amaggi / waveloc

Migration code for seismic event location

waveloc

Migration code for seismic event location. Unstable development version. The latest stable version is available on the Downloads page. The documentation is available here: http://amaggi.github.com/waveloc.

waveloc's Issues

re-introduce a channel-file for selecting data for analysis and migration

For complex networks, data selection becomes a tricky business. Introduce some sort of file containing

NET STA CHA

triplets.

• introduce file
• check for existence (in options.py): if not present, accept sta_list``net_list etc. with all combinations thereof
• use it in the processing phase

Test effect of S-wave kurtosis maxima

Migration tests require processing tests to be run first (so that kurt files exist)

This behavior is not ideal. Tests that must be run in sequence do not seem to be a good idea. Try to fix this by providing pro-forma kurtosis data for migration tests.

Create an examples directory

Add a gaussian filter to filters to deal with post-kurtogram filtering

Add figures to documentation introduction

Clean up synthetic migration - use sta_list

• use <sta_list> to decide on which stations to use for synthetic tests instead of looking for real data

Small and full res migration tests are broken

Looks like the files used for comparison are bad. Needs to be fixed with new files (and propagated back to downloadable test files).

Create a .gitattributes file to deal with line endings

See this for example https://help.github.com/articles/dealing-with-line-endings

Match NLL code for uncertainty ellipsoids

Add tests for all OP_waveforms methods

profile sw_kurtosis2 and optimize for memory use

Code is fully documented (api)

The api is fully documented

Add h5py requirement

• on online documentation
• in setup.py file

Update OP_waveforms to most recent obspy

base_path : allow both environment variable and option setting

The option setting should override the environment variable setting. If neither are set, should exit gracefully.

use maximum travel-time to set duration of signal for small migrations

Add simple parallelization to processing

Simple parallelization will make waveloc run faster. The simplest task to start with is the waveform processing, as each waveform is processed independently of the others.

when using gaussian convolution, consider postponing decimation to after convolution

add real migration example to waveloc_examples

Base it on the full migration test.

Test 3D prob density volume at origin time

Rewrite to use unstructured grids

Using rectangular grids for search regions involves doing a lot of calculations that are actually unnecessary. The code would be speeded up significantly if an un-ordered sequence of search points were used instead. This upgrade would also make it possible to craft a seismicity zone that more closely matches the observed pattern.

rt-waveloc development already uses such an unstructured grid.

In stack volume plots, have depth increasing downwards

Should be simple. Matplotlib probably has a flag for this.

Optimize sliding window kurtosis

Set up a 2D array where each row is a window, and send to ss.kurtosos . To set up sliding window, try following code:

def window(seq, n=2):
    "Returns a sliding window (of width n) over data from the iterable"
    "   s -> (s0,s1,...s[n-1]), (s1,s2,...,sn), ...                   "
    it = iter(seq)
    result = tuple(islice(it, n))
    if len(result) == n:
        yield result    
    for elem in it:
        result = result[1:] + (elem,)
        yield result

Output of resolution test should go into the fig directory.

For consistency, all figures should go in the out/name/fig directory. Requires updating example and its documentation.

Deal with deprecation warning in kurtogram.py

Systematically sort after glob readings

glob.glob reads files sorted in alphabetical order on a mac, but not on a linux system. To avoid problems, systematically check that all glob.glob calls are followed by sort calls.

Add uncertainty bounds to plots

Shade in uncertainty bounds on locations on output plots

Deal with encoding issue in test_processing.py

Code should be flake8 compatible

Modify synthetic example to use real coordinates

Update doc for waveloc example

Add output figures for test

Tutorial is completed

Migration is broken :(

Migration is broken - gives very stupid results. Spend some time fixing it !!

Rewrite tutorial to correspond to current version

Deconvolve network response from stack?

Is there any way to deconvolve the network response to clean up a stack and maybe isolate a source of error?

Trigger for reloc is broken

Need unittests for correlation and clustering

@nlanget

When you've finished writing your unit-tests, we'll put them in the test directory.

Re-design layout for grid plot figures

See

https://github.com/amaggi/latex-waveloc-method/issues/5

Add simple parallelization to migration

This may be a little tricky, but conceptually at least each point can be migrated on its own. Don't attempt this until the unstructured grids are in place.

Add magnitude calculation (simply using obspy)

• Was requested in reviews of previous version of the paper.
• Requires a DATALESS volume

refactor options.py

verification code in options.py is dreadful and poised to break - refactor it !!

Doc strings for correlation.py incomplete

Probability density location

Requires coding from scratch.

Fix location plotting

Plotting is broken. Fix it!

Doc-strings for clustering.py incomplete

I've creates proto-doc-strings for this module, but I don't know enough about it to write correct documentation.

VOIE1.mat missing from test_data directory

Improve memory usage by going back to staking integer arrays.

This facility was removed during a big bug search in dev branch, and has not yet been put back. Consider using int32 arrays instead of int16.

Look into metropolis algorithm to sample the 4D volume around local maxima

Recursive kurtosis broken for variances != 1

Nadège reports that the recursive kurtosis gives values of -3 for gaussian data whose variance is not equal to 1.0.

[ ] Reproduce the error in a unit test
[ ] Look up better ways of estimating means and variances recursively
[ ] Test
[ ] Rerun