Validation of Regional Seismic Travel Time (RSTT) Predictions and - - PowerPoint PPT Presentation

▶
validation of regional seismic travel time rstt
SMART_READER_LITE
LIVE PREVIEW

Validation of Regional Seismic Travel Time (RSTT) Predictions and - - PowerPoint PPT Presentation

Nov 04, 2022 532 likes •833 views

Ground-Based Nuclear Explosion Monitoring R&D Validation of Regional Seismic Travel Time (RSTT) Predictions and Use in Event Location Stephen C. Myers 1 Michael L. Begnaud 2 Sanford Ballard 3 Abelardo L. Ramirez 1 , Michael E. Pasyanos 1 , W.

slide-1
SLIDE 1

Ground-Based Nuclear Explosion Monitoring R&D

LLNL-PRES-638493

Validation of Regional Seismic Travel Time (RSTT) Predictions and Use in Event Location

Abelardo L. Ramirez1, Michael E. Pasyanos1, W. Scott Phillips2 + RSTT Team (LLNL, LANL, SNL)

1Lawrence Livermore National Laboratory, 2Los Alamos National Laboratory, 3Sandia National Laboratories

Stephen C. Myers1 Michael L. Begnaud2 Sanford Ballard3

Sponsored by: US Department of State

This work was performed under the auspices of the U.S. Department of Energy by Lawrence Livermore National Security, LLC, Lawrence Livermore National Laboratory under contract DE-AC52-07NA27344. The views expressed here do not necessarily reflect the views of the United States Government, the United States Department of Energy, or the Lawrence Livermore National Laboratory

slide-2
SLIDE 2

Ground-Based Nuclear Explosion Monitoring R&D

LLNL-PRES-638493

Introduction

  • Goal of the RSTT project:

– Improve travel time prediction accuracy with respect to global base models for the regional phases Pn, Pg, Sn and Lg using a real-time calculation.

  • Real-time computation must be achieved with commonly available

computers, as exist at any CTBT National Data Center (NDC).

  • Real-time computation enables application to flexible networks and data

centers who may wish to combine IMS and other networks to achieve

  • ptimal location accuracy.

Pn Pg Sn Lg

slide-3
SLIDE 3

Ground-Based Nuclear Explosion Monitoring R&D

LLNL-PRES-638493

IASPEI Definitions for Regional Phases

  • Computes travel times for Pn, Pg, Sn, Lg phases commonly

used for routine event location.

Figures from Stochak et al., 2003 IASPEI definitions for phases computed by RSTT RSTT follows the convention used by explosion monitoring data centers, whereby Pg is not distinguished from Pb and Lg is not distinguished from Sb . Pn Pg Sn Lg

IASPEI is the International Association of Seismology and Physics of the Earth’s Interior

slide-4
SLIDE 4

Ground-Based Nuclear Explosion Monitoring R&D

LLNL-PRES-638493

Regional Phases are a Significant Percentage of REB Data for Small Events

Teleseismic Regional Local Teleseismic Regional Local Time-defining phases in REB 2009-2010. Distances 0° to 95° Total Percentage 83% teleseismic 16% Regional 0.6% Local

  • S. Myers, LLNL

REB is the Reviewed Event Bulletin

slide-5
SLIDE 5

Ground-Based Nuclear Explosion Monitoring R&D

LLNL-PRES-638493

Median Error Degradation (km)

51

Degradation in location error when 1 Pn is added (RSTT Validation Data, ak135 (1D) model)

Average number of phases used in REB (2008 to present) to locate events of varying magnitude mb=3.5 mb=4.0 mb=4.5 mb=5.

Number of P phases

  • Location of low-magnitude

events may rely on 1 or more regional stations.

  • When using a simple 1D

velocity model, adding even 1 regional (Pn) arrival time tends to degrade location accuracy.

  • Need a regional phase velocity

model that will not degrade locations when regional phases are combined with teleseismic.

