The  San  Andreas  Fault  in  finedetail:  seismology  and  rock  physics

Eun-­‐Sun  Chong  (ASU)John  D.  West  (ASU)Arizona  State  UniversityEarthScope  Seminar  Class

Figures  and  informaGon  referencedfrom  various  sources.

April  20,  2010

Discussion

1

1.  MigraGng  tremors  illuminate  complex  deformaGonbeneath  the  seismogenic  San  Andreas  FaultDavid  R.  Shelly    Nature  (2010)  Vol.  463  (7281)  pp.  648  -­‐  653

Shelly  (2010)

Data:  ConGnuous  seismic  datafrom  mid-­‐2001  to  2008.

Result:  Evidence  supportsdeep  fault  shear  failure  as  thebasic  mechanism  of  tremor.Physical  condiGon  thatcontrols  the  spaGaldistribuGon  of  such  tremorsremain  in  quesGon.  Fluidpressure  in  the  fault  zone  atdepth  is  suggested.

2

Shelly  (2010)3

*Lack  of  legend  makes  this  figure  difficult  to  comprehend.

This  figure  is  said  to  represent  tremormigraGon.a)Extended  tremor  migraGon  episode,approximately  25km  along  fault  strikefor  over  90  minutes.  17  km  representsaverage  propagaGon  rate  in  a  givenepisode.b)Three  HRSN  borehole  staGons,waveforms  and  event  family  matchesduring  the  same  period.c)Zoomed  view  of  the  highest-­‐amplitude  and  fastest-­‐migraGngporGon  of  the  sequence.

A  lot  seems  to  be  said  throughdisplay  of  this  figure  but  veryuncertain  on  how  suchinformaGon  was  derived.

Event  Families?

4

QuesGons:

1.  What  does  it  mean  when  cumulaGve  eventsare  “normalized”?

2.  What  do  the  numbers  on  that  parGcularaxis  represent?

Shelly  (2010)

3.  What  is  the  purpose  of  this  figure?

5

Shelly  (2010)

Figure  4.  Recurrence  paeerns  for21  event  families.

a)  Mid  2001  to  2008.  Black  dashlines  indicate  San  Simeon  andParkfield  earthquakes.  Gray  dashlines  indicate  when  HRSNstaGons  experienced  operaGonalchanges.

b)  Basically  a  zoomed  in  versionof  (a)  with  different  Gme  scale.

Methods

6

1. SelecGon  of  waveform  templates.a. SelecGon  of  relaGvely  strong  and  isolated  events

as  templates.b. Used  the  template  that  appeared  to  have  the

highest  signal-­‐to-­‐noise  raGo.

2. Template  event  locaGon.a) An  absolute  locaGon  for  each  template  event

based  upon  P  and  S  wave  arrival  Gmes  of  a  lowfrequency  earthquake  with  a  good  signal-­‐to-­‐noise  raGo  at  permanent  and  temporarystaGons.

b) A  grid  search  for  locaGon  of  hypocenters  with  aVp/Vs  raGo  of  1.78.

3. Event  detecGon.a) 20  samples/second,  bandpass  filtered  at  2-­‐8  Hz.b) Cross-­‐correlaGon  at  lag  increments  of  one

sample.  Measure  similarity  by  correlaGon  sumacross  all  channels  and  staGons  of  the  network.

c) Fixed  correlaGon  sum  threshold  of  4.0  for  the  25selected  channels,  with  a  maximum  of  one  eventevery  4  seconds.

Cross-­‐correlaGon  in  a  mulG-­‐channel  matched  filter.

4. StaGons  and  components  used  forconsistent  detecGon.a) 10  sets  of  staGons  and  channels  used  to  compare

tremor  rates  over  Gme.

5. StaGons  used  for  template  event  locaGons.a) AddiGonal  staGons  to  constrain  the  data.b) Four  temporary  arrays  of  ten  staGons  each.c) Six  individual  three-­‐component  staGons  from  the

PBO  and  HRSN.d) Two  addiGonal  staGons  from  the  Southern

California  Seismic  Network  to  help  locatesouthernmost  event  family.

