Questions about location outputs

[ev-anderson]

Hi Anthony,

I have produced an initial location catalog but had a few questions about some of the outputs.

My first question pertains to the estimated output VpVs ratio for events. I defined a VpVs ratio of 1.8 under LOCMETH, but noticed the estimated output VpVs ratios are very low, in many cases less than 1. What would be the cause of such low ratios?

Next, I was curious about the RMS_Max parameter under LOCPHSTAT. I set an RMS_Max of 0.5, yet some of the outputted locations exceed this threshold in the summary catalog. Is there something I overlooked in LOCPHSTAT's ability to curtail the resultant catalog?

Attached is my control file to give you an idea of the settings I am using.

Thank you!

cayman_month_3d.txt

[alomax]

Hi @ev-anderson

estimated output VpVs ratio for events

Do you mean VpVs ratios from the lines in the output hypocenter files like:
VPVSRATIO VpVsRatio 1.64365 Npair 14 Diff 10.59 ?

These VpVs estimates are calculated for each event from the P and S pick times following the algorithm in Chapter 5 of HYPOELLIPSE, Lahr, 1989: https://pubs.er.usgs.gov/publication/ofr72224
So they are calculated from the data and are independent of the LOCMETH VpVs program parameter.

However, I think that a VpVs less then 1 is very unusual - S velocity is greater than P velocity!
What are the typical values of VpVs you get? The VpVs estimate can be poor if Npair is small (e.g. < 4-8???) or Diff is small (this is the difference between the largest and smallest Tp-Ts in the observations - the Lahr algorithm is fitting a line to the Ts-Tp pairs, so a bigger Diff gives more precision in the fit).

the RMS_Max parameter under LOCPHSTAT

This does not filter the output locations, but instead only filters which locations are used to accumulate statistics for output in *.sum.*.stat and *.sum.*.stat_totcorr, which show mean residuals and other statistics and can be used to re-run NLLoc with station static corrections.

[ev-anderson]

Hi @alomax,

Thanks again for the quick responses. The mean for my estimated Vp/Vs is ~0.92. I've attached a histogram of my Vp/Vs values, as well as some plots of Vp/Vs vs Npairs and Vp/Vs vs Diff.

It does seem my Diff values are generally low, suggesting that is the culprit for the overall lower Vp/Vs values...

[alomax]

It does seem my Diff values are generally low, suggesting that is the culprit for the overall lower Vp/Vs values...

Yes, these do seem like very low values. Numerically the Vp/Vs algorithm should still work as long as there are stations with P ans S picks at a proportionally large range of distances. Are most stations at similar distance from the events?
The Vp/Vs is reasonable rise for many events with Diff > ~2-3.
But I do not understand negative values for Vp/Vs, let alone Vp/Vs < 1. Maybe some noise or bias coming from somewhere?

Anthony

[ev-anderson]

The network is dense consisting of ~40 instruments all spaced ~3-5 km apart from each other around and within the target seismicity area, so it should be sampling an adequate range of distances for arrivals.

The deployment is in a deep marine setting, so noise related to this environment is a possibility.

Could my velocity model be affecting these low Diff values? I'm using a 1-D velocity model obtained through previous active source experiments draped beneath the seafloor bathymetry. This setting has fast material near the surface, with the model reflecting this with high Vp near the surface. Additionally the model includes a velocity inversion ~ 5km below the seafloor. Could these be contributing to the low Diff values?