Hypocenter of an explosion

[Cthuulhaa]

Hello Anthony,

I am testing NLL for the automatic location of local events. I tried to locate some quarry blasts, which depth should be close to 0 km. Using the Crust1.0 crustal model, most of the quarry blasts are located successfully, but for one of them in particular I get 15 km as depth, which I think is not acceptable and I try to find the reason.

This event was detected on 3 stations, producing P- and S-picks that were manually/visually inspected. The stations have epicentral distances between 14 - 21 km.

I followed the example (sample) that locates the Alaskan events to automatically produce the files necessary to run NLL. These are the files:

For the station coordinates:

GTSRCE GRA1   LATLON   49.690776  11.220436  0.0  0.4995
GTSRCE GRA3   LATLON   49.761076  11.31735  0.0  0.455
GTSRCE GRB4   LATLON   49.467841  11.55948  0.0  0.507

Then the observed travel times:

GRA3   ?   HHZ   ?   P   ?   20220120   0822   17.255   GAU   9.00e-02   -1.00e+00   -1.00e+00   -1.00e+00   1
GRA3   ?   HHZ   ?   S   ?   20220120   0822   20.175   GAU   6.00e-02   -1.00e+00   -1.00e+00   -1.00e+00   1
GRB4   ?   HHZ   ?   P   ?   20220120   0822   18.009   GAU   1.00e-02   -1.00e+00   -1.00e+00   -1.00e+00   1
GRB4   ?   HHZ   ?   S   ?   20220120   0822   21.269   GAU   3.00e-02   -1.00e+00   -1.00e+00   -1.00e+00   1
GRA1   ?   HHZ   ?   P   ?   20220120   0822   16.999   GAU   3.00e-02   -1.00e+00   -1.00e+00   -1.00e+00   1
GRA1   ?   HHZ   ?   S   ?   20220120   0822   19.659   GAU   8.00e-02   -1.00e+00   -1.00e+00   -1.00e+00   1

And finally the control file (here the command Grid2Time is executed once for travel time calculation for P-waves and once for S-waves, by creating a similar file which I don't show here):

CONTROL   0   54321
TRANS   LAMBERT   Clarke-1880   49.62   11.38   48.62   50.620.0
VGOUT   2022-01-20T08:22:17.000000Z_tiebenn_loc/model_layer
VGTYPE   P
VGTYPE   S
VGGRID   2   401   102   0.0   0.0   -5.0   1   1   1   SLOW_LEN
LAYER   -0.46   2.5   0.0   1.07   0.0   2.11   0.0
LAYER   0.34   6.1   0.0   3.55   0.0   2.74   0.0
LAYER   10.05   6.3   0.0   3.65   0.0   2.78   0.0
LAYER   19.76   6.6   0.0   3.6   0.0   2.86   0.0
LAYER   29.77   8.18   0.0   4.54   0.0   3.37   0.0
LAYER   35.0   8.04   0.0   4.48   0.0   3.38   0.0
LAYER   77.5   8.045   0.0   4.49   0.0   3.38   0.0
LAYER   120.0   8.05   0.0   4.5   0.0   3.36   0.0
GTFILES   2022-01-20T08:22:17.000000Z/model_layer   2022-01-20T08:22:17.000000Z/time_layer   P
GTMODE   GRID2D   ANGLES_YES
INCLUDE   2022-01-20T08:22:17.000000Z/station_coordinates.txt
GT_PLFD   1.0e-3   0
EQFILES   2022-01-20T08:22:17.000000Z/time_layer   2022-01-20T08:22:17.000000Z/obs_synth.obs
EQMODE   SRCE_TO_STA
EQSRCE   2022-1-20_08:22:17   LATLON   49.62   11.38   0.0   0.0
EQSTA   GRA1   P  GAU 0.00   GAU 0.05
EQSTA   GRA1   S  GAU 0.00   GAU 0.05
EQSTA   GRA3   P  GAU 0.00   GAU 0.05
EQSTA   GRA3   S  GAU 0.00   GAU 0.05
EQSTA   GRB4   P  GAU 0.00   GAU 0.05
EQSTA   GRB4   S  GAU 0.00   GAU 0.05
EQVPVS    1.68
EQQUAL2ERR   0.05   0.1   0.5   1.0   99999.9
LOCSIG  NonLinLoc
LOCCOM   2022-1-20 08:22:17
LOCFILES   2022-01-20T08:22:17.000000Z/obs_ttimes.obs   NLLOC_OBS   2022-01-20T08:22:17.000000Z/time_layer   2022-01-20T08:22:17.000000Z/loc_eqdatetime
LOCHYPOUT   SAVE_NLLOC_ALL   NLL_FORMAT_VER_2
LOCSEARCH   OCT   10   10   4   0.01   20000   5000   0   1
LOCGRID   201   201   102   -100.0   -100.0   -5.0   1   1   1   PROB_DENSITY   SAVE
LOCMETH   EDT_OT_WT   9999.0   6   -1   3   -1   6   -1.0   1
LOCGAU   0.2   0.0
LOCGAU2   0.02   0.05   2.0
LOCQUAL2ERR   0.1   0.5   1.0   2.0   99999.9
LOCANGLES   ANGLES_YES   5
LOCPHSTAT   9999.0   -1   9999.0   1.0   1.0   9999.9   -9999.9   9999.9

Do you see anything anomalous in any of the automatically produced files? I must mention that I have not changed the parameters used in the sample/example, so I ignore if there is a more optimal parameter selection.

Best, C.

[alomax]

Hello Cthuulhaa

I tried your location. With P and S readings all looks fine with the configuration and the solution. But the depth is indeed well constrained at ~15km.

I then tried with P readings only - the epicenter is almost the same but the depth is shallower (~6km) and the pdf is smeared vertically, indicating no depth constraint.

I then tried removing each of the S readings in turn. In each case, the maximum likelihood hypocenter is shallow, but the pdf still shows a strong secondary solution at ~15km depth.

I then tried increasing the S pick uncertainty by X10. The The maximum likelihood hypocenter is shallow, and the pdf gives a smearing in depth with a secondary maximum solution at ~15km.

So I am not sure what is happening, but some possible ideas:

1. The thin, very low velocity (2.5km/s P) shallow layer over a constant (6.1km/s P) velocity crust is correct, but tends to allow an ambiguous, double solution (shallow near surface and deeper), and small changes or error in the (S?) picks make the maximum likelihood solution jump between the two depths.
2. The thin, very low velocity (2.5km/s P) shallow layer over a constant (6.1km/s P) velocity crust is incorrect - are there alternative, smoother models? Is Vp/Vs correct?
3. The S picks are surface waves, not S. Or other problem with S picks (they may be delayed).
4. ???

Interesting problem...

Best regards, Anthony

PS - I use SeismicityViewer for most of the above analysis, especially for visualization of the pdf.