Key  Points

7

oTremor  has  been  idenGfied  beneath  the  strike  slip  San  Andreas  fault  near  Parkfield,  California.Tremor  locaGons  aligned  with  fault  strike  and  triggering  by  Gdally  induced  right  lateral  shear  stressstrongly  suggest  that  San  Andreas  fault  tremor  is  generated  by  tectonically  driven  slip  on  the  deepfault.Similar  to  subducGon  zone  tremor.Dominant  frequency  content  of  2-­‐8  Hz.

oSeveral  advantages  over  amplitude-­‐based  analysis  of  tremor.Ability  to  idenGfy  short  duraGon  tremor  and  low  amplitude  tremor.Shape  and  Gming  of  waveforms.  Analysis  on  much  shorter  spaGal  and  temporal  scales,  differenGaGngacGvity  on  many  small  patches  of  the  deep  fault.DifferenGates  between  tremor  and  other  seismic  signals  (alershocks,  cultural  noise).No  need  to  exclude  the  Gmes  immediately  following  significant  earthquakes.

oSuggesGon  of  fluid  pressure  in  the  fault  zone.Possible  sources  of  fluid:

1. A  fossil  slab2. Subducted  sediments  proposed  to  exist  near  the  tremor  zone  by  Trehu  et  al.  (1987)3. SerpenGnite  or  fluid-­‐saturated  schist  on  basis  of  receiver  funcGon  study  by  Ozacar  et  al.  (2009)

2.  Tremor-­‐Gde  correlaGons  and  near-­‐lithostaGc  porepressure  on  the  deep  San  Andreas  FaultThomas  et  al.  Nature  (2009)  vol.  462  (7276)  pp.  1048-­‐1051

8Thomas  et  al.  (2009)

Data:  Catalogue  of  1,777  non-­‐volcanic  tremorsdetected  over  an  eight  year  period.  Regionalcatalogue  of  earthquakes  within  0.5°  ofCholame.  Another  catalogue  ofcharacterisGcally  repeaGng  micro-­‐earthquakeslocated  along  the  creeping  segment  of  the  SanAndreas  fault  (NW  of  Cholame)

Results:  Tremor  occurs  preferenGally  whensubjected  to  Gdal  shear  stresses  that  promoteright  lateral  failure  along  San  Andreas  fault.  Lowstress  perturbaGons  from  solid-­‐Earth  Gdes  areresponsible  for  significant  tremor  rate  increasesand  effecGve  normal  stress  in  the  trmor  sourceregion  is  required  to  explain  apparent  Gdaltriggering.  Capable  of  producing  seismicradiaGon.  LithostaGc  pore  fluids  are  present  inthe  tremor  source  region.

Figure  1.  Example  one-­‐day  tremor  Gme  series  withsuperimposed  Gdal  stresses.

9

Thomas  et  al.  (2009)

Figure  2.  Results  of  Chi-­‐square  significance  tests.•Results  correlate  the  tremor,  regional  and  repeaGng  earthquake  catalogues  withGdally  induced  ΔFNS,  ΔRLSS,  ΔCS.•Tremor  start  Gmes  were  separated  into  four  “quadrants”  based  on  magnitude  andrate  of  change  of  the  stresses.•Dashed  line  shows  99%  confidence  level.  Larger  values  indicate  that  the  hypothesisof  random  event  occurrence  can  be  rejected.

10

Thomas  et  al.  (2009)Most  compelling  correlaGon

Figure  3.  Tidal  stress  magnitude  and  rate  distribuGons.

Most  compelling  correlaGon  because  disGnct  increases  intremor  acGvity  that  correspond  to  posiGve  right  lateralshear  stresses  parallel  to  the  San  Andreas  fault  andequally  apparent  decreases  when  values  are  negaGve.

11

Figure  4.  Percentage  of  excess  events  versusfricGon  coefficient.

Above  long  term  average  during  Gmes  of  posiGve  ΔCSversus  effecGve  coefficient  of  fricGon.

σ  =  EffecGve  normal  stress  =  0.035  to  0.009  MPa

τ  =  Shear  stress  =  177  Pa  for  maximum  Gdalinduced  shear  stress

a  =  rate  consGtuGve  parameter  =  0.005  to0.02

Ra  =  (Rmax  –  Rmin)/Ravg

EffecGve  normal  stress  is  found  to  be0.035  to  0.009  Mpa.  These  values  areorders  of  magnitude  lower  than  thelithostaGc  overburden  pressure  at  givendepth,  suggesGng  that  lithostaGc  porefluids  are  present  in  the  tremor  sourceregion.

Methods

12

1. Event  Catalogues• 1,777  events  between  July  2001  to  May

2008  using  detecGon  methodology.• Tidal  stress  Gme  series  were  computed  for

Cholame  only.• RepeaGng  earthquake  catalogue  constains

2,594  events  between  1984  and  1999  onthe  creeping  secGon  of  San  Andreas  fault.

• Regional  earthquake  catalogue  consists  ofall  earthquakes  from  the  Advanced  NaGonalSeismic  System  catalogue  between  July2001  to  May  2008.

2. Tidal  Stress  CalculaGon• Tidally  induced  stresses  in  the  lithosphere

were  computed  using  SPOTL  code  byAgnew.

• Green’s  funcGons  to  compute  azimuthal  andverGcal  strains  that  can  be  converted  tostress.

• Tidal  model  predicGons  were  comparedwith  strainmeter  records  in  Pinon  Flat  usingperturbaGon  matrix.

