Lecture 19 Notes: Earthquakes

Earthquakes   12.001   In  our  lab  on  plate  tectonics,  one  of  the  main  datasets  we  drew  on  was  earthquake   epicenter  locations  and  depths.  We  saw  that  both  correlate  strongly  (though  not   exclusively)  with  plate  boundary  locations,  and  inferred  that  most  seismicity  is   associated  somehow  with  plate  motions.  Today  we’ll  examine  the  details  of  that   connection,  and  in  the  process  will  draw  on  our  discussions  of  rheology  and  rock   deformation.     Why  do  EQs  happen?   • Pre-­‐  plate  tectonics,  it  was  recognized  that  EQs  were  associated  with  slip  on   faults,  but  the  ultimate  reason  for  fault  motion  was  not  understood.  Now  we   know  that  most  large  EQs  are  associated  with  relative  plate  motions.   • How  is  plate  motion  accommodated?   o At  surface,  crust  behaves  brittlely   o At  depth,  crust  behaves  ductilely  (recall  brittle-­‐ductile  transition).   Deformation  is  still  localized,  but  occurs  as  creep  rather  than  frictional   sliding.     o B-­‐D  transition  in  continental  crust  ~  15  km,  so  most  EQs  are   shallower  than  this.  But  not  all  –  where  might  deeper  EQs  occur?   Subduction  zones,  where  cold  crust  behaves  brittlely  to  greater  depth.   © Mike Norton. Some rights reserved. License: CC-BY-SA. • •     Most  relative  plate  motion  at  surface  accommodated  on  faults   o Almost  always  multiple  faults   o But  with  a  distribution  in  which  most  motion  occurs  on  a  few  big   faults,  with  the  rest  spread  across  a  larger  number  of  smaller  faults.     o This  means  that,  in  a  time-­‐averaged  sense,  the  major  faults  must   accommodate  cm/yr  of  offset.  Ex)  Relative  motion  between  Pacific  &   NA  plates  ~  50mm/yr.  Time-­‐averaged  offset  on  SAF  ~35mm/yr.   So  why  don’t  faults  just  glide  along  smoothly  and  happily  at  a  few  cm/yr?   o Sometimes  they  do  and  sometimes  they  don’t:  creeping  section  of  SAF,   bounded  by  “locked”  sections.  [PPT:  SAF  map]  What’s  the  difference?   o Faults  are  rough  -­‐-­‐>  friction   1   o Static  and  dynamic  friction  forces  can  differ.  The  first-­‐order   explanation  is  that  creeping  faults  have  fd  >  fs,  whereas  faults  that   exhibit  stick-­‐slip  behavior  have  fd  <  fs.  [DEMO:  friction  block  attached   to  spring]   o What  would  make  fd  >  fs,  such  that  faults  creep?  A  big  research   question.  Probably  lowering  of  static  friction  due  to  reduced  normal   stress  or  increased  fluid  pressure.     What  controls  when  an  EQ  happens  (i.e.,  when  faults  suddenly  unlock)?   © source unknown. All rights reserved. . • As  plates  move  WRT  one  another,  rocks  deform  elastically.     © source unknown. All rights reserved. . • •   This  elastic  deformation  must  be  caused  by  the  buildup  of  stress  (recall   Hooke’s  law).  Eventually,  this  stress  will  exceed  the  frictional  strength  of  a   fault  within  the  plate  boundary  zone,  and  the  fault  will  begin  to  slip.   Slip  occurs  along  a  finite  section  of  the  fault,  releasing  the  energy  stored  in   the  elastic  strain.   © source unknown. All rights reserved. • This  hypothesized  cycle  of  repeated  buildup  of  stress,  accumulation  of   distributed  elastic  strain,  and  rapid  release  of  the  energy  stored  in  that   elastic  strain,  has  been  labeled  the  EQ  cycle.     © source unknown. All rights reserved.   2   o Does  it  really  work  this  way?  Obviously,  not  exactly:  if  it  did,  we’d  be   able  to  predict  EQs.   o There  are  some  places  that  are  close:  we  know  that  some  faults   repeatedly  cause  EQs  with  a  characteristic  size  and  recurrence   interval.  Ex)  SAF  near  Parkfield,  site  of  drilling  project.     o This  is  how  most  probabilistic  attempts  at  EQ  hazard  forecasting   work.  [PPT:  USGS  “forecast”  for  SF  Bay  Area]   o But  in  reality,  there  are  several  factors  that  prevent  us  from  predicting   EQs  this  way:   § Change  in  rate  of  stress  accumulation  (variations  in  plate   motion)  –  add  a  curved  line  to  the  sketch   § Incomplete  stress  release   § Change  in  fault  strength  (rock  WX,  water  pressure,  “damage”   due  to  previous  EQs)   § Changes  in  stress  state  due  to  nearby  EQs   § The  actual  EQ  time  series  often  looks  like  this:   © source unknown. All rights reserved. .   What  happens  during  an  EQ?     • Fault  rupture  nucleates  at  a  point  (focus  or  hypocenter).  These  are  the  first   seismic  waves  emitted.   • Rupture  propagates  outward  along  the  fault  surface  at  a  few  km/s  (~  speed   of  sound  in  rock),  continuing  to  emit  seismic  waves.  (Note,  though,  that  the   majority  of  the  energy  released  usually  goes  into  frictionally  generated  heat.)     • In  large  EQs,  rupture  propagation  lasts  seconds  to  over  a  minute,  rupture   length  can  be  100s  of  km,  and  total  slip  can  be  meters.   © source unknown. All rights reserved. . Rupture  stops  when  stress  drops  below  fault  strength,  or  when  it  encounters   material  that  behaves  ductilely  rather  than  brittlely.   • Why  doesn’t  rupture  stop  propagating  immediately  after  it  starts,  dropping   stress  back  just  below  the  fault  strength?  Again,  the  first-­‐order  answer  is   because  fd  <  fs.  Otherwise  it  would  be  a  creeping  fault.     What  happens  after  an  EQ?   3   •   • • Aftershocks   o Smaller  than  mainshock   o Sections  on  fault  that  didn’t  slip,  or  didn’t  completely  relieve  stress,  or   where  stress  increased.     o How  can  an  EQ  increase  stress?  Release  of  elastic  strain  alters  stress   on  surrounding  faults  (example  later).  There  is  evidence  that  this  can   trigger  large  nearby  EQs  in  addition  to  small  aftershocks   Seismic  waves  propagate…we’ll  discuss  this  shortly.     How  do  we  describe  EQ  size,  sense  of  slip  and  stress  release,  etc?   • Size   o Richter  Scale  (1930s)   § Log10  scale  of  amplitude  of  surface  ground  motion  (each   Richter  unit  is  10x  the  ground  motion)   § Each  Richter  unit  corresponds  to  a  ~32x  (=101.5)  increase  in   energy   § Mostly  obsolete   o Moment  Magnitude   § Seismic  moment,  Mo   • Proportional  to  rupture  area  A,  displacement  D,  and   shear  modulus  (aka  rigidity)  µ:  Mo  =  µAD.  Rigidity  has   units  of  stress  (typically  tens  of  GPa  for  rocks),  so   moment  has  units  of  force*displacement,  or  work.   • Mo  is  determined  from  modeling  of  seismograms   § Moment  magnitude,  Mw  =  ⅔  log10  Mo  –  10.73   • The  w  subscript  stands  for  work   • Mo  is  in  dyne  cm  (=  10-­‐7  J)   • Constants  chosen  so  Mw  values  correspond   approximately  to  old  Richter  scale.  