (thanks to TravisCI) & Windows (thanks to Appveyor), for Python versions 2.7 and 3.5. With MSNoise 1.5 we also added the MacOSX tests on TravisCI. With these tests, we can guarantee MSNoise works on different platforms and Anaconda (or miniconda) python versions.

This version has benefited from outputs/ideas/pull requests/questions from several users/friends (listed alphabetically):

  • Raphael De Plaen

  • Clare Donaldson

  • Robert Green

  • Aurelien Mordret

  • Lukas Preiswerk

  • The participants to the NERC MSNoise Liverpool Workshop in January 2017

  • all others (don’t be mad :-) )

Thanks to all for using MSNoise, and please, let us know why/how you use it (and please cite it!)!

To date, we found/are aware of 25 publications using MSNoise ! That’s the best validation of our project ever ! See the full list on the MSNoise website.

Thomas, Corentin and others


PS: if you use MSNoise for your research and prepare publications, please consider citing it:

Lecocq, T., C. Caudron, et F. Brenguier (2014), MSNoise, a Python Package for Monitoring Seismic Velocity Changes Using Ambient Seismic Noise, Seismological Research Letters, 85(3), 715‑726, doi:10.1785/0220130073.

Requirements

If you have any package older than those mentioned here, some things might not work as expected.

  • Pandas >= 0.18.0

  • obspy >= 1.1

No longer needed:

  • scikits-samplerate

  • statsmodels

Web-based Admin Interface Changes

  • Bugfix: Stations can now be added manually

  • Bugfix: Bulk operations on Job based on their jobtype is now possible

  • Bugfix: Set debug-mode to False to increase performance and remove the risk of Werkzeug failing.

  • Feature: Calling admin/data_availability.png should serve the image directly

  • Feature: It’s now possible to “Brand” MSNoise: The name and the logo of the page can be overriden by setting an environment variable with a name and the HTML tag of the logo image:

    set msnoise_brand="ROB|<img src='http://www.seismologie.be/img/oma/ROB-logo.svg' width=200 height=200>"
    

    and then starting msnoise admin:

    ../_images/branding1.png

Configuration Parameters

  • REMOVED: decimation_factor: it’s now computed automatically from the stream’s sampling rate. If the decimation factor is non-integer, the routine exits and the user is advised to use the Lanczos resampler.

  • REMOVED: ZZ, RR, TT, etc in favor of components_to_compute.

  • ADDED: components_to_compute: it’s accepting a comma-separated list of components.

  • ADDED: whitening method for disabling spectral whitening in specific cases.

Populate Station Table and Scan Archive

  • Because of platform-dependent issues, we dropped the need for the “find” command to be present on each machine. The “cron” scan_archive procedure is now a simple python routine. It might take a little longer, but much safer. This also means the crondays parameter can be set to a floating point value, “-1.0” meaning “files modified in the last 1 day”.

  • The custom archive structure can now directly be input in the config. This is still quite preliminary and needs at least one slash in the parameter value to be identified as such by the code.

  • A change in the database structure (added an index on the “path”+”file”+”net”+”sta”+”comp”) allows running msnoise scan_archive a little more efficiently than before.

Warning

The change in the database (adding an index) requires that you msnoise upgrade_db every project!

Expert/Lazy mode

  • Added an Expert/Lazy mode to scan files directly by passing the path of their containing folder to scan_archive --path (see here).

  • Added an Expert/Lazy mode to input network/stations directly from the data scanned using the new populate --fromDA procedure ((see here)).

  • The scan_archive is now capable of handling multiplexed files and inputs one line in DataAvailability per unique “trace id” in files.

Practically, the three changes above allow users to:

  1. Download a single or few mseed file with N stations, for example from webdc or using FDSN webservices in ObsPy or else.

  2. Run msnoise install

  3. Run msnoise scan_archive --path /path/to/the/big/file/

  4. Run msnoise populate --fromDA

  5. Run msnoise new_jobs --init

  6. Run msnoise admin, define (pre-)processing and filter parameters

  7. Run msnoise compute_cc

New Jobs

  • The msnoise new_jobs routine has been rewritten to correct a few bugs linked to start/end dates (missed days if a file of day 2 begins on day 1 and ends on day 3, jobs were only created for day 1 and 3. This is corrected.

  • A change in the database structure (added an index on the “day”+”pair”+”jobtype”) allows running msnoise new_jobs much faster than before. It took 5m40s for a test with 105381 jobs, while it would have taken hours before. For the first run, it is still faster (20s for the example above) to use msnoise new_jobs --init as this doesn’t check for existing jobs.

Warning

The change in the database (adding an index) requires that you msnoise upgrade_db every project!

Preprocessing and Cross-Correlation

Preprocessing

  • A complete rewrite of the preprocessing function to avoid padding and merging with zeros. The preprocessing function is now separated from the compute_cc code and can be called by external plugins. It returns a Stream object that can easily be filtered or slided.

  • The instrument response removal has been accelerated by doing it after decimation, on fewer data points (Thanks to Robert Green). The response removal is done without the evalresp stuff from ObsPy, it’s faster but potentially a little less safe.

  • The default decimation tool is now Lanczos (builtin in ObsPy) and scikits.samplerate is no longer needed.