Global data used to evaluate location accuracy

Regional Data Can Degrade Location Accuracy When Using a Simple Radially Symmetric Model

  • S. Myers, LLNL
slide-6
SLIDE 6

Ground-Based Nuclear Explosion Monitoring R&D

LLNL-PRES-638493

Regional Phase Travel Times Vary More Than Travel Times at Other Distances

Regional Teleseismic Local

95

Average Pn Travel Time Error

Area of distance ranges

Modified from Rodi and Myers (2013)

slide-7
SLIDE 7

Ground-Based Nuclear Explosion Monitoring R&D

LLNL-PRES-638493

Lateral Variation in Wave Speed is the Cause of Poor Regional Travel Time Prediction

Pn P

  • S. Myers, LLNL
slide-8
SLIDE 8

Ground-Based Nuclear Explosion Monitoring R&D

LLNL-PRES-638493

RSTT Accounts for Lateral Variability in Seismic Velocity to Improve Prediction Accuracy

From Previous Presentation (T1-06)

  • M. Begnaud, LANL
slide-9
SLIDE 9

Ground-Based Nuclear Explosion Monitoring R&D

LLNL-PRES-638493

Travel Time Validation, Pn

Procedure 1) Randomly sample 10% of the tomography data set for validation 2) Repeat tomographic inversion, withholding the validation data 3) Difference observed and predicted travel times

  • Account for unknown origin time by

removing the median residual for each event 355K summary paths 11K stations 25K events 77K paths 6K stations 19K events

  • S. Myers, LLNL
slide-10
SLIDE 10

Ground-Based Nuclear Explosion Monitoring R&D

LLNL-PRES-638493

RSTT Reduces Residual Trends (Biases), Pn

Curves include measurement errors

AK135 RSTT

slide-11
SLIDE 11

Ground-Based Nuclear Explosion Monitoring R&D

LLNL-PRES-638493

Travel Time Validation, Pg

Procedure 1) Randomly sample 10% of the tomography data set for validation 2) Repeat tomographic inversion, withholding the validation data 3) Difference observed and predicted travel times

  • Account for unknown origin time by

removing the median residual for each event 32K summary paths 6K stations 8K events 7.8K paths 2.3K stations 4.5K events

slide-12
SLIDE 12

Ground-Based Nuclear Explosion Monitoring R&D

LLNL-PRES-638493

Travel Time Validation, Pg

AK135 RSTT

slide-13
SLIDE 13

Ground-Based Nuclear Explosion Monitoring R&D

LLNL-PRES-638493

Travel Time Validation, Sn

Procedure 1) Randomly sample 10% of the tomography data set for validation 2) Repeat tomographic inversion, withholding the validation data 3) Difference observed and predicted travel times

  • Account for unknown origin time by

removing the median residual for each event 84K summary paths 6K stations 17K events 15K paths 2.8K stations 7.9K events

slide-14
SLIDE 14

Ground-Based Nuclear Explosion Monitoring R&D

LLNL-PRES-638493

Travel Time Validation, Sn

AK135 RSTT

slide-15
SLIDE 15

Ground-Based Nuclear Explosion Monitoring R&D

LLNL-PRES-638493

Travel Time Validation, Lg

Procedure 1) Randomly sample 10% of the tomography data set for validation 2) Repeat tomographic inversion, withholding the validation data 3) Difference observed and predicted travel times

  • Account for unknown origin time by

removing the median residual for each event 16.0K summary paths 3.0K stations 5.0K events 2.8K paths 0.8K stations 1.9K events

slide-16
SLIDE 16

Ground-Based Nuclear Explosion Monitoring R&D

LLNL-PRES-638493

Travel Time Validation, Lg

AK135 RSTT

slide-17
SLIDE 17

Ground-Based Nuclear Explosion Monitoring R&D

LLNL-PRES-638493

Location Validation Events

Validation events used to test location accuracy

These events not used in tomography to create RSTT validation model

GT0 (1) GT1(14) GT2(2) GT5(69)

slide-18
SLIDE 18

Ground-Based Nuclear Explosion Monitoring R&D

LLNL-PRES-638493

Network Sampling

Event and network 10 random samples of network stations 14 of 23 stations chosen Chile Argentina Bolivia

  • S. Myers, LLNL
  • S. Myers, LLNL
slide-19
SLIDE 19

Ground-Based Nuclear Explosion Monitoring R&D

LLNL-PRES-638493

Examples of Network Realizations

Network coverage (azimuthal gap) can differ significantly between realizations AzGap=91° AzGap=33° AzGap=117°