3. StaGsGcal  Methods• Pearson’s  Chi-­‐square  test  is  designed  to

test  the  similarity  between  two  frequencydistrubiGons  knowns  as  chi-­‐squaredstaGsGc.

4. Uncertainty  esGmates  from  Fig  3.• Two-­‐standard-­‐deviaGon  error  esGmates

are  determined  using  a  bootstrapmethod.

• Randomly  selects  an  individual  tremorfrom  the  original  tremor  catalogue.Repeated  1,777  Gmes  for  random  sample.

• Above  process  repeated  50  Gmes  for  50random  versions  of  the  original  catalogue.

• Tremor  rate  distribuGons  as  a  funcGon  ofthe  Gdally  induced  stresses  are  thencalculated  for  each  of  the  50  randomlysampled  catalogues

• MulGplied  by  2.• Errors  arise  from  variability  in  tremor

detecGon  sensiGvity.

Key  Points

13

1. The  lack  of  correlaGon  in  the  regional  and  repeaGng  earthquake  catalogues  is  not  surprisinggiven  the  size  of  the  catalogues  and  results  from  previous  efforts  to  establish  a  significantGdal  triggering  of  earthquakes.

2. Couloumb  stresses  exhibit  less  correlaGon  than  the  shear  stress  alone.

3. Assuming  a  fricGonal  Coulomb  failure  process  is  an  appropriate  model  for  non-­‐volcanictremor,  the  opGmal  fricGon  coefficient  is  the  value  that  maximizes  the  number  of  eventsthat  occur  during  Gmes  of  encouraged  failure  stress.

4. Tremor  appears  to  be  associated  with  the  presence  of  fluids  at  near-­‐lithostaGc  pressures,and  given  similar  observaGons  in  variable  tectonic  environments,  the  same  mechanism  isprobably  responsible  for  non-­‐volcanic  tremor  elsewhere.

3.  FricGonal  behavior  of  materials  inthe  3D  SAFOD  volume

Carpenter  et  al.  Geophysical  Research  Leeers  (2009)  vol.  36  (5)  pp.  L05302

14

Data:

Outcrop  samples  of  lithologies  thought  to  comprise  andabut  the  fault  zone  at  depth  on  the  basis  of:1.Detailed  geologic  cross  secGons  for  the  SAFOD  sites.2.Published  geologic  maps3.Cuwngs  retrieved  during  SAFOD  Phase  I  and  II  drilling.

Samples  included:1.Natural  serpenGnite  from  New  Idria  Mine2.Granodiorite  from  Phase  I3.Arkosic  metasandstone  from  Phase  I4.Siltstone  from  Phase  II

Results:

1.SerpenGnite  exihibits  low  strength  but  is  not  weak  enoughto  completely  saGsfy  weak  falut  models.2.All  other  samples  are  consistent  with  a  strong  fault  andcrust.3.If  the  fault  is  weak  (μ<0.2)  due  to  the  presence  ofserpenGnite  or  talk,  these  minerals  would  likely  need  toconsGtute  over  50%  by  weight  of  the  shear  zone.

Purpose:  To  resolve  following

1.Debate  concerning  the  absolute  strength  of  the  San  Andreas  Fault.2.Controls  on  the  stability  of  fricGonal  sliding.

Methods

15

1. Rock  and  core  samples  were  crushed  in  a  roll  crusher  in  a  shaeer  box  to  a  grain  size  of<125μm.

2. Samples  were  sheared  in  layers  3  to  5  mm  thick  constructed  by  pouring  and  smoothing  theexperimental  gouge  onto  grooved,  steel  forcing  blocks.

3. Conducted  shearing  experiments  in  double  direct  shear  in  a  true  triaxial  pressure  vessel.Stresses  are  same  and  nominal  contact  area  remains  constant  at  0.025m2.

4. Constant  normal  stress  is  applied  and  maintained  by  a  hydraulic  servo-­‐control  system.5. Displacing  the  center  block  between  the  two  staGonary  side  blocks  causes  shear.6. Forces  are  measured  with  strain  gauge  load  cells  to  ±5  N,  and  verGcal  and  horizontal

displacements  are  measured  by  DCDTs  to  ±0.1μm.7. Recorded  data  conGnuously  at  10kHz  and  averaged  to  values  ranging  from  1  Hz  to  1kHz

depending  on  shearing  velocity.8. Stable  shear  fabric  in  the  layer  was  established  to  aeain  steady-­‐state  fricGon.  Used  5mm

shear  displacement  ‘run  in’  in  the  beginning  of  all  experiments.9. Coulomb  failure  envelope  and  fricGon  consGtuGve  properGes  under  a  variety  of  stresses

using  normal  stress  stepping  procedure.