Cross-Correlation

  • It is now possible to do the Cross-Correlation (classic “CC”), the Auto- Correlation (“AC”) or the Cross-Components within the same station (“SC”). To achieve this, we removed the ZZ, ZT, etc parameters from the configuration and replaced it with components_to_compute which takes a list: e.g. ZZ,ZE,ZN,EZ,EE,EN,NZ,NE,NN for the full non-rotated tensor between stations. If autocorr is set to “Y”, then the cross-components (SC) of each station will also be computed (of course, ZE and EZ are identical).

  • The cross-correlation is done on sliding windows on the available data. If one trace contains a gap, the window is skipped. This corrects previous errors linked with gaps synchronised in time that lead to perfect sinc autocorr functions. The windows should have a duration of at least 2 times the maxlag configuration value.

  • The whitening procedure can be skipped by setting the whitening configuration to None. The two other whitening modes are “[A]ll except for auto-correlation” or “Only if [C]omponents are different”. This allows skipping the whitening when, for example, computing ZZ components for very close by stations (much closer than the wavelength sampled), leading to spatial autocorrelation issues.

Command Line changes

  • msnoise config --sync will try to parse the dataless files for existing stations in the Station table and, if found, will input the coordinates.

  • msnoise scan_archive --path will only scan the path independently of its structure. It will only read files, not “walk” in subfolders.

  • msnoise populate --fromDA will populate the station table from the existing data_availability table.

  • msnoise p is a lazy alias to msnoise plugin

  • add a default delay of 1 second (customizable) to start parallel threads (using -t)

Note, all commands are documented: Help on the msnoise commands.

API Changes

  • New make_same_length API method to return common data for two traces, this is necessary to compute the rotation of the horizontal traces to Radial/ Transverse.

  • New clean_scipy_cache API method to reduce the memory imprint caused by scipy’s automatic caching of FFT arrays.

See MSNoise API.

Plugin support

  • Plugins can now declare their own templates using the msnoise.plugins.templates entry point.

  • Plugins can override the “components_to_compute” config bit and the MSNoise API get_components_to_compute(session, plugin=None) works like get_config.

See Extending MSNoise with Plugins.

Other Bugfixes

  • Removed the call to scipy.stats.nanmean and replaced by numpy.nanmean

  • Better error message in compute_cc when the content of the slice is only zeros or smaller than rms_threshold

  • Checked all “integer” - “float” warnings from numpy/scipy

  • crondays were hardcoded to -2, now taking the crondays value from the DB

  • Py3 error when msnoise scan_archive in cron mode

Plot Updates

  • msnoise plot ccftime now accepts -e (–envelope) and will plot the envelope of the ccfs.

  • msnoise plot ccftime, msnoise plot interferogram and msnoise plot distance now accept -r (–refilter) to refilter the CCFs before plotting. The argument must be a column-separated string (e.g. -r 4:8 for filtering between 4.0 and 8.0 Hz).

  • msnoise plot distance accepts a new --virtual-source NET.STA parameter to only plot the pairs including this station.

  • Most plots have better titles (filter details, etc).

  • The dv/v plot now allows averaging over components by passing them as comma- separated values.

See Plotting.

Performance and Code improvements

Improvements in terms of performances have also been done for MSNoise 1.5:

  • Added fftpack optimized nfft (scipy’s next_fast_len). This could lead to some small differences in the final result of the MWCS procedure, because of the number of points used for smoothing the (cross-)spectra.

  • Replaced binarization (sign) and windsorizing (clip) by standard numpy functions operating directly inplace on the arrays, avoiding unecessary copies.

  • The preprocessing only reads files that should contain the right component.

  • The stretching code has been improved (thanks to Clare Donaldson)

Documentation

  • New tutorial for installing and configuring MySQL and MySQL workbench.

  • The Workflow is separated in steps (See Workflow))

  • The Documentation is now available in PDF format (PDF).

Upgrading an existing project to MSNoise 1.5

Some users will want to keep their current project without recomputing everything. This requires adding a few configuration parameters to the database

Running the following command will take care of the upgrade from 1.4 to 1.5:

msnoise upgrade_db

Warning

Upgradind the database will not remove deprecated configuration bits, so users should remember to define, for example, the components_to_compute parameter if anything else than ZZ was set before.

A final note about development pace and choices

  • MSNoise team is

    • 1 developper (Thomas)

    • 1 dedicated debugger (Corentin)

    • less than 10 really active users, providing feedback and/or lines of codes (Esteban, Raphaël, Aurélien, Carmelo, Clare, Rob …)

  • All software engineering ideas are coming from too infrequent beerstormings between Thomas & others

  • The web-interface and the plugin support were developed during Thomas’ holidays

If you need help, please ask your questions on the mailing list. Don’t be afraid to ask. If you have ideas, please share them. If you develop codes to supplement MSNoise, please share them, even if very small, even if you don’t master gitHub. If you have complaints, post them too, but remember that the package you are using has been coded by 1 person, and that it’s not his full time job. So MSNoise is provided “as-is”, carefully written and tested, but there will be bugs, issues, incompatibility with certain python installations, OS or module versions. If you want or need developments made, contact Thomas via email directly. If these developments are within the focus of the developers’ research, then a collaboration, i.e. resulting in a co-authored peer reviewed publication, can be an option. Otherwise, you can contract us for paid-developments.