1 Wavefield Calibration Using Regional Network Data R. B. Herrmann Saint Louis University
2 Questions What is a good Earth model for regional studies? –Location –Source Parameters –Structure What can be done with very good regional data in a simple region?
3 Answer It is difficult to obtain the Earth model
4 Presentation Tools to obtain earth models –Surface wave dispersion inversion –Receiver function inversion –Waveform inversion Adequacy of data set –Tests
5 Joint Inversion
6 Rayleigh Wave Sensitivity
7 RFTN Partials RFTN
8 Postulated Advantages of Joint Inversion Receiver function depends upon travel time and fine detail of structure related to conversions Surface wave is smoothly affected by velocities So Advantages of one overcome deficiencies of the other
9 Purpose of models Assist location by correctly predicting first arrivals Properly characterize dynamic wavefield to obtain quantitative estimates of source mechanism and strength
10 Receiver Function Sensitivity to Structure Perturb simple crust/mantle model Examine effect of gradient Design model to have same vertical travel time
11 Red = sharp / Blue = strong gradient
12 Gravity Anomaly Imagine sampling different structures within a region What would be seen in Bouguer anomaly
mGal variation among models
14 Love Rayleigh
15 Surface waves Subtle differences in dispersion for fundamental mode in second period range For surface waves to really contribute structure information, need dispersion for a fine grid of periods Need short periods to focus on upper crust
16 Receiver Functions Slides for different filter parameter - alpha =1.0 corresponds to a lowpass corner of about 1/3.14 Hz Focus on effect of Moho transition on nature of P-wave receiver function
17
18
19
20
21 Comments 1st peak controlled by shallow structure Gradient indicated by absence of signal for high alpha, character by low alpha Sharp moho is indicated by distinct bounce arrivals for all alpha, especially higher Simultaneous fit to several alpha robust
22 KOREA Can events be located within 5 km? Can Earth model be defined that can –fit receiver function data –model regional waveforms or Is the joint inversion model of any value?
23
24
25 21 NOV 2001
26 21 NOV 2001 WVFGRD
27 bp c np 2
28 bp c np 2
29 bp c np 2
30 Excellent fit, even at high frequencies but used Central US model - not model derived specifically or even valid for Korea so perform joint inversion surface - wave/receiver function for a Korea model
31
32 Receiver functions Two filter parameters Stacked RFTN’s Arranged by similarity in shape Last 3 are from island stations Similarity in RFTN’s - > similarity in structure
33 Phase velocity (Herrmann, 2001) - Group velocity (Stevens, 1999) SNU (InvSNU)
34 Stacked receiver function for alpha 1 and stations Dispersion Stevens (1999) and a few phase velocity points joint96 same script (iterations controls ) for all AK135 (modified)
35 SNU (InvSNU)
36 All models Average of layer velocities, not slowness
37 P-wave 1st Arrival Surface focus
38 Good fits to both data sets Subtle differences in P-wave first arrival times Models do not fit 21 NOV 01 earthquake data - surface wave arrival too late - S-P times wrong Augment with CUS dispersion
39 SNU (nInvSNU)
40 Strong constraint on upper crust
41 Fits using CUS dispersion
42 P-wave 1st arrival surface focus
43 SNU Model comparison Red - first Blue - CUS
44 Discussion RFTN need very good dispersion Model requires independent test - waveform modeling?
45 New Dispersion Data Harvard group velocities Colorado group velocities Phase velocities from Korea –Treat BB network as array –Optionally apply match filter –Apply McMechan and Yedlin p-tau implemented as sacpom96
46 01 Jan 2001 Alaska Event - phase match output used from 10 stations
47 Blue - Colorado Green - Harvard Orange - Stevens Red - Korea phase velocity
48 Starting Model AK135 - depths > 50 km Upper 50 km is a halfspace with velocity of z=50 km Invert new dispersion Use stacked RFTN’s Use same script
49
50
51 Test Compare data, CUS and Korea model Add modes to save time - so no P Bandpass Hz
54 Focal Mechanisms 21 NOV waveform Mw= NOV waveform Mw= DEC surface-wave radiation pattern Mw=4.0 –does not require precise location –does not require precise Earth model –does not require broadband signal –works with high noise
55 Rayleigh Wave
56 Love Wave
57 SEO spectra
58 New Mechanisms
59 KOREA Good data sets but little local activity Waveforms exhibit distinct P and sP –together these are good depth indicators Data are available to improve dispersion estimates Some events may have good azimuthal distribution and stations to 200 km
60 Future work Continued interaction with SNU and KMRI on –Ground motion scaling with distance –Location –Source mechanisms –Seismic Hazard maps
61 Organize and use waveform data set Get more data, especially events within 20 degrees away in NW and SE sector from Korea –Why? Get shorter period dispersion Digital data available from KMA website (if one reads Korean)
62 Other Data Teleseismic P-wave residuals –Effect of subduction zone –Usefulness for crustal wave propagation Detailed travel times - Pn, Pg, SmS 1 Hz dispersion for upper 1 km Calibration shots - not likely High frequency phase attenuation
63 Questions What degree of calibration is required? What can be done with a very good regional data set?