OT_WT exceeds EDT_OT_WT_FLOOR

[fontiela]

I’m currently running NLLoc to relocate seismic activity on a small volcanic island (30 Km X 20Km) with the following settings:

VGGRID  2 121 60  0.0 0.0 -1.0  1.0 1.0 1.0

LOCHYPOUT SAVE_NLLOC_ALL NLL_FORMAT_VER_2 SAVE_HYPOINV_SUM

LOCGRID  41 41 21  -20.0 -20.0 -1.0  1.0 1.0 1.0   PROB_DENSITY  SAVE 

LOCSEARCH  OCT 6 6 3 0.0001 50000 5000 

LOCMETH EDT_OT_WT 20.0 6 -1 4 1.76 -1 -1.0 1

LOCGAU 0.2 0.0

LOCGAU2 0.05 0.1 2.0

LOCPHASEID  P   P Pb Pn Pg

LOCPHASEID  S   S Sb Sn Sg

LOCQUAL2ERR 0.05 0.11 0.225 0.5 99999.9

LOCANGLES ANGLES_YES 5

LOCPHSTAT 3 -1 -1 3 3 70 70 -1 

LOCELEVCORR 1 3.8 2.15

While running the program, I received the following output:

Reading next set of observations (Files open: Tot:3 Buf:0 Hdr:0  Alloc: 0) ...

... 10 observations read, 10 will be used for location (./loc_2024/local.20240115.014431).
LOCGAU param CorrLen is zero, will not be used: 0.000000
Locating... (Files open: Tot:3 Buf:0 Hdr:0  Alloc: 20  3DMem: used:0/avail:0/load:0) ...

Applying Octtree search within Grid 0:
INFO: EDT_otime_weight activated, OT_WT exceeds EDT_OT_WT_FLOOR.
OctTree num samples = 50000 / 50000
Octree oct_node_value_max= 5.315795e+01 oct_tree_integral= 1.526792e+00
OCTREE nInitial 108 nEvaluated 50020 smallestNodeSide 0.026042/0.026042/0.026042 oct_tree_integral 1.526792e+00
INFO: EDT_otime_weight: ot_ml_std 0.097717

ellipsoid_volume = 1.387858e+02

Finished location: ./loc_2024/local.20240115.014431.grid0.loc.hyp

I’m particularly curious about this line: INFO: EDT_otime_weight activated, OT_WT exceeds EDT_OT_WT_FLOOR. Does this indicate any potential effect on my results, and could it impact the accuracy of the location? Any insights would be greatly appreciated.

[alomax]

Hello @fontiela,

This is only an information message and can be ignored.

At each tested hypocenter point, LOCMETH EDT_OT_WT multiplies the EDT location likelihood by a weight that depends on the agreement between the origin time estimates over different EDT pick pairs. In certain rare cases with poor station coverage or outlier pick readings, EDT can find an optimal solution where the estimated origin times are very different for different station pairs. This is not a reasonable solution, so the OT_WT is included. The OT_WT also increases the likelihood in general for areas where the origin time estimates agree best over observation pairs. The information message just indicates that this additional weighting is activated.

With regards to your settings:

Since memory is now vast, I generally define fairly thin layers with 2D velocity models. So I might use for your setup: VGGRID 2 1201 601 0.0 0.0 -1.0 0.1 0.1 0.1 This helps guarantee that the resulting travel-time grids will better capture curved travel-time fronts. This is also important if your velocity model has interfaces that do not fall on, e.g., integer km depths.

Given your LOCGRID would also use something like: LOCSEARCH OCT 30 30 15 0.0001 50000 5000 so the initial octree grid is finer, has ~cubic cells (num x,y,z of intial octree grid is proportional to num x,y,z of LOCGRID), and the number of initial octree cells is ~1/4 - 1/3 the total number of samples taken (50000).

Anthony

[fontiela]

HI @alomax

Thank you for your detailed explanation regarding the warning message, as well as your helpful advice for improving my results. I’ll implement your suggestions and will update you once I have some results to share.