  • S. Myers, LLNL
slide-20
SLIDE 20

Ground-Based Nuclear Explosion Monitoring R&D

LLNL-PRES-638493

Realizations of Locations

Chile Argentina ak135 (51 km median err) Reference (GT5) RSTT (10 km median err) 14 Pn arrivals

  • S. Myers, LLNL
slide-21
SLIDE 21

Ground-Based Nuclear Explosion Monitoring R&D

LLNL-PRES-638493

Average Location Accuracy

ak135 RSTT Using RSTT Pn travel times results in more accuracy event locations on average. Improvement in epicenter accuracy is consistent regardless of the number of phases used. The plateau in location accuracy for ~20 phases and more is thought to be pick- error related. Results based on averaging 10 network realizations for each number of picks for 58 global test events Over 10,000 locations

slide-22
SLIDE 22

Ground-Based Nuclear Explosion Monitoring R&D

LLNL-PRES-638493

IMS Locations Typically Utilize Both Regional and Teleseismic Data

Dashed lines: Location accuracy using varying numbers of Pn data as in previous slide. Solid line: Location accuracy using ak135 travel times and a mixture

  • f regional (3 Pn) and varying

numbers of teleseismic P data. ak135 teleseismic + regional ak135 regional RSTT regional

Median Error Degradation (km)

mb=3.5 mb=4.0 mb=4.5 mb=5.0

Number of P phases

P Pn Number of Picks

Recall number of P for average REB event

slide-23
SLIDE 23

Ground-Based Nuclear Explosion Monitoring R&D

LLNL-PRES-638493

RSTT Enables the Use of a Mixed Regional and Teleseismic Data Set

ak135 teleseismic+ regional ak135 regional RSTT regional Using RSTT for the 3 regional phases calculations Improves location accuracy for small events for which regional data make up a larger percentage of the data set. Allows NDCs to add their regional data to further improve location accuracy.

mb=3.5, 27% regional mb=4.5, 10% mb=5.0, 5% mb=4.0, 20 %

ak135 teleseismic+ RSTT regional Number of Picks P Pn

slide-24
SLIDE 24

Ground-Based Nuclear Explosion Monitoring R&D

LLNL-PRES-638493

Efforts to Improve the RSTT Global Model

slide-25
SLIDE 25

Ground-Based Nuclear Explosion Monitoring R&D

LLNL-PRES-638493

Pn Travel Time Residuals Geographically

North America RSTT -> 0.64 (s) Ak135-> 0.95 (s) Eurasia RSTT -> 0.71 (s) Ak135-> 0.84 (s) Latin America RSTT -> 0.86 (s) Ak135-> 0.98(s) Median Residuals

  • S. Myers, LLNL
slide-26
SLIDE 26

Ground-Based Nuclear Explosion Monitoring R&D

LLNL-PRES-638493

Location Error For Latin America

Latin America 7 test events Globe-Latin_America 51 test events For this test data set Location are less accurate than the global average RSTT improves location accuracy, but still considerable room for improvement

  • Tomography and validation data are few

Additional data have been contributed More is needed!

slide-27
SLIDE 27

Ground-Based Nuclear Explosion Monitoring R&D

LLNL-PRES-638493

A CTBTO Workshop was Held in the Vanuatu The AusREM model was converted to RSTT format. Australia contribute a number of GT events

  • S. Myers, LLNL
slide-28
SLIDE 28

Ground-Based Nuclear Explosion Monitoring R&D

LLNL-PRES-638493

Conclusions

  • Regional seismic travel times are improved by accounting

for 3-D crustal structure and laterally variable wave velocity in the upper mantle.

  • Improved travel time prediction accuracy results in more

accurate event locations.

  • New RSTT tomography uses a global data set

Chile Argentina ak135 (51 km median err) Reference (GT5) RSTT (10 km median err) 14 Pn arrivals

The RSTT model can be improved by augmenting publically available data with contributions from regional

  • networks. Efforts underway in:
  • South America
  • Australia and South Pacific