Use nonlinloc with 1d model

[JuanJose9611]

Hi,

Is it possible to use nonlinloc with a 1d velocity model instead of a 3d model? If so, are there any parameters in the input control file that need to be changed to work with 1d instead of 3d models?

Juan

[alomax]

Hi Juan,

Yes, the basic running of NLL uses a 1D model. See https://github.com/alomax/NonLinLoc/blob/dev/nlloc_sample/run/nlloc_sample.in in the sample/tutorial https://github.com/alomax/NonLinLoc/tree/dev/nlloc_sample for control file specifications and running with a 1D model.

Best regards,
Anthony

[JuanJose9611]

To run Nlloc and locate the events is it necessary to first run the program Time2EQ?

Juan

[alomax]

Hello Juan,

No, not at all. Time2EQ is a (not so easy to use!) separate tool for generating synthetic picks.

Anthony

[JuanJose9611]

I created my own control file to run a sample with my own .obs file.

The programs returns a message that says X amount of events where read and it finishes without warnings or errors however there are 0 located events.

Would it be possible for you to take a look at my file and see if there is something wrong with it?

Juan

[alomax]

Hello Juan,

A common reason for no events locating is that the LOCGRID is not contained within the travel-time grid (determined by VGGRID, if travel-times generated by Grid2Time). You can check for this case (and see if there are other problems) by setting CONTROL 3 ....

If this does not help, you are welcome to send your control file, but I may also need the observations and any other files that running NLLoc requires. If you do not want to post this information, you can send to alomaxfree.fr

Best regards,

Anthony

[JuanJose9611]

Hi,

After changing to CONTROL 3 I noticed that I set the offset values incorrectly (should have been 0). After properly selecting the origin now the program does locate the events, however, the events appear to be on a L shape near the pre-locations. I will attach the .obs and .in files. If you have some time please take a look. My guess is that i am not properly setting the grids, both for the model and for location.

test_location_run.zip

Juan

[alomax]

Hi Juan,

I took a quick look. The main problem was that the LOCGRID was not centered on the seismicity so the events were locating at the corner of the LOCGRID.

I made a few other changes to the control parameters, see my comments ### in
test_runfile_ALomax.in.txt

Many of these changes are optional.

Most events have hypocenter at the top or bottom of the LOCGRID, this is probably due to a lack of depth constraint since the stations are along a line and all to the NE of the seismicity:

One event inside the station network has reasonable depth and moderate depth constraint:

Anthony

[JuanJose9611]

How did you determine the values to properly center the grid?

Juan

[alomax]

As I do in general, I center the projection near the middle of the seismicity:
TRANS AZIMUTHAL_EQUIDIST WGS-84 30.79 -116.39 0

And then make the LOCGRID
LOCGRID 301 301 51 -75.0 -75.0 0.0 0.5 0.5 0.5
larger then the extent of the seismicity 300x, 300y * 0.5 km, and put the grid origin (SW corner) half of the grid width in -x and half the grid height in -y -75.0 -75.0:

Anthony

[JuanJose9611]

Thank you for answering my questions.
Some of the located earthquakes show stripes, if i wanted better resolution should i modify the LOCGRID or the VGGRID?

Juan

[alomax]

Hi Juan,
This type of gridding usually is due to configuration/search settings or limitations of the available data or stations.
Please see the following comments:
#52 (comment)
#17 (comment)
There may be others, this issue arises often!

Best regards,
Anthony

[JuanJose9611]

Dear Anthony,

I am tryint to use seismicty viewer howeve when i select my sum.hyp file it seems that there is an error in the projection. The stations and hypocenters are placed on on a straight line.

[alomax]

Hi Juan,
It looks like all the events have the same or very similar epicenter?
Otherwise, can you send me an example sum*hyp file that gives this pattern?
Anthony

[JuanJose9611]

Here are the input file, obs and hyp files.
test_run.tar.gz

[alomax]

Here is what I get with a few changes to the input test_runfile.in.zip :

The stations are off to one side of the seismicity and the model is thick, constant velocity layers - it looks like the hypocenters are not constrained in depth so are getting stuck to the top or bottom of the search volume. Most of the event PDF's look like this one:

Plotting your locations it looks like the hypocenters are getting stuck on or near layer boundaries:

Also try the latest SeismicityViewer here: http://alomax.free.fr/seismicity/beta/

Anthony

[JuanJose9611]

Do you think there is perhaps something that can be done to fix the near layer boundaries issue? or is it impossible due to the location of the stations? perhaps a 0d model?

The local seismic network uses a traditional method for automatic detection (STA/LTA), each automatic detection is then confirmed and each pick is fixed by an analyst and the locations are done with SEISAN (same model i am using).

RESNOM LOCATIONS (https://resnom.cicese.mx/sitio/):

PyOcto (https://github.com/yetinam/pyocto) pre-location with ML picks (Earthquake Transformer):

Location with NonLinLoc:

I wonder if it would be possible to manage better locations or the small amount of stations makes it impossible.

[alomax]

With stations only to one side (gap < 180deg), you would need a very, very good P and S model, and S readings. But even then, there could be an origin-time / distance / depth trade-off problem.
The best would be to find additional stations to the west, but I suppose not possible - all ocean?

Perhaps fixing the depth would at least give epicenters with useful relative positions. This is done with a very small depth range in LOCGRID, something like:

#  small depth range around 5 km depth
LOCSEARCH  OCT 26 26 1 0.0001 50000 5000 0 1
LOCGRID  130 130 2  -60.0 -65.0 4.5  1 1 1   PROB_DENSITY  SAVE

Best regards,
Anthony