Potential BUG in `NLLoc` inversion with 3D velocity model

[JUNZHU-SEIS]

Hi, below is an output file (*.hyp) of NLLoc inversion with 3D velocity model.
It is strange to me that the Dist parameter is incorrect and so large (1e+30) in this file, but the resulting location seems correct. This is a synthetic test actually. The resulting location (153.125, -310.93, 11.3417) in this fle is close to the synthetic location (154.65, -313.23, 9.0). This synthetic location is close to the boundary of the 3D model.

I have some questions:

• Why the Dist parameter is not TRUE, and why it is so large?
• Does this error affect the inversion result? Should I trust it?
• Why are the observations from the closer stations excluded in inversion, e.g., the cloest station (SC_JLI)?

Any reply is much appreciated.

PS: In this synthetic test, the 3D model used by Time2EQ is identical to the model used by NLLoc. This is not the only anomalous case. There exist some normal cases as well. I can provide more code snippet if necessary. I'm sorry I have to remove the exact locations of the seismic stations as they are sensitive data.

[JUNZHU-SEIS]

NLLOC "nlloc/3Dloc/nlloc.19000209.011002.grid0" "LOCATED" "Location completed."
PUBLIC_ID 59026
SIGNATURE "NonLinLoc - Jun Zhu obs:nlloc/obs/eew.obs NLLoc:v7.1.04(05Sep2024) run:02Sep2025 11h51m38"
COMMENT "Test 3D velocity heterogenity"
GRID 561 646 156 -280 -320 -5 1 1 1 PROB_DENSITY
SEARCH OCTREE nInitial 400 nEvaluated 20016 smallestNodeSide 0.437500/0.503906/0.302734 oct_tree_integral 5.440299e+02 scatter_volume 5.379697e+02
HYPOCENTER x 153.125 y -310.93 z 11.3477 OT 0.437175 ix -1 iy -1 iz -1 MAXIMUM_LIKELIHOOD
GEOGRAPHIC OT 1900 02 09 01 10 0.437175 Lat 28.185953 Long 104.557869 Depth 11.347656
QUALITY Pmax 3.3039e+127 MFmin 28.224 MFmax 43.9023 RMS 0.147252 Nphs 44 Gap 281.002 Dist 1e+30 Mamp -9.90 0 Mdur -9.90 0
VPVSRATIO VpVsRatio 1.73 Npair 96 Diff 8.9668
STATISTICS ExpectX 152.768522 Y -310.232829 Z 12.583823 CovXX 3.28029 XY -1.47845 XZ -2.05071 YY 4.9822 YZ 7.28654 ZZ 87.3483 EllAz1 56.6926 Dip1 -1.54198 Len1 2.89541 Az2 326.557 Dip2 -5.01563 Len2 4.2812 Len3 1.762934e+01
STAT_GEOG ExpectLat 28.192279 Long 104.554341 Depth 12.583823
TRANSFORM LAMBERT RefEllipsoid WGS-84 LatOrig 31.000000 LongOrig 103.000000 FirstStdParal 30.000000 SecondStdParal 32.000000 RotCW 0.000000
QML_OriginQuality assocPhCt 196 usedPhCt 44 assocStaCt -1 usedStaCt 22 depthPhCt -1 stdErr 0.147252 azGap 281.002 secAzGap 290.554 gtLevel - minDist 1e+30 maxDist -1 medDist 1e+30
QML_OriginUncertainty horUnc -1 minHorUnc 2.36186 maxHorUnc 3.66406 azMaxHorUnc 149.962
QML_ConfidenceEllipsoid semiMajorAxisLength 17.6293 semiMinorAxisLength 2.89541 semiIntermediateAxisLength 4.2812 majorAxisPlunge -84.7515 majorAxisAzimuth 163.733 majorAxisRotation 162.618
FOCALMECH Hyp 28.185953 104.557869 11.347656 Mech 0 0 0 mf 0 nObs -1
PHASE ID Ins Cmp On Pha FM Date HrMn Sec Err ErrMag Coda Amp Per PriorWt > TTpred Res Weight StaLoc(X Y Z) SDist SAzim RAz RDip RQual Tcorr TTerr
SC_JLI ? ? ? P ? 19000209 0110 2.2140 GAU 5.00e-02 -1.00e+00 -1.00e+00 -1.00e+00 1.0000 > 2.4641 -0.6872 0.0000 xx yy zz 4.2293 az 243.0 158.9 7 0.0000 0.0000
SC_JLI ? ? ? S ? 19000209 0110 3.8301 GAU 5.00e-02 -1.00e+00 -1.00e+00 -1.00e+00 1.0000 > 4.2628 -0.8699 0.0000 xx yy zz 4.2293 az 243.0 158.9 7 0.0000 0.0000
...
END_PHASE
END_NLLOC

[alomax]

Hi @JUNZHU-SEIS

Why the Dist parameter is not TRUE, and why it is so large?

I do not recall seeing this problem before. I do see in the two phase lines you provided that Weight=0 (this is posterior weight derived from optimal hypocenter location). The Dist parameter is calculated only using phases with Weight>0, so apparently all your phases have Weight=0 and no Dist is determined. This would explain the Dist=1e+30, but then why do all your phases have Weight=0?

Weight=0 could be due to:

1. zero prior (not you case)
2. very large ErrMag (not your case)
3. something to do with zero TTerr? I think this is OK, but maybe...
4. very, very large residual
5. Use of LOCEXCLUDE or LOCINCLUDE statements
6. ...???

Does this error affect the inversion result? Should I trust it?

It seems that maybe the result is reasonable, but I think it best to explain and understand what is happening.

Why are the observations from the closer stations excluded in inversion, e.g., the cloest station (SC_JLI)?

See above, what could give Weight=0.

Best regards,
Anthony

[JUNZHU-SEIS]

Thanks for your explanation.

• Actually NOT all the phases are zero weight. Take the station YN_CS001 as an example, it has a weight of 1.1066 and 0.8761 on P- and S- phase, respectively. Its epicentral distance is 49.7661 km.

YN_CS001     ?    ?    ? P      ? 19000209 0110   10.5006 GAU  5.00e-02 -1.00e+00 -1.00e+00 -1.00e+00    1.0000 >   10.1115   -0.0481    1.1066  xx yy zz 49.7661 az 341.6 100.5  9     0.0000    0.5056
YN_CS001     ?    ?    ? S      ? 19000209 0110   18.1660 GAU  5.00e-02 -1.00e+00 -1.00e+00 -1.00e+00    1.0000 >   17.4930    0.2359    0.8761  xx yy zz 49.7661 az 341.6 100.5  9     0.0000    0.8746

• LOCEXCLUDE or LOCINCLUDE seems NOT stated in the control file of NonLinLoc (please see the attched file)
  control_3D.zip

[alomax]

Hi @JUNZHU-SEIS

I can take a better look if you send me your control file and some event (not sum) output *.hyp for one or a few events where there are zero and non-zero weights? You can send privately to anthony@alomax.net and hide any sensitive information.

Thanks,
Anthony

[JUNZHU-SEIS]

Sure. I have just sent you an email via zhujun2316@mail.ustc.edu.cn.