As  with  Richter,  each   Mw  unit  corresponds  to  101.5  x  more  energy   • Ex)  Solve  for  Mo  to  get  expression  for  relative  energy   release:  Mo  =  103/2(Mw  +  10.73).  So  ratio  of  Mo  for  two   quakes  is  103/2(Mw1  –  Mw2).     o For  Japan  2011  (Mw9.0)  vs.  Haiti  2010  (Mw7.0),   the  ratio  is  103/2*2=1000   o For  Chile  1960  (Mw9.5)  vs.  Japan  2011  (Mw9.0)   the  ratio  is  103/2*0.5=5.6   § Use  of  logarithmic  scales  is  efficient  for  science,  bad  for  PR   (M9.5  doesn’t  sound  that  much  worse  than  a  M7)   o Mercalli  intensity   § Written  in  Roman  numeral   § Qualitative  measure  of  severity  based  on  observed  effects  on   humans  &  structures   § Source  of  estimates  of  EQ  location  &  magnitude  prior  to   seismometers     4   • Sense  of  slip   o Slip  on  a  finite  rupture  creates  zones  of  compression  and  zones  of   dilatation   o This  pattern  will  dictate  the  sense  of  first  motion  at  seismic  stations  in   the  different  zones.   o The  fault  displacement  in  an  EQ  is  constrained  by  examining  the   seismograms  (including,  but  not  limited  to,  first  arrivals)  at  multiple   stations,  and  determining  the  best-­‐fit  solution  (“focal  mechanism   solution”)   © source unknown. All rights reserved. Focal Mechanisms:   geometry of faulting o To  describe  the  pattern   of  energy   radiated  during during  aan n  Eearthquake Q  (which  is  a   tensor  called  the  “moment  tensor  solution”),  we  use  stereonets   (“beachballs”)   § Take  the  intersection  of  the  fault  with  a  lower  hemispherical   surface  centered  on  the  focus   § Mark  the  quadrants  of  compression  (colored)  and  dilatation   § Project  the  result  onto  a  horizontal  plane   Focal Mechanisms: geometry of faulting during an earthquake © source unknown. All rights reserved.   Note  that  there  is  an  ambiguity  in  the  sense  of  slip.  This  is  not   just  a  limitation  of  the  stereonet;  it  is  a  real  limitation  of  our   ability  to  determine  the  fault  motion.   Examples  from  recent  quakes:   o Sumatra  2004   § Mw  =  9.0   § [PPT:  Moment  tensor]   5   § •   Dimensions:  ~200km  in  dip  dir,  1500  km  in  strike  dir  (about   the  size  of  California),  max  ~20m  slip   o Haiti  2010   § Mw  =  7.0   § [PPT:  Moment  tensor]   § Dimensions:  ~15km  x  40  km,  max  ~5  m  /  avg  2m  slip   § Enriquillo fault last ruptured on October 18 and November 21, 1751, and June 3, 1770   § [PPT:  Stress  change  on  nearby  Septentrional  fault]   o Japan  2011   § Mw  =  9.0   § [PPT:  Tectonic  context]   § [PPT:  Moment  tensor  +  depth  profile]   § [PPT:  Finite  fault  model].  Yes,  up  to  18  m  of  slip!   § Dimensions:  ~150km  x  500  km,  max  ~18  m  slip   o Chile  1960   § Mw  =  9.5  (largest  EQ  recorded  w/modern  instruments)   § Dimensions:  1000  km  along  strike,  max  ~40m  slip   § Mw8.2  foreshock(!)  25  hrs  before  main  shock,  aftershocks  as   large  as  Mw7.9!   §   Seismic  waves:  2  classes     • Body  waves   o Compressional  waves  (P-­‐waves)   § Particle  displacement  parallel  to  propagation  direction   § Faster  (P=  “primary”).  Typical  crustal  VP  ~  6  km/s   § VP  =  [(K  +  4/3µ)/ρ]1/2,  where  µ  is  shear  modulus  (“rigidity”,   ratio  of  shear  stress  to  shear  strain),  K  is  bulk  modulus   (“incompressibility”,  ratio  of  pressure  change  to  volume   change),  and  ρ  is  density.     § In  liquid,  µà0,  and  VP  =  [K/ρ]1/2   o Shear  waves  (S-­‐waves)   § Particle  displacement  normal  to  propagation  direction   § Slower  (S  =  “secondary”).  Typical  crustal  VS  ~3-­‐4  km/s   § VS  =  (µ/ρ)1/2.     § In  liquid,  µà0,  and  VS  =  0   • Surface  waves   o Slower  than  body  waves  (~3  km/s),  but  more  destructive   o [PPT:  animations  of  surface  waves]   o Rayleigh  waves   § “Rolling”  motion   § Particle  paths  are  elliptical  in  plane  normal  to  surface  and   parallel  to  propagation  direction   o Love  waves   § Sideways  shaking  motion     6   Particle  displacements  are  in  the  surface  plane,  normal  to   propagation  direction.  Essentially  a  horizontally  polarized   shear  wave  guided  by  the  surface.   § Slightly  faster  than  Rayleigh  waves   § [PPT:  Love  waves  in  1995  Kobe  EQ]   Why  are  there  variations  in  K,  µ,  ρ  that  influence  seismic  wave  speeds?   o Mineral  composition   o Mineral  phase  (even  for  same  elemental  composition,  crystal   structure  changes  with  P  and  T)   o State  of  matter:  Solid,  liquid  (melt),  or  a  combination   o Fabric:  in  this  case,  orientation  of  crystals   o Stress  state  (you  will  explore  this  effect  in  your  lab!)   o These  are  all  things  that  seismologists  try  to  infer  about  the  Earth,   based  on  their  observations  of  wave  propagation   Wave  propagation  and  arrival  patterns   o Surface  waves  travel  along  the  surface.  What  path  will  body  waves   follow  through  Earth’s  interior?   § [Start  drawing  hypothetical  ray  paths  on  board]  à  Depends  on   how  wave  speed  varies  with  depth,  because  changes  in  speed   will  cause  refraction.   § • • © source unknown. All rights reserved.   What  controls  wave  speed?  Moduli  and  density.   How  does  density  vary  with  depth?  Increases  à  slower  speed!   What  kind  of  ray  paths  would  this  predict?     § How  do  moduli  vary  with  depth?  Not  as  easy  to  intuit…you  will   discover  this  in  your  lab.   o Typical  seismogram:   § § © source unknown. All rights reserved. . •     Interior  Earth  structure  worked  out  by  seismologists  in  early  20th  c.:  crust,   mantle,  outer  &  inner  core.  How?   o Velocity  of  a  medium  +  path  of  wave  -­‐-­‐>  travel  time   7   o Differences  in  velocity  between  layers  cause  refraction,  which  also   influences  when  (or  even  whether  or  not)  a  wave  arrives  at  a  given   location   § Ex)  Liquid  outer  core  inferred  from  “shadow  zone”  caused  by   gap  between  wave  that  just  misses  outer  core,  and  refraction  of   wave  that  hits  it.   o Differences  in  velocity  can  also  cause  reflection   § Ex)  Inner  core  was  originally  suggested  because  there  were   actually  some  weak  P-­‐wave  arrivals  in  shadow  zones,  which   could  be  explained  by  a  solid  reflector  inside  the  core.  (Solid   inner  core  was  later  demonstrated  based  on  free  oscillations.)     Other  earthquake-­‐related  topics,  time  permitting   • Change  in  MOI  due  to  EQs  (especially  when  dense  subducting  slabs  move   deeper  into  the  mantle),  TPW,  and  length-­‐of-­‐day   • Earthquake  hazards  mitigation   o Difference  between  impact  of  Haiti  2010  and  Japan  2011  EQs   o Use  this  to  illustrate  the  difference  between  hazard  and  